blob: 093284c5c03be73c6871fd6ee3dc5d1c0f1ad59d [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
Joe Perches60c04ae2014-12-01 20:29:06 -0800394static inline int compute_score(struct sock *sk, struct net *net,
395 __be32 saddr, unsigned short hnum, __be16 sport,
396 __be32 daddr, __be16 dport, 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 Dumazet5051ebd2009-11-08 10:18:11 +0000437/*
438 * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num)
439 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000440static inline int compute_score2(struct sock *sk, struct net *net,
441 __be32 saddr, __be16 sport,
442 __be32 daddr, unsigned int hnum, int dif)
443{
Joe Perches60c04ae2014-12-01 20:29:06 -0800444 int score;
445 struct inet_sock *inet;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000446
Joe Perches60c04ae2014-12-01 20:29:06 -0800447 if (!net_eq(sock_net(sk), net) ||
448 ipv6_only_sock(sk))
449 return -1;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000450
Joe Perches60c04ae2014-12-01 20:29:06 -0800451 inet = inet_sk(sk);
452
453 if (inet->inet_rcv_saddr != daddr ||
454 inet->inet_num != hnum)
455 return -1;
456
457 score = (sk->sk_family == PF_INET) ? 2 : 1;
458
459 if (inet->inet_daddr) {
460 if (inet->inet_daddr != saddr)
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000461 return -1;
Joe Perches60c04ae2014-12-01 20:29:06 -0800462 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000463 }
Joe Perches60c04ae2014-12-01 20:29:06 -0800464
465 if (inet->inet_dport) {
466 if (inet->inet_dport != sport)
467 return -1;
468 score += 4;
469 }
470
471 if (sk->sk_bound_dev_if) {
472 if (sk->sk_bound_dev_if != dif)
473 return -1;
474 score += 4;
475 }
476
Eric Dumazet70da2682015-10-08 19:33:21 -0700477 if (sk->sk_incoming_cpu == raw_smp_processor_id())
478 score++;
479
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000480 return score;
481}
482
Eric Dumazet6eada012015-03-18 14:05:33 -0700483static u32 udp_ehashfn(const struct net *net, const __be32 laddr,
484 const __u16 lport, const __be32 faddr,
485 const __be16 fport)
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200486{
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200487 static u32 udp_ehash_secret __read_mostly;
488
489 net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret));
490
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200491 return __inet_ehashfn(laddr, lport, faddr, fport,
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200492 udp_ehash_secret + net_hash_mix(net));
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200493}
494
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000495/* called with read_rcu_lock() */
496static struct sock *udp4_lib_lookup2(struct net *net,
497 __be32 saddr, __be16 sport,
498 __be32 daddr, unsigned int hnum, int dif,
Craig Gallek11341582016-01-05 15:08:07 -0500499 struct udp_hslot *hslot2, unsigned int slot2,
500 struct sk_buff *skb)
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000501{
502 struct sock *sk, *result;
Tom Herbertba418fa2013-01-22 09:50:32 +0000503 int score, badness, matches = 0, reuseport = 0;
504 u32 hash = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000505
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000506 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000507 badness = 0;
Eric Dumazetca065d02016-04-01 08:52:13 -0700508 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000509 score = compute_score2(sk, net, saddr, sport,
510 daddr, hnum, dif);
511 if (score > badness) {
Tom Herbertba418fa2013-01-22 09:50:32 +0000512 reuseport = sk->sk_reuseport;
513 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200514 hash = udp_ehashfn(net, daddr, hnum,
515 saddr, sport);
Eric Dumazetca065d02016-04-01 08:52:13 -0700516 result = reuseport_select_sock(sk, hash, skb,
Eric Dumazeted0dfff2016-01-19 08:36:43 -0800517 sizeof(struct udphdr));
Eric Dumazetca065d02016-04-01 08:52:13 -0700518 if (result)
519 return result;
Tom Herbertba418fa2013-01-22 09:50:32 +0000520 matches = 1;
521 }
Eric Dumazetca065d02016-04-01 08:52:13 -0700522 badness = score;
523 result = sk;
Tom Herbertba418fa2013-01-22 09:50:32 +0000524 } else if (score == badness && reuseport) {
525 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200526 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000527 result = sk;
528 hash = next_pseudo_random32(hash);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000529 }
530 }
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000531 return result;
532}
533
David S. Millerdb8dac22008-03-06 16:22:02 -0800534/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
535 * harder than this. -DaveM
536 */
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000537struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -0800538 __be16 sport, __be32 daddr, __be16 dport,
Craig Gallek538950a2016-01-04 17:41:47 -0500539 int dif, struct udp_table *udptable, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800540{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700541 struct sock *sk, *result;
David S. Millerdb8dac22008-03-06 16:22:02 -0800542 unsigned short hnum = ntohs(dport);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000543 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
544 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
Tom Herbertba418fa2013-01-22 09:50:32 +0000545 int score, badness, matches = 0, reuseport = 0;
546 u32 hash = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800547
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000548 if (hslot->count > 10) {
549 hash2 = udp4_portaddr_hash(net, daddr, hnum);
550 slot2 = hash2 & udptable->mask;
551 hslot2 = &udptable->hash2[slot2];
552 if (hslot->count < hslot2->count)
553 goto begin;
554
555 result = udp4_lib_lookup2(net, saddr, sport,
556 daddr, hnum, dif,
Craig Gallek11341582016-01-05 15:08:07 -0500557 hslot2, slot2, skb);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000558 if (!result) {
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700559 hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000560 slot2 = hash2 & udptable->mask;
561 hslot2 = &udptable->hash2[slot2];
562 if (hslot->count < hslot2->count)
563 goto begin;
564
Jorge Boncompte [DTI2]1223c672010-04-08 04:56:48 +0000565 result = udp4_lib_lookup2(net, saddr, sport,
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700566 htonl(INADDR_ANY), hnum, dif,
Craig Gallek11341582016-01-05 15:08:07 -0500567 hslot2, slot2, skb);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000568 }
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000569 return result;
570 }
Eric Dumazet271b72c2008-10-29 02:11:14 -0700571begin:
572 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000573 badness = 0;
Eric Dumazetca065d02016-04-01 08:52:13 -0700574 sk_for_each_rcu(sk, &hslot->head) {
Eric Dumazet645ca702008-10-29 01:41:45 -0700575 score = compute_score(sk, net, saddr, hnum, sport,
576 daddr, dport, dif);
577 if (score > badness) {
Tom Herbertba418fa2013-01-22 09:50:32 +0000578 reuseport = sk->sk_reuseport;
579 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200580 hash = udp_ehashfn(net, daddr, hnum,
581 saddr, sport);
Eric Dumazetca065d02016-04-01 08:52:13 -0700582 result = reuseport_select_sock(sk, hash, skb,
Craig Gallek538950a2016-01-04 17:41:47 -0500583 sizeof(struct udphdr));
Eric Dumazetca065d02016-04-01 08:52:13 -0700584 if (result)
585 return result;
Tom Herbertba418fa2013-01-22 09:50:32 +0000586 matches = 1;
587 }
Eric Dumazetca065d02016-04-01 08:52:13 -0700588 result = sk;
589 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000590 } else if (score == badness && reuseport) {
591 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200592 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000593 result = sk;
594 hash = next_pseudo_random32(hash);
David S. Millerdb8dac22008-03-06 16:22:02 -0800595 }
596 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800597 return result;
598}
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000599EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
David S. Millerdb8dac22008-03-06 16:22:02 -0800600
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700601static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
602 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700603 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700604{
605 const struct iphdr *iph = ip_hdr(skb);
606
Eric Dumazet8afdd992013-12-10 18:07:23 -0800607 return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport,
608 iph->daddr, dport, inet_iif(skb),
Craig Gallek538950a2016-01-04 17:41:47 -0500609 udptable, skb);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700610}
611
Tom Herbert63058302016-04-05 08:22:50 -0700612struct sock *udp4_lib_lookup_skb(struct sk_buff *skb,
613 __be16 sport, __be16 dport)
614{
615 const struct iphdr *iph = ip_hdr(skb);
616 const struct net_device *dev =
617 skb_dst(skb) ? skb_dst(skb)->dev : skb->dev;
618
619 return __udp4_lib_lookup(dev_net(dev), iph->saddr, sport,
620 iph->daddr, dport, inet_iif(skb),
621 &udp_table, skb);
622}
623EXPORT_SYMBOL_GPL(udp4_lib_lookup_skb);
624
Eric Dumazetca065d02016-04-01 08:52:13 -0700625/* Must be called under rcu_read_lock().
626 * Does increment socket refcount.
627 */
628#if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \
629 IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY)
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700630struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
631 __be32 daddr, __be16 dport, int dif)
632{
Eric Dumazetca065d02016-04-01 08:52:13 -0700633 struct sock *sk;
634
635 sk = __udp4_lib_lookup(net, saddr, sport, daddr, dport,
636 dif, &udp_table, NULL);
637 if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
638 sk = NULL;
639 return sk;
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700640}
641EXPORT_SYMBOL_GPL(udp4_lib_lookup);
Eric Dumazetca065d02016-04-01 08:52:13 -0700642#endif
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700643
Shawn Bohrer421b3882013-10-07 11:01:39 -0500644static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk,
645 __be16 loc_port, __be32 loc_addr,
646 __be16 rmt_port, __be32 rmt_addr,
647 int dif, unsigned short hnum)
648{
649 struct inet_sock *inet = inet_sk(sk);
650
651 if (!net_eq(sock_net(sk), net) ||
652 udp_sk(sk)->udp_port_hash != hnum ||
653 (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
654 (inet->inet_dport != rmt_port && inet->inet_dport) ||
655 (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) ||
656 ipv6_only_sock(sk) ||
657 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
658 return false;
659 if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif))
660 return false;
661 return true;
662}
663
David S. Millerdb8dac22008-03-06 16:22:02 -0800664/*
665 * This routine is called by the ICMP module when it gets some
666 * sort of error condition. If err < 0 then the socket should
667 * be closed and the error returned to the user. If err > 0
668 * it's just the icmp type << 8 | icmp code.
669 * Header points to the ip header of the error packet. We move
670 * on past this. Then (as it used to claim before adjustment)
671 * header points to the first 8 bytes of the udp header. We need
672 * to find the appropriate port.
673 */
674
Eric Dumazet645ca702008-10-29 01:41:45 -0700675void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800676{
677 struct inet_sock *inet;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000678 const struct iphdr *iph = (const struct iphdr *)skb->data;
Eric Dumazetc482c562009-07-17 00:26:32 +0000679 struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
David S. Millerdb8dac22008-03-06 16:22:02 -0800680 const int type = icmp_hdr(skb)->type;
681 const int code = icmp_hdr(skb)->code;
682 struct sock *sk;
683 int harderr;
684 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700685 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800686
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700687 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
Craig Gallek538950a2016-01-04 17:41:47 -0500688 iph->saddr, uh->source, skb->dev->ifindex, udptable,
689 NULL);
Ian Morris51456b22015-04-03 09:17:26 +0100690 if (!sk) {
Eric Dumazet5d3848b2016-04-27 16:44:29 -0700691 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800692 return; /* No socket for error */
693 }
694
695 err = 0;
696 harderr = 0;
697 inet = inet_sk(sk);
698
699 switch (type) {
700 default:
701 case ICMP_TIME_EXCEEDED:
702 err = EHOSTUNREACH;
703 break;
704 case ICMP_SOURCE_QUENCH:
705 goto out;
706 case ICMP_PARAMETERPROB:
707 err = EPROTO;
708 harderr = 1;
709 break;
710 case ICMP_DEST_UNREACH:
711 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
David S. Miller36393392012-06-14 22:21:46 -0700712 ipv4_sk_update_pmtu(skb, sk, info);
David S. Millerdb8dac22008-03-06 16:22:02 -0800713 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
714 err = EMSGSIZE;
715 harderr = 1;
716 break;
717 }
718 goto out;
719 }
720 err = EHOSTUNREACH;
721 if (code <= NR_ICMP_UNREACH) {
722 harderr = icmp_err_convert[code].fatal;
723 err = icmp_err_convert[code].errno;
724 }
725 break;
David S. Miller55be7a92012-07-11 21:27:49 -0700726 case ICMP_REDIRECT:
727 ipv4_sk_redirect(skb, sk);
Duan Jiong1a462d12013-09-20 18:20:28 +0800728 goto out;
David S. Millerdb8dac22008-03-06 16:22:02 -0800729 }
730
731 /*
732 * RFC1122: OK. Passes ICMP errors back to application, as per
733 * 4.1.3.3.
734 */
735 if (!inet->recverr) {
736 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
737 goto out;
Eric Dumazetb1faf562010-05-31 23:44:05 -0700738 } else
Eric Dumazetc482c562009-07-17 00:26:32 +0000739 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
Eric Dumazetb1faf562010-05-31 23:44:05 -0700740
David S. Millerdb8dac22008-03-06 16:22:02 -0800741 sk->sk_err = err;
742 sk->sk_error_report(sk);
743out:
Eric Dumazetca065d02016-04-01 08:52:13 -0700744 return;
David S. Millerdb8dac22008-03-06 16:22:02 -0800745}
746
747void udp_err(struct sk_buff *skb, u32 info)
748{
Eric Dumazet645ca702008-10-29 01:41:45 -0700749 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800750}
751
752/*
753 * Throw away all pending data and cancel the corking. Socket is locked.
754 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400755void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800756{
757 struct udp_sock *up = udp_sk(sk);
758
759 if (up->pending) {
760 up->len = 0;
761 up->pending = 0;
762 ip_flush_pending_frames(sk);
763 }
764}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400765EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800766
767/**
Herbert Xuf6b96642011-03-01 02:36:48 +0000768 * udp4_hwcsum - handle outgoing HW checksumming
David S. Millerdb8dac22008-03-06 16:22:02 -0800769 * @skb: sk_buff containing the filled-in UDP header
770 * (checksum field must be zeroed out)
Herbert Xuf6b96642011-03-01 02:36:48 +0000771 * @src: source IP address
772 * @dst: destination IP address
David S. Millerdb8dac22008-03-06 16:22:02 -0800773 */
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200774void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
David S. Millerdb8dac22008-03-06 16:22:02 -0800775{
David S. Millerdb8dac22008-03-06 16:22:02 -0800776 struct udphdr *uh = udp_hdr(skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000777 int offset = skb_transport_offset(skb);
778 int len = skb->len - offset;
779 int hlen = len;
David S. Millerdb8dac22008-03-06 16:22:02 -0800780 __wsum csum = 0;
781
WANG Congebbe4952014-06-02 16:12:02 -0700782 if (!skb_has_frag_list(skb)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800783 /*
784 * Only one fragment on the socket.
785 */
786 skb->csum_start = skb_transport_header(skb) - skb->head;
787 skb->csum_offset = offsetof(struct udphdr, check);
Herbert Xuf6b96642011-03-01 02:36:48 +0000788 uh->check = ~csum_tcpudp_magic(src, dst, len,
789 IPPROTO_UDP, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800790 } else {
WANG Congebbe4952014-06-02 16:12:02 -0700791 struct sk_buff *frags;
792
David S. Millerdb8dac22008-03-06 16:22:02 -0800793 /*
794 * HW-checksum won't work as there are two or more
795 * fragments on the socket so that all csums of sk_buffs
796 * should be together
797 */
WANG Congebbe4952014-06-02 16:12:02 -0700798 skb_walk_frags(skb, frags) {
Herbert Xuf6b96642011-03-01 02:36:48 +0000799 csum = csum_add(csum, frags->csum);
800 hlen -= frags->len;
WANG Congebbe4952014-06-02 16:12:02 -0700801 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800802
Herbert Xuf6b96642011-03-01 02:36:48 +0000803 csum = skb_checksum(skb, offset, hlen, csum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800804 skb->ip_summed = CHECKSUM_NONE;
805
David S. Millerdb8dac22008-03-06 16:22:02 -0800806 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
807 if (uh->check == 0)
808 uh->check = CSUM_MANGLED_0;
809 }
810}
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200811EXPORT_SYMBOL_GPL(udp4_hwcsum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800812
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700813/* Function to set UDP checksum for an IPv4 UDP packet. This is intended
814 * for the simple case like when setting the checksum for a UDP tunnel.
815 */
816void udp_set_csum(bool nocheck, struct sk_buff *skb,
817 __be32 saddr, __be32 daddr, int len)
818{
819 struct udphdr *uh = udp_hdr(skb);
820
Edward Cree179bc672016-02-11 20:48:04 +0000821 if (nocheck) {
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700822 uh->check = 0;
Edward Cree179bc672016-02-11 20:48:04 +0000823 } else if (skb_is_gso(skb)) {
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700824 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
Edward Cree179bc672016-02-11 20:48:04 +0000825 } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
826 uh->check = 0;
827 uh->check = udp_v4_check(len, saddr, daddr, lco_csum(skb));
828 if (uh->check == 0)
829 uh->check = CSUM_MANGLED_0;
Edward Creed75f1302016-02-11 20:49:40 +0000830 } else {
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700831 skb->ip_summed = CHECKSUM_PARTIAL;
832 skb->csum_start = skb_transport_header(skb) - skb->head;
833 skb->csum_offset = offsetof(struct udphdr, check);
834 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700835 }
836}
837EXPORT_SYMBOL(udp_set_csum);
838
David S. Miller79ab0532011-05-09 13:31:04 -0700839static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
Herbert Xuf6b96642011-03-01 02:36:48 +0000840{
841 struct sock *sk = skb->sk;
842 struct inet_sock *inet = inet_sk(sk);
843 struct udphdr *uh;
Herbert Xuf6b96642011-03-01 02:36:48 +0000844 int err = 0;
845 int is_udplite = IS_UDPLITE(sk);
846 int offset = skb_transport_offset(skb);
847 int len = skb->len - offset;
848 __wsum csum = 0;
849
850 /*
851 * Create a UDP header
852 */
853 uh = udp_hdr(skb);
854 uh->source = inet->inet_sport;
David S. Miller79ab0532011-05-09 13:31:04 -0700855 uh->dest = fl4->fl4_dport;
Herbert Xuf6b96642011-03-01 02:36:48 +0000856 uh->len = htons(len);
857 uh->check = 0;
858
859 if (is_udplite) /* UDP-Lite */
860 csum = udplite_csum(skb);
861
Tom Herbert28448b82014-05-23 08:47:19 -0700862 else if (sk->sk_no_check_tx) { /* UDP csum disabled */
Herbert Xuf6b96642011-03-01 02:36:48 +0000863
864 skb->ip_summed = CHECKSUM_NONE;
865 goto send;
866
867 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
868
David S. Miller79ab0532011-05-09 13:31:04 -0700869 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
Herbert Xuf6b96642011-03-01 02:36:48 +0000870 goto send;
871
872 } else
873 csum = udp_csum(skb);
874
875 /* add protocol-dependent pseudo-header */
David S. Miller79ab0532011-05-09 13:31:04 -0700876 uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
Herbert Xuf6b96642011-03-01 02:36:48 +0000877 sk->sk_protocol, csum);
878 if (uh->check == 0)
879 uh->check = CSUM_MANGLED_0;
880
881send:
Eric Dumazetb5ec8ee2012-08-10 02:22:47 +0000882 err = ip_send_skb(sock_net(sk), skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000883 if (err) {
884 if (err == -ENOBUFS && !inet->recverr) {
Eric Dumazet6aef70a2016-04-27 16:44:27 -0700885 UDP_INC_STATS(sock_net(sk),
886 UDP_MIB_SNDBUFERRORS, is_udplite);
Herbert Xuf6b96642011-03-01 02:36:48 +0000887 err = 0;
888 }
889 } else
Eric Dumazet6aef70a2016-04-27 16:44:27 -0700890 UDP_INC_STATS(sock_net(sk),
891 UDP_MIB_OUTDATAGRAMS, is_udplite);
Herbert Xuf6b96642011-03-01 02:36:48 +0000892 return err;
893}
894
David S. Millerdb8dac22008-03-06 16:22:02 -0800895/*
896 * Push out all pending data as one UDP datagram. Socket is locked.
897 */
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200898int udp_push_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800899{
900 struct udp_sock *up = udp_sk(sk);
901 struct inet_sock *inet = inet_sk(sk);
David S. Millerb6f21b22011-03-12 02:09:18 -0500902 struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800903 struct sk_buff *skb;
David S. Millerdb8dac22008-03-06 16:22:02 -0800904 int err = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800905
David S. Miller77968b72011-05-08 17:12:19 -0700906 skb = ip_finish_skb(sk, fl4);
Herbert Xuf6b96642011-03-01 02:36:48 +0000907 if (!skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800908 goto out;
909
David S. Miller79ab0532011-05-09 13:31:04 -0700910 err = udp_send_skb(skb, fl4);
David S. Millerdb8dac22008-03-06 16:22:02 -0800911
David S. Millerdb8dac22008-03-06 16:22:02 -0800912out:
913 up->len = 0;
914 up->pending = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800915 return err;
916}
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200917EXPORT_SYMBOL(udp_push_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800918
Ying Xue1b784142015-03-02 15:37:48 +0800919int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
David S. Millerdb8dac22008-03-06 16:22:02 -0800920{
921 struct inet_sock *inet = inet_sk(sk);
922 struct udp_sock *up = udp_sk(sk);
David S. Millere4749952011-05-08 16:38:45 -0700923 struct flowi4 fl4_stack;
David S. Millerb6f21b22011-03-12 02:09:18 -0500924 struct flowi4 *fl4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800925 int ulen = len;
926 struct ipcm_cookie ipc;
927 struct rtable *rt = NULL;
928 int free = 0;
929 int connected = 0;
930 __be32 daddr, faddr, saddr;
931 __be16 dport;
932 u8 tos;
933 int err, is_udplite = IS_UDPLITE(sk);
934 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
935 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
Herbert Xu903ab862011-03-01 02:36:48 +0000936 struct sk_buff *skb;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000937 struct ip_options_data opt_copy;
David S. Millerdb8dac22008-03-06 16:22:02 -0800938
939 if (len > 0xFFFF)
940 return -EMSGSIZE;
941
942 /*
943 * Check the flags.
944 */
945
Eric Dumazetc482c562009-07-17 00:26:32 +0000946 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
David S. Millerdb8dac22008-03-06 16:22:02 -0800947 return -EOPNOTSUPP;
948
949 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000950 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200951 ipc.ttl = 0;
952 ipc.tos = -1;
David S. Millerdb8dac22008-03-06 16:22:02 -0800953
Herbert Xu903ab862011-03-01 02:36:48 +0000954 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
955
David S. Millerf5fca602011-05-08 17:24:10 -0700956 fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800957 if (up->pending) {
958 /*
959 * There are pending frames.
960 * The socket lock must be held while it's corked.
961 */
962 lock_sock(sk);
963 if (likely(up->pending)) {
964 if (unlikely(up->pending != AF_INET)) {
965 release_sock(sk);
966 return -EINVAL;
967 }
968 goto do_append_data;
969 }
970 release_sock(sk);
971 }
972 ulen += sizeof(struct udphdr);
973
974 /*
975 * Get and verify the address.
976 */
977 if (msg->msg_name) {
Steffen Hurrle342dfc32014-01-17 22:53:15 +0100978 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -0800979 if (msg->msg_namelen < sizeof(*usin))
980 return -EINVAL;
981 if (usin->sin_family != AF_INET) {
982 if (usin->sin_family != AF_UNSPEC)
983 return -EAFNOSUPPORT;
984 }
985
986 daddr = usin->sin_addr.s_addr;
987 dport = usin->sin_port;
988 if (dport == 0)
989 return -EINVAL;
990 } else {
991 if (sk->sk_state != TCP_ESTABLISHED)
992 return -EDESTADDRREQ;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000993 daddr = inet->inet_daddr;
994 dport = inet->inet_dport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800995 /* Open fast path for connected socket.
996 Route will not be used, if at least one option is set.
997 */
998 connected = 1;
999 }
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001000
1001 ipc.sockc.tsflags = sk->sk_tsflags;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001002 ipc.addr = inet->inet_saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001003 ipc.oif = sk->sk_bound_dev_if;
Daniel Borkmannbf84a012013-04-14 08:08:13 +00001004
David S. Millerdb8dac22008-03-06 16:22:02 -08001005 if (msg->msg_controllen) {
Soheil Hassas Yeganeh24025c42016-04-02 23:08:10 -04001006 err = ip_cmsg_send(sk, msg, &ipc, sk->sk_family == AF_INET6);
Eric Dumazet91948302016-02-04 06:23:28 -08001007 if (unlikely(err)) {
1008 kfree(ipc.opt);
David S. Millerdb8dac22008-03-06 16:22:02 -08001009 return err;
Eric Dumazet91948302016-02-04 06:23:28 -08001010 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001011 if (ipc.opt)
1012 free = 1;
1013 connected = 0;
1014 }
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001015 if (!ipc.opt) {
1016 struct ip_options_rcu *inet_opt;
1017
1018 rcu_read_lock();
1019 inet_opt = rcu_dereference(inet->inet_opt);
1020 if (inet_opt) {
1021 memcpy(&opt_copy, inet_opt,
1022 sizeof(*inet_opt) + inet_opt->opt.optlen);
1023 ipc.opt = &opt_copy.opt;
1024 }
1025 rcu_read_unlock();
1026 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001027
1028 saddr = ipc.addr;
1029 ipc.addr = faddr = daddr;
1030
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04001031 sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
1032
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001033 if (ipc.opt && ipc.opt->opt.srr) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001034 if (!daddr)
1035 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001036 faddr = ipc.opt->opt.faddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001037 connected = 0;
1038 }
Francesco Fuscoaa661582013-09-24 15:43:09 +02001039 tos = get_rttos(&ipc, inet);
David S. Millerdb8dac22008-03-06 16:22:02 -08001040 if (sock_flag(sk, SOCK_LOCALROUTE) ||
1041 (msg->msg_flags & MSG_DONTROUTE) ||
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001042 (ipc.opt && ipc.opt->opt.is_strictroute)) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001043 tos |= RTO_ONLINK;
1044 connected = 0;
1045 }
1046
1047 if (ipv4_is_multicast(daddr)) {
1048 if (!ipc.oif)
1049 ipc.oif = inet->mc_index;
1050 if (!saddr)
1051 saddr = inet->mc_addr;
1052 connected = 0;
Erich E. Hoover76e21052012-02-08 09:11:07 +00001053 } else if (!ipc.oif)
1054 ipc.oif = inet->uc_index;
David S. Millerdb8dac22008-03-06 16:22:02 -08001055
1056 if (connected)
Eric Dumazetc482c562009-07-17 00:26:32 +00001057 rt = (struct rtable *)sk_dst_check(sk, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001058
Ian Morris51456b22015-04-03 09:17:26 +01001059 if (!rt) {
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001060 struct net *net = sock_net(sk);
David Ahern9a24abf2015-08-13 14:59:03 -06001061 __u8 flow_flags = inet_sk_flowi_flags(sk);
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001062
David S. Millere4749952011-05-08 16:38:45 -07001063 fl4 = &fl4_stack;
David Ahern9a24abf2015-08-13 14:59:03 -06001064
David S. Millere4749952011-05-08 16:38:45 -07001065 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
David S. Millerc0951cb2011-03-31 04:54:27 -07001066 RT_SCOPE_UNIVERSE, sk->sk_protocol,
David Ahern9a24abf2015-08-13 14:59:03 -06001067 flow_flags,
David S. Millerc0951cb2011-03-31 04:54:27 -07001068 faddr, saddr, dport, inet->inet_sport);
1069
David Ahernb5bdacf2016-01-04 09:09:27 -08001070 if (!saddr && ipc.oif) {
1071 err = l3mdev_get_saddr(net, ipc.oif, fl4);
1072 if (err < 0)
1073 goto out;
1074 }
David Ahern8cbb512c2015-10-05 08:51:26 -07001075
David S. Millere4749952011-05-08 16:38:45 -07001076 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
1077 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001078 if (IS_ERR(rt)) {
1079 err = PTR_ERR(rt);
David S. Miller06dc94b2011-03-03 10:38:01 -08001080 rt = NULL;
David S. Millerdb8dac22008-03-06 16:22:02 -08001081 if (err == -ENETUNREACH)
Eric Dumazetf1d8cba2013-11-28 09:51:22 -08001082 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -08001083 goto out;
1084 }
1085
1086 err = -EACCES;
1087 if ((rt->rt_flags & RTCF_BROADCAST) &&
1088 !sock_flag(sk, SOCK_BROADCAST))
1089 goto out;
1090 if (connected)
Changli Gaod8d1f302010-06-10 23:31:35 -07001091 sk_dst_set(sk, dst_clone(&rt->dst));
David S. Millerdb8dac22008-03-06 16:22:02 -08001092 }
1093
1094 if (msg->msg_flags&MSG_CONFIRM)
1095 goto do_confirm;
1096back_from_confirm:
1097
David S. Millere4749952011-05-08 16:38:45 -07001098 saddr = fl4->saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001099 if (!ipc.addr)
David S. Millere4749952011-05-08 16:38:45 -07001100 daddr = ipc.addr = fl4->daddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001101
Herbert Xu903ab862011-03-01 02:36:48 +00001102 /* Lockless fast path for the non-corking case. */
1103 if (!corkreq) {
Al Virof69e6d12014-11-24 13:23:40 -05001104 skb = ip_make_skb(sk, fl4, getfrag, msg, ulen,
Herbert Xu903ab862011-03-01 02:36:48 +00001105 sizeof(struct udphdr), &ipc, &rt,
1106 msg->msg_flags);
1107 err = PTR_ERR(skb);
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001108 if (!IS_ERR_OR_NULL(skb))
David S. Miller79ab0532011-05-09 13:31:04 -07001109 err = udp_send_skb(skb, fl4);
Herbert Xu903ab862011-03-01 02:36:48 +00001110 goto out;
1111 }
1112
David S. Millerdb8dac22008-03-06 16:22:02 -08001113 lock_sock(sk);
1114 if (unlikely(up->pending)) {
1115 /* The socket is already corked while preparing it. */
1116 /* ... which is an evident application bug. --ANK */
1117 release_sock(sk);
1118
Joe Perchesba7a46f2014-11-11 10:59:17 -08001119 net_dbg_ratelimited("cork app bug 2\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001120 err = -EINVAL;
1121 goto out;
1122 }
1123 /*
1124 * Now cork the socket to pend data.
1125 */
David S. Millerb6f21b22011-03-12 02:09:18 -05001126 fl4 = &inet->cork.fl.u.ip4;
1127 fl4->daddr = daddr;
1128 fl4->saddr = saddr;
David S. Miller9cce96d2011-03-12 03:00:33 -05001129 fl4->fl4_dport = dport;
1130 fl4->fl4_sport = inet->inet_sport;
David S. Millerdb8dac22008-03-06 16:22:02 -08001131 up->pending = AF_INET;
1132
1133do_append_data:
1134 up->len += ulen;
Al Virof69e6d12014-11-24 13:23:40 -05001135 err = ip_append_data(sk, fl4, getfrag, msg, ulen,
David S. Millerf5fca602011-05-08 17:24:10 -07001136 sizeof(struct udphdr), &ipc, &rt,
1137 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001138 if (err)
1139 udp_flush_pending_frames(sk);
1140 else if (!corkreq)
1141 err = udp_push_pending_frames(sk);
1142 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1143 up->pending = 0;
1144 release_sock(sk);
1145
1146out:
1147 ip_rt_put(rt);
1148 if (free)
1149 kfree(ipc.opt);
1150 if (!err)
1151 return len;
1152 /*
1153 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1154 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1155 * we don't have a good statistic (IpOutDiscards but it can be too many
1156 * things). We could add another new stat but at least for now that
1157 * seems like overkill.
1158 */
1159 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Eric Dumazet6aef70a2016-04-27 16:44:27 -07001160 UDP_INC_STATS(sock_net(sk),
1161 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001162 }
1163 return err;
1164
1165do_confirm:
Changli Gaod8d1f302010-06-10 23:31:35 -07001166 dst_confirm(&rt->dst);
David S. Millerdb8dac22008-03-06 16:22:02 -08001167 if (!(msg->msg_flags&MSG_PROBE) || len)
1168 goto back_from_confirm;
1169 err = 0;
1170 goto out;
1171}
Eric Dumazetc482c562009-07-17 00:26:32 +00001172EXPORT_SYMBOL(udp_sendmsg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001173
1174int udp_sendpage(struct sock *sk, struct page *page, int offset,
1175 size_t size, int flags)
1176{
David S. Millerf5fca602011-05-08 17:24:10 -07001177 struct inet_sock *inet = inet_sk(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001178 struct udp_sock *up = udp_sk(sk);
1179 int ret;
1180
Shawn Landdend3f7d562013-11-24 22:36:28 -08001181 if (flags & MSG_SENDPAGE_NOTLAST)
1182 flags |= MSG_MORE;
1183
David S. Millerdb8dac22008-03-06 16:22:02 -08001184 if (!up->pending) {
1185 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
1186
1187 /* Call udp_sendmsg to specify destination address which
1188 * sendpage interface can't pass.
1189 * This will succeed only when the socket is connected.
1190 */
Ying Xue1b784142015-03-02 15:37:48 +08001191 ret = udp_sendmsg(sk, &msg, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001192 if (ret < 0)
1193 return ret;
1194 }
1195
1196 lock_sock(sk);
1197
1198 if (unlikely(!up->pending)) {
1199 release_sock(sk);
1200
Joe Perchesba7a46f2014-11-11 10:59:17 -08001201 net_dbg_ratelimited("udp cork app bug 3\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001202 return -EINVAL;
1203 }
1204
David S. Millerf5fca602011-05-08 17:24:10 -07001205 ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1206 page, offset, size, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001207 if (ret == -EOPNOTSUPP) {
1208 release_sock(sk);
1209 return sock_no_sendpage(sk->sk_socket, page, offset,
1210 size, flags);
1211 }
1212 if (ret < 0) {
1213 udp_flush_pending_frames(sk);
1214 goto out;
1215 }
1216
1217 up->len += size;
1218 if (!(up->corkflag || (flags&MSG_MORE)))
1219 ret = udp_push_pending_frames(sk);
1220 if (!ret)
1221 ret = size;
1222out:
1223 release_sock(sk);
1224 return ret;
1225}
1226
Eric Dumazet85584672009-10-09 04:43:40 +00001227/**
1228 * first_packet_length - return length of first packet in receive queue
1229 * @sk: socket
1230 *
1231 * Drops all bad checksum frames, until a valid one is found.
1232 * Returns the length of found skb, or 0 if none is found.
1233 */
1234static unsigned int first_packet_length(struct sock *sk)
1235{
1236 struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1237 struct sk_buff *skb;
1238 unsigned int res;
1239
1240 __skb_queue_head_init(&list_kill);
1241
1242 spin_lock_bh(&rcvq->lock);
1243 while ((skb = skb_peek(rcvq)) != NULL &&
1244 udp_lib_checksum_complete(skb)) {
Eric Dumazet02c22342016-04-27 16:44:30 -07001245 __UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS,
1246 IS_UDPLITE(sk));
1247 __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
1248 IS_UDPLITE(sk));
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001249 atomic_inc(&sk->sk_drops);
Eric Dumazet85584672009-10-09 04:43:40 +00001250 __skb_unlink(skb, rcvq);
1251 __skb_queue_tail(&list_kill, skb);
1252 }
1253 res = skb ? skb->len : 0;
1254 spin_unlock_bh(&rcvq->lock);
1255
1256 if (!skb_queue_empty(&list_kill)) {
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001257 bool slow = lock_sock_fast(sk);
1258
Eric Dumazet85584672009-10-09 04:43:40 +00001259 __skb_queue_purge(&list_kill);
1260 sk_mem_reclaim_partial(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001261 unlock_sock_fast(sk, slow);
Eric Dumazet85584672009-10-09 04:43:40 +00001262 }
1263 return res;
1264}
1265
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266/*
1267 * IOCTL requests applicable to the UDP protocol
1268 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001269
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1271{
Stephen Hemminger6516c652007-03-08 20:41:55 -08001272 switch (cmd) {
1273 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 {
Eric Dumazet31e6d362009-06-17 19:05:41 -07001275 int amount = sk_wmem_alloc_get(sk);
1276
Stephen Hemminger6516c652007-03-08 20:41:55 -08001277 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 }
Stephen Hemminger6516c652007-03-08 20:41:55 -08001279
1280 case SIOCINQ:
1281 {
Eric Dumazet85584672009-10-09 04:43:40 +00001282 unsigned int amount = first_packet_length(sk);
Stephen Hemminger6516c652007-03-08 20:41:55 -08001283
Stephen Hemminger6516c652007-03-08 20:41:55 -08001284 return put_user(amount, (int __user *)arg);
1285 }
1286
1287 default:
1288 return -ENOIOCTLCMD;
1289 }
1290
1291 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292}
Eric Dumazetc482c562009-07-17 00:26:32 +00001293EXPORT_SYMBOL(udp_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294
David S. Millerdb8dac22008-03-06 16:22:02 -08001295/*
1296 * This should be easy, if there is something there we
1297 * return it, otherwise we block.
1298 */
1299
Ying Xue1b784142015-03-02 15:37:48 +08001300int udp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
1301 int flags, int *addr_len)
David S. Millerdb8dac22008-03-06 16:22:02 -08001302{
1303 struct inet_sock *inet = inet_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001304 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -08001305 struct sk_buff *skb;
David S. Miller59c2cda2011-12-01 14:12:55 -05001306 unsigned int ulen, copied;
samanthakumar627d2d62016-04-05 12:41:16 -04001307 int peeked, peeking, off;
David S. Millerdb8dac22008-03-06 16:22:02 -08001308 int err;
1309 int is_udplite = IS_UDPLITE(sk);
Eric Dumazet197c9492015-12-30 08:51:12 -05001310 bool checksum_valid = false;
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001311 bool slow;
David S. Millerdb8dac22008-03-06 16:22:02 -08001312
David S. Millerdb8dac22008-03-06 16:22:02 -08001313 if (flags & MSG_ERRQUEUE)
Hannes Frederic Sowa85fbaa72013-11-23 00:46:12 +01001314 return ip_recv_error(sk, msg, len, addr_len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001315
1316try_again:
samanthakumar627d2d62016-04-05 12:41:16 -04001317 peeking = off = sk_peek_offset(sk, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001318 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001319 &peeked, &off, &err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001320 if (!skb)
samanthakumar627d2d62016-04-05 12:41:16 -04001321 return err;
David S. Millerdb8dac22008-03-06 16:22:02 -08001322
samanthakumare6afc8a2016-04-05 12:41:15 -04001323 ulen = skb->len;
David S. Miller59c2cda2011-12-01 14:12:55 -05001324 copied = len;
samanthakumar627d2d62016-04-05 12:41:16 -04001325 if (copied > ulen - off)
1326 copied = ulen - off;
David S. Miller59c2cda2011-12-01 14:12:55 -05001327 else if (copied < ulen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001328 msg->msg_flags |= MSG_TRUNC;
1329
1330 /*
1331 * If checksum is needed at all, try to do it while copying the
1332 * data. If the data is truncated, or if we only want a partial
1333 * coverage checksum (UDP-Lite), do it before the copy.
1334 */
1335
samanthakumar627d2d62016-04-05 12:41:16 -04001336 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov || peeking) {
Eric Dumazet197c9492015-12-30 08:51:12 -05001337 checksum_valid = !udp_lib_checksum_complete(skb);
1338 if (!checksum_valid)
David S. Millerdb8dac22008-03-06 16:22:02 -08001339 goto csum_copy_err;
1340 }
1341
Eric Dumazet197c9492015-12-30 08:51:12 -05001342 if (checksum_valid || skb_csum_unnecessary(skb))
samanthakumar627d2d62016-04-05 12:41:16 -04001343 err = skb_copy_datagram_msg(skb, off, msg, copied);
David S. Millerdb8dac22008-03-06 16:22:02 -08001344 else {
samanthakumar627d2d62016-04-05 12:41:16 -04001345 err = skb_copy_and_csum_datagram_msg(skb, off, msg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001346
1347 if (err == -EINVAL)
1348 goto csum_copy_err;
1349 }
1350
Eric Dumazet22911fc2012-06-27 00:23:44 +00001351 if (unlikely(err)) {
1352 trace_kfree_skb(skb, udp_recvmsg);
Eric Dumazet979402b2012-09-05 23:34:44 +00001353 if (!peeked) {
1354 atomic_inc(&sk->sk_drops);
Eric Dumazet6aef70a2016-04-27 16:44:27 -07001355 UDP_INC_STATS(sock_net(sk),
1356 UDP_MIB_INERRORS, is_udplite);
Eric Dumazet979402b2012-09-05 23:34:44 +00001357 }
samanthakumar627d2d62016-04-05 12:41:16 -04001358 skb_free_datagram_locked(sk, skb);
1359 return err;
Eric Dumazet22911fc2012-06-27 00:23:44 +00001360 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001361
1362 if (!peeked)
Eric Dumazet6aef70a2016-04-27 16:44:27 -07001363 UDP_INC_STATS(sock_net(sk),
1364 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001365
Neil Horman3b885782009-10-12 13:26:31 -07001366 sock_recv_ts_and_drops(msg, sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001367
1368 /* Copy the address. */
Eric Dumazetc482c562009-07-17 00:26:32 +00001369 if (sin) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001370 sin->sin_family = AF_INET;
1371 sin->sin_port = udp_hdr(skb)->source;
1372 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1373 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
Hannes Frederic Sowabceaa902013-11-18 04:20:45 +01001374 *addr_len = sizeof(*sin);
David S. Millerdb8dac22008-03-06 16:22:02 -08001375 }
1376 if (inet->cmsg_flags)
Willem de Bruijn31c2e492016-04-07 18:12:59 -04001377 ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr) + off);
David S. Millerdb8dac22008-03-06 16:22:02 -08001378
David S. Miller59c2cda2011-12-01 14:12:55 -05001379 err = copied;
David S. Millerdb8dac22008-03-06 16:22:02 -08001380 if (flags & MSG_TRUNC)
1381 err = ulen;
1382
samanthakumar627d2d62016-04-05 12:41:16 -04001383 __skb_free_datagram_locked(sk, skb, peeking ? -err : err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001384 return err;
1385
1386csum_copy_err:
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001387 slow = lock_sock_fast(sk);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001388 if (!skb_kill_datagram(sk, skb, flags)) {
Eric Dumazet6aef70a2016-04-27 16:44:27 -07001389 UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
1390 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001391 }
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001392 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001393
Eric Dumazetbeb39db2015-05-30 09:16:53 -07001394 /* starting over for a new packet, but check if we need to yield */
1395 cond_resched();
Xufeng Zhang9cfaa8d2011-06-21 10:43:40 +00001396 msg->msg_flags &= ~MSG_TRUNC;
David S. Millerdb8dac22008-03-06 16:22:02 -08001397 goto try_again;
1398}
1399
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400int udp_disconnect(struct sock *sk, int flags)
1401{
1402 struct inet_sock *inet = inet_sk(sk);
1403 /*
1404 * 1003.1g - break association.
1405 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001406
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 sk->sk_state = TCP_CLOSE;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001408 inet->inet_daddr = 0;
1409 inet->inet_dport = 0;
Tom Herbertbdeab992011-08-14 19:45:55 +00001410 sock_rps_reset_rxhash(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 sk->sk_bound_dev_if = 0;
1412 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1413 inet_reset_saddr(sk);
1414
1415 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1416 sk->sk_prot->unhash(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001417 inet->inet_sport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 }
1419 sk_dst_reset(sk);
1420 return 0;
1421}
Eric Dumazetc482c562009-07-17 00:26:32 +00001422EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423
Eric Dumazet645ca702008-10-29 01:41:45 -07001424void udp_lib_unhash(struct sock *sk)
1425{
Eric Dumazet723b4612008-11-25 13:55:15 -08001426 if (sk_hashed(sk)) {
1427 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Eric Dumazet512615b2009-11-08 10:17:58 +00001428 struct udp_hslot *hslot, *hslot2;
1429
1430 hslot = udp_hashslot(udptable, sock_net(sk),
1431 udp_sk(sk)->udp_port_hash);
1432 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
Eric Dumazet645ca702008-10-29 01:41:45 -07001433
Eric Dumazet723b4612008-11-25 13:55:15 -08001434 spin_lock_bh(&hslot->lock);
Craig Galleke32ea7e2016-01-04 17:41:46 -05001435 if (rcu_access_pointer(sk->sk_reuseport_cb))
1436 reuseport_detach_sock(sk);
Eric Dumazetca065d02016-04-01 08:52:13 -07001437 if (sk_del_node_init_rcu(sk)) {
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00001438 hslot->count--;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001439 inet_sk(sk)->inet_num = 0;
Eric Dumazet723b4612008-11-25 13:55:15 -08001440 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet512615b2009-11-08 10:17:58 +00001441
1442 spin_lock(&hslot2->lock);
Eric Dumazetca065d02016-04-01 08:52:13 -07001443 hlist_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
Eric Dumazet512615b2009-11-08 10:17:58 +00001444 hslot2->count--;
1445 spin_unlock(&hslot2->lock);
Eric Dumazet723b4612008-11-25 13:55:15 -08001446 }
1447 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001448 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001449}
1450EXPORT_SYMBOL(udp_lib_unhash);
1451
Eric Dumazet719f8352010-09-08 05:08:44 +00001452/*
1453 * inet_rcv_saddr was changed, we must rehash secondary hash
1454 */
1455void udp_lib_rehash(struct sock *sk, u16 newhash)
1456{
1457 if (sk_hashed(sk)) {
1458 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1459 struct udp_hslot *hslot, *hslot2, *nhslot2;
1460
1461 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1462 nhslot2 = udp_hashslot2(udptable, newhash);
1463 udp_sk(sk)->udp_portaddr_hash = newhash;
Craig Galleke32ea7e2016-01-04 17:41:46 -05001464
1465 if (hslot2 != nhslot2 ||
1466 rcu_access_pointer(sk->sk_reuseport_cb)) {
Eric Dumazet719f8352010-09-08 05:08:44 +00001467 hslot = udp_hashslot(udptable, sock_net(sk),
1468 udp_sk(sk)->udp_port_hash);
1469 /* we must lock primary chain too */
1470 spin_lock_bh(&hslot->lock);
Craig Galleke32ea7e2016-01-04 17:41:46 -05001471 if (rcu_access_pointer(sk->sk_reuseport_cb))
1472 reuseport_detach_sock(sk);
Eric Dumazet719f8352010-09-08 05:08:44 +00001473
Craig Galleke32ea7e2016-01-04 17:41:46 -05001474 if (hslot2 != nhslot2) {
1475 spin_lock(&hslot2->lock);
Eric Dumazetca065d02016-04-01 08:52:13 -07001476 hlist_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
Craig Galleke32ea7e2016-01-04 17:41:46 -05001477 hslot2->count--;
1478 spin_unlock(&hslot2->lock);
Eric Dumazet719f8352010-09-08 05:08:44 +00001479
Craig Galleke32ea7e2016-01-04 17:41:46 -05001480 spin_lock(&nhslot2->lock);
Eric Dumazetca065d02016-04-01 08:52:13 -07001481 hlist_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
Craig Galleke32ea7e2016-01-04 17:41:46 -05001482 &nhslot2->head);
1483 nhslot2->count++;
1484 spin_unlock(&nhslot2->lock);
1485 }
Eric Dumazet719f8352010-09-08 05:08:44 +00001486
1487 spin_unlock_bh(&hslot->lock);
1488 }
1489 }
1490}
1491EXPORT_SYMBOL(udp_lib_rehash);
1492
1493static void udp_v4_rehash(struct sock *sk)
1494{
1495 u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1496 inet_sk(sk)->inet_rcv_saddr,
1497 inet_sk(sk)->inet_num);
1498 udp_lib_rehash(sk, new_hash);
1499}
1500
Herbert Xu93821772008-09-15 11:48:46 -07001501static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1502{
Tom Herbertfec5e652010-04-16 16:01:27 -07001503 int rc;
Herbert Xu93821772008-09-15 11:48:46 -07001504
Shawn Bohrer005ec972013-10-07 11:01:38 -05001505 if (inet_sk(sk)->inet_daddr) {
Tom Herbertbdeab992011-08-14 19:45:55 +00001506 sock_rps_save_rxhash(sk, skb);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001507 sk_mark_napi_id(sk, skb);
Eric Dumazet2c8c56e2014-11-11 05:54:28 -08001508 sk_incoming_cpu_update(sk);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001509 }
Tom Herbertfec5e652010-04-16 16:01:27 -07001510
samanthakumare6afc8a2016-04-05 12:41:15 -04001511 rc = __sock_queue_rcv_skb(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -07001512 if (rc < 0) {
1513 int is_udplite = IS_UDPLITE(sk);
1514
Herbert Xu93821772008-09-15 11:48:46 -07001515 /* Note that an ENOMEM error is charged twice */
Eric Dumazet766e90372009-10-14 20:40:11 -07001516 if (rc == -ENOMEM)
Eric Dumazet02c22342016-04-27 16:44:30 -07001517 __UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1518 is_udplite);
1519 __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet766e90372009-10-14 20:40:11 -07001520 kfree_skb(skb);
Satoru Moriya296f7ea2011-06-17 11:58:39 +00001521 trace_udp_fail_queue_rcv_skb(rc, sk);
Eric Dumazet766e90372009-10-14 20:40:11 -07001522 return -1;
Herbert Xu93821772008-09-15 11:48:46 -07001523 }
1524
1525 return 0;
1526
Herbert Xu93821772008-09-15 11:48:46 -07001527}
1528
Eric Dumazet447167b2012-04-11 23:05:28 +00001529static struct static_key udp_encap_needed __read_mostly;
1530void udp_encap_enable(void)
1531{
1532 if (!static_key_enabled(&udp_encap_needed))
1533 static_key_slow_inc(&udp_encap_needed);
1534}
1535EXPORT_SYMBOL(udp_encap_enable);
1536
David S. Millerdb8dac22008-03-06 16:22:02 -08001537/* returns:
1538 * -1: error
1539 * 0: success
1540 * >0: "udp encap" protocol resubmission
1541 *
1542 * Note that in the success and error cases, the skb is assumed to
1543 * have either been requeued or freed.
1544 */
Eric Dumazetc482c562009-07-17 00:26:32 +00001545int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -08001546{
1547 struct udp_sock *up = udp_sk(sk);
1548 int rc;
1549 int is_udplite = IS_UDPLITE(sk);
1550
1551 /*
1552 * Charge it to the socket, dropping if the queue is full.
1553 */
1554 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1555 goto drop;
1556 nf_reset(skb);
1557
Eric Dumazet447167b2012-04-11 23:05:28 +00001558 if (static_key_false(&udp_encap_needed) && up->encap_type) {
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001559 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1560
David S. Millerdb8dac22008-03-06 16:22:02 -08001561 /*
1562 * This is an encapsulation socket so pass the skb to
1563 * the socket's udp_encap_rcv() hook. Otherwise, just
1564 * fall through and pass this up the UDP socket.
1565 * up->encap_rcv() returns the following value:
1566 * =0 if skb was successfully passed to the encap
1567 * handler or was discarded by it.
1568 * >0 if skb should be passed on to UDP.
1569 * <0 if skb should be resubmitted as proto -N
1570 */
1571
1572 /* if we're overly short, let UDP handle it */
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001573 encap_rcv = ACCESS_ONCE(up->encap_rcv);
Ian Morris00db4122015-04-03 09:17:27 +01001574 if (skb->len > sizeof(struct udphdr) && encap_rcv) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001575 int ret;
1576
Tom Herbert0a809662014-05-07 16:52:39 -07001577 /* Verify checksum before giving to encap */
1578 if (udp_lib_checksum_complete(skb))
1579 goto csum_error;
1580
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001581 ret = encap_rcv(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001582 if (ret <= 0) {
Eric Dumazet02c22342016-04-27 16:44:30 -07001583 __UDP_INC_STATS(sock_net(sk),
1584 UDP_MIB_INDATAGRAMS,
1585 is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001586 return -ret;
1587 }
1588 }
1589
1590 /* FALLTHROUGH -- it's a UDP Packet */
1591 }
1592
1593 /*
1594 * UDP-Lite specific tests, ignored on UDP sockets
1595 */
1596 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1597
1598 /*
1599 * MIB statistics other than incrementing the error count are
1600 * disabled for the following two types of errors: these depend
1601 * on the application settings, not on the functioning of the
1602 * protocol stack as such.
1603 *
1604 * RFC 3828 here recommends (sec 3.3): "There should also be a
1605 * way ... to ... at least let the receiving application block
1606 * delivery of packets with coverage values less than a value
1607 * provided by the application."
1608 */
1609 if (up->pcrlen == 0) { /* full coverage was set */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001610 net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n",
1611 UDP_SKB_CB(skb)->cscov, skb->len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001612 goto drop;
1613 }
1614 /* The next case involves violating the min. coverage requested
1615 * by the receiver. This is subtle: if receiver wants x and x is
1616 * greater than the buffersize/MTU then receiver will complain
1617 * that it wants x while sender emits packets of smaller size y.
1618 * Therefore the above ...()->partial_cov statement is essential.
1619 */
1620 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
Joe Perchesba7a46f2014-11-11 10:59:17 -08001621 net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n",
1622 UDP_SKB_CB(skb)->cscov, up->pcrlen);
David S. Millerdb8dac22008-03-06 16:22:02 -08001623 goto drop;
1624 }
1625 }
1626
samanthakumare6afc8a2016-04-05 12:41:15 -04001627 if (rcu_access_pointer(sk->sk_filter)) {
1628 if (udp_lib_checksum_complete(skb))
1629 goto csum_error;
1630 if (sk_filter(sk, skb))
1631 goto drop;
1632 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001633
samanthakumare6afc8a2016-04-05 12:41:15 -04001634 udp_csum_pull_header(skb);
Sorin Dumitru274f4822014-07-22 21:16:51 +03001635 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
Eric Dumazet02c22342016-04-27 16:44:30 -07001636 __UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1637 is_udplite);
Eric Dumazetc3774112010-04-27 15:13:20 -07001638 goto drop;
James M Leddy3e215c82014-06-25 17:38:13 -04001639 }
Eric Dumazetc3774112010-04-27 15:13:20 -07001640
Herbert Xu93821772008-09-15 11:48:46 -07001641 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001642
Shawn Bohrerfbf88662013-10-07 11:01:40 -05001643 ipv4_pktinfo_prepare(sk, skb);
Herbert Xu93821772008-09-15 11:48:46 -07001644 bh_lock_sock(sk);
1645 if (!sock_owned_by_user(sk))
1646 rc = __udp_queue_rcv_skb(sk, skb);
Eric Dumazetf545a382012-04-22 23:34:26 +00001647 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
Zhu Yi55349792010-03-04 18:01:42 +00001648 bh_unlock_sock(sk);
1649 goto drop;
1650 }
Herbert Xu93821772008-09-15 11:48:46 -07001651 bh_unlock_sock(sk);
1652
1653 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001654
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001655csum_error:
Eric Dumazet02c22342016-04-27 16:44:30 -07001656 __UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001657drop:
Eric Dumazet02c22342016-04-27 16:44:30 -07001658 __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001659 atomic_inc(&sk->sk_drops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001660 kfree_skb(skb);
1661 return -1;
1662}
1663
Eric Dumazet97502232013-12-11 14:46:51 -08001664/* For TCP sockets, sk_rx_dst is protected by socket lock
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001665 * For UDP, we use xchg() to guard against concurrent changes.
Eric Dumazet97502232013-12-11 14:46:51 -08001666 */
1667static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
Shawn Bohrer421b3882013-10-07 11:01:39 -05001668{
Eric Dumazet97502232013-12-11 14:46:51 -08001669 struct dst_entry *old;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001670
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001671 dst_hold(dst);
1672 old = xchg(&sk->sk_rx_dst, dst);
1673 dst_release(old);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001674}
1675
David S. Millerdb8dac22008-03-06 16:22:02 -08001676/*
1677 * Multicasts and broadcasts go to each listener.
1678 *
Eric Dumazet1240d132009-11-08 10:18:44 +00001679 * Note: called only from the BH handler context.
David S. Millerdb8dac22008-03-06 16:22:02 -08001680 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001681static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001682 struct udphdr *uh,
1683 __be32 saddr, __be32 daddr,
Rick Jones36cbb242014-11-06 10:37:54 -08001684 struct udp_table *udptable,
1685 int proto)
David S. Millerdb8dac22008-03-06 16:22:02 -08001686{
Eric Dumazetca065d02016-04-01 08:52:13 -07001687 struct sock *sk, *first = NULL;
David Held5cf3d462014-07-15 23:28:31 -04001688 unsigned short hnum = ntohs(uh->dest);
1689 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
David Held2dc41cf2014-07-15 23:28:32 -04001690 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
Eric Dumazetca065d02016-04-01 08:52:13 -07001691 unsigned int offset = offsetof(typeof(*sk), sk_node);
1692 int dif = skb->dev->ifindex;
1693 struct hlist_node *node;
1694 struct sk_buff *nskb;
David Held2dc41cf2014-07-15 23:28:32 -04001695
1696 if (use_hash2) {
1697 hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) &
1698 udp_table.mask;
1699 hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask;
1700start_lookup:
1701 hslot = &udp_table.hash2[hash2];
1702 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
1703 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001704
Eric Dumazetca065d02016-04-01 08:52:13 -07001705 sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
1706 if (!__udp_is_mcast_sock(net, sk, uh->dest, daddr,
1707 uh->source, saddr, dif, hnum))
1708 continue;
David S. Millerdb8dac22008-03-06 16:22:02 -08001709
Eric Dumazetca065d02016-04-01 08:52:13 -07001710 if (!first) {
1711 first = sk;
1712 continue;
1713 }
1714 nskb = skb_clone(skb, GFP_ATOMIC);
1715
1716 if (unlikely(!nskb)) {
1717 atomic_inc(&sk->sk_drops);
Eric Dumazet02c22342016-04-27 16:44:30 -07001718 __UDP_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
1719 IS_UDPLITE(sk));
1720 __UDP_INC_STATS(net, UDP_MIB_INERRORS,
1721 IS_UDPLITE(sk));
Eric Dumazetca065d02016-04-01 08:52:13 -07001722 continue;
1723 }
1724 if (udp_queue_rcv_skb(sk, nskb) > 0)
1725 consume_skb(nskb);
1726 }
Eric Dumazet1240d132009-11-08 10:18:44 +00001727
David Held2dc41cf2014-07-15 23:28:32 -04001728 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
1729 if (use_hash2 && hash2 != hash2_any) {
1730 hash2 = hash2_any;
1731 goto start_lookup;
1732 }
1733
Eric Dumazetca065d02016-04-01 08:52:13 -07001734 if (first) {
1735 if (udp_queue_rcv_skb(first, skb) > 0)
1736 consume_skb(skb);
Eric Dumazet1240d132009-11-08 10:18:44 +00001737 } else {
Eric Dumazetca065d02016-04-01 08:52:13 -07001738 kfree_skb(skb);
Eric Dumazet02c22342016-04-27 16:44:30 -07001739 __UDP_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
1740 proto == IPPROTO_UDPLITE);
Eric Dumazet1240d132009-11-08 10:18:44 +00001741 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001742 return 0;
1743}
1744
1745/* Initialize UDP checksum. If exited with zero value (success),
1746 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1747 * Otherwise, csum completion requires chacksumming packet body,
1748 * including udp header and folding it to skb->csum.
1749 */
1750static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1751 int proto)
1752{
David S. Millerdb8dac22008-03-06 16:22:02 -08001753 int err;
1754
1755 UDP_SKB_CB(skb)->partial_cov = 0;
1756 UDP_SKB_CB(skb)->cscov = skb->len;
1757
1758 if (proto == IPPROTO_UDPLITE) {
1759 err = udplite_checksum_init(skb, uh);
1760 if (err)
1761 return err;
1762 }
1763
Tom Herberted70fcf2014-05-02 16:29:38 -07001764 return skb_checksum_init_zero_check(skb, proto, uh->check,
1765 inet_compute_pseudo);
David S. Millerdb8dac22008-03-06 16:22:02 -08001766}
1767
1768/*
1769 * All we need to do is get the socket, and then do a checksum.
1770 */
1771
Eric Dumazet645ca702008-10-29 01:41:45 -07001772int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001773 int proto)
1774{
1775 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001776 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001777 unsigned short ulen;
Eric Dumazetadf30902009-06-02 05:19:30 +00001778 struct rtable *rt = skb_rtable(skb);
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001779 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001780 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001781
1782 /*
1783 * Validate the packet.
1784 */
1785 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1786 goto drop; /* No space for header. */
1787
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001788 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001789 ulen = ntohs(uh->len);
Bjørn Morkccc2d972010-05-06 03:44:34 +00001790 saddr = ip_hdr(skb)->saddr;
1791 daddr = ip_hdr(skb)->daddr;
1792
David S. Millerdb8dac22008-03-06 16:22:02 -08001793 if (ulen > skb->len)
1794 goto short_packet;
1795
1796 if (proto == IPPROTO_UDP) {
1797 /* UDP validates ulen. */
1798 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1799 goto short_packet;
1800 uh = udp_hdr(skb);
1801 }
1802
1803 if (udp4_csum_init(skb, uh, proto))
1804 goto csum_error;
1805
Eric Dumazet8afdd992013-12-10 18:07:23 -08001806 sk = skb_steal_sock(skb);
1807 if (sk) {
Eric Dumazet97502232013-12-11 14:46:51 -08001808 struct dst_entry *dst = skb_dst(skb);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001809 int ret;
David S. Millerdb8dac22008-03-06 16:22:02 -08001810
Eric Dumazet97502232013-12-11 14:46:51 -08001811 if (unlikely(sk->sk_rx_dst != dst))
1812 udp_sk_rx_dst_set(sk, dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001813
1814 ret = udp_queue_rcv_skb(sk, skb);
Eric Dumazet8afdd992013-12-10 18:07:23 -08001815 sock_put(sk);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001816 /* a return value > 0 means to resubmit the input, but
1817 * it wants the return to be -protocol, or 0
1818 */
1819 if (ret > 0)
1820 return -ret;
1821 return 0;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001822 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001823
Fabian Frederickc18450a2014-11-04 20:48:41 +01001824 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1825 return __udp4_lib_mcast_deliver(net, skb, uh,
Rick Jones36cbb242014-11-06 10:37:54 -08001826 saddr, daddr, udptable, proto);
Fabian Frederickc18450a2014-11-04 20:48:41 +01001827
1828 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
Ian Morris00db4122015-04-03 09:17:27 +01001829 if (sk) {
Eliezer Tamira5b50472013-06-10 11:40:00 +03001830 int ret;
1831
Tom Herbert224d0192015-01-05 13:56:14 -08001832 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
Tom Herbert2abb7cd2014-08-31 15:12:43 -07001833 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
1834 inet_compute_pseudo);
1835
Eliezer Tamira5b50472013-06-10 11:40:00 +03001836 ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001837
1838 /* a return value > 0 means to resubmit the input, but
1839 * it wants the return to be -protocol, or 0
1840 */
1841 if (ret > 0)
1842 return -ret;
1843 return 0;
1844 }
1845
1846 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1847 goto drop;
1848 nf_reset(skb);
1849
1850 /* No socket. Drop packet silently, if checksum is wrong */
1851 if (udp_lib_checksum_complete(skb))
1852 goto csum_error;
1853
Eric Dumazet02c22342016-04-27 16:44:30 -07001854 __UDP_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001855 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1856
1857 /*
1858 * Hmm. We got an UDP packet to a port to which we
1859 * don't wanna listen. Ignore it.
1860 */
1861 kfree_skb(skb);
1862 return 0;
1863
1864short_packet:
Joe Perchesba7a46f2014-11-11 10:59:17 -08001865 net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
1866 proto == IPPROTO_UDPLITE ? "Lite" : "",
1867 &saddr, ntohs(uh->source),
1868 ulen, skb->len,
1869 &daddr, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001870 goto drop;
1871
1872csum_error:
1873 /*
1874 * RFC1122: OK. Discards the bad packet silently (as far as
1875 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1876 */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001877 net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
1878 proto == IPPROTO_UDPLITE ? "Lite" : "",
1879 &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
1880 ulen);
Eric Dumazet02c22342016-04-27 16:44:30 -07001881 __UDP_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001882drop:
Eric Dumazet02c22342016-04-27 16:44:30 -07001883 __UDP_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001884 kfree_skb(skb);
1885 return 0;
1886}
1887
Shawn Bohrer421b3882013-10-07 11:01:39 -05001888/* We can only early demux multicast if there is a single matching socket.
1889 * If more than one socket found returns NULL
1890 */
1891static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net,
1892 __be16 loc_port, __be32 loc_addr,
1893 __be16 rmt_port, __be32 rmt_addr,
1894 int dif)
1895{
1896 struct sock *sk, *result;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001897 unsigned short hnum = ntohs(loc_port);
Eric Dumazetca065d02016-04-01 08:52:13 -07001898 unsigned int slot = udp_hashfn(net, hnum, udp_table.mask);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001899 struct udp_hslot *hslot = &udp_table.hash[slot];
1900
Eric Dumazet63c6f812014-06-12 16:13:06 -07001901 /* Do not bother scanning a too big list */
1902 if (hslot->count > 10)
1903 return NULL;
1904
Shawn Bohrer421b3882013-10-07 11:01:39 -05001905 result = NULL;
Eric Dumazetca065d02016-04-01 08:52:13 -07001906 sk_for_each_rcu(sk, &hslot->head) {
1907 if (__udp_is_mcast_sock(net, sk, loc_port, loc_addr,
1908 rmt_port, rmt_addr, dif, hnum)) {
1909 if (result)
1910 return NULL;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001911 result = sk;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001912 }
1913 }
Shawn Bohrer421b3882013-10-07 11:01:39 -05001914
Shawn Bohrer421b3882013-10-07 11:01:39 -05001915 return result;
1916}
1917
1918/* For unicast we should only early demux connected sockets or we can
1919 * break forwarding setups. The chains here can be long so only check
1920 * if the first socket is an exact match and if not move on.
1921 */
1922static struct sock *__udp4_lib_demux_lookup(struct net *net,
1923 __be16 loc_port, __be32 loc_addr,
1924 __be16 rmt_port, __be32 rmt_addr,
1925 int dif)
1926{
Shawn Bohrer421b3882013-10-07 11:01:39 -05001927 unsigned short hnum = ntohs(loc_port);
1928 unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum);
1929 unsigned int slot2 = hash2 & udp_table.mask;
1930 struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
Joe Perchesc7228312014-05-13 20:30:07 -07001931 INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001932 const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
Eric Dumazetca065d02016-04-01 08:52:13 -07001933 struct sock *sk;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001934
Eric Dumazetca065d02016-04-01 08:52:13 -07001935 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
1936 if (INET_MATCH(sk, net, acookie, rmt_addr,
1937 loc_addr, ports, dif))
1938 return sk;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001939 /* Only check first socket in chain */
1940 break;
1941 }
Eric Dumazetca065d02016-04-01 08:52:13 -07001942 return NULL;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001943}
1944
1945void udp_v4_early_demux(struct sk_buff *skb)
1946{
Eric Dumazet610438b2013-12-11 08:10:05 -08001947 struct net *net = dev_net(skb->dev);
1948 const struct iphdr *iph;
1949 const struct udphdr *uh;
Eric Dumazetca065d02016-04-01 08:52:13 -07001950 struct sock *sk = NULL;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001951 struct dst_entry *dst;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001952 int dif = skb->dev->ifindex;
Shawn Bohrer6e540302015-06-03 16:27:38 -05001953 int ours;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001954
1955 /* validate the packet */
1956 if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr)))
1957 return;
1958
Eric Dumazet610438b2013-12-11 08:10:05 -08001959 iph = ip_hdr(skb);
1960 uh = udp_hdr(skb);
1961
Shawn Bohrer421b3882013-10-07 11:01:39 -05001962 if (skb->pkt_type == PACKET_BROADCAST ||
Shawn Bohrer6e540302015-06-03 16:27:38 -05001963 skb->pkt_type == PACKET_MULTICAST) {
1964 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
1965
1966 if (!in_dev)
1967 return;
1968
Paolo Abeniad0ea192016-03-22 09:19:38 +01001969 /* we are supposed to accept bcast packets */
1970 if (skb->pkt_type == PACKET_MULTICAST) {
1971 ours = ip_check_mc_rcu(in_dev, iph->daddr, iph->saddr,
1972 iph->protocol);
1973 if (!ours)
1974 return;
1975 }
1976
Shawn Bohrer421b3882013-10-07 11:01:39 -05001977 sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr,
1978 uh->source, iph->saddr, dif);
Shawn Bohrer6e540302015-06-03 16:27:38 -05001979 } else if (skb->pkt_type == PACKET_HOST) {
Shawn Bohrer421b3882013-10-07 11:01:39 -05001980 sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr,
1981 uh->source, iph->saddr, dif);
Shawn Bohrer6e540302015-06-03 16:27:38 -05001982 }
Shawn Bohrer421b3882013-10-07 11:01:39 -05001983
Eric Dumazetca065d02016-04-01 08:52:13 -07001984 if (!sk || !atomic_inc_not_zero_hint(&sk->sk_refcnt, 2))
Shawn Bohrer421b3882013-10-07 11:01:39 -05001985 return;
1986
1987 skb->sk = sk;
Alexander Duyck82eabd92014-09-04 13:32:11 -04001988 skb->destructor = sock_efree;
Eric Dumazet10e2eb82015-08-01 12:14:33 +02001989 dst = READ_ONCE(sk->sk_rx_dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001990
1991 if (dst)
1992 dst = dst_check(dst, 0);
Eric Dumazet10e2eb82015-08-01 12:14:33 +02001993 if (dst) {
1994 /* DST_NOCACHE can not be used without taking a reference */
1995 if (dst->flags & DST_NOCACHE) {
1996 if (likely(atomic_inc_not_zero(&dst->__refcnt)))
1997 skb_dst_set(skb, dst);
1998 } else {
1999 skb_dst_set_noref(skb, dst);
2000 }
2001 }
Shawn Bohrer421b3882013-10-07 11:01:39 -05002002}
2003
David S. Millerdb8dac22008-03-06 16:22:02 -08002004int udp_rcv(struct sk_buff *skb)
2005{
Eric Dumazet645ca702008-10-29 01:41:45 -07002006 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08002007}
2008
Brian Haley7d06b2e2008-06-14 17:04:49 -07002009void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08002010{
Tom Parkin44046a52013-03-19 06:11:12 +00002011 struct udp_sock *up = udp_sk(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00002012 bool slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08002013 udp_flush_pending_frames(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00002014 unlock_sock_fast(sk, slow);
Tom Parkin44046a52013-03-19 06:11:12 +00002015 if (static_key_false(&udp_encap_needed) && up->encap_type) {
2016 void (*encap_destroy)(struct sock *sk);
2017 encap_destroy = ACCESS_ONCE(up->encap_destroy);
2018 if (encap_destroy)
2019 encap_destroy(sk);
2020 }
David S. Millerdb8dac22008-03-06 16:22:02 -08002021}
2022
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023/*
2024 * Socket option code for UDP
2025 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002026int udp_lib_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002027 char __user *optval, unsigned int optlen,
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002028 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029{
2030 struct udp_sock *up = udp_sk(sk);
Tom Herbert1c194482014-05-23 08:47:32 -07002031 int val, valbool;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11002033 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034
Eric Dumazetc482c562009-07-17 00:26:32 +00002035 if (optlen < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036 return -EINVAL;
2037
2038 if (get_user(val, (int __user *)optval))
2039 return -EFAULT;
2040
Tom Herbert1c194482014-05-23 08:47:32 -07002041 valbool = val ? 1 : 0;
2042
Stephen Hemminger6516c652007-03-08 20:41:55 -08002043 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044 case UDP_CORK:
2045 if (val != 0) {
2046 up->corkflag = 1;
2047 } else {
2048 up->corkflag = 0;
2049 lock_sock(sk);
Joe Perches4243cdc2014-11-11 21:59:20 -08002050 push_pending_frames(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051 release_sock(sk);
2052 }
2053 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002054
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 case UDP_ENCAP:
2056 switch (val) {
2057 case 0:
2058 case UDP_ENCAP_ESPINUDP:
2059 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07002060 up->encap_rcv = xfrm4_udp_encap_rcv;
2061 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07002062 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063 up->encap_type = val;
Eric Dumazet447167b2012-04-11 23:05:28 +00002064 udp_encap_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065 break;
2066 default:
2067 err = -ENOPROTOOPT;
2068 break;
2069 }
2070 break;
2071
Tom Herbert1c194482014-05-23 08:47:32 -07002072 case UDP_NO_CHECK6_TX:
2073 up->no_check6_tx = valbool;
2074 break;
2075
2076 case UDP_NO_CHECK6_RX:
2077 up->no_check6_rx = valbool;
2078 break;
2079
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002080 /*
2081 * UDP-Lite's partial checksum coverage (RFC 3828).
2082 */
2083 /* The sender sets actual checksum coverage length via this option.
2084 * The case coverage > packet length is handled by send module. */
2085 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002086 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002087 return -ENOPROTOOPT;
2088 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
2089 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002090 else if (val > USHRT_MAX)
2091 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002092 up->pcslen = val;
2093 up->pcflag |= UDPLITE_SEND_CC;
2094 break;
2095
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002096 /* The receiver specifies a minimum checksum coverage value. To make
2097 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002098 * used, this again means full checksum coverage. */
2099 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002100 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002101 return -ENOPROTOOPT;
2102 if (val != 0 && val < 8) /* Avoid silly minimal values. */
2103 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002104 else if (val > USHRT_MAX)
2105 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002106 up->pcrlen = val;
2107 up->pcflag |= UDPLITE_RECV_CC;
2108 break;
2109
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110 default:
2111 err = -ENOPROTOOPT;
2112 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002113 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114
2115 return err;
2116}
Eric Dumazetc482c562009-07-17 00:26:32 +00002117EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118
David S. Millerdb8dac22008-03-06 16:22:02 -08002119int udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002120 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002121{
2122 if (level == SOL_UDP || level == SOL_UDPLITE)
2123 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2124 udp_push_pending_frames);
2125 return ip_setsockopt(sk, level, optname, optval, optlen);
2126}
2127
2128#ifdef CONFIG_COMPAT
2129int compat_udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002130 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002131{
2132 if (level == SOL_UDP || level == SOL_UDPLITE)
2133 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2134 udp_push_pending_frames);
2135 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
2136}
2137#endif
2138
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002139int udp_lib_getsockopt(struct sock *sk, int level, int optname,
2140 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141{
2142 struct udp_sock *up = udp_sk(sk);
2143 int val, len;
2144
Eric Dumazetc482c562009-07-17 00:26:32 +00002145 if (get_user(len, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146 return -EFAULT;
2147
2148 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002149
Stephen Hemminger6516c652007-03-08 20:41:55 -08002150 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151 return -EINVAL;
2152
Stephen Hemminger6516c652007-03-08 20:41:55 -08002153 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 case UDP_CORK:
2155 val = up->corkflag;
2156 break;
2157
2158 case UDP_ENCAP:
2159 val = up->encap_type;
2160 break;
2161
Tom Herbert1c194482014-05-23 08:47:32 -07002162 case UDP_NO_CHECK6_TX:
2163 val = up->no_check6_tx;
2164 break;
2165
2166 case UDP_NO_CHECK6_RX:
2167 val = up->no_check6_rx;
2168 break;
2169
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002170 /* The following two cannot be changed on UDP sockets, the return is
2171 * always 0 (which corresponds to the full checksum coverage of UDP). */
2172 case UDPLITE_SEND_CSCOV:
2173 val = up->pcslen;
2174 break;
2175
2176 case UDPLITE_RECV_CSCOV:
2177 val = up->pcrlen;
2178 break;
2179
Linus Torvalds1da177e2005-04-16 15:20:36 -07002180 default:
2181 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002182 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183
Stephen Hemminger6516c652007-03-08 20:41:55 -08002184 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002185 return -EFAULT;
Eric Dumazetc482c562009-07-17 00:26:32 +00002186 if (copy_to_user(optval, &val, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002187 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002188 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002189}
Eric Dumazetc482c562009-07-17 00:26:32 +00002190EXPORT_SYMBOL(udp_lib_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191
David S. Millerdb8dac22008-03-06 16:22:02 -08002192int udp_getsockopt(struct sock *sk, int level, int optname,
2193 char __user *optval, int __user *optlen)
2194{
2195 if (level == SOL_UDP || level == SOL_UDPLITE)
2196 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2197 return ip_getsockopt(sk, level, optname, optval, optlen);
2198}
2199
2200#ifdef CONFIG_COMPAT
2201int compat_udp_getsockopt(struct sock *sk, int level, int optname,
2202 char __user *optval, int __user *optlen)
2203{
2204 if (level == SOL_UDP || level == SOL_UDPLITE)
2205 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2206 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
2207}
2208#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209/**
2210 * udp_poll - wait for a UDP event.
2211 * @file - file struct
2212 * @sock - socket
2213 * @wait - poll table
2214 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002215 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216 * blocking sockets. If application is using a blocking fd
2217 * and a packet with checksum error is in the queue;
2218 * then it could get return from select indicating data available
2219 * but then block when reading it. Add special case code
2220 * to work around these arguably broken applications.
2221 */
2222unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
2223{
2224 unsigned int mask = datagram_poll(file, sock, wait);
2225 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002226
David Majnemerc3f1dba2013-05-31 13:15:38 +00002227 sock_rps_record_flow(sk);
2228
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229 /* Check for false positives due to checksum errors */
Eric Dumazet85584672009-10-09 04:43:40 +00002230 if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
2231 !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
2232 mask &= ~(POLLIN | POLLRDNORM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233
2234 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002235
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236}
Eric Dumazetc482c562009-07-17 00:26:32 +00002237EXPORT_SYMBOL(udp_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238
David S. Millerdb8dac22008-03-06 16:22:02 -08002239struct proto udp_prot = {
2240 .name = "UDP",
2241 .owner = THIS_MODULE,
2242 .close = udp_lib_close,
2243 .connect = ip4_datagram_connect,
2244 .disconnect = udp_disconnect,
2245 .ioctl = udp_ioctl,
2246 .destroy = udp_destroy_sock,
2247 .setsockopt = udp_setsockopt,
2248 .getsockopt = udp_getsockopt,
2249 .sendmsg = udp_sendmsg,
2250 .recvmsg = udp_recvmsg,
2251 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07002252 .backlog_rcv = __udp_queue_rcv_skb,
Steffen Klassert8141ed92013-01-21 02:00:03 +00002253 .release_cb = ip4_datagram_release_cb,
David S. Millerdb8dac22008-03-06 16:22:02 -08002254 .hash = udp_lib_hash,
2255 .unhash = udp_lib_unhash,
Eric Dumazet719f8352010-09-08 05:08:44 +00002256 .rehash = udp_v4_rehash,
David S. Millerdb8dac22008-03-06 16:22:02 -08002257 .get_port = udp_v4_get_port,
2258 .memory_allocated = &udp_memory_allocated,
2259 .sysctl_mem = sysctl_udp_mem,
2260 .sysctl_wmem = &sysctl_udp_wmem_min,
2261 .sysctl_rmem = &sysctl_udp_rmem_min,
2262 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07002263 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07002264 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08002265#ifdef CONFIG_COMPAT
2266 .compat_setsockopt = compat_udp_setsockopt,
2267 .compat_getsockopt = compat_udp_getsockopt,
2268#endif
Octavian Purdilafcbdf092010-12-16 14:26:56 -08002269 .clear_sk = sk_prot_clear_portaddr_nulls,
David S. Millerdb8dac22008-03-06 16:22:02 -08002270};
Eric Dumazetc482c562009-07-17 00:26:32 +00002271EXPORT_SYMBOL(udp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002272
2273/* ------------------------------------------------------------------------ */
2274#ifdef CONFIG_PROC_FS
2275
Eric Dumazet645ca702008-10-29 01:41:45 -07002276static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002277{
2278 struct sock *sk;
2279 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002280 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002282 for (state->bucket = start; state->bucket <= state->udp_table->mask;
2283 ++state->bucket) {
Eric Dumazet645ca702008-10-29 01:41:45 -07002284 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002285
Eric Dumazetca065d02016-04-01 08:52:13 -07002286 if (hlist_empty(&hslot->head))
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002287 continue;
2288
Eric Dumazet645ca702008-10-29 01:41:45 -07002289 spin_lock_bh(&hslot->lock);
Eric Dumazetca065d02016-04-01 08:52:13 -07002290 sk_for_each(sk, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002291 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002292 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293 if (sk->sk_family == state->family)
2294 goto found;
2295 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002296 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297 }
2298 sk = NULL;
2299found:
2300 return sk;
2301}
2302
2303static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2304{
2305 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002306 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002307
2308 do {
Eric Dumazetca065d02016-04-01 08:52:13 -07002309 sk = sk_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002310 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311
Eric Dumazet645ca702008-10-29 01:41:45 -07002312 if (!sk) {
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002313 if (state->bucket <= state->udp_table->mask)
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002314 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07002315 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002316 }
2317 return sk;
2318}
2319
2320static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2321{
Eric Dumazet645ca702008-10-29 01:41:45 -07002322 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323
2324 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08002325 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326 --pos;
2327 return pos ? NULL : sk;
2328}
2329
2330static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2331{
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002332 struct udp_iter_state *state = seq->private;
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002333 state->bucket = MAX_UDP_PORTS;
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002334
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002335 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336}
2337
2338static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2339{
2340 struct sock *sk;
2341
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002342 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 sk = udp_get_idx(seq, 0);
2344 else
2345 sk = udp_get_next(seq, v);
2346
2347 ++*pos;
2348 return sk;
2349}
2350
2351static void udp_seq_stop(struct seq_file *seq, void *v)
2352{
Eric Dumazet645ca702008-10-29 01:41:45 -07002353 struct udp_iter_state *state = seq->private;
2354
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002355 if (state->bucket <= state->udp_table->mask)
Eric Dumazet645ca702008-10-29 01:41:45 -07002356 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002357}
2358
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002359int udp_seq_open(struct inode *inode, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360{
Al Virod9dda782013-03-31 18:16:14 -04002361 struct udp_seq_afinfo *afinfo = PDE_DATA(inode);
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002362 struct udp_iter_state *s;
2363 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002365 err = seq_open_net(inode, file, &afinfo->seq_ops,
2366 sizeof(struct udp_iter_state));
2367 if (err < 0)
2368 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002369
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002370 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002371 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07002372 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002373 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002374}
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002375EXPORT_SYMBOL(udp_seq_open);
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002376
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002378int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002379{
2380 struct proc_dir_entry *p;
2381 int rc = 0;
2382
Denis V. Lunevdda61922008-03-28 18:24:26 -07002383 afinfo->seq_ops.start = udp_seq_start;
2384 afinfo->seq_ops.next = udp_seq_next;
2385 afinfo->seq_ops.stop = udp_seq_stop;
2386
Denis V. Lunev84841c32008-05-02 04:10:08 -07002387 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002388 afinfo->seq_fops, afinfo);
Denis V. Lunev84841c32008-05-02 04:10:08 -07002389 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 rc = -ENOMEM;
2391 return rc;
2392}
Eric Dumazetc482c562009-07-17 00:26:32 +00002393EXPORT_SYMBOL(udp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002394
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002395void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396{
Gao fengece31ff2013-02-18 01:34:56 +00002397 remove_proc_entry(afinfo->name, net->proc_net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002398}
Eric Dumazetc482c562009-07-17 00:26:32 +00002399EXPORT_SYMBOL(udp_proc_unregister);
David S. Millerdb8dac22008-03-06 16:22:02 -08002400
2401/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002402static void udp4_format_sock(struct sock *sp, struct seq_file *f,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002403 int bucket)
David S. Millerdb8dac22008-03-06 16:22:02 -08002404{
2405 struct inet_sock *inet = inet_sk(sp);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002406 __be32 dest = inet->inet_daddr;
2407 __be32 src = inet->inet_rcv_saddr;
2408 __u16 destp = ntohs(inet->inet_dport);
2409 __u16 srcp = ntohs(inet->inet_sport);
David S. Millerdb8dac22008-03-06 16:22:02 -08002410
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002411 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
Tetsuo Handa652586d2013-11-14 14:31:57 -08002412 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
David S. Millerdb8dac22008-03-06 16:22:02 -08002413 bucket, src, srcp, dest, destp, sp->sk_state,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002414 sk_wmem_alloc_get(sp),
2415 sk_rmem_alloc_get(sp),
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06002416 0, 0L, 0,
2417 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2418 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002419 atomic_read(&sp->sk_refcnt), sp,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002420 atomic_read(&sp->sk_drops));
David S. Millerdb8dac22008-03-06 16:22:02 -08002421}
2422
2423int udp4_seq_show(struct seq_file *seq, void *v)
2424{
Tetsuo Handa652586d2013-11-14 14:31:57 -08002425 seq_setwidth(seq, 127);
David S. Millerdb8dac22008-03-06 16:22:02 -08002426 if (v == SEQ_START_TOKEN)
Tetsuo Handa652586d2013-11-14 14:31:57 -08002427 seq_puts(seq, " sl local_address rem_address st tx_queue "
David S. Millerdb8dac22008-03-06 16:22:02 -08002428 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002429 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08002430 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08002431 struct udp_iter_state *state = seq->private;
2432
Tetsuo Handa652586d2013-11-14 14:31:57 -08002433 udp4_format_sock(v, seq, state->bucket);
David S. Millerdb8dac22008-03-06 16:22:02 -08002434 }
Tetsuo Handa652586d2013-11-14 14:31:57 -08002435 seq_pad(seq, '\n');
David S. Millerdb8dac22008-03-06 16:22:02 -08002436 return 0;
2437}
2438
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002439static const struct file_operations udp_afinfo_seq_fops = {
2440 .owner = THIS_MODULE,
2441 .open = udp_seq_open,
2442 .read = seq_read,
2443 .llseek = seq_lseek,
2444 .release = seq_release_net
2445};
2446
David S. Millerdb8dac22008-03-06 16:22:02 -08002447/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08002448static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08002449 .name = "udp",
2450 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07002451 .udp_table = &udp_table,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002452 .seq_fops = &udp_afinfo_seq_fops,
Denis V. Lunevdda61922008-03-28 18:24:26 -07002453 .seq_ops = {
2454 .show = udp4_seq_show,
2455 },
David S. Millerdb8dac22008-03-06 16:22:02 -08002456};
2457
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002458static int __net_init udp4_proc_init_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002459{
2460 return udp_proc_register(net, &udp4_seq_afinfo);
2461}
2462
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002463static void __net_exit udp4_proc_exit_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002464{
2465 udp_proc_unregister(net, &udp4_seq_afinfo);
2466}
2467
2468static struct pernet_operations udp4_net_ops = {
2469 .init = udp4_proc_init_net,
2470 .exit = udp4_proc_exit_net,
2471};
2472
David S. Millerdb8dac22008-03-06 16:22:02 -08002473int __init udp4_proc_init(void)
2474{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002475 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002476}
2477
2478void udp4_proc_exit(void)
2479{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002480 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002481}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482#endif /* CONFIG_PROC_FS */
2483
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002484static __initdata unsigned long uhash_entries;
2485static int __init set_uhash_entries(char *str)
Eric Dumazet645ca702008-10-29 01:41:45 -07002486{
Eldad Zack413c27d2012-05-19 14:13:18 +00002487 ssize_t ret;
2488
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002489 if (!str)
2490 return 0;
Eldad Zack413c27d2012-05-19 14:13:18 +00002491
2492 ret = kstrtoul(str, 0, &uhash_entries);
2493 if (ret)
2494 return 0;
2495
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002496 if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2497 uhash_entries = UDP_HTABLE_SIZE_MIN;
2498 return 1;
2499}
2500__setup("uhash_entries=", set_uhash_entries);
Eric Dumazet645ca702008-10-29 01:41:45 -07002501
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002502void __init udp_table_init(struct udp_table *table, const char *name)
2503{
2504 unsigned int i;
2505
Tim Bird31fe62b2012-05-23 13:33:35 +00002506 table->hash = alloc_large_system_hash(name,
2507 2 * sizeof(struct udp_hslot),
2508 uhash_entries,
2509 21, /* one slot per 2 MB */
2510 0,
2511 &table->log,
2512 &table->mask,
2513 UDP_HTABLE_SIZE_MIN,
2514 64 * 1024);
2515
Eric Dumazet512615b2009-11-08 10:17:58 +00002516 table->hash2 = table->hash + (table->mask + 1);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002517 for (i = 0; i <= table->mask; i++) {
Eric Dumazetca065d02016-04-01 08:52:13 -07002518 INIT_HLIST_HEAD(&table->hash[i].head);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00002519 table->hash[i].count = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -07002520 spin_lock_init(&table->hash[i].lock);
2521 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002522 for (i = 0; i <= table->mask; i++) {
Eric Dumazetca065d02016-04-01 08:52:13 -07002523 INIT_HLIST_HEAD(&table->hash2[i].head);
Eric Dumazet512615b2009-11-08 10:17:58 +00002524 table->hash2[i].count = 0;
2525 spin_lock_init(&table->hash2[i].lock);
2526 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002527}
2528
Tom Herbert723b8e42015-02-24 09:17:31 -08002529u32 udp_flow_hashrnd(void)
2530{
2531 static u32 hashrnd __read_mostly;
2532
2533 net_get_random_once(&hashrnd, sizeof(hashrnd));
2534
2535 return hashrnd;
2536}
2537EXPORT_SYMBOL(udp_flow_hashrnd);
2538
Hideo Aoki95766ff2007-12-31 00:29:24 -08002539void __init udp_init(void)
2540{
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002541 unsigned long limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002542
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002543 udp_table_init(&udp_table, "UDP");
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002544 limit = nr_free_buffer_pages() / 8;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002545 limit = max(limit, 128UL);
2546 sysctl_udp_mem[0] = limit / 4 * 3;
2547 sysctl_udp_mem[1] = limit;
2548 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2549
2550 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2551 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2552}