blob: 4aed8fc23d328592f8cf267fca70582f62fc6a3e [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;
Tom Herbertba418fa2013-01-22 09:50:32 +0000146 kuid_t uid = sock_i_uid(sk);
Gerrit Renker25030a72006-08-26 20:06:05 -0700147
Eric Dumazetca065d02016-04-01 08:52:13 -0700148 sk_for_each(sk2, &hslot->head) {
Joe Perches9d4fb272009-11-23 10:41:23 -0800149 if (net_eq(sock_net(sk2), net) &&
150 sk2 != sk &&
Eric Dumazetd4cada42009-11-08 10:17:30 +0000151 (bitmap || udp_sk(sk2)->udp_port_hash == num) &&
Joe Perches9d4fb272009-11-23 10:41:23 -0800152 (!sk2->sk_reuse || !sk->sk_reuse) &&
153 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
154 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000155 (!sk2->sk_reuseport || !sk->sk_reuseport ||
Craig Galleke32ea7e2016-01-04 17:41:46 -0500156 rcu_access_pointer(sk->sk_reuseport_cb) ||
Joe Perches4243cdc2014-11-11 21:59:20 -0800157 !uid_eq(uid, sock_i_uid(sk2))) &&
Craig Galleke32ea7e2016-01-04 17:41:46 -0500158 saddr_comp(sk, sk2, true)) {
Joe Perches4243cdc2014-11-11 21:59:20 -0800159 if (!bitmap)
Eric Dumazet98322f22009-01-26 21:35:35 -0800160 return 1;
Joe Perches4243cdc2014-11-11 21:59:20 -0800161 __set_bit(udp_sk(sk2)->udp_port_hash >> log, bitmap);
Eric Dumazet98322f22009-01-26 21:35:35 -0800162 }
Joe Perches4243cdc2014-11-11 21:59:20 -0800163 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700164 return 0;
165}
166
Eric Dumazet30fff922009-11-09 05:26:33 +0000167/*
168 * Note: we still hold spinlock of primary hash chain, so no other writer
169 * can insert/delete a socket with local_port == num
170 */
171static int udp_lib_lport_inuse2(struct net *net, __u16 num,
Joe Perches4243cdc2014-11-11 21:59:20 -0800172 struct udp_hslot *hslot2,
173 struct sock *sk,
174 int (*saddr_comp)(const struct sock *sk1,
Craig Galleke32ea7e2016-01-04 17:41:46 -0500175 const struct sock *sk2,
176 bool match_wildcard))
Eric Dumazet30fff922009-11-09 05:26:33 +0000177{
178 struct sock *sk2;
Tom Herbertba418fa2013-01-22 09:50:32 +0000179 kuid_t uid = sock_i_uid(sk);
Eric Dumazet30fff922009-11-09 05:26:33 +0000180 int res = 0;
181
182 spin_lock(&hslot2->lock);
Eric Dumazetca065d02016-04-01 08:52:13 -0700183 udp_portaddr_for_each_entry(sk2, &hslot2->head) {
Joe Perches9d4fb272009-11-23 10:41:23 -0800184 if (net_eq(sock_net(sk2), net) &&
185 sk2 != sk &&
186 (udp_sk(sk2)->udp_port_hash == num) &&
187 (!sk2->sk_reuse || !sk->sk_reuse) &&
188 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
189 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000190 (!sk2->sk_reuseport || !sk->sk_reuseport ||
Craig Galleke32ea7e2016-01-04 17:41:46 -0500191 rcu_access_pointer(sk->sk_reuseport_cb) ||
Joe Perches4243cdc2014-11-11 21:59:20 -0800192 !uid_eq(uid, sock_i_uid(sk2))) &&
Craig Galleke32ea7e2016-01-04 17:41:46 -0500193 saddr_comp(sk, sk2, true)) {
Eric Dumazet30fff922009-11-09 05:26:33 +0000194 res = 1;
195 break;
196 }
Joe Perches4243cdc2014-11-11 21:59:20 -0800197 }
Eric Dumazet30fff922009-11-09 05:26:33 +0000198 spin_unlock(&hslot2->lock);
199 return res;
200}
201
Craig Galleke32ea7e2016-01-04 17:41:46 -0500202static int udp_reuseport_add_sock(struct sock *sk, struct udp_hslot *hslot,
203 int (*saddr_same)(const struct sock *sk1,
204 const struct sock *sk2,
205 bool match_wildcard))
206{
207 struct net *net = sock_net(sk);
Craig Galleke32ea7e2016-01-04 17:41:46 -0500208 kuid_t uid = sock_i_uid(sk);
209 struct sock *sk2;
210
Eric Dumazetca065d02016-04-01 08:52:13 -0700211 sk_for_each(sk2, &hslot->head) {
Craig Galleke32ea7e2016-01-04 17:41:46 -0500212 if (net_eq(sock_net(sk2), net) &&
213 sk2 != sk &&
214 sk2->sk_family == sk->sk_family &&
215 ipv6_only_sock(sk2) == ipv6_only_sock(sk) &&
216 (udp_sk(sk2)->udp_port_hash == udp_sk(sk)->udp_port_hash) &&
217 (sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
218 sk2->sk_reuseport && uid_eq(uid, sock_i_uid(sk2)) &&
219 (*saddr_same)(sk, sk2, false)) {
220 return reuseport_add_sock(sk, sk2);
221 }
222 }
223
224 /* Initial allocation may have already happened via setsockopt */
225 if (!rcu_access_pointer(sk->sk_reuseport_cb))
226 return reuseport_alloc(sk);
227 return 0;
228}
229
Gerrit Renker25030a72006-08-26 20:06:05 -0700230/**
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700231 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
Gerrit Renker25030a72006-08-26 20:06:05 -0700232 *
233 * @sk: socket struct in question
234 * @snum: port number to look up
David S. Millerdf2bc452007-06-05 15:18:43 -0700235 * @saddr_comp: AF-dependent comparison of bound local IP addresses
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300236 * @hash2_nulladdr: AF-dependent hash value in secondary hash chains,
Eric Dumazet30fff922009-11-09 05:26:33 +0000237 * with NULL address
Gerrit Renker25030a72006-08-26 20:06:05 -0700238 */
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700239int udp_lib_get_port(struct sock *sk, unsigned short snum,
Joe Perches4243cdc2014-11-11 21:59:20 -0800240 int (*saddr_comp)(const struct sock *sk1,
Craig Galleke32ea7e2016-01-04 17:41:46 -0500241 const struct sock *sk2,
242 bool match_wildcard),
Eric Dumazet30fff922009-11-09 05:26:33 +0000243 unsigned int hash2_nulladdr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244{
Eric Dumazet512615b2009-11-08 10:17:58 +0000245 struct udp_hslot *hslot, *hslot2;
Eric Dumazet645ca702008-10-29 01:41:45 -0700246 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Gerrit Renker25030a72006-08-26 20:06:05 -0700247 int error = 1;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900248 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700250 if (!snum) {
Eric Dumazet9088c562008-10-08 11:44:17 -0700251 int low, high, remaining;
Eric Dumazet95c96172012-04-15 05:58:06 +0000252 unsigned int rand;
Eric Dumazet98322f22009-01-26 21:35:35 -0800253 unsigned short first, last;
254 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -0700256 inet_get_local_port_range(net, &low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700257 remaining = (high - low) + 1;
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700258
Aruna-Hewapathirane63862b52014-01-11 07:15:59 -0500259 rand = prandom_u32();
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200260 first = reciprocal_scale(rand, remaining) + low;
Eric Dumazet98322f22009-01-26 21:35:35 -0800261 /*
262 * force rand to be an odd multiple of UDP_HTABLE_SIZE
263 */
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000264 rand = (rand | 1) * (udptable->mask + 1);
Eric Dumazet5781b232009-12-13 19:32:39 -0800265 last = first + udptable->mask + 1;
266 do {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000267 hslot = udp_hashslot(udptable, net, first);
Eric Dumazet98322f22009-01-26 21:35:35 -0800268 bitmap_zero(bitmap, PORTS_PER_CHAIN);
Eric Dumazet645ca702008-10-29 01:41:45 -0700269 spin_lock_bh(&hslot->lock);
Eric Dumazet98322f22009-01-26 21:35:35 -0800270 udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000271 saddr_comp, udptable->log);
Eric Dumazet98322f22009-01-26 21:35:35 -0800272
273 snum = first;
274 /*
275 * Iterate on all possible values of snum for this hash.
276 * Using steps of an odd multiple of UDP_HTABLE_SIZE
277 * give us randomization and full range coverage.
278 */
Eric Dumazet9088c562008-10-08 11:44:17 -0700279 do {
Eric Dumazet98322f22009-01-26 21:35:35 -0800280 if (low <= snum && snum <= high &&
Amerigo Wange3826f12010-05-05 00:27:06 +0000281 !test_bit(snum >> udptable->log, bitmap) &&
WANG Cong122ff242014-05-12 16:04:53 -0700282 !inet_is_local_reserved_port(net, snum))
Eric Dumazet98322f22009-01-26 21:35:35 -0800283 goto found;
284 snum += rand;
285 } while (snum != first);
286 spin_unlock_bh(&hslot->lock);
Eric Dumazet5781b232009-12-13 19:32:39 -0800287 } while (++first != last);
Eric Dumazet98322f22009-01-26 21:35:35 -0800288 goto fail;
Eric Dumazet645ca702008-10-29 01:41:45 -0700289 } else {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000290 hslot = udp_hashslot(udptable, net, snum);
Eric Dumazet645ca702008-10-29 01:41:45 -0700291 spin_lock_bh(&hslot->lock);
Eric Dumazet30fff922009-11-09 05:26:33 +0000292 if (hslot->count > 10) {
293 int exist;
294 unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum;
295
296 slot2 &= udptable->mask;
297 hash2_nulladdr &= udptable->mask;
298
299 hslot2 = udp_hashslot2(udptable, slot2);
300 if (hslot->count < hslot2->count)
301 goto scan_primary_hash;
302
303 exist = udp_lib_lport_inuse2(net, snum, hslot2,
304 sk, saddr_comp);
305 if (!exist && (hash2_nulladdr != slot2)) {
306 hslot2 = udp_hashslot2(udptable, hash2_nulladdr);
307 exist = udp_lib_lport_inuse2(net, snum, hslot2,
308 sk, saddr_comp);
309 }
310 if (exist)
311 goto fail_unlock;
312 else
313 goto found;
314 }
315scan_primary_hash:
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000316 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk,
317 saddr_comp, 0))
Eric Dumazet645ca702008-10-29 01:41:45 -0700318 goto fail_unlock;
319 }
Eric Dumazet98322f22009-01-26 21:35:35 -0800320found:
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000321 inet_sk(sk)->inet_num = snum;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000322 udp_sk(sk)->udp_port_hash = snum;
323 udp_sk(sk)->udp_portaddr_hash ^= snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 if (sk_unhashed(sk)) {
Craig Galleke32ea7e2016-01-04 17:41:46 -0500325 if (sk->sk_reuseport &&
326 udp_reuseport_add_sock(sk, hslot, saddr_comp)) {
327 inet_sk(sk)->inet_num = 0;
328 udp_sk(sk)->udp_port_hash = 0;
329 udp_sk(sk)->udp_portaddr_hash ^= snum;
330 goto fail_unlock;
331 }
332
Eric Dumazetca065d02016-04-01 08:52:13 -0700333 sk_add_node_rcu(sk, &hslot->head);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +0000334 hslot->count++;
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700335 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet512615b2009-11-08 10:17:58 +0000336
337 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
338 spin_lock(&hslot2->lock);
Craig Gallekd894ba12016-04-12 13:11:25 -0400339 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
David S. Miller1602f492016-04-23 18:26:24 -0400340 sk->sk_family == AF_INET6)
341 hlist_add_tail_rcu(&udp_sk(sk)->udp_portaddr_node,
342 &hslot2->head);
Craig Gallekd894ba12016-04-12 13:11:25 -0400343 else
David S. Miller1602f492016-04-23 18:26:24 -0400344 hlist_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
345 &hslot2->head);
Eric Dumazet512615b2009-11-08 10:17:58 +0000346 hslot2->count++;
347 spin_unlock(&hslot2->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 }
Eric Dumazetca065d02016-04-01 08:52:13 -0700349 sock_set_flag(sk, SOCK_RCU_FREE);
Gerrit Renker25030a72006-08-26 20:06:05 -0700350 error = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -0700351fail_unlock:
352 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353fail:
Gerrit Renker25030a72006-08-26 20:06:05 -0700354 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355}
Eric Dumazetc482c562009-07-17 00:26:32 +0000356EXPORT_SYMBOL(udp_lib_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357
Craig Galleke32ea7e2016-01-04 17:41:46 -0500358/* match_wildcard == true: 0.0.0.0 equals to any IPv4 addresses
359 * match_wildcard == false: addresses must be exactly the same, i.e.
360 * 0.0.0.0 only equals to 0.0.0.0
361 */
Craig Gallekc125e802016-02-10 11:50:40 -0500362int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2,
363 bool match_wildcard)
David S. Millerdb8dac22008-03-06 16:22:02 -0800364{
365 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
366
Craig Galleke32ea7e2016-01-04 17:41:46 -0500367 if (!ipv6_only_sock(sk2)) {
368 if (inet1->inet_rcv_saddr == inet2->inet_rcv_saddr)
369 return 1;
370 if (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr)
371 return match_wildcard;
372 }
373 return 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800374}
375
Eric Dumazet6eada012015-03-18 14:05:33 -0700376static u32 udp4_portaddr_hash(const struct net *net, __be32 saddr,
377 unsigned int port)
Eric Dumazetd4cada42009-11-08 10:17:30 +0000378{
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700379 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000380}
381
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700382int udp_v4_get_port(struct sock *sk, unsigned short snum)
David S. Millerdb8dac22008-03-06 16:22:02 -0800383{
Eric Dumazet30fff922009-11-09 05:26:33 +0000384 unsigned int hash2_nulladdr =
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700385 udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum);
Eric Dumazet30fff922009-11-09 05:26:33 +0000386 unsigned int hash2_partial =
387 udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0);
388
Eric Dumazetd4cada42009-11-08 10:17:30 +0000389 /* precompute partial secondary hash */
Eric Dumazet30fff922009-11-09 05:26:33 +0000390 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
391 return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr);
David S. Millerdb8dac22008-03-06 16:22:02 -0800392}
393
Su, Xuemind1e37282016-06-13 11:02:50 +0800394static int compute_score(struct sock *sk, struct net *net,
395 __be32 saddr, __be16 sport,
396 __be32 daddr, unsigned short hnum, int dif)
Eric Dumazet645ca702008-10-29 01:41:45 -0700397{
Joe Perches60c04ae2014-12-01 20:29:06 -0800398 int score;
399 struct inet_sock *inet;
Eric Dumazet645ca702008-10-29 01:41:45 -0700400
Joe Perches60c04ae2014-12-01 20:29:06 -0800401 if (!net_eq(sock_net(sk), net) ||
402 udp_sk(sk)->udp_port_hash != hnum ||
403 ipv6_only_sock(sk))
404 return -1;
Eric Dumazet645ca702008-10-29 01:41:45 -0700405
Joe Perches60c04ae2014-12-01 20:29:06 -0800406 score = (sk->sk_family == PF_INET) ? 2 : 1;
407 inet = inet_sk(sk);
408
409 if (inet->inet_rcv_saddr) {
410 if (inet->inet_rcv_saddr != daddr)
411 return -1;
412 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700413 }
Joe Perches60c04ae2014-12-01 20:29:06 -0800414
415 if (inet->inet_daddr) {
416 if (inet->inet_daddr != saddr)
417 return -1;
418 score += 4;
419 }
420
421 if (inet->inet_dport) {
422 if (inet->inet_dport != sport)
423 return -1;
424 score += 4;
425 }
426
427 if (sk->sk_bound_dev_if) {
428 if (sk->sk_bound_dev_if != dif)
429 return -1;
430 score += 4;
431 }
Eric Dumazet70da2682015-10-08 19:33:21 -0700432 if (sk->sk_incoming_cpu == raw_smp_processor_id())
433 score++;
Eric Dumazet645ca702008-10-29 01:41:45 -0700434 return score;
435}
436
Eric Dumazet6eada012015-03-18 14:05:33 -0700437static u32 udp_ehashfn(const struct net *net, const __be32 laddr,
438 const __u16 lport, const __be32 faddr,
439 const __be16 fport)
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200440{
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200441 static u32 udp_ehash_secret __read_mostly;
442
443 net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret));
444
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200445 return __inet_ehashfn(laddr, lport, faddr, fport,
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200446 udp_ehash_secret + net_hash_mix(net));
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200447}
448
Su, Xuemind1e37282016-06-13 11:02:50 +0800449/* called with rcu_read_lock() */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000450static struct sock *udp4_lib_lookup2(struct net *net,
451 __be32 saddr, __be16 sport,
452 __be32 daddr, unsigned int hnum, int dif,
Su, Xuemind1e37282016-06-13 11:02:50 +0800453 struct udp_hslot *hslot2,
Craig Gallek11341582016-01-05 15:08:07 -0500454 struct sk_buff *skb)
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000455{
456 struct sock *sk, *result;
Tom Herbertba418fa2013-01-22 09:50:32 +0000457 int score, badness, matches = 0, reuseport = 0;
458 u32 hash = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000459
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000460 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000461 badness = 0;
Eric Dumazetca065d02016-04-01 08:52:13 -0700462 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
Su, Xuemind1e37282016-06-13 11:02:50 +0800463 score = compute_score(sk, net, saddr, sport,
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000464 daddr, hnum, dif);
465 if (score > badness) {
Tom Herbertba418fa2013-01-22 09:50:32 +0000466 reuseport = sk->sk_reuseport;
467 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200468 hash = udp_ehashfn(net, daddr, hnum,
469 saddr, sport);
Eric Dumazetca065d02016-04-01 08:52:13 -0700470 result = reuseport_select_sock(sk, hash, skb,
Eric Dumazeted0dfff2016-01-19 08:36:43 -0800471 sizeof(struct udphdr));
Eric Dumazetca065d02016-04-01 08:52:13 -0700472 if (result)
473 return result;
Tom Herbertba418fa2013-01-22 09:50:32 +0000474 matches = 1;
475 }
Eric Dumazetca065d02016-04-01 08:52:13 -0700476 badness = score;
477 result = sk;
Tom Herbertba418fa2013-01-22 09:50:32 +0000478 } else if (score == badness && reuseport) {
479 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200480 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000481 result = sk;
482 hash = next_pseudo_random32(hash);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000483 }
484 }
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000485 return result;
486}
487
David S. Millerdb8dac22008-03-06 16:22:02 -0800488/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
489 * harder than this. -DaveM
490 */
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000491struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -0800492 __be16 sport, __be32 daddr, __be16 dport,
Craig Gallek538950a2016-01-04 17:41:47 -0500493 int dif, struct udp_table *udptable, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800494{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700495 struct sock *sk, *result;
David S. Millerdb8dac22008-03-06 16:22:02 -0800496 unsigned short hnum = ntohs(dport);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000497 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
498 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
Tom Herbertba418fa2013-01-22 09:50:32 +0000499 int score, badness, matches = 0, reuseport = 0;
500 u32 hash = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800501
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000502 if (hslot->count > 10) {
503 hash2 = udp4_portaddr_hash(net, daddr, hnum);
504 slot2 = hash2 & udptable->mask;
505 hslot2 = &udptable->hash2[slot2];
506 if (hslot->count < hslot2->count)
507 goto begin;
508
509 result = udp4_lib_lookup2(net, saddr, sport,
510 daddr, hnum, dif,
Su, Xuemind1e37282016-06-13 11:02:50 +0800511 hslot2, skb);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000512 if (!result) {
Su, Xuemind1e37282016-06-13 11:02:50 +0800513 unsigned int old_slot2 = slot2;
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700514 hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000515 slot2 = hash2 & udptable->mask;
Su, Xuemind1e37282016-06-13 11:02:50 +0800516 /* avoid searching the same slot again. */
517 if (unlikely(slot2 == old_slot2))
518 return result;
519
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000520 hslot2 = &udptable->hash2[slot2];
521 if (hslot->count < hslot2->count)
522 goto begin;
523
Jorge Boncompte [DTI2]1223c672010-04-08 04:56:48 +0000524 result = udp4_lib_lookup2(net, saddr, sport,
Su, Xuemind1e37282016-06-13 11:02:50 +0800525 daddr, hnum, dif,
526 hslot2, skb);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000527 }
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000528 return result;
529 }
Eric Dumazet271b72c2008-10-29 02:11:14 -0700530begin:
531 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000532 badness = 0;
Eric Dumazetca065d02016-04-01 08:52:13 -0700533 sk_for_each_rcu(sk, &hslot->head) {
Su, Xuemind1e37282016-06-13 11:02:50 +0800534 score = compute_score(sk, net, saddr, sport,
535 daddr, hnum, dif);
Eric Dumazet645ca702008-10-29 01:41:45 -0700536 if (score > badness) {
Tom Herbertba418fa2013-01-22 09:50:32 +0000537 reuseport = sk->sk_reuseport;
538 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200539 hash = udp_ehashfn(net, daddr, hnum,
540 saddr, sport);
Eric Dumazetca065d02016-04-01 08:52:13 -0700541 result = reuseport_select_sock(sk, hash, skb,
Craig Gallek538950a2016-01-04 17:41:47 -0500542 sizeof(struct udphdr));
Eric Dumazetca065d02016-04-01 08:52:13 -0700543 if (result)
544 return result;
Tom Herbertba418fa2013-01-22 09:50:32 +0000545 matches = 1;
546 }
Eric Dumazetca065d02016-04-01 08:52:13 -0700547 result = sk;
548 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000549 } else if (score == badness && reuseport) {
550 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200551 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000552 result = sk;
553 hash = next_pseudo_random32(hash);
David S. Millerdb8dac22008-03-06 16:22:02 -0800554 }
555 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800556 return result;
557}
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000558EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
David S. Millerdb8dac22008-03-06 16:22:02 -0800559
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700560static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
561 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700562 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700563{
564 const struct iphdr *iph = ip_hdr(skb);
565
Alexander Duycked7cbbc2016-05-12 16:23:44 -0700566 return __udp4_lib_lookup(dev_net(skb->dev), iph->saddr, sport,
Eric Dumazet8afdd992013-12-10 18:07:23 -0800567 iph->daddr, dport, inet_iif(skb),
Craig Gallek538950a2016-01-04 17:41:47 -0500568 udptable, skb);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700569}
570
Tom Herbert63058302016-04-05 08:22:50 -0700571struct sock *udp4_lib_lookup_skb(struct sk_buff *skb,
572 __be16 sport, __be16 dport)
573{
Alexander Duycked7cbbc2016-05-12 16:23:44 -0700574 return __udp4_lib_lookup_skb(skb, sport, dport, &udp_table);
Tom Herbert63058302016-04-05 08:22:50 -0700575}
576EXPORT_SYMBOL_GPL(udp4_lib_lookup_skb);
577
Eric Dumazetca065d02016-04-01 08:52:13 -0700578/* Must be called under rcu_read_lock().
579 * Does increment socket refcount.
580 */
581#if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \
582 IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY)
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700583struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
584 __be32 daddr, __be16 dport, int dif)
585{
Eric Dumazetca065d02016-04-01 08:52:13 -0700586 struct sock *sk;
587
588 sk = __udp4_lib_lookup(net, saddr, sport, daddr, dport,
589 dif, &udp_table, NULL);
590 if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
591 sk = NULL;
592 return sk;
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700593}
594EXPORT_SYMBOL_GPL(udp4_lib_lookup);
Eric Dumazetca065d02016-04-01 08:52:13 -0700595#endif
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700596
Shawn Bohrer421b3882013-10-07 11:01:39 -0500597static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk,
598 __be16 loc_port, __be32 loc_addr,
599 __be16 rmt_port, __be32 rmt_addr,
600 int dif, unsigned short hnum)
601{
602 struct inet_sock *inet = inet_sk(sk);
603
604 if (!net_eq(sock_net(sk), net) ||
605 udp_sk(sk)->udp_port_hash != hnum ||
606 (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
607 (inet->inet_dport != rmt_port && inet->inet_dport) ||
608 (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) ||
609 ipv6_only_sock(sk) ||
610 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
611 return false;
612 if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif))
613 return false;
614 return true;
615}
616
David S. Millerdb8dac22008-03-06 16:22:02 -0800617/*
618 * This routine is called by the ICMP module when it gets some
619 * sort of error condition. If err < 0 then the socket should
620 * be closed and the error returned to the user. If err > 0
621 * it's just the icmp type << 8 | icmp code.
622 * Header points to the ip header of the error packet. We move
623 * on past this. Then (as it used to claim before adjustment)
624 * header points to the first 8 bytes of the udp header. We need
625 * to find the appropriate port.
626 */
627
Eric Dumazet645ca702008-10-29 01:41:45 -0700628void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800629{
630 struct inet_sock *inet;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000631 const struct iphdr *iph = (const struct iphdr *)skb->data;
Eric Dumazetc482c562009-07-17 00:26:32 +0000632 struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
David S. Millerdb8dac22008-03-06 16:22:02 -0800633 const int type = icmp_hdr(skb)->type;
634 const int code = icmp_hdr(skb)->code;
635 struct sock *sk;
636 int harderr;
637 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700638 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800639
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700640 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
Craig Gallek538950a2016-01-04 17:41:47 -0500641 iph->saddr, uh->source, skb->dev->ifindex, udptable,
642 NULL);
Ian Morris51456b22015-04-03 09:17:26 +0100643 if (!sk) {
Eric Dumazet5d3848b2016-04-27 16:44:29 -0700644 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800645 return; /* No socket for error */
646 }
647
648 err = 0;
649 harderr = 0;
650 inet = inet_sk(sk);
651
652 switch (type) {
653 default:
654 case ICMP_TIME_EXCEEDED:
655 err = EHOSTUNREACH;
656 break;
657 case ICMP_SOURCE_QUENCH:
658 goto out;
659 case ICMP_PARAMETERPROB:
660 err = EPROTO;
661 harderr = 1;
662 break;
663 case ICMP_DEST_UNREACH:
664 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
David S. Miller36393392012-06-14 22:21:46 -0700665 ipv4_sk_update_pmtu(skb, sk, info);
David S. Millerdb8dac22008-03-06 16:22:02 -0800666 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
667 err = EMSGSIZE;
668 harderr = 1;
669 break;
670 }
671 goto out;
672 }
673 err = EHOSTUNREACH;
674 if (code <= NR_ICMP_UNREACH) {
675 harderr = icmp_err_convert[code].fatal;
676 err = icmp_err_convert[code].errno;
677 }
678 break;
David S. Miller55be7a92012-07-11 21:27:49 -0700679 case ICMP_REDIRECT:
680 ipv4_sk_redirect(skb, sk);
Duan Jiong1a462d12013-09-20 18:20:28 +0800681 goto out;
David S. Millerdb8dac22008-03-06 16:22:02 -0800682 }
683
684 /*
685 * RFC1122: OK. Passes ICMP errors back to application, as per
686 * 4.1.3.3.
687 */
688 if (!inet->recverr) {
689 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
690 goto out;
Eric Dumazetb1faf562010-05-31 23:44:05 -0700691 } else
Eric Dumazetc482c562009-07-17 00:26:32 +0000692 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
Eric Dumazetb1faf562010-05-31 23:44:05 -0700693
David S. Millerdb8dac22008-03-06 16:22:02 -0800694 sk->sk_err = err;
695 sk->sk_error_report(sk);
696out:
Eric Dumazetca065d02016-04-01 08:52:13 -0700697 return;
David S. Millerdb8dac22008-03-06 16:22:02 -0800698}
699
700void udp_err(struct sk_buff *skb, u32 info)
701{
Eric Dumazet645ca702008-10-29 01:41:45 -0700702 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800703}
704
705/*
706 * Throw away all pending data and cancel the corking. Socket is locked.
707 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400708void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800709{
710 struct udp_sock *up = udp_sk(sk);
711
712 if (up->pending) {
713 up->len = 0;
714 up->pending = 0;
715 ip_flush_pending_frames(sk);
716 }
717}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400718EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800719
720/**
Herbert Xuf6b96642011-03-01 02:36:48 +0000721 * udp4_hwcsum - handle outgoing HW checksumming
David S. Millerdb8dac22008-03-06 16:22:02 -0800722 * @skb: sk_buff containing the filled-in UDP header
723 * (checksum field must be zeroed out)
Herbert Xuf6b96642011-03-01 02:36:48 +0000724 * @src: source IP address
725 * @dst: destination IP address
David S. Millerdb8dac22008-03-06 16:22:02 -0800726 */
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200727void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
David S. Millerdb8dac22008-03-06 16:22:02 -0800728{
David S. Millerdb8dac22008-03-06 16:22:02 -0800729 struct udphdr *uh = udp_hdr(skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000730 int offset = skb_transport_offset(skb);
731 int len = skb->len - offset;
732 int hlen = len;
David S. Millerdb8dac22008-03-06 16:22:02 -0800733 __wsum csum = 0;
734
WANG Congebbe4952014-06-02 16:12:02 -0700735 if (!skb_has_frag_list(skb)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800736 /*
737 * Only one fragment on the socket.
738 */
739 skb->csum_start = skb_transport_header(skb) - skb->head;
740 skb->csum_offset = offsetof(struct udphdr, check);
Herbert Xuf6b96642011-03-01 02:36:48 +0000741 uh->check = ~csum_tcpudp_magic(src, dst, len,
742 IPPROTO_UDP, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800743 } else {
WANG Congebbe4952014-06-02 16:12:02 -0700744 struct sk_buff *frags;
745
David S. Millerdb8dac22008-03-06 16:22:02 -0800746 /*
747 * HW-checksum won't work as there are two or more
748 * fragments on the socket so that all csums of sk_buffs
749 * should be together
750 */
WANG Congebbe4952014-06-02 16:12:02 -0700751 skb_walk_frags(skb, frags) {
Herbert Xuf6b96642011-03-01 02:36:48 +0000752 csum = csum_add(csum, frags->csum);
753 hlen -= frags->len;
WANG Congebbe4952014-06-02 16:12:02 -0700754 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800755
Herbert Xuf6b96642011-03-01 02:36:48 +0000756 csum = skb_checksum(skb, offset, hlen, csum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800757 skb->ip_summed = CHECKSUM_NONE;
758
David S. Millerdb8dac22008-03-06 16:22:02 -0800759 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
760 if (uh->check == 0)
761 uh->check = CSUM_MANGLED_0;
762 }
763}
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200764EXPORT_SYMBOL_GPL(udp4_hwcsum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800765
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700766/* Function to set UDP checksum for an IPv4 UDP packet. This is intended
767 * for the simple case like when setting the checksum for a UDP tunnel.
768 */
769void udp_set_csum(bool nocheck, struct sk_buff *skb,
770 __be32 saddr, __be32 daddr, int len)
771{
772 struct udphdr *uh = udp_hdr(skb);
773
Edward Cree179bc672016-02-11 20:48:04 +0000774 if (nocheck) {
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700775 uh->check = 0;
Edward Cree179bc672016-02-11 20:48:04 +0000776 } else if (skb_is_gso(skb)) {
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700777 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
Edward Cree179bc672016-02-11 20:48:04 +0000778 } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
779 uh->check = 0;
780 uh->check = udp_v4_check(len, saddr, daddr, lco_csum(skb));
781 if (uh->check == 0)
782 uh->check = CSUM_MANGLED_0;
Edward Creed75f1302016-02-11 20:49:40 +0000783 } else {
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700784 skb->ip_summed = CHECKSUM_PARTIAL;
785 skb->csum_start = skb_transport_header(skb) - skb->head;
786 skb->csum_offset = offsetof(struct udphdr, check);
787 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700788 }
789}
790EXPORT_SYMBOL(udp_set_csum);
791
David S. Miller79ab0532011-05-09 13:31:04 -0700792static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
Herbert Xuf6b96642011-03-01 02:36:48 +0000793{
794 struct sock *sk = skb->sk;
795 struct inet_sock *inet = inet_sk(sk);
796 struct udphdr *uh;
Herbert Xuf6b96642011-03-01 02:36:48 +0000797 int err = 0;
798 int is_udplite = IS_UDPLITE(sk);
799 int offset = skb_transport_offset(skb);
800 int len = skb->len - offset;
801 __wsum csum = 0;
802
803 /*
804 * Create a UDP header
805 */
806 uh = udp_hdr(skb);
807 uh->source = inet->inet_sport;
David S. Miller79ab0532011-05-09 13:31:04 -0700808 uh->dest = fl4->fl4_dport;
Herbert Xuf6b96642011-03-01 02:36:48 +0000809 uh->len = htons(len);
810 uh->check = 0;
811
812 if (is_udplite) /* UDP-Lite */
813 csum = udplite_csum(skb);
814
Tom Herbert28448b82014-05-23 08:47:19 -0700815 else if (sk->sk_no_check_tx) { /* UDP csum disabled */
Herbert Xuf6b96642011-03-01 02:36:48 +0000816
817 skb->ip_summed = CHECKSUM_NONE;
818 goto send;
819
820 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
821
David S. Miller79ab0532011-05-09 13:31:04 -0700822 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
Herbert Xuf6b96642011-03-01 02:36:48 +0000823 goto send;
824
825 } else
826 csum = udp_csum(skb);
827
828 /* add protocol-dependent pseudo-header */
David S. Miller79ab0532011-05-09 13:31:04 -0700829 uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
Herbert Xuf6b96642011-03-01 02:36:48 +0000830 sk->sk_protocol, csum);
831 if (uh->check == 0)
832 uh->check = CSUM_MANGLED_0;
833
834send:
Eric Dumazetb5ec8ee2012-08-10 02:22:47 +0000835 err = ip_send_skb(sock_net(sk), skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000836 if (err) {
837 if (err == -ENOBUFS && !inet->recverr) {
Eric Dumazet6aef70a2016-04-27 16:44:27 -0700838 UDP_INC_STATS(sock_net(sk),
839 UDP_MIB_SNDBUFERRORS, is_udplite);
Herbert Xuf6b96642011-03-01 02:36:48 +0000840 err = 0;
841 }
842 } else
Eric Dumazet6aef70a2016-04-27 16:44:27 -0700843 UDP_INC_STATS(sock_net(sk),
844 UDP_MIB_OUTDATAGRAMS, is_udplite);
Herbert Xuf6b96642011-03-01 02:36:48 +0000845 return err;
846}
847
David S. Millerdb8dac22008-03-06 16:22:02 -0800848/*
849 * Push out all pending data as one UDP datagram. Socket is locked.
850 */
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200851int udp_push_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800852{
853 struct udp_sock *up = udp_sk(sk);
854 struct inet_sock *inet = inet_sk(sk);
David S. Millerb6f21b22011-03-12 02:09:18 -0500855 struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800856 struct sk_buff *skb;
David S. Millerdb8dac22008-03-06 16:22:02 -0800857 int err = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800858
David S. Miller77968b72011-05-08 17:12:19 -0700859 skb = ip_finish_skb(sk, fl4);
Herbert Xuf6b96642011-03-01 02:36:48 +0000860 if (!skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800861 goto out;
862
David S. Miller79ab0532011-05-09 13:31:04 -0700863 err = udp_send_skb(skb, fl4);
David S. Millerdb8dac22008-03-06 16:22:02 -0800864
David S. Millerdb8dac22008-03-06 16:22:02 -0800865out:
866 up->len = 0;
867 up->pending = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800868 return err;
869}
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200870EXPORT_SYMBOL(udp_push_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800871
Ying Xue1b784142015-03-02 15:37:48 +0800872int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
David S. Millerdb8dac22008-03-06 16:22:02 -0800873{
874 struct inet_sock *inet = inet_sk(sk);
875 struct udp_sock *up = udp_sk(sk);
David S. Millere4749952011-05-08 16:38:45 -0700876 struct flowi4 fl4_stack;
David S. Millerb6f21b22011-03-12 02:09:18 -0500877 struct flowi4 *fl4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800878 int ulen = len;
879 struct ipcm_cookie ipc;
880 struct rtable *rt = NULL;
881 int free = 0;
882 int connected = 0;
883 __be32 daddr, faddr, saddr;
884 __be16 dport;
885 u8 tos;
886 int err, is_udplite = IS_UDPLITE(sk);
887 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
888 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
Herbert Xu903ab862011-03-01 02:36:48 +0000889 struct sk_buff *skb;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000890 struct ip_options_data opt_copy;
David S. Millerdb8dac22008-03-06 16:22:02 -0800891
892 if (len > 0xFFFF)
893 return -EMSGSIZE;
894
895 /*
896 * Check the flags.
897 */
898
Eric Dumazetc482c562009-07-17 00:26:32 +0000899 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
David S. Millerdb8dac22008-03-06 16:22:02 -0800900 return -EOPNOTSUPP;
901
902 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000903 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200904 ipc.ttl = 0;
905 ipc.tos = -1;
David S. Millerdb8dac22008-03-06 16:22:02 -0800906
Herbert Xu903ab862011-03-01 02:36:48 +0000907 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
908
David S. Millerf5fca602011-05-08 17:24:10 -0700909 fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800910 if (up->pending) {
911 /*
912 * There are pending frames.
913 * The socket lock must be held while it's corked.
914 */
915 lock_sock(sk);
916 if (likely(up->pending)) {
917 if (unlikely(up->pending != AF_INET)) {
918 release_sock(sk);
919 return -EINVAL;
920 }
921 goto do_append_data;
922 }
923 release_sock(sk);
924 }
925 ulen += sizeof(struct udphdr);
926
927 /*
928 * Get and verify the address.
929 */
930 if (msg->msg_name) {
Steffen Hurrle342dfc32014-01-17 22:53:15 +0100931 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -0800932 if (msg->msg_namelen < sizeof(*usin))
933 return -EINVAL;
934 if (usin->sin_family != AF_INET) {
935 if (usin->sin_family != AF_UNSPEC)
936 return -EAFNOSUPPORT;
937 }
938
939 daddr = usin->sin_addr.s_addr;
940 dport = usin->sin_port;
941 if (dport == 0)
942 return -EINVAL;
943 } else {
944 if (sk->sk_state != TCP_ESTABLISHED)
945 return -EDESTADDRREQ;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000946 daddr = inet->inet_daddr;
947 dport = inet->inet_dport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800948 /* Open fast path for connected socket.
949 Route will not be used, if at least one option is set.
950 */
951 connected = 1;
952 }
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400953
954 ipc.sockc.tsflags = sk->sk_tsflags;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000955 ipc.addr = inet->inet_saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800956 ipc.oif = sk->sk_bound_dev_if;
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000957
David S. Millerdb8dac22008-03-06 16:22:02 -0800958 if (msg->msg_controllen) {
Soheil Hassas Yeganeh24025c42016-04-02 23:08:10 -0400959 err = ip_cmsg_send(sk, msg, &ipc, sk->sk_family == AF_INET6);
Eric Dumazet91948302016-02-04 06:23:28 -0800960 if (unlikely(err)) {
961 kfree(ipc.opt);
David S. Millerdb8dac22008-03-06 16:22:02 -0800962 return err;
Eric Dumazet91948302016-02-04 06:23:28 -0800963 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800964 if (ipc.opt)
965 free = 1;
966 connected = 0;
967 }
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000968 if (!ipc.opt) {
969 struct ip_options_rcu *inet_opt;
970
971 rcu_read_lock();
972 inet_opt = rcu_dereference(inet->inet_opt);
973 if (inet_opt) {
974 memcpy(&opt_copy, inet_opt,
975 sizeof(*inet_opt) + inet_opt->opt.optlen);
976 ipc.opt = &opt_copy.opt;
977 }
978 rcu_read_unlock();
979 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800980
981 saddr = ipc.addr;
982 ipc.addr = faddr = daddr;
983
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400984 sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
985
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000986 if (ipc.opt && ipc.opt->opt.srr) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800987 if (!daddr)
988 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000989 faddr = ipc.opt->opt.faddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800990 connected = 0;
991 }
Francesco Fuscoaa661582013-09-24 15:43:09 +0200992 tos = get_rttos(&ipc, inet);
David S. Millerdb8dac22008-03-06 16:22:02 -0800993 if (sock_flag(sk, SOCK_LOCALROUTE) ||
994 (msg->msg_flags & MSG_DONTROUTE) ||
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000995 (ipc.opt && ipc.opt->opt.is_strictroute)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800996 tos |= RTO_ONLINK;
997 connected = 0;
998 }
999
1000 if (ipv4_is_multicast(daddr)) {
1001 if (!ipc.oif)
1002 ipc.oif = inet->mc_index;
1003 if (!saddr)
1004 saddr = inet->mc_addr;
1005 connected = 0;
Erich E. Hoover76e21052012-02-08 09:11:07 +00001006 } else if (!ipc.oif)
1007 ipc.oif = inet->uc_index;
David S. Millerdb8dac22008-03-06 16:22:02 -08001008
1009 if (connected)
Eric Dumazetc482c562009-07-17 00:26:32 +00001010 rt = (struct rtable *)sk_dst_check(sk, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001011
Ian Morris51456b22015-04-03 09:17:26 +01001012 if (!rt) {
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001013 struct net *net = sock_net(sk);
David Ahern9a24abf2015-08-13 14:59:03 -06001014 __u8 flow_flags = inet_sk_flowi_flags(sk);
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001015
David S. Millere4749952011-05-08 16:38:45 -07001016 fl4 = &fl4_stack;
David Ahern9a24abf2015-08-13 14:59:03 -06001017
David S. Millere4749952011-05-08 16:38:45 -07001018 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
David S. Millerc0951cb2011-03-31 04:54:27 -07001019 RT_SCOPE_UNIVERSE, sk->sk_protocol,
David Ahern9a24abf2015-08-13 14:59:03 -06001020 flow_flags,
David S. Millerc0951cb2011-03-31 04:54:27 -07001021 faddr, saddr, dport, inet->inet_sport);
1022
David Ahernb5bdacf2016-01-04 09:09:27 -08001023 if (!saddr && ipc.oif) {
1024 err = l3mdev_get_saddr(net, ipc.oif, fl4);
1025 if (err < 0)
1026 goto out;
1027 }
David Ahern8cbb512c2015-10-05 08:51:26 -07001028
David S. Millere4749952011-05-08 16:38:45 -07001029 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
1030 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001031 if (IS_ERR(rt)) {
1032 err = PTR_ERR(rt);
David S. Miller06dc94b2011-03-03 10:38:01 -08001033 rt = NULL;
David S. Millerdb8dac22008-03-06 16:22:02 -08001034 if (err == -ENETUNREACH)
Eric Dumazetf1d8cba2013-11-28 09:51:22 -08001035 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -08001036 goto out;
1037 }
1038
1039 err = -EACCES;
1040 if ((rt->rt_flags & RTCF_BROADCAST) &&
1041 !sock_flag(sk, SOCK_BROADCAST))
1042 goto out;
1043 if (connected)
Changli Gaod8d1f302010-06-10 23:31:35 -07001044 sk_dst_set(sk, dst_clone(&rt->dst));
David S. Millerdb8dac22008-03-06 16:22:02 -08001045 }
1046
1047 if (msg->msg_flags&MSG_CONFIRM)
1048 goto do_confirm;
1049back_from_confirm:
1050
David S. Millere4749952011-05-08 16:38:45 -07001051 saddr = fl4->saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001052 if (!ipc.addr)
David S. Millere4749952011-05-08 16:38:45 -07001053 daddr = ipc.addr = fl4->daddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001054
Herbert Xu903ab862011-03-01 02:36:48 +00001055 /* Lockless fast path for the non-corking case. */
1056 if (!corkreq) {
Al Virof69e6d12014-11-24 13:23:40 -05001057 skb = ip_make_skb(sk, fl4, getfrag, msg, ulen,
Herbert Xu903ab862011-03-01 02:36:48 +00001058 sizeof(struct udphdr), &ipc, &rt,
1059 msg->msg_flags);
1060 err = PTR_ERR(skb);
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001061 if (!IS_ERR_OR_NULL(skb))
David S. Miller79ab0532011-05-09 13:31:04 -07001062 err = udp_send_skb(skb, fl4);
Herbert Xu903ab862011-03-01 02:36:48 +00001063 goto out;
1064 }
1065
David S. Millerdb8dac22008-03-06 16:22:02 -08001066 lock_sock(sk);
1067 if (unlikely(up->pending)) {
1068 /* The socket is already corked while preparing it. */
1069 /* ... which is an evident application bug. --ANK */
1070 release_sock(sk);
1071
Joe Perchesba7a46f2014-11-11 10:59:17 -08001072 net_dbg_ratelimited("cork app bug 2\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001073 err = -EINVAL;
1074 goto out;
1075 }
1076 /*
1077 * Now cork the socket to pend data.
1078 */
David S. Millerb6f21b22011-03-12 02:09:18 -05001079 fl4 = &inet->cork.fl.u.ip4;
1080 fl4->daddr = daddr;
1081 fl4->saddr = saddr;
David S. Miller9cce96d2011-03-12 03:00:33 -05001082 fl4->fl4_dport = dport;
1083 fl4->fl4_sport = inet->inet_sport;
David S. Millerdb8dac22008-03-06 16:22:02 -08001084 up->pending = AF_INET;
1085
1086do_append_data:
1087 up->len += ulen;
Al Virof69e6d12014-11-24 13:23:40 -05001088 err = ip_append_data(sk, fl4, getfrag, msg, ulen,
David S. Millerf5fca602011-05-08 17:24:10 -07001089 sizeof(struct udphdr), &ipc, &rt,
1090 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001091 if (err)
1092 udp_flush_pending_frames(sk);
1093 else if (!corkreq)
1094 err = udp_push_pending_frames(sk);
1095 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1096 up->pending = 0;
1097 release_sock(sk);
1098
1099out:
1100 ip_rt_put(rt);
1101 if (free)
1102 kfree(ipc.opt);
1103 if (!err)
1104 return len;
1105 /*
1106 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1107 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1108 * we don't have a good statistic (IpOutDiscards but it can be too many
1109 * things). We could add another new stat but at least for now that
1110 * seems like overkill.
1111 */
1112 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Eric Dumazet6aef70a2016-04-27 16:44:27 -07001113 UDP_INC_STATS(sock_net(sk),
1114 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001115 }
1116 return err;
1117
1118do_confirm:
Changli Gaod8d1f302010-06-10 23:31:35 -07001119 dst_confirm(&rt->dst);
David S. Millerdb8dac22008-03-06 16:22:02 -08001120 if (!(msg->msg_flags&MSG_PROBE) || len)
1121 goto back_from_confirm;
1122 err = 0;
1123 goto out;
1124}
Eric Dumazetc482c562009-07-17 00:26:32 +00001125EXPORT_SYMBOL(udp_sendmsg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001126
1127int udp_sendpage(struct sock *sk, struct page *page, int offset,
1128 size_t size, int flags)
1129{
David S. Millerf5fca602011-05-08 17:24:10 -07001130 struct inet_sock *inet = inet_sk(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001131 struct udp_sock *up = udp_sk(sk);
1132 int ret;
1133
Shawn Landdend3f7d562013-11-24 22:36:28 -08001134 if (flags & MSG_SENDPAGE_NOTLAST)
1135 flags |= MSG_MORE;
1136
David S. Millerdb8dac22008-03-06 16:22:02 -08001137 if (!up->pending) {
1138 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
1139
1140 /* Call udp_sendmsg to specify destination address which
1141 * sendpage interface can't pass.
1142 * This will succeed only when the socket is connected.
1143 */
Ying Xue1b784142015-03-02 15:37:48 +08001144 ret = udp_sendmsg(sk, &msg, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001145 if (ret < 0)
1146 return ret;
1147 }
1148
1149 lock_sock(sk);
1150
1151 if (unlikely(!up->pending)) {
1152 release_sock(sk);
1153
Joe Perchesba7a46f2014-11-11 10:59:17 -08001154 net_dbg_ratelimited("udp cork app bug 3\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001155 return -EINVAL;
1156 }
1157
David S. Millerf5fca602011-05-08 17:24:10 -07001158 ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1159 page, offset, size, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001160 if (ret == -EOPNOTSUPP) {
1161 release_sock(sk);
1162 return sock_no_sendpage(sk->sk_socket, page, offset,
1163 size, flags);
1164 }
1165 if (ret < 0) {
1166 udp_flush_pending_frames(sk);
1167 goto out;
1168 }
1169
1170 up->len += size;
1171 if (!(up->corkflag || (flags&MSG_MORE)))
1172 ret = udp_push_pending_frames(sk);
1173 if (!ret)
1174 ret = size;
1175out:
1176 release_sock(sk);
1177 return ret;
1178}
1179
Eric Dumazet85584672009-10-09 04:43:40 +00001180/**
1181 * first_packet_length - return length of first packet in receive queue
1182 * @sk: socket
1183 *
1184 * Drops all bad checksum frames, until a valid one is found.
1185 * Returns the length of found skb, or 0 if none is found.
1186 */
1187static unsigned int first_packet_length(struct sock *sk)
1188{
1189 struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1190 struct sk_buff *skb;
1191 unsigned int res;
1192
1193 __skb_queue_head_init(&list_kill);
1194
1195 spin_lock_bh(&rcvq->lock);
1196 while ((skb = skb_peek(rcvq)) != NULL &&
1197 udp_lib_checksum_complete(skb)) {
Eric Dumazet02c22342016-04-27 16:44:30 -07001198 __UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS,
1199 IS_UDPLITE(sk));
1200 __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
1201 IS_UDPLITE(sk));
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001202 atomic_inc(&sk->sk_drops);
Eric Dumazet85584672009-10-09 04:43:40 +00001203 __skb_unlink(skb, rcvq);
1204 __skb_queue_tail(&list_kill, skb);
1205 }
1206 res = skb ? skb->len : 0;
1207 spin_unlock_bh(&rcvq->lock);
1208
1209 if (!skb_queue_empty(&list_kill)) {
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001210 bool slow = lock_sock_fast(sk);
1211
Eric Dumazet85584672009-10-09 04:43:40 +00001212 __skb_queue_purge(&list_kill);
1213 sk_mem_reclaim_partial(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001214 unlock_sock_fast(sk, slow);
Eric Dumazet85584672009-10-09 04:43:40 +00001215 }
1216 return res;
1217}
1218
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219/*
1220 * IOCTL requests applicable to the UDP protocol
1221 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001222
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1224{
Stephen Hemminger6516c652007-03-08 20:41:55 -08001225 switch (cmd) {
1226 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 {
Eric Dumazet31e6d362009-06-17 19:05:41 -07001228 int amount = sk_wmem_alloc_get(sk);
1229
Stephen Hemminger6516c652007-03-08 20:41:55 -08001230 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 }
Stephen Hemminger6516c652007-03-08 20:41:55 -08001232
1233 case SIOCINQ:
1234 {
Eric Dumazet85584672009-10-09 04:43:40 +00001235 unsigned int amount = first_packet_length(sk);
Stephen Hemminger6516c652007-03-08 20:41:55 -08001236
Stephen Hemminger6516c652007-03-08 20:41:55 -08001237 return put_user(amount, (int __user *)arg);
1238 }
1239
1240 default:
1241 return -ENOIOCTLCMD;
1242 }
1243
1244 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245}
Eric Dumazetc482c562009-07-17 00:26:32 +00001246EXPORT_SYMBOL(udp_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247
David S. Millerdb8dac22008-03-06 16:22:02 -08001248/*
1249 * This should be easy, if there is something there we
1250 * return it, otherwise we block.
1251 */
1252
Ying Xue1b784142015-03-02 15:37:48 +08001253int udp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
1254 int flags, int *addr_len)
David S. Millerdb8dac22008-03-06 16:22:02 -08001255{
1256 struct inet_sock *inet = inet_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001257 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -08001258 struct sk_buff *skb;
David S. Miller59c2cda2011-12-01 14:12:55 -05001259 unsigned int ulen, copied;
samanthakumar627d2d62016-04-05 12:41:16 -04001260 int peeked, peeking, off;
David S. Millerdb8dac22008-03-06 16:22:02 -08001261 int err;
1262 int is_udplite = IS_UDPLITE(sk);
Eric Dumazet197c9492015-12-30 08:51:12 -05001263 bool checksum_valid = false;
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001264 bool slow;
David S. Millerdb8dac22008-03-06 16:22:02 -08001265
David S. Millerdb8dac22008-03-06 16:22:02 -08001266 if (flags & MSG_ERRQUEUE)
Hannes Frederic Sowa85fbaa72013-11-23 00:46:12 +01001267 return ip_recv_error(sk, msg, len, addr_len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001268
1269try_again:
samanthakumar627d2d62016-04-05 12:41:16 -04001270 peeking = off = sk_peek_offset(sk, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001271 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001272 &peeked, &off, &err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001273 if (!skb)
samanthakumar627d2d62016-04-05 12:41:16 -04001274 return err;
David S. Millerdb8dac22008-03-06 16:22:02 -08001275
samanthakumare6afc8a2016-04-05 12:41:15 -04001276 ulen = skb->len;
David S. Miller59c2cda2011-12-01 14:12:55 -05001277 copied = len;
samanthakumar627d2d62016-04-05 12:41:16 -04001278 if (copied > ulen - off)
1279 copied = ulen - off;
David S. Miller59c2cda2011-12-01 14:12:55 -05001280 else if (copied < ulen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001281 msg->msg_flags |= MSG_TRUNC;
1282
1283 /*
1284 * If checksum is needed at all, try to do it while copying the
1285 * data. If the data is truncated, or if we only want a partial
1286 * coverage checksum (UDP-Lite), do it before the copy.
1287 */
1288
samanthakumar627d2d62016-04-05 12:41:16 -04001289 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov || peeking) {
Eric Dumazet197c9492015-12-30 08:51:12 -05001290 checksum_valid = !udp_lib_checksum_complete(skb);
1291 if (!checksum_valid)
David S. Millerdb8dac22008-03-06 16:22:02 -08001292 goto csum_copy_err;
1293 }
1294
Eric Dumazet197c9492015-12-30 08:51:12 -05001295 if (checksum_valid || skb_csum_unnecessary(skb))
samanthakumar627d2d62016-04-05 12:41:16 -04001296 err = skb_copy_datagram_msg(skb, off, msg, copied);
David S. Millerdb8dac22008-03-06 16:22:02 -08001297 else {
samanthakumar627d2d62016-04-05 12:41:16 -04001298 err = skb_copy_and_csum_datagram_msg(skb, off, msg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001299
1300 if (err == -EINVAL)
1301 goto csum_copy_err;
1302 }
1303
Eric Dumazet22911fc2012-06-27 00:23:44 +00001304 if (unlikely(err)) {
1305 trace_kfree_skb(skb, udp_recvmsg);
Eric Dumazet979402b2012-09-05 23:34:44 +00001306 if (!peeked) {
1307 atomic_inc(&sk->sk_drops);
Eric Dumazet6aef70a2016-04-27 16:44:27 -07001308 UDP_INC_STATS(sock_net(sk),
1309 UDP_MIB_INERRORS, is_udplite);
Eric Dumazet979402b2012-09-05 23:34:44 +00001310 }
samanthakumar627d2d62016-04-05 12:41:16 -04001311 skb_free_datagram_locked(sk, skb);
1312 return err;
Eric Dumazet22911fc2012-06-27 00:23:44 +00001313 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001314
1315 if (!peeked)
Eric Dumazet6aef70a2016-04-27 16:44:27 -07001316 UDP_INC_STATS(sock_net(sk),
1317 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001318
Neil Horman3b885782009-10-12 13:26:31 -07001319 sock_recv_ts_and_drops(msg, sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001320
1321 /* Copy the address. */
Eric Dumazetc482c562009-07-17 00:26:32 +00001322 if (sin) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001323 sin->sin_family = AF_INET;
1324 sin->sin_port = udp_hdr(skb)->source;
1325 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1326 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
Hannes Frederic Sowabceaa902013-11-18 04:20:45 +01001327 *addr_len = sizeof(*sin);
David S. Millerdb8dac22008-03-06 16:22:02 -08001328 }
1329 if (inet->cmsg_flags)
Willem de Bruijn31c2e492016-04-07 18:12:59 -04001330 ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr) + off);
David S. Millerdb8dac22008-03-06 16:22:02 -08001331
David S. Miller59c2cda2011-12-01 14:12:55 -05001332 err = copied;
David S. Millerdb8dac22008-03-06 16:22:02 -08001333 if (flags & MSG_TRUNC)
1334 err = ulen;
1335
samanthakumar627d2d62016-04-05 12:41:16 -04001336 __skb_free_datagram_locked(sk, skb, peeking ? -err : err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001337 return err;
1338
1339csum_copy_err:
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001340 slow = lock_sock_fast(sk);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001341 if (!skb_kill_datagram(sk, skb, flags)) {
Eric Dumazet6aef70a2016-04-27 16:44:27 -07001342 UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
1343 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001344 }
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001345 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001346
Eric Dumazetbeb39db2015-05-30 09:16:53 -07001347 /* starting over for a new packet, but check if we need to yield */
1348 cond_resched();
Xufeng Zhang9cfaa8d2011-06-21 10:43:40 +00001349 msg->msg_flags &= ~MSG_TRUNC;
David S. Millerdb8dac22008-03-06 16:22:02 -08001350 goto try_again;
1351}
1352
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353int udp_disconnect(struct sock *sk, int flags)
1354{
1355 struct inet_sock *inet = inet_sk(sk);
1356 /*
1357 * 1003.1g - break association.
1358 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001359
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 sk->sk_state = TCP_CLOSE;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001361 inet->inet_daddr = 0;
1362 inet->inet_dport = 0;
Tom Herbertbdeab992011-08-14 19:45:55 +00001363 sock_rps_reset_rxhash(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 sk->sk_bound_dev_if = 0;
1365 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1366 inet_reset_saddr(sk);
1367
1368 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1369 sk->sk_prot->unhash(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001370 inet->inet_sport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 }
1372 sk_dst_reset(sk);
1373 return 0;
1374}
Eric Dumazetc482c562009-07-17 00:26:32 +00001375EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376
Eric Dumazet645ca702008-10-29 01:41:45 -07001377void udp_lib_unhash(struct sock *sk)
1378{
Eric Dumazet723b4612008-11-25 13:55:15 -08001379 if (sk_hashed(sk)) {
1380 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Eric Dumazet512615b2009-11-08 10:17:58 +00001381 struct udp_hslot *hslot, *hslot2;
1382
1383 hslot = udp_hashslot(udptable, sock_net(sk),
1384 udp_sk(sk)->udp_port_hash);
1385 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
Eric Dumazet645ca702008-10-29 01:41:45 -07001386
Eric Dumazet723b4612008-11-25 13:55:15 -08001387 spin_lock_bh(&hslot->lock);
Craig Galleke32ea7e2016-01-04 17:41:46 -05001388 if (rcu_access_pointer(sk->sk_reuseport_cb))
1389 reuseport_detach_sock(sk);
Eric Dumazetca065d02016-04-01 08:52:13 -07001390 if (sk_del_node_init_rcu(sk)) {
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00001391 hslot->count--;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001392 inet_sk(sk)->inet_num = 0;
Eric Dumazet723b4612008-11-25 13:55:15 -08001393 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet512615b2009-11-08 10:17:58 +00001394
1395 spin_lock(&hslot2->lock);
Eric Dumazetca065d02016-04-01 08:52:13 -07001396 hlist_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
Eric Dumazet512615b2009-11-08 10:17:58 +00001397 hslot2->count--;
1398 spin_unlock(&hslot2->lock);
Eric Dumazet723b4612008-11-25 13:55:15 -08001399 }
1400 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001401 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001402}
1403EXPORT_SYMBOL(udp_lib_unhash);
1404
Eric Dumazet719f8352010-09-08 05:08:44 +00001405/*
1406 * inet_rcv_saddr was changed, we must rehash secondary hash
1407 */
1408void udp_lib_rehash(struct sock *sk, u16 newhash)
1409{
1410 if (sk_hashed(sk)) {
1411 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1412 struct udp_hslot *hslot, *hslot2, *nhslot2;
1413
1414 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1415 nhslot2 = udp_hashslot2(udptable, newhash);
1416 udp_sk(sk)->udp_portaddr_hash = newhash;
Craig Galleke32ea7e2016-01-04 17:41:46 -05001417
1418 if (hslot2 != nhslot2 ||
1419 rcu_access_pointer(sk->sk_reuseport_cb)) {
Eric Dumazet719f8352010-09-08 05:08:44 +00001420 hslot = udp_hashslot(udptable, sock_net(sk),
1421 udp_sk(sk)->udp_port_hash);
1422 /* we must lock primary chain too */
1423 spin_lock_bh(&hslot->lock);
Craig Galleke32ea7e2016-01-04 17:41:46 -05001424 if (rcu_access_pointer(sk->sk_reuseport_cb))
1425 reuseport_detach_sock(sk);
Eric Dumazet719f8352010-09-08 05:08:44 +00001426
Craig Galleke32ea7e2016-01-04 17:41:46 -05001427 if (hslot2 != nhslot2) {
1428 spin_lock(&hslot2->lock);
Eric Dumazetca065d02016-04-01 08:52:13 -07001429 hlist_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
Craig Galleke32ea7e2016-01-04 17:41:46 -05001430 hslot2->count--;
1431 spin_unlock(&hslot2->lock);
Eric Dumazet719f8352010-09-08 05:08:44 +00001432
Craig Galleke32ea7e2016-01-04 17:41:46 -05001433 spin_lock(&nhslot2->lock);
Eric Dumazetca065d02016-04-01 08:52:13 -07001434 hlist_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
Craig Galleke32ea7e2016-01-04 17:41:46 -05001435 &nhslot2->head);
1436 nhslot2->count++;
1437 spin_unlock(&nhslot2->lock);
1438 }
Eric Dumazet719f8352010-09-08 05:08:44 +00001439
1440 spin_unlock_bh(&hslot->lock);
1441 }
1442 }
1443}
1444EXPORT_SYMBOL(udp_lib_rehash);
1445
1446static void udp_v4_rehash(struct sock *sk)
1447{
1448 u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1449 inet_sk(sk)->inet_rcv_saddr,
1450 inet_sk(sk)->inet_num);
1451 udp_lib_rehash(sk, new_hash);
1452}
1453
Herbert Xu93821772008-09-15 11:48:46 -07001454static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1455{
Tom Herbertfec5e652010-04-16 16:01:27 -07001456 int rc;
Herbert Xu93821772008-09-15 11:48:46 -07001457
Shawn Bohrer005ec972013-10-07 11:01:38 -05001458 if (inet_sk(sk)->inet_daddr) {
Tom Herbertbdeab992011-08-14 19:45:55 +00001459 sock_rps_save_rxhash(sk, skb);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001460 sk_mark_napi_id(sk, skb);
Eric Dumazet2c8c56e2014-11-11 05:54:28 -08001461 sk_incoming_cpu_update(sk);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001462 }
Tom Herbertfec5e652010-04-16 16:01:27 -07001463
samanthakumare6afc8a2016-04-05 12:41:15 -04001464 rc = __sock_queue_rcv_skb(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -07001465 if (rc < 0) {
1466 int is_udplite = IS_UDPLITE(sk);
1467
Herbert Xu93821772008-09-15 11:48:46 -07001468 /* Note that an ENOMEM error is charged twice */
Eric Dumazet766e90372009-10-14 20:40:11 -07001469 if (rc == -ENOMEM)
Eric Dumazete61da9e2016-04-29 14:16:50 -07001470 UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS,
Eric Dumazet02c22342016-04-27 16:44:30 -07001471 is_udplite);
Eric Dumazete61da9e2016-04-29 14:16:50 -07001472 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet766e90372009-10-14 20:40:11 -07001473 kfree_skb(skb);
Satoru Moriya296f7ea2011-06-17 11:58:39 +00001474 trace_udp_fail_queue_rcv_skb(rc, sk);
Eric Dumazet766e90372009-10-14 20:40:11 -07001475 return -1;
Herbert Xu93821772008-09-15 11:48:46 -07001476 }
1477
1478 return 0;
1479
Herbert Xu93821772008-09-15 11:48:46 -07001480}
1481
Eric Dumazet447167b2012-04-11 23:05:28 +00001482static struct static_key udp_encap_needed __read_mostly;
1483void udp_encap_enable(void)
1484{
1485 if (!static_key_enabled(&udp_encap_needed))
1486 static_key_slow_inc(&udp_encap_needed);
1487}
1488EXPORT_SYMBOL(udp_encap_enable);
1489
David S. Millerdb8dac22008-03-06 16:22:02 -08001490/* returns:
1491 * -1: error
1492 * 0: success
1493 * >0: "udp encap" protocol resubmission
1494 *
1495 * Note that in the success and error cases, the skb is assumed to
1496 * have either been requeued or freed.
1497 */
Eric Dumazetc482c562009-07-17 00:26:32 +00001498int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -08001499{
1500 struct udp_sock *up = udp_sk(sk);
1501 int rc;
1502 int is_udplite = IS_UDPLITE(sk);
1503
1504 /*
1505 * Charge it to the socket, dropping if the queue is full.
1506 */
1507 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1508 goto drop;
1509 nf_reset(skb);
1510
Eric Dumazet447167b2012-04-11 23:05:28 +00001511 if (static_key_false(&udp_encap_needed) && up->encap_type) {
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001512 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1513
David S. Millerdb8dac22008-03-06 16:22:02 -08001514 /*
1515 * This is an encapsulation socket so pass the skb to
1516 * the socket's udp_encap_rcv() hook. Otherwise, just
1517 * fall through and pass this up the UDP socket.
1518 * up->encap_rcv() returns the following value:
1519 * =0 if skb was successfully passed to the encap
1520 * handler or was discarded by it.
1521 * >0 if skb should be passed on to UDP.
1522 * <0 if skb should be resubmitted as proto -N
1523 */
1524
1525 /* if we're overly short, let UDP handle it */
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001526 encap_rcv = ACCESS_ONCE(up->encap_rcv);
Hannes Frederic Sowae5aed002016-05-19 15:58:33 +02001527 if (encap_rcv) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001528 int ret;
1529
Tom Herbert0a809662014-05-07 16:52:39 -07001530 /* Verify checksum before giving to encap */
1531 if (udp_lib_checksum_complete(skb))
1532 goto csum_error;
1533
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001534 ret = encap_rcv(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001535 if (ret <= 0) {
Eric Dumazet02c22342016-04-27 16:44:30 -07001536 __UDP_INC_STATS(sock_net(sk),
1537 UDP_MIB_INDATAGRAMS,
1538 is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001539 return -ret;
1540 }
1541 }
1542
1543 /* FALLTHROUGH -- it's a UDP Packet */
1544 }
1545
1546 /*
1547 * UDP-Lite specific tests, ignored on UDP sockets
1548 */
1549 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1550
1551 /*
1552 * MIB statistics other than incrementing the error count are
1553 * disabled for the following two types of errors: these depend
1554 * on the application settings, not on the functioning of the
1555 * protocol stack as such.
1556 *
1557 * RFC 3828 here recommends (sec 3.3): "There should also be a
1558 * way ... to ... at least let the receiving application block
1559 * delivery of packets with coverage values less than a value
1560 * provided by the application."
1561 */
1562 if (up->pcrlen == 0) { /* full coverage was set */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001563 net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n",
1564 UDP_SKB_CB(skb)->cscov, skb->len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001565 goto drop;
1566 }
1567 /* The next case involves violating the min. coverage requested
1568 * by the receiver. This is subtle: if receiver wants x and x is
1569 * greater than the buffersize/MTU then receiver will complain
1570 * that it wants x while sender emits packets of smaller size y.
1571 * Therefore the above ...()->partial_cov statement is essential.
1572 */
1573 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
Joe Perchesba7a46f2014-11-11 10:59:17 -08001574 net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n",
1575 UDP_SKB_CB(skb)->cscov, up->pcrlen);
David S. Millerdb8dac22008-03-06 16:22:02 -08001576 goto drop;
1577 }
1578 }
1579
Eric Dumazetce25d662016-06-02 14:52:43 -07001580 if (rcu_access_pointer(sk->sk_filter) &&
1581 udp_lib_checksum_complete(skb))
samanthakumare6afc8a2016-04-05 12:41:15 -04001582 goto csum_error;
Eric Dumazetce25d662016-06-02 14:52:43 -07001583
1584 if (sk_filter(sk, skb))
1585 goto drop;
Michal Kubečeka6127692016-07-08 17:52:33 +02001586 if (unlikely(skb->len < sizeof(struct udphdr)))
1587 goto drop;
David S. Millerdb8dac22008-03-06 16:22:02 -08001588
samanthakumare6afc8a2016-04-05 12:41:15 -04001589 udp_csum_pull_header(skb);
Sorin Dumitru274f4822014-07-22 21:16:51 +03001590 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
Eric Dumazet02c22342016-04-27 16:44:30 -07001591 __UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1592 is_udplite);
Eric Dumazetc3774112010-04-27 15:13:20 -07001593 goto drop;
James M Leddy3e215c82014-06-25 17:38:13 -04001594 }
Eric Dumazetc3774112010-04-27 15:13:20 -07001595
Herbert Xu93821772008-09-15 11:48:46 -07001596 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001597
Shawn Bohrerfbf88662013-10-07 11:01:40 -05001598 ipv4_pktinfo_prepare(sk, skb);
Herbert Xu93821772008-09-15 11:48:46 -07001599 bh_lock_sock(sk);
1600 if (!sock_owned_by_user(sk))
1601 rc = __udp_queue_rcv_skb(sk, skb);
Eric Dumazetf545a382012-04-22 23:34:26 +00001602 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
Zhu Yi55349792010-03-04 18:01:42 +00001603 bh_unlock_sock(sk);
1604 goto drop;
1605 }
Herbert Xu93821772008-09-15 11:48:46 -07001606 bh_unlock_sock(sk);
1607
1608 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001609
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001610csum_error:
Eric Dumazet02c22342016-04-27 16:44:30 -07001611 __UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001612drop:
Eric Dumazet02c22342016-04-27 16:44:30 -07001613 __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001614 atomic_inc(&sk->sk_drops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001615 kfree_skb(skb);
1616 return -1;
1617}
1618
Eric Dumazet97502232013-12-11 14:46:51 -08001619/* For TCP sockets, sk_rx_dst is protected by socket lock
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001620 * For UDP, we use xchg() to guard against concurrent changes.
Eric Dumazet97502232013-12-11 14:46:51 -08001621 */
1622static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
Shawn Bohrer421b3882013-10-07 11:01:39 -05001623{
Eric Dumazet97502232013-12-11 14:46:51 -08001624 struct dst_entry *old;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001625
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001626 dst_hold(dst);
1627 old = xchg(&sk->sk_rx_dst, dst);
1628 dst_release(old);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001629}
1630
David S. Millerdb8dac22008-03-06 16:22:02 -08001631/*
1632 * Multicasts and broadcasts go to each listener.
1633 *
Eric Dumazet1240d132009-11-08 10:18:44 +00001634 * Note: called only from the BH handler context.
David S. Millerdb8dac22008-03-06 16:22:02 -08001635 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001636static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001637 struct udphdr *uh,
1638 __be32 saddr, __be32 daddr,
Rick Jones36cbb242014-11-06 10:37:54 -08001639 struct udp_table *udptable,
1640 int proto)
David S. Millerdb8dac22008-03-06 16:22:02 -08001641{
Eric Dumazetca065d02016-04-01 08:52:13 -07001642 struct sock *sk, *first = NULL;
David Held5cf3d462014-07-15 23:28:31 -04001643 unsigned short hnum = ntohs(uh->dest);
1644 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
David Held2dc41cf2014-07-15 23:28:32 -04001645 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
Eric Dumazetca065d02016-04-01 08:52:13 -07001646 unsigned int offset = offsetof(typeof(*sk), sk_node);
1647 int dif = skb->dev->ifindex;
1648 struct hlist_node *node;
1649 struct sk_buff *nskb;
David Held2dc41cf2014-07-15 23:28:32 -04001650
1651 if (use_hash2) {
1652 hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) &
1653 udp_table.mask;
1654 hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask;
1655start_lookup:
1656 hslot = &udp_table.hash2[hash2];
1657 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
1658 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001659
Eric Dumazetca065d02016-04-01 08:52:13 -07001660 sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
1661 if (!__udp_is_mcast_sock(net, sk, uh->dest, daddr,
1662 uh->source, saddr, dif, hnum))
1663 continue;
David S. Millerdb8dac22008-03-06 16:22:02 -08001664
Eric Dumazetca065d02016-04-01 08:52:13 -07001665 if (!first) {
1666 first = sk;
1667 continue;
1668 }
1669 nskb = skb_clone(skb, GFP_ATOMIC);
1670
1671 if (unlikely(!nskb)) {
1672 atomic_inc(&sk->sk_drops);
Eric Dumazet02c22342016-04-27 16:44:30 -07001673 __UDP_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
1674 IS_UDPLITE(sk));
1675 __UDP_INC_STATS(net, UDP_MIB_INERRORS,
1676 IS_UDPLITE(sk));
Eric Dumazetca065d02016-04-01 08:52:13 -07001677 continue;
1678 }
1679 if (udp_queue_rcv_skb(sk, nskb) > 0)
1680 consume_skb(nskb);
1681 }
Eric Dumazet1240d132009-11-08 10:18:44 +00001682
David Held2dc41cf2014-07-15 23:28:32 -04001683 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
1684 if (use_hash2 && hash2 != hash2_any) {
1685 hash2 = hash2_any;
1686 goto start_lookup;
1687 }
1688
Eric Dumazetca065d02016-04-01 08:52:13 -07001689 if (first) {
1690 if (udp_queue_rcv_skb(first, skb) > 0)
1691 consume_skb(skb);
Eric Dumazet1240d132009-11-08 10:18:44 +00001692 } else {
Eric Dumazetca065d02016-04-01 08:52:13 -07001693 kfree_skb(skb);
Eric Dumazet02c22342016-04-27 16:44:30 -07001694 __UDP_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
1695 proto == IPPROTO_UDPLITE);
Eric Dumazet1240d132009-11-08 10:18:44 +00001696 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001697 return 0;
1698}
1699
1700/* Initialize UDP checksum. If exited with zero value (success),
1701 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1702 * Otherwise, csum completion requires chacksumming packet body,
1703 * including udp header and folding it to skb->csum.
1704 */
1705static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1706 int proto)
1707{
David S. Millerdb8dac22008-03-06 16:22:02 -08001708 int err;
1709
1710 UDP_SKB_CB(skb)->partial_cov = 0;
1711 UDP_SKB_CB(skb)->cscov = skb->len;
1712
1713 if (proto == IPPROTO_UDPLITE) {
1714 err = udplite_checksum_init(skb, uh);
1715 if (err)
1716 return err;
1717 }
1718
Hannes Frederic Sowab46d9f62016-06-12 12:02:46 +02001719 /* Note, we are only interested in != 0 or == 0, thus the
1720 * force to int.
1721 */
1722 return (__force int)skb_checksum_init_zero_check(skb, proto, uh->check,
1723 inet_compute_pseudo);
David S. Millerdb8dac22008-03-06 16:22:02 -08001724}
1725
1726/*
1727 * All we need to do is get the socket, and then do a checksum.
1728 */
1729
Eric Dumazet645ca702008-10-29 01:41:45 -07001730int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001731 int proto)
1732{
1733 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001734 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001735 unsigned short ulen;
Eric Dumazetadf30902009-06-02 05:19:30 +00001736 struct rtable *rt = skb_rtable(skb);
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001737 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001738 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001739
1740 /*
1741 * Validate the packet.
1742 */
1743 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1744 goto drop; /* No space for header. */
1745
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001746 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001747 ulen = ntohs(uh->len);
Bjørn Morkccc2d972010-05-06 03:44:34 +00001748 saddr = ip_hdr(skb)->saddr;
1749 daddr = ip_hdr(skb)->daddr;
1750
David S. Millerdb8dac22008-03-06 16:22:02 -08001751 if (ulen > skb->len)
1752 goto short_packet;
1753
1754 if (proto == IPPROTO_UDP) {
1755 /* UDP validates ulen. */
1756 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1757 goto short_packet;
1758 uh = udp_hdr(skb);
1759 }
1760
1761 if (udp4_csum_init(skb, uh, proto))
1762 goto csum_error;
1763
Eric Dumazet8afdd992013-12-10 18:07:23 -08001764 sk = skb_steal_sock(skb);
1765 if (sk) {
Eric Dumazet97502232013-12-11 14:46:51 -08001766 struct dst_entry *dst = skb_dst(skb);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001767 int ret;
David S. Millerdb8dac22008-03-06 16:22:02 -08001768
Eric Dumazet97502232013-12-11 14:46:51 -08001769 if (unlikely(sk->sk_rx_dst != dst))
1770 udp_sk_rx_dst_set(sk, dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001771
1772 ret = udp_queue_rcv_skb(sk, skb);
Eric Dumazet8afdd992013-12-10 18:07:23 -08001773 sock_put(sk);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001774 /* a return value > 0 means to resubmit the input, but
1775 * it wants the return to be -protocol, or 0
1776 */
1777 if (ret > 0)
1778 return -ret;
1779 return 0;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001780 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001781
Fabian Frederickc18450a2014-11-04 20:48:41 +01001782 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1783 return __udp4_lib_mcast_deliver(net, skb, uh,
Rick Jones36cbb242014-11-06 10:37:54 -08001784 saddr, daddr, udptable, proto);
Fabian Frederickc18450a2014-11-04 20:48:41 +01001785
1786 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
Ian Morris00db4122015-04-03 09:17:27 +01001787 if (sk) {
Eliezer Tamira5b50472013-06-10 11:40:00 +03001788 int ret;
1789
Tom Herbert224d0192015-01-05 13:56:14 -08001790 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
Tom Herbert2abb7cd2014-08-31 15:12:43 -07001791 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
1792 inet_compute_pseudo);
1793
Eliezer Tamira5b50472013-06-10 11:40:00 +03001794 ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001795
1796 /* a return value > 0 means to resubmit the input, but
1797 * it wants the return to be -protocol, or 0
1798 */
1799 if (ret > 0)
1800 return -ret;
1801 return 0;
1802 }
1803
1804 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1805 goto drop;
1806 nf_reset(skb);
1807
1808 /* No socket. Drop packet silently, if checksum is wrong */
1809 if (udp_lib_checksum_complete(skb))
1810 goto csum_error;
1811
Eric Dumazet02c22342016-04-27 16:44:30 -07001812 __UDP_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001813 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1814
1815 /*
1816 * Hmm. We got an UDP packet to a port to which we
1817 * don't wanna listen. Ignore it.
1818 */
1819 kfree_skb(skb);
1820 return 0;
1821
1822short_packet:
Joe Perchesba7a46f2014-11-11 10:59:17 -08001823 net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
1824 proto == IPPROTO_UDPLITE ? "Lite" : "",
1825 &saddr, ntohs(uh->source),
1826 ulen, skb->len,
1827 &daddr, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001828 goto drop;
1829
1830csum_error:
1831 /*
1832 * RFC1122: OK. Discards the bad packet silently (as far as
1833 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1834 */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001835 net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
1836 proto == IPPROTO_UDPLITE ? "Lite" : "",
1837 &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
1838 ulen);
Eric Dumazet02c22342016-04-27 16:44:30 -07001839 __UDP_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001840drop:
Eric Dumazet02c22342016-04-27 16:44:30 -07001841 __UDP_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001842 kfree_skb(skb);
1843 return 0;
1844}
1845
Shawn Bohrer421b3882013-10-07 11:01:39 -05001846/* We can only early demux multicast if there is a single matching socket.
1847 * If more than one socket found returns NULL
1848 */
1849static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net,
1850 __be16 loc_port, __be32 loc_addr,
1851 __be16 rmt_port, __be32 rmt_addr,
1852 int dif)
1853{
1854 struct sock *sk, *result;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001855 unsigned short hnum = ntohs(loc_port);
Eric Dumazetca065d02016-04-01 08:52:13 -07001856 unsigned int slot = udp_hashfn(net, hnum, udp_table.mask);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001857 struct udp_hslot *hslot = &udp_table.hash[slot];
1858
Eric Dumazet63c6f812014-06-12 16:13:06 -07001859 /* Do not bother scanning a too big list */
1860 if (hslot->count > 10)
1861 return NULL;
1862
Shawn Bohrer421b3882013-10-07 11:01:39 -05001863 result = NULL;
Eric Dumazetca065d02016-04-01 08:52:13 -07001864 sk_for_each_rcu(sk, &hslot->head) {
1865 if (__udp_is_mcast_sock(net, sk, loc_port, loc_addr,
1866 rmt_port, rmt_addr, dif, hnum)) {
1867 if (result)
1868 return NULL;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001869 result = sk;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001870 }
1871 }
Shawn Bohrer421b3882013-10-07 11:01:39 -05001872
Shawn Bohrer421b3882013-10-07 11:01:39 -05001873 return result;
1874}
1875
1876/* For unicast we should only early demux connected sockets or we can
1877 * break forwarding setups. The chains here can be long so only check
1878 * if the first socket is an exact match and if not move on.
1879 */
1880static struct sock *__udp4_lib_demux_lookup(struct net *net,
1881 __be16 loc_port, __be32 loc_addr,
1882 __be16 rmt_port, __be32 rmt_addr,
1883 int dif)
1884{
Shawn Bohrer421b3882013-10-07 11:01:39 -05001885 unsigned short hnum = ntohs(loc_port);
1886 unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum);
1887 unsigned int slot2 = hash2 & udp_table.mask;
1888 struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
Joe Perchesc7228312014-05-13 20:30:07 -07001889 INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001890 const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
Eric Dumazetca065d02016-04-01 08:52:13 -07001891 struct sock *sk;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001892
Eric Dumazetca065d02016-04-01 08:52:13 -07001893 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
1894 if (INET_MATCH(sk, net, acookie, rmt_addr,
1895 loc_addr, ports, dif))
1896 return sk;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001897 /* Only check first socket in chain */
1898 break;
1899 }
Eric Dumazetca065d02016-04-01 08:52:13 -07001900 return NULL;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001901}
1902
1903void udp_v4_early_demux(struct sk_buff *skb)
1904{
Eric Dumazet610438b2013-12-11 08:10:05 -08001905 struct net *net = dev_net(skb->dev);
1906 const struct iphdr *iph;
1907 const struct udphdr *uh;
Eric Dumazetca065d02016-04-01 08:52:13 -07001908 struct sock *sk = NULL;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001909 struct dst_entry *dst;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001910 int dif = skb->dev->ifindex;
Shawn Bohrer6e540302015-06-03 16:27:38 -05001911 int ours;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001912
1913 /* validate the packet */
1914 if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr)))
1915 return;
1916
Eric Dumazet610438b2013-12-11 08:10:05 -08001917 iph = ip_hdr(skb);
1918 uh = udp_hdr(skb);
1919
Shawn Bohrer421b3882013-10-07 11:01:39 -05001920 if (skb->pkt_type == PACKET_BROADCAST ||
Shawn Bohrer6e540302015-06-03 16:27:38 -05001921 skb->pkt_type == PACKET_MULTICAST) {
1922 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
1923
1924 if (!in_dev)
1925 return;
1926
Paolo Abeniad0ea192016-03-22 09:19:38 +01001927 /* we are supposed to accept bcast packets */
1928 if (skb->pkt_type == PACKET_MULTICAST) {
1929 ours = ip_check_mc_rcu(in_dev, iph->daddr, iph->saddr,
1930 iph->protocol);
1931 if (!ours)
1932 return;
1933 }
1934
Shawn Bohrer421b3882013-10-07 11:01:39 -05001935 sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr,
1936 uh->source, iph->saddr, dif);
Shawn Bohrer6e540302015-06-03 16:27:38 -05001937 } else if (skb->pkt_type == PACKET_HOST) {
Shawn Bohrer421b3882013-10-07 11:01:39 -05001938 sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr,
1939 uh->source, iph->saddr, dif);
Shawn Bohrer6e540302015-06-03 16:27:38 -05001940 }
Shawn Bohrer421b3882013-10-07 11:01:39 -05001941
Eric Dumazetca065d02016-04-01 08:52:13 -07001942 if (!sk || !atomic_inc_not_zero_hint(&sk->sk_refcnt, 2))
Shawn Bohrer421b3882013-10-07 11:01:39 -05001943 return;
1944
1945 skb->sk = sk;
Alexander Duyck82eabd92014-09-04 13:32:11 -04001946 skb->destructor = sock_efree;
Eric Dumazet10e2eb82015-08-01 12:14:33 +02001947 dst = READ_ONCE(sk->sk_rx_dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001948
1949 if (dst)
1950 dst = dst_check(dst, 0);
Eric Dumazet10e2eb82015-08-01 12:14:33 +02001951 if (dst) {
1952 /* DST_NOCACHE can not be used without taking a reference */
1953 if (dst->flags & DST_NOCACHE) {
1954 if (likely(atomic_inc_not_zero(&dst->__refcnt)))
1955 skb_dst_set(skb, dst);
1956 } else {
1957 skb_dst_set_noref(skb, dst);
1958 }
1959 }
Shawn Bohrer421b3882013-10-07 11:01:39 -05001960}
1961
David S. Millerdb8dac22008-03-06 16:22:02 -08001962int udp_rcv(struct sk_buff *skb)
1963{
Eric Dumazet645ca702008-10-29 01:41:45 -07001964 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08001965}
1966
Brian Haley7d06b2e2008-06-14 17:04:49 -07001967void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08001968{
Tom Parkin44046a52013-03-19 06:11:12 +00001969 struct udp_sock *up = udp_sk(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001970 bool slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001971 udp_flush_pending_frames(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001972 unlock_sock_fast(sk, slow);
Tom Parkin44046a52013-03-19 06:11:12 +00001973 if (static_key_false(&udp_encap_needed) && up->encap_type) {
1974 void (*encap_destroy)(struct sock *sk);
1975 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1976 if (encap_destroy)
1977 encap_destroy(sk);
1978 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001979}
1980
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981/*
1982 * Socket option code for UDP
1983 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001984int udp_lib_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001985 char __user *optval, unsigned int optlen,
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001986 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987{
1988 struct udp_sock *up = udp_sk(sk);
Tom Herbert1c194482014-05-23 08:47:32 -07001989 int val, valbool;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11001991 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992
Eric Dumazetc482c562009-07-17 00:26:32 +00001993 if (optlen < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994 return -EINVAL;
1995
1996 if (get_user(val, (int __user *)optval))
1997 return -EFAULT;
1998
Tom Herbert1c194482014-05-23 08:47:32 -07001999 valbool = val ? 1 : 0;
2000
Stephen Hemminger6516c652007-03-08 20:41:55 -08002001 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 case UDP_CORK:
2003 if (val != 0) {
2004 up->corkflag = 1;
2005 } else {
2006 up->corkflag = 0;
2007 lock_sock(sk);
Joe Perches4243cdc2014-11-11 21:59:20 -08002008 push_pending_frames(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 release_sock(sk);
2010 }
2011 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002012
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 case UDP_ENCAP:
2014 switch (val) {
2015 case 0:
2016 case UDP_ENCAP_ESPINUDP:
2017 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07002018 up->encap_rcv = xfrm4_udp_encap_rcv;
2019 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07002020 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 up->encap_type = val;
Eric Dumazet447167b2012-04-11 23:05:28 +00002022 udp_encap_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 break;
2024 default:
2025 err = -ENOPROTOOPT;
2026 break;
2027 }
2028 break;
2029
Tom Herbert1c194482014-05-23 08:47:32 -07002030 case UDP_NO_CHECK6_TX:
2031 up->no_check6_tx = valbool;
2032 break;
2033
2034 case UDP_NO_CHECK6_RX:
2035 up->no_check6_rx = valbool;
2036 break;
2037
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002038 /*
2039 * UDP-Lite's partial checksum coverage (RFC 3828).
2040 */
2041 /* The sender sets actual checksum coverage length via this option.
2042 * The case coverage > packet length is handled by send module. */
2043 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002044 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002045 return -ENOPROTOOPT;
2046 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
2047 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002048 else if (val > USHRT_MAX)
2049 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002050 up->pcslen = val;
2051 up->pcflag |= UDPLITE_SEND_CC;
2052 break;
2053
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002054 /* The receiver specifies a minimum checksum coverage value. To make
2055 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002056 * used, this again means full checksum coverage. */
2057 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002058 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002059 return -ENOPROTOOPT;
2060 if (val != 0 && val < 8) /* Avoid silly minimal values. */
2061 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002062 else if (val > USHRT_MAX)
2063 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002064 up->pcrlen = val;
2065 up->pcflag |= UDPLITE_RECV_CC;
2066 break;
2067
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068 default:
2069 err = -ENOPROTOOPT;
2070 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002071 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072
2073 return err;
2074}
Eric Dumazetc482c562009-07-17 00:26:32 +00002075EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076
David S. Millerdb8dac22008-03-06 16:22:02 -08002077int udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002078 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002079{
2080 if (level == SOL_UDP || level == SOL_UDPLITE)
2081 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2082 udp_push_pending_frames);
2083 return ip_setsockopt(sk, level, optname, optval, optlen);
2084}
2085
2086#ifdef CONFIG_COMPAT
2087int compat_udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002088 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002089{
2090 if (level == SOL_UDP || level == SOL_UDPLITE)
2091 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2092 udp_push_pending_frames);
2093 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
2094}
2095#endif
2096
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002097int udp_lib_getsockopt(struct sock *sk, int level, int optname,
2098 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099{
2100 struct udp_sock *up = udp_sk(sk);
2101 int val, len;
2102
Eric Dumazetc482c562009-07-17 00:26:32 +00002103 if (get_user(len, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104 return -EFAULT;
2105
2106 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002107
Stephen Hemminger6516c652007-03-08 20:41:55 -08002108 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109 return -EINVAL;
2110
Stephen Hemminger6516c652007-03-08 20:41:55 -08002111 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112 case UDP_CORK:
2113 val = up->corkflag;
2114 break;
2115
2116 case UDP_ENCAP:
2117 val = up->encap_type;
2118 break;
2119
Tom Herbert1c194482014-05-23 08:47:32 -07002120 case UDP_NO_CHECK6_TX:
2121 val = up->no_check6_tx;
2122 break;
2123
2124 case UDP_NO_CHECK6_RX:
2125 val = up->no_check6_rx;
2126 break;
2127
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002128 /* The following two cannot be changed on UDP sockets, the return is
2129 * always 0 (which corresponds to the full checksum coverage of UDP). */
2130 case UDPLITE_SEND_CSCOV:
2131 val = up->pcslen;
2132 break;
2133
2134 case UDPLITE_RECV_CSCOV:
2135 val = up->pcrlen;
2136 break;
2137
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138 default:
2139 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002140 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141
Stephen Hemminger6516c652007-03-08 20:41:55 -08002142 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002143 return -EFAULT;
Eric Dumazetc482c562009-07-17 00:26:32 +00002144 if (copy_to_user(optval, &val, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002146 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147}
Eric Dumazetc482c562009-07-17 00:26:32 +00002148EXPORT_SYMBOL(udp_lib_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149
David S. Millerdb8dac22008-03-06 16:22:02 -08002150int udp_getsockopt(struct sock *sk, int level, int optname,
2151 char __user *optval, int __user *optlen)
2152{
2153 if (level == SOL_UDP || level == SOL_UDPLITE)
2154 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2155 return ip_getsockopt(sk, level, optname, optval, optlen);
2156}
2157
2158#ifdef CONFIG_COMPAT
2159int compat_udp_getsockopt(struct sock *sk, int level, int optname,
2160 char __user *optval, int __user *optlen)
2161{
2162 if (level == SOL_UDP || level == SOL_UDPLITE)
2163 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2164 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
2165}
2166#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167/**
2168 * udp_poll - wait for a UDP event.
2169 * @file - file struct
2170 * @sock - socket
2171 * @wait - poll table
2172 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002173 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07002174 * blocking sockets. If application is using a blocking fd
2175 * and a packet with checksum error is in the queue;
2176 * then it could get return from select indicating data available
2177 * but then block when reading it. Add special case code
2178 * to work around these arguably broken applications.
2179 */
2180unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
2181{
2182 unsigned int mask = datagram_poll(file, sock, wait);
2183 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002184
David Majnemerc3f1dba2013-05-31 13:15:38 +00002185 sock_rps_record_flow(sk);
2186
Linus Torvalds1da177e2005-04-16 15:20:36 -07002187 /* Check for false positives due to checksum errors */
Eric Dumazet85584672009-10-09 04:43:40 +00002188 if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
2189 !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
2190 mask &= ~(POLLIN | POLLRDNORM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191
2192 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002193
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194}
Eric Dumazetc482c562009-07-17 00:26:32 +00002195EXPORT_SYMBOL(udp_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196
David S. Millerdb8dac22008-03-06 16:22:02 -08002197struct proto udp_prot = {
2198 .name = "UDP",
2199 .owner = THIS_MODULE,
2200 .close = udp_lib_close,
2201 .connect = ip4_datagram_connect,
2202 .disconnect = udp_disconnect,
2203 .ioctl = udp_ioctl,
2204 .destroy = udp_destroy_sock,
2205 .setsockopt = udp_setsockopt,
2206 .getsockopt = udp_getsockopt,
2207 .sendmsg = udp_sendmsg,
2208 .recvmsg = udp_recvmsg,
2209 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07002210 .backlog_rcv = __udp_queue_rcv_skb,
Steffen Klassert8141ed92013-01-21 02:00:03 +00002211 .release_cb = ip4_datagram_release_cb,
David S. Millerdb8dac22008-03-06 16:22:02 -08002212 .hash = udp_lib_hash,
2213 .unhash = udp_lib_unhash,
Eric Dumazet719f8352010-09-08 05:08:44 +00002214 .rehash = udp_v4_rehash,
David S. Millerdb8dac22008-03-06 16:22:02 -08002215 .get_port = udp_v4_get_port,
2216 .memory_allocated = &udp_memory_allocated,
2217 .sysctl_mem = sysctl_udp_mem,
2218 .sysctl_wmem = &sysctl_udp_wmem_min,
2219 .sysctl_rmem = &sysctl_udp_rmem_min,
2220 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07002221 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07002222 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08002223#ifdef CONFIG_COMPAT
2224 .compat_setsockopt = compat_udp_setsockopt,
2225 .compat_getsockopt = compat_udp_getsockopt,
2226#endif
Octavian Purdilafcbdf092010-12-16 14:26:56 -08002227 .clear_sk = sk_prot_clear_portaddr_nulls,
David S. Millerdb8dac22008-03-06 16:22:02 -08002228};
Eric Dumazetc482c562009-07-17 00:26:32 +00002229EXPORT_SYMBOL(udp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230
2231/* ------------------------------------------------------------------------ */
2232#ifdef CONFIG_PROC_FS
2233
Eric Dumazet645ca702008-10-29 01:41:45 -07002234static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235{
2236 struct sock *sk;
2237 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002238 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002239
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002240 for (state->bucket = start; state->bucket <= state->udp_table->mask;
2241 ++state->bucket) {
Eric Dumazet645ca702008-10-29 01:41:45 -07002242 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002243
Eric Dumazetca065d02016-04-01 08:52:13 -07002244 if (hlist_empty(&hslot->head))
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002245 continue;
2246
Eric Dumazet645ca702008-10-29 01:41:45 -07002247 spin_lock_bh(&hslot->lock);
Eric Dumazetca065d02016-04-01 08:52:13 -07002248 sk_for_each(sk, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002249 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002250 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002251 if (sk->sk_family == state->family)
2252 goto found;
2253 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002254 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002255 }
2256 sk = NULL;
2257found:
2258 return sk;
2259}
2260
2261static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2262{
2263 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002264 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002265
2266 do {
Eric Dumazetca065d02016-04-01 08:52:13 -07002267 sk = sk_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002268 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002269
Eric Dumazet645ca702008-10-29 01:41:45 -07002270 if (!sk) {
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002271 if (state->bucket <= state->udp_table->mask)
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002272 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07002273 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274 }
2275 return sk;
2276}
2277
2278static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2279{
Eric Dumazet645ca702008-10-29 01:41:45 -07002280 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281
2282 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08002283 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284 --pos;
2285 return pos ? NULL : sk;
2286}
2287
2288static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2289{
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002290 struct udp_iter_state *state = seq->private;
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002291 state->bucket = MAX_UDP_PORTS;
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002292
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002293 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294}
2295
2296static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2297{
2298 struct sock *sk;
2299
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002300 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301 sk = udp_get_idx(seq, 0);
2302 else
2303 sk = udp_get_next(seq, v);
2304
2305 ++*pos;
2306 return sk;
2307}
2308
2309static void udp_seq_stop(struct seq_file *seq, void *v)
2310{
Eric Dumazet645ca702008-10-29 01:41:45 -07002311 struct udp_iter_state *state = seq->private;
2312
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002313 if (state->bucket <= state->udp_table->mask)
Eric Dumazet645ca702008-10-29 01:41:45 -07002314 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315}
2316
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002317int udp_seq_open(struct inode *inode, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318{
Al Virod9dda782013-03-31 18:16:14 -04002319 struct udp_seq_afinfo *afinfo = PDE_DATA(inode);
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002320 struct udp_iter_state *s;
2321 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002322
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002323 err = seq_open_net(inode, file, &afinfo->seq_ops,
2324 sizeof(struct udp_iter_state));
2325 if (err < 0)
2326 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002327
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002328 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07002330 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002331 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002332}
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002333EXPORT_SYMBOL(udp_seq_open);
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002334
Linus Torvalds1da177e2005-04-16 15:20:36 -07002335/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002336int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337{
2338 struct proc_dir_entry *p;
2339 int rc = 0;
2340
Denis V. Lunevdda61922008-03-28 18:24:26 -07002341 afinfo->seq_ops.start = udp_seq_start;
2342 afinfo->seq_ops.next = udp_seq_next;
2343 afinfo->seq_ops.stop = udp_seq_stop;
2344
Denis V. Lunev84841c32008-05-02 04:10:08 -07002345 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002346 afinfo->seq_fops, afinfo);
Denis V. Lunev84841c32008-05-02 04:10:08 -07002347 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348 rc = -ENOMEM;
2349 return rc;
2350}
Eric Dumazetc482c562009-07-17 00:26:32 +00002351EXPORT_SYMBOL(udp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002353void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002354{
Gao fengece31ff2013-02-18 01:34:56 +00002355 remove_proc_entry(afinfo->name, net->proc_net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356}
Eric Dumazetc482c562009-07-17 00:26:32 +00002357EXPORT_SYMBOL(udp_proc_unregister);
David S. Millerdb8dac22008-03-06 16:22:02 -08002358
2359/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002360static void udp4_format_sock(struct sock *sp, struct seq_file *f,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002361 int bucket)
David S. Millerdb8dac22008-03-06 16:22:02 -08002362{
2363 struct inet_sock *inet = inet_sk(sp);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002364 __be32 dest = inet->inet_daddr;
2365 __be32 src = inet->inet_rcv_saddr;
2366 __u16 destp = ntohs(inet->inet_dport);
2367 __u16 srcp = ntohs(inet->inet_sport);
David S. Millerdb8dac22008-03-06 16:22:02 -08002368
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002369 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
Tetsuo Handa652586d2013-11-14 14:31:57 -08002370 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
David S. Millerdb8dac22008-03-06 16:22:02 -08002371 bucket, src, srcp, dest, destp, sp->sk_state,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002372 sk_wmem_alloc_get(sp),
2373 sk_rmem_alloc_get(sp),
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06002374 0, 0L, 0,
2375 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2376 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002377 atomic_read(&sp->sk_refcnt), sp,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002378 atomic_read(&sp->sk_drops));
David S. Millerdb8dac22008-03-06 16:22:02 -08002379}
2380
2381int udp4_seq_show(struct seq_file *seq, void *v)
2382{
Tetsuo Handa652586d2013-11-14 14:31:57 -08002383 seq_setwidth(seq, 127);
David S. Millerdb8dac22008-03-06 16:22:02 -08002384 if (v == SEQ_START_TOKEN)
Tetsuo Handa652586d2013-11-14 14:31:57 -08002385 seq_puts(seq, " sl local_address rem_address st tx_queue "
David S. Millerdb8dac22008-03-06 16:22:02 -08002386 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002387 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08002388 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08002389 struct udp_iter_state *state = seq->private;
2390
Tetsuo Handa652586d2013-11-14 14:31:57 -08002391 udp4_format_sock(v, seq, state->bucket);
David S. Millerdb8dac22008-03-06 16:22:02 -08002392 }
Tetsuo Handa652586d2013-11-14 14:31:57 -08002393 seq_pad(seq, '\n');
David S. Millerdb8dac22008-03-06 16:22:02 -08002394 return 0;
2395}
2396
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002397static const struct file_operations udp_afinfo_seq_fops = {
2398 .owner = THIS_MODULE,
2399 .open = udp_seq_open,
2400 .read = seq_read,
2401 .llseek = seq_lseek,
2402 .release = seq_release_net
2403};
2404
David S. Millerdb8dac22008-03-06 16:22:02 -08002405/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08002406static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08002407 .name = "udp",
2408 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07002409 .udp_table = &udp_table,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002410 .seq_fops = &udp_afinfo_seq_fops,
Denis V. Lunevdda61922008-03-28 18:24:26 -07002411 .seq_ops = {
2412 .show = udp4_seq_show,
2413 },
David S. Millerdb8dac22008-03-06 16:22:02 -08002414};
2415
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002416static int __net_init udp4_proc_init_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002417{
2418 return udp_proc_register(net, &udp4_seq_afinfo);
2419}
2420
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002421static void __net_exit udp4_proc_exit_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002422{
2423 udp_proc_unregister(net, &udp4_seq_afinfo);
2424}
2425
2426static struct pernet_operations udp4_net_ops = {
2427 .init = udp4_proc_init_net,
2428 .exit = udp4_proc_exit_net,
2429};
2430
David S. Millerdb8dac22008-03-06 16:22:02 -08002431int __init udp4_proc_init(void)
2432{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002433 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002434}
2435
2436void udp4_proc_exit(void)
2437{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002438 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002439}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440#endif /* CONFIG_PROC_FS */
2441
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002442static __initdata unsigned long uhash_entries;
2443static int __init set_uhash_entries(char *str)
Eric Dumazet645ca702008-10-29 01:41:45 -07002444{
Eldad Zack413c27d2012-05-19 14:13:18 +00002445 ssize_t ret;
2446
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002447 if (!str)
2448 return 0;
Eldad Zack413c27d2012-05-19 14:13:18 +00002449
2450 ret = kstrtoul(str, 0, &uhash_entries);
2451 if (ret)
2452 return 0;
2453
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002454 if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2455 uhash_entries = UDP_HTABLE_SIZE_MIN;
2456 return 1;
2457}
2458__setup("uhash_entries=", set_uhash_entries);
Eric Dumazet645ca702008-10-29 01:41:45 -07002459
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002460void __init udp_table_init(struct udp_table *table, const char *name)
2461{
2462 unsigned int i;
2463
Tim Bird31fe62b2012-05-23 13:33:35 +00002464 table->hash = alloc_large_system_hash(name,
2465 2 * sizeof(struct udp_hslot),
2466 uhash_entries,
2467 21, /* one slot per 2 MB */
2468 0,
2469 &table->log,
2470 &table->mask,
2471 UDP_HTABLE_SIZE_MIN,
2472 64 * 1024);
2473
Eric Dumazet512615b2009-11-08 10:17:58 +00002474 table->hash2 = table->hash + (table->mask + 1);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002475 for (i = 0; i <= table->mask; i++) {
Eric Dumazetca065d02016-04-01 08:52:13 -07002476 INIT_HLIST_HEAD(&table->hash[i].head);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00002477 table->hash[i].count = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -07002478 spin_lock_init(&table->hash[i].lock);
2479 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002480 for (i = 0; i <= table->mask; i++) {
Eric Dumazetca065d02016-04-01 08:52:13 -07002481 INIT_HLIST_HEAD(&table->hash2[i].head);
Eric Dumazet512615b2009-11-08 10:17:58 +00002482 table->hash2[i].count = 0;
2483 spin_lock_init(&table->hash2[i].lock);
2484 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002485}
2486
Tom Herbert723b8e42015-02-24 09:17:31 -08002487u32 udp_flow_hashrnd(void)
2488{
2489 static u32 hashrnd __read_mostly;
2490
2491 net_get_random_once(&hashrnd, sizeof(hashrnd));
2492
2493 return hashrnd;
2494}
2495EXPORT_SYMBOL(udp_flow_hashrnd);
2496
Hideo Aoki95766ff2007-12-31 00:29:24 -08002497void __init udp_init(void)
2498{
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002499 unsigned long limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002500
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002501 udp_table_init(&udp_table, "UDP");
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002502 limit = nr_free_buffer_pages() / 8;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002503 limit = max(limit, 128UL);
2504 sysctl_udp_mem[0] = limit / 4 * 3;
2505 sysctl_udp_mem[1] = limit;
2506 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2507
2508 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2509 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2510}