blob: d92f94b7e4025dd4779e75e6a75f2de560713778 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * The User Datagram Protocol (UDP).
7 *
Jesper Juhl02c30a82005-05-05 16:16:16 -07008 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
Alan Cox113aa832008-10-13 19:01:08 -070011 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012 * Hirokazu Takahashi, <taka@valinux.co.jp>
13 *
14 * Fixes:
15 * Alan Cox : verify_area() calls
16 * Alan Cox : stopped close while in use off icmp
17 * messages. Not a fix but a botch that
18 * for udp at least is 'valid'.
19 * Alan Cox : Fixed icmp handling properly
20 * Alan Cox : Correct error for oversized datagrams
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090021 * Alan Cox : Tidied select() semantics.
22 * Alan Cox : udp_err() fixed properly, also now
Linus Torvalds1da177e2005-04-16 15:20:36 -070023 * select and read wake correctly on errors
24 * Alan Cox : udp_send verify_area moved to avoid mem leak
25 * Alan Cox : UDP can count its memory
26 * Alan Cox : send to an unknown connection causes
27 * an ECONNREFUSED off the icmp, but
28 * does NOT close.
29 * Alan Cox : Switched to new sk_buff handlers. No more backlog!
30 * Alan Cox : Using generic datagram code. Even smaller and the PEEK
31 * bug no longer crashes it.
32 * Fred Van Kempen : Net2e support for sk->broadcast.
33 * Alan Cox : Uses skb_free_datagram
34 * Alan Cox : Added get/set sockopt support.
35 * Alan Cox : Broadcasting without option set returns EACCES.
36 * Alan Cox : No wakeup calls. Instead we now use the callbacks.
37 * Alan Cox : Use ip_tos and ip_ttl
38 * Alan Cox : SNMP Mibs
39 * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support.
40 * Matt Dillon : UDP length checks.
41 * Alan Cox : Smarter af_inet used properly.
42 * Alan Cox : Use new kernel side addressing.
43 * Alan Cox : Incorrect return on truncated datagram receive.
44 * Arnt Gulbrandsen : New udp_send and stuff
45 * Alan Cox : Cache last socket
46 * Alan Cox : Route cache
47 * Jon Peatfield : Minor efficiency fix to sendto().
48 * Mike Shaver : RFC1122 checks.
49 * Alan Cox : Nonblocking error fix.
50 * Willy Konynenberg : Transparent proxying support.
51 * Mike McLagan : Routing by source
52 * David S. Miller : New socket lookup architecture.
53 * Last socket cache retained as it
54 * does have a high hit rate.
55 * Olaf Kirch : Don't linearise iovec on sendmsg.
56 * Andi Kleen : Some cleanups, cache destination entry
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090057 * for connect.
Linus Torvalds1da177e2005-04-16 15:20:36 -070058 * Vitaly E. Lavrov : Transparent proxy revived after year coma.
59 * Melvin Smith : Check msg_name not msg_namelen in sendto(),
60 * return ENOTCONN for unconnected sockets (POSIX)
61 * Janos Farkas : don't deliver multi/broadcasts to a different
62 * bound-to-device socket
63 * Hirokazu Takahashi : HW checksumming for outgoing UDP
64 * datagrams.
65 * Hirokazu Takahashi : sendfile() on UDP works now.
66 * Arnaldo C. Melo : convert /proc/net/udp to seq_file
67 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
68 * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind
69 * a single port at the same time.
70 * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support
James Chapman342f0232007-06-27 15:37:46 -070071 * James Chapman : Add L2TP encapsulation type.
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 *
73 *
74 * This program is free software; you can redistribute it and/or
75 * modify it under the terms of the GNU General Public License
76 * as published by the Free Software Foundation; either version
77 * 2 of the License, or (at your option) any later version.
78 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090079
Joe Perchesafd465032012-03-12 07:03:32 +000080#define pr_fmt(fmt) "UDP: " fmt
81
Linus Torvalds1da177e2005-04-16 15:20:36 -070082#include <asm/uaccess.h>
83#include <asm/ioctls.h>
Hideo Aoki95766ff2007-12-31 00:29:24 -080084#include <linux/bootmem.h>
Eric Dumazet8203efb2008-10-29 02:32:32 -070085#include <linux/highmem.h>
86#include <linux/swap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070087#include <linux/types.h>
88#include <linux/fcntl.h>
89#include <linux/module.h>
90#include <linux/socket.h>
91#include <linux/sockios.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020092#include <linux/igmp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070093#include <linux/in.h>
94#include <linux/errno.h>
95#include <linux/timer.h>
96#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097#include <linux/inet.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <linux/netdevice.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090099#include <linux/slab.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -0700100#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101#include <linux/skbuff.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200104#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#include <net/icmp.h>
Shawn Bohrer421b3882013-10-07 11:01:39 -0500106#include <net/inet_hashtables.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#include <net/route.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108#include <net/checksum.h>
109#include <net/xfrm.h>
Satoru Moriya296f7ea2011-06-17 11:58:39 +0000110#include <trace/events/udp.h>
Eric Dumazet447167b2012-04-11 23:05:28 +0000111#include <linux/static_key.h>
Eric Dumazet22911fc2012-06-27 00:23:44 +0000112#include <trace/events/skb.h>
Eliezer Tamir076bb0c2013-07-10 17:13:17 +0300113#include <net/busy_poll.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800114#include "udp_impl.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000116struct udp_table udp_table __read_mostly;
Eric Dumazet645ca702008-10-29 01:41:45 -0700117EXPORT_SYMBOL(udp_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118
Eric Dumazet8d987e52010-11-09 23:24:26 +0000119long sysctl_udp_mem[3] __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800120EXPORT_SYMBOL(sysctl_udp_mem);
Eric Dumazetc482c562009-07-17 00:26:32 +0000121
122int sysctl_udp_rmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800123EXPORT_SYMBOL(sysctl_udp_rmem_min);
Eric Dumazetc482c562009-07-17 00:26:32 +0000124
125int sysctl_udp_wmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800126EXPORT_SYMBOL(sysctl_udp_wmem_min);
127
Eric Dumazet8d987e52010-11-09 23:24:26 +0000128atomic_long_t udp_memory_allocated;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800129EXPORT_SYMBOL(udp_memory_allocated);
130
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000131#define MAX_UDP_PORTS 65536
132#define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN)
Eric Dumazet98322f22009-01-26 21:35:35 -0800133
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700134static int udp_lib_lport_inuse(struct net *net, __u16 num,
Eric Dumazet645ca702008-10-29 01:41:45 -0700135 const struct udp_hslot *hslot,
Eric Dumazet98322f22009-01-26 21:35:35 -0800136 unsigned long *bitmap,
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700137 struct sock *sk,
138 int (*saddr_comp)(const struct sock *sk1,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000139 const struct sock *sk2),
140 unsigned int log)
Gerrit Renker25030a72006-08-26 20:06:05 -0700141{
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700142 struct sock *sk2;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800143 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000144 kuid_t uid = sock_i_uid(sk);
Gerrit Renker25030a72006-08-26 20:06:05 -0700145
Eric Dumazet88ab1932008-11-16 19:39:21 -0800146 sk_nulls_for_each(sk2, node, &hslot->head)
Joe Perches9d4fb272009-11-23 10:41:23 -0800147 if (net_eq(sock_net(sk2), net) &&
148 sk2 != sk &&
Eric Dumazetd4cada42009-11-08 10:17:30 +0000149 (bitmap || udp_sk(sk2)->udp_port_hash == num) &&
Joe Perches9d4fb272009-11-23 10:41:23 -0800150 (!sk2->sk_reuse || !sk->sk_reuse) &&
151 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
152 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000153 (!sk2->sk_reuseport || !sk->sk_reuseport ||
154 !uid_eq(uid, sock_i_uid(sk2))) &&
Eric Dumazet98322f22009-01-26 21:35:35 -0800155 (*saddr_comp)(sk, sk2)) {
156 if (bitmap)
Eric Dumazetd4cada42009-11-08 10:17:30 +0000157 __set_bit(udp_sk(sk2)->udp_port_hash >> log,
158 bitmap);
Eric Dumazet98322f22009-01-26 21:35:35 -0800159 else
160 return 1;
161 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700162 return 0;
163}
164
Eric Dumazet30fff922009-11-09 05:26:33 +0000165/*
166 * Note: we still hold spinlock of primary hash chain, so no other writer
167 * can insert/delete a socket with local_port == num
168 */
169static int udp_lib_lport_inuse2(struct net *net, __u16 num,
170 struct udp_hslot *hslot2,
171 struct sock *sk,
172 int (*saddr_comp)(const struct sock *sk1,
173 const struct sock *sk2))
174{
175 struct sock *sk2;
176 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000177 kuid_t uid = sock_i_uid(sk);
Eric Dumazet30fff922009-11-09 05:26:33 +0000178 int res = 0;
179
180 spin_lock(&hslot2->lock);
181 udp_portaddr_for_each_entry(sk2, node, &hslot2->head)
Joe Perches9d4fb272009-11-23 10:41:23 -0800182 if (net_eq(sock_net(sk2), net) &&
183 sk2 != sk &&
184 (udp_sk(sk2)->udp_port_hash == num) &&
185 (!sk2->sk_reuse || !sk->sk_reuse) &&
186 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
187 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000188 (!sk2->sk_reuseport || !sk->sk_reuseport ||
189 !uid_eq(uid, sock_i_uid(sk2))) &&
Eric Dumazet30fff922009-11-09 05:26:33 +0000190 (*saddr_comp)(sk, sk2)) {
191 res = 1;
192 break;
193 }
194 spin_unlock(&hslot2->lock);
195 return res;
196}
197
Gerrit Renker25030a72006-08-26 20:06:05 -0700198/**
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700199 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
Gerrit Renker25030a72006-08-26 20:06:05 -0700200 *
201 * @sk: socket struct in question
202 * @snum: port number to look up
David S. Millerdf2bc452007-06-05 15:18:43 -0700203 * @saddr_comp: AF-dependent comparison of bound local IP addresses
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300204 * @hash2_nulladdr: AF-dependent hash value in secondary hash chains,
Eric Dumazet30fff922009-11-09 05:26:33 +0000205 * with NULL address
Gerrit Renker25030a72006-08-26 20:06:05 -0700206 */
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700207int udp_lib_get_port(struct sock *sk, unsigned short snum,
David S. Millerdf2bc452007-06-05 15:18:43 -0700208 int (*saddr_comp)(const struct sock *sk1,
Eric Dumazet30fff922009-11-09 05:26:33 +0000209 const struct sock *sk2),
210 unsigned int hash2_nulladdr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211{
Eric Dumazet512615b2009-11-08 10:17:58 +0000212 struct udp_hslot *hslot, *hslot2;
Eric Dumazet645ca702008-10-29 01:41:45 -0700213 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Gerrit Renker25030a72006-08-26 20:06:05 -0700214 int error = 1;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900215 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700217 if (!snum) {
Eric Dumazet9088c562008-10-08 11:44:17 -0700218 int low, high, remaining;
Eric Dumazet95c96172012-04-15 05:58:06 +0000219 unsigned int rand;
Eric Dumazet98322f22009-01-26 21:35:35 -0800220 unsigned short first, last;
221 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -0700223 inet_get_local_port_range(net, &low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700224 remaining = (high - low) + 1;
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700225
Aruna-Hewapathirane63862b52014-01-11 07:15:59 -0500226 rand = prandom_u32();
Eric Dumazet98322f22009-01-26 21:35:35 -0800227 first = (((u64)rand * remaining) >> 32) + low;
228 /*
229 * force rand to be an odd multiple of UDP_HTABLE_SIZE
230 */
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000231 rand = (rand | 1) * (udptable->mask + 1);
Eric Dumazet5781b232009-12-13 19:32:39 -0800232 last = first + udptable->mask + 1;
233 do {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000234 hslot = udp_hashslot(udptable, net, first);
Eric Dumazet98322f22009-01-26 21:35:35 -0800235 bitmap_zero(bitmap, PORTS_PER_CHAIN);
Eric Dumazet645ca702008-10-29 01:41:45 -0700236 spin_lock_bh(&hslot->lock);
Eric Dumazet98322f22009-01-26 21:35:35 -0800237 udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000238 saddr_comp, udptable->log);
Eric Dumazet98322f22009-01-26 21:35:35 -0800239
240 snum = first;
241 /*
242 * Iterate on all possible values of snum for this hash.
243 * Using steps of an odd multiple of UDP_HTABLE_SIZE
244 * give us randomization and full range coverage.
245 */
Eric Dumazet9088c562008-10-08 11:44:17 -0700246 do {
Eric Dumazet98322f22009-01-26 21:35:35 -0800247 if (low <= snum && snum <= high &&
Amerigo Wange3826f12010-05-05 00:27:06 +0000248 !test_bit(snum >> udptable->log, bitmap) &&
WANG Cong122ff242014-05-12 16:04:53 -0700249 !inet_is_local_reserved_port(net, snum))
Eric Dumazet98322f22009-01-26 21:35:35 -0800250 goto found;
251 snum += rand;
252 } while (snum != first);
253 spin_unlock_bh(&hslot->lock);
Eric Dumazet5781b232009-12-13 19:32:39 -0800254 } while (++first != last);
Eric Dumazet98322f22009-01-26 21:35:35 -0800255 goto fail;
Eric Dumazet645ca702008-10-29 01:41:45 -0700256 } else {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000257 hslot = udp_hashslot(udptable, net, snum);
Eric Dumazet645ca702008-10-29 01:41:45 -0700258 spin_lock_bh(&hslot->lock);
Eric Dumazet30fff922009-11-09 05:26:33 +0000259 if (hslot->count > 10) {
260 int exist;
261 unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum;
262
263 slot2 &= udptable->mask;
264 hash2_nulladdr &= udptable->mask;
265
266 hslot2 = udp_hashslot2(udptable, slot2);
267 if (hslot->count < hslot2->count)
268 goto scan_primary_hash;
269
270 exist = udp_lib_lport_inuse2(net, snum, hslot2,
271 sk, saddr_comp);
272 if (!exist && (hash2_nulladdr != slot2)) {
273 hslot2 = udp_hashslot2(udptable, hash2_nulladdr);
274 exist = udp_lib_lport_inuse2(net, snum, hslot2,
275 sk, saddr_comp);
276 }
277 if (exist)
278 goto fail_unlock;
279 else
280 goto found;
281 }
282scan_primary_hash:
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000283 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk,
284 saddr_comp, 0))
Eric Dumazet645ca702008-10-29 01:41:45 -0700285 goto fail_unlock;
286 }
Eric Dumazet98322f22009-01-26 21:35:35 -0800287found:
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000288 inet_sk(sk)->inet_num = snum;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000289 udp_sk(sk)->udp_port_hash = snum;
290 udp_sk(sk)->udp_portaddr_hash ^= snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 if (sk_unhashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800292 sk_nulls_add_node_rcu(sk, &hslot->head);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +0000293 hslot->count++;
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700294 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet512615b2009-11-08 10:17:58 +0000295
296 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
297 spin_lock(&hslot2->lock);
298 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
299 &hslot2->head);
300 hslot2->count++;
301 spin_unlock(&hslot2->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700303 error = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -0700304fail_unlock:
305 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306fail:
Gerrit Renker25030a72006-08-26 20:06:05 -0700307 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308}
Eric Dumazetc482c562009-07-17 00:26:32 +0000309EXPORT_SYMBOL(udp_lib_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310
Vlad Yasevich499923c2009-04-09 17:37:33 +0000311static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
David S. Millerdb8dac22008-03-06 16:22:02 -0800312{
313 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
314
Eric Dumazetc482c562009-07-17 00:26:32 +0000315 return (!ipv6_only_sock(sk2) &&
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000316 (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr ||
317 inet1->inet_rcv_saddr == inet2->inet_rcv_saddr));
David S. Millerdb8dac22008-03-06 16:22:02 -0800318}
319
Eric Dumazetd4cada42009-11-08 10:17:30 +0000320static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr,
321 unsigned int port)
322{
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700323 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000324}
325
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700326int udp_v4_get_port(struct sock *sk, unsigned short snum)
David S. Millerdb8dac22008-03-06 16:22:02 -0800327{
Eric Dumazet30fff922009-11-09 05:26:33 +0000328 unsigned int hash2_nulladdr =
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700329 udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum);
Eric Dumazet30fff922009-11-09 05:26:33 +0000330 unsigned int hash2_partial =
331 udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0);
332
Eric Dumazetd4cada42009-11-08 10:17:30 +0000333 /* precompute partial secondary hash */
Eric Dumazet30fff922009-11-09 05:26:33 +0000334 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
335 return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr);
David S. Millerdb8dac22008-03-06 16:22:02 -0800336}
337
Eric Dumazet645ca702008-10-29 01:41:45 -0700338static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr,
339 unsigned short hnum,
340 __be16 sport, __be32 daddr, __be16 dport, int dif)
341{
342 int score = -1;
343
Eric Dumazetd4cada42009-11-08 10:17:30 +0000344 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
Eric Dumazet645ca702008-10-29 01:41:45 -0700345 !ipv6_only_sock(sk)) {
346 struct inet_sock *inet = inet_sk(sk);
347
Tom Herbertba418fa2013-01-22 09:50:32 +0000348 score = (sk->sk_family == PF_INET ? 2 : 1);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000349 if (inet->inet_rcv_saddr) {
350 if (inet->inet_rcv_saddr != daddr)
Eric Dumazet645ca702008-10-29 01:41:45 -0700351 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000352 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700353 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000354 if (inet->inet_daddr) {
355 if (inet->inet_daddr != saddr)
Eric Dumazet645ca702008-10-29 01:41:45 -0700356 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000357 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700358 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000359 if (inet->inet_dport) {
360 if (inet->inet_dport != sport)
Eric Dumazet645ca702008-10-29 01:41:45 -0700361 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000362 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700363 }
364 if (sk->sk_bound_dev_if) {
365 if (sk->sk_bound_dev_if != dif)
366 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000367 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700368 }
369 }
370 return score;
371}
372
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000373/*
374 * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num)
375 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000376static inline int compute_score2(struct sock *sk, struct net *net,
377 __be32 saddr, __be16 sport,
378 __be32 daddr, unsigned int hnum, int dif)
379{
380 int score = -1;
381
382 if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) {
383 struct inet_sock *inet = inet_sk(sk);
384
385 if (inet->inet_rcv_saddr != daddr)
386 return -1;
387 if (inet->inet_num != hnum)
388 return -1;
389
Tom Herbertba418fa2013-01-22 09:50:32 +0000390 score = (sk->sk_family == PF_INET ? 2 : 1);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000391 if (inet->inet_daddr) {
392 if (inet->inet_daddr != saddr)
393 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000394 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000395 }
396 if (inet->inet_dport) {
397 if (inet->inet_dport != sport)
398 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000399 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000400 }
401 if (sk->sk_bound_dev_if) {
402 if (sk->sk_bound_dev_if != dif)
403 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000404 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000405 }
406 }
407 return score;
408}
409
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200410static unsigned int udp_ehashfn(struct net *net, const __be32 laddr,
411 const __u16 lport, const __be32 faddr,
412 const __be16 fport)
413{
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200414 static u32 udp_ehash_secret __read_mostly;
415
416 net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret));
417
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200418 return __inet_ehashfn(laddr, lport, faddr, fport,
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200419 udp_ehash_secret + net_hash_mix(net));
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200420}
421
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000422
423/* called with read_rcu_lock() */
424static struct sock *udp4_lib_lookup2(struct net *net,
425 __be32 saddr, __be16 sport,
426 __be32 daddr, unsigned int hnum, int dif,
427 struct udp_hslot *hslot2, unsigned int slot2)
428{
429 struct sock *sk, *result;
430 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000431 int score, badness, matches = 0, reuseport = 0;
432 u32 hash = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000433
434begin:
435 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000436 badness = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000437 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
438 score = compute_score2(sk, net, saddr, sport,
439 daddr, hnum, dif);
440 if (score > badness) {
441 result = sk;
442 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000443 reuseport = sk->sk_reuseport;
444 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200445 hash = udp_ehashfn(net, daddr, hnum,
446 saddr, sport);
Tom Herbertba418fa2013-01-22 09:50:32 +0000447 matches = 1;
448 }
449 } else if (score == badness && reuseport) {
450 matches++;
451 if (((u64)hash * matches) >> 32 == 0)
452 result = sk;
453 hash = next_pseudo_random32(hash);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000454 }
455 }
456 /*
457 * if the nulls value we got at the end of this lookup is
458 * not the expected one, we must restart lookup.
459 * We probably met an item that was moved to another chain.
460 */
461 if (get_nulls_value(node) != slot2)
462 goto begin;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000463 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000464 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000465 result = NULL;
466 else if (unlikely(compute_score2(result, net, saddr, sport,
467 daddr, hnum, dif) < badness)) {
468 sock_put(result);
469 goto begin;
470 }
471 }
472 return result;
473}
474
David S. Millerdb8dac22008-03-06 16:22:02 -0800475/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
476 * harder than this. -DaveM
477 */
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000478struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -0800479 __be16 sport, __be32 daddr, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700480 int dif, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800481{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700482 struct sock *sk, *result;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800483 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800484 unsigned short hnum = ntohs(dport);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000485 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
486 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
Tom Herbertba418fa2013-01-22 09:50:32 +0000487 int score, badness, matches = 0, reuseport = 0;
488 u32 hash = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800489
Eric Dumazet271b72c2008-10-29 02:11:14 -0700490 rcu_read_lock();
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000491 if (hslot->count > 10) {
492 hash2 = udp4_portaddr_hash(net, daddr, hnum);
493 slot2 = hash2 & udptable->mask;
494 hslot2 = &udptable->hash2[slot2];
495 if (hslot->count < hslot2->count)
496 goto begin;
497
498 result = udp4_lib_lookup2(net, saddr, sport,
499 daddr, hnum, dif,
500 hslot2, slot2);
501 if (!result) {
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700502 hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000503 slot2 = hash2 & udptable->mask;
504 hslot2 = &udptable->hash2[slot2];
505 if (hslot->count < hslot2->count)
506 goto begin;
507
Jorge Boncompte [DTI2]1223c672010-04-08 04:56:48 +0000508 result = udp4_lib_lookup2(net, saddr, sport,
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700509 htonl(INADDR_ANY), hnum, dif,
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000510 hslot2, slot2);
511 }
512 rcu_read_unlock();
513 return result;
514 }
Eric Dumazet271b72c2008-10-29 02:11:14 -0700515begin:
516 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000517 badness = 0;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800518 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
Eric Dumazet645ca702008-10-29 01:41:45 -0700519 score = compute_score(sk, net, saddr, hnum, sport,
520 daddr, dport, dif);
521 if (score > badness) {
522 result = sk;
523 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000524 reuseport = sk->sk_reuseport;
525 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200526 hash = udp_ehashfn(net, daddr, hnum,
527 saddr, sport);
Tom Herbertba418fa2013-01-22 09:50:32 +0000528 matches = 1;
529 }
530 } else if (score == badness && reuseport) {
531 matches++;
532 if (((u64)hash * matches) >> 32 == 0)
533 result = sk;
534 hash = next_pseudo_random32(hash);
David S. Millerdb8dac22008-03-06 16:22:02 -0800535 }
536 }
Eric Dumazet88ab1932008-11-16 19:39:21 -0800537 /*
538 * if the nulls value we got at the end of this lookup is
539 * not the expected one, we must restart lookup.
540 * We probably met an item that was moved to another chain.
541 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000542 if (get_nulls_value(node) != slot)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800543 goto begin;
544
Eric Dumazet271b72c2008-10-29 02:11:14 -0700545 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000546 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet271b72c2008-10-29 02:11:14 -0700547 result = NULL;
548 else if (unlikely(compute_score(result, net, saddr, hnum, sport,
549 daddr, dport, dif) < badness)) {
550 sock_put(result);
551 goto begin;
552 }
553 }
554 rcu_read_unlock();
David S. Millerdb8dac22008-03-06 16:22:02 -0800555 return result;
556}
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000557EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
David S. Millerdb8dac22008-03-06 16:22:02 -0800558
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700559static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
560 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700561 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700562{
563 const struct iphdr *iph = ip_hdr(skb);
564
Eric Dumazet8afdd992013-12-10 18:07:23 -0800565 return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport,
566 iph->daddr, dport, inet_iif(skb),
567 udptable);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700568}
569
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700570struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
571 __be32 daddr, __be16 dport, int dif)
572{
Eric Dumazet645ca702008-10-29 01:41:45 -0700573 return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700574}
575EXPORT_SYMBOL_GPL(udp4_lib_lookup);
576
Shawn Bohrer421b3882013-10-07 11:01:39 -0500577static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk,
578 __be16 loc_port, __be32 loc_addr,
579 __be16 rmt_port, __be32 rmt_addr,
580 int dif, unsigned short hnum)
581{
582 struct inet_sock *inet = inet_sk(sk);
583
584 if (!net_eq(sock_net(sk), net) ||
585 udp_sk(sk)->udp_port_hash != hnum ||
586 (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
587 (inet->inet_dport != rmt_port && inet->inet_dport) ||
588 (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) ||
589 ipv6_only_sock(sk) ||
590 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
591 return false;
592 if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif))
593 return false;
594 return true;
595}
596
Eric Dumazet920a4612008-11-01 21:22:23 -0700597static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk,
David S. Millerdb8dac22008-03-06 16:22:02 -0800598 __be16 loc_port, __be32 loc_addr,
599 __be16 rmt_port, __be32 rmt_addr,
600 int dif)
601{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800602 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800603 struct sock *s = sk;
604 unsigned short hnum = ntohs(loc_port);
605
Eric Dumazet88ab1932008-11-16 19:39:21 -0800606 sk_nulls_for_each_from(s, node) {
Shawn Bohrer421b3882013-10-07 11:01:39 -0500607 if (__udp_is_mcast_sock(net, s,
608 loc_port, loc_addr,
609 rmt_port, rmt_addr,
610 dif, hnum))
611 goto found;
David S. Millerdb8dac22008-03-06 16:22:02 -0800612 }
613 s = NULL;
614found:
615 return s;
616}
617
618/*
619 * This routine is called by the ICMP module when it gets some
620 * sort of error condition. If err < 0 then the socket should
621 * be closed and the error returned to the user. If err > 0
622 * it's just the icmp type << 8 | icmp code.
623 * Header points to the ip header of the error packet. We move
624 * on past this. Then (as it used to claim before adjustment)
625 * header points to the first 8 bytes of the udp header. We need
626 * to find the appropriate port.
627 */
628
Eric Dumazet645ca702008-10-29 01:41:45 -0700629void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800630{
631 struct inet_sock *inet;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000632 const struct iphdr *iph = (const struct iphdr *)skb->data;
Eric Dumazetc482c562009-07-17 00:26:32 +0000633 struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
David S. Millerdb8dac22008-03-06 16:22:02 -0800634 const int type = icmp_hdr(skb)->type;
635 const int code = icmp_hdr(skb)->code;
636 struct sock *sk;
637 int harderr;
638 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700639 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800640
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700641 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
David S. Millerdb8dac22008-03-06 16:22:02 -0800642 iph->saddr, uh->source, skb->dev->ifindex, udptable);
643 if (sk == NULL) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700644 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800645 return; /* No socket for error */
646 }
647
648 err = 0;
649 harderr = 0;
650 inet = inet_sk(sk);
651
652 switch (type) {
653 default:
654 case ICMP_TIME_EXCEEDED:
655 err = EHOSTUNREACH;
656 break;
657 case ICMP_SOURCE_QUENCH:
658 goto out;
659 case ICMP_PARAMETERPROB:
660 err = EPROTO;
661 harderr = 1;
662 break;
663 case ICMP_DEST_UNREACH:
664 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
David S. Miller36393392012-06-14 22:21:46 -0700665 ipv4_sk_update_pmtu(skb, sk, info);
David S. Millerdb8dac22008-03-06 16:22:02 -0800666 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
667 err = EMSGSIZE;
668 harderr = 1;
669 break;
670 }
671 goto out;
672 }
673 err = EHOSTUNREACH;
674 if (code <= NR_ICMP_UNREACH) {
675 harderr = icmp_err_convert[code].fatal;
676 err = icmp_err_convert[code].errno;
677 }
678 break;
David S. Miller55be7a92012-07-11 21:27:49 -0700679 case ICMP_REDIRECT:
680 ipv4_sk_redirect(skb, sk);
Duan Jiong1a462d12013-09-20 18:20:28 +0800681 goto out;
David S. Millerdb8dac22008-03-06 16:22:02 -0800682 }
683
684 /*
685 * RFC1122: OK. Passes ICMP errors back to application, as per
686 * 4.1.3.3.
687 */
688 if (!inet->recverr) {
689 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
690 goto out;
Eric Dumazetb1faf562010-05-31 23:44:05 -0700691 } else
Eric Dumazetc482c562009-07-17 00:26:32 +0000692 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
Eric Dumazetb1faf562010-05-31 23:44:05 -0700693
David S. Millerdb8dac22008-03-06 16:22:02 -0800694 sk->sk_err = err;
695 sk->sk_error_report(sk);
696out:
697 sock_put(sk);
698}
699
700void udp_err(struct sk_buff *skb, u32 info)
701{
Eric Dumazet645ca702008-10-29 01:41:45 -0700702 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800703}
704
705/*
706 * Throw away all pending data and cancel the corking. Socket is locked.
707 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400708void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800709{
710 struct udp_sock *up = udp_sk(sk);
711
712 if (up->pending) {
713 up->len = 0;
714 up->pending = 0;
715 ip_flush_pending_frames(sk);
716 }
717}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400718EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800719
720/**
Herbert Xuf6b96642011-03-01 02:36:48 +0000721 * udp4_hwcsum - handle outgoing HW checksumming
David S. Millerdb8dac22008-03-06 16:22:02 -0800722 * @skb: sk_buff containing the filled-in UDP header
723 * (checksum field must be zeroed out)
Herbert Xuf6b96642011-03-01 02:36:48 +0000724 * @src: source IP address
725 * @dst: destination IP address
David S. Millerdb8dac22008-03-06 16:22:02 -0800726 */
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200727void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
David S. Millerdb8dac22008-03-06 16:22:02 -0800728{
David S. Millerdb8dac22008-03-06 16:22:02 -0800729 struct udphdr *uh = udp_hdr(skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000730 int offset = skb_transport_offset(skb);
731 int len = skb->len - offset;
732 int hlen = len;
David S. Millerdb8dac22008-03-06 16:22:02 -0800733 __wsum csum = 0;
734
WANG Congebbe4952014-06-02 16:12:02 -0700735 if (!skb_has_frag_list(skb)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800736 /*
737 * Only one fragment on the socket.
738 */
739 skb->csum_start = skb_transport_header(skb) - skb->head;
740 skb->csum_offset = offsetof(struct udphdr, check);
Herbert Xuf6b96642011-03-01 02:36:48 +0000741 uh->check = ~csum_tcpudp_magic(src, dst, len,
742 IPPROTO_UDP, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800743 } else {
WANG Congebbe4952014-06-02 16:12:02 -0700744 struct sk_buff *frags;
745
David S. Millerdb8dac22008-03-06 16:22:02 -0800746 /*
747 * HW-checksum won't work as there are two or more
748 * fragments on the socket so that all csums of sk_buffs
749 * should be together
750 */
WANG Congebbe4952014-06-02 16:12:02 -0700751 skb_walk_frags(skb, frags) {
Herbert Xuf6b96642011-03-01 02:36:48 +0000752 csum = csum_add(csum, frags->csum);
753 hlen -= frags->len;
WANG Congebbe4952014-06-02 16:12:02 -0700754 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800755
Herbert Xuf6b96642011-03-01 02:36:48 +0000756 csum = skb_checksum(skb, offset, hlen, csum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800757 skb->ip_summed = CHECKSUM_NONE;
758
David S. Millerdb8dac22008-03-06 16:22:02 -0800759 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
760 if (uh->check == 0)
761 uh->check = CSUM_MANGLED_0;
762 }
763}
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200764EXPORT_SYMBOL_GPL(udp4_hwcsum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800765
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700766/* Function to set UDP checksum for an IPv4 UDP packet. This is intended
767 * for the simple case like when setting the checksum for a UDP tunnel.
768 */
769void udp_set_csum(bool nocheck, struct sk_buff *skb,
770 __be32 saddr, __be32 daddr, int len)
771{
772 struct udphdr *uh = udp_hdr(skb);
773
774 if (nocheck)
775 uh->check = 0;
776 else if (skb_is_gso(skb))
777 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
778 else if (skb_dst(skb) && skb_dst(skb)->dev &&
779 (skb_dst(skb)->dev->features & NETIF_F_V4_CSUM)) {
780
781 BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL);
782
783 skb->ip_summed = CHECKSUM_PARTIAL;
784 skb->csum_start = skb_transport_header(skb) - skb->head;
785 skb->csum_offset = offsetof(struct udphdr, check);
786 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
787 } else {
788 __wsum csum;
789
790 BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL);
791
792 uh->check = 0;
793 csum = skb_checksum(skb, 0, len, 0);
794 uh->check = udp_v4_check(len, saddr, daddr, csum);
795 if (uh->check == 0)
796 uh->check = CSUM_MANGLED_0;
797
798 skb->ip_summed = CHECKSUM_UNNECESSARY;
799 }
800}
801EXPORT_SYMBOL(udp_set_csum);
802
David S. Miller79ab0532011-05-09 13:31:04 -0700803static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
Herbert Xuf6b96642011-03-01 02:36:48 +0000804{
805 struct sock *sk = skb->sk;
806 struct inet_sock *inet = inet_sk(sk);
807 struct udphdr *uh;
Herbert Xuf6b96642011-03-01 02:36:48 +0000808 int err = 0;
809 int is_udplite = IS_UDPLITE(sk);
810 int offset = skb_transport_offset(skb);
811 int len = skb->len - offset;
812 __wsum csum = 0;
813
814 /*
815 * Create a UDP header
816 */
817 uh = udp_hdr(skb);
818 uh->source = inet->inet_sport;
David S. Miller79ab0532011-05-09 13:31:04 -0700819 uh->dest = fl4->fl4_dport;
Herbert Xuf6b96642011-03-01 02:36:48 +0000820 uh->len = htons(len);
821 uh->check = 0;
822
823 if (is_udplite) /* UDP-Lite */
824 csum = udplite_csum(skb);
825
Tom Herbert28448b82014-05-23 08:47:19 -0700826 else if (sk->sk_no_check_tx) { /* UDP csum disabled */
Herbert Xuf6b96642011-03-01 02:36:48 +0000827
828 skb->ip_summed = CHECKSUM_NONE;
829 goto send;
830
831 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
832
David S. Miller79ab0532011-05-09 13:31:04 -0700833 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
Herbert Xuf6b96642011-03-01 02:36:48 +0000834 goto send;
835
836 } else
837 csum = udp_csum(skb);
838
839 /* add protocol-dependent pseudo-header */
David S. Miller79ab0532011-05-09 13:31:04 -0700840 uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
Herbert Xuf6b96642011-03-01 02:36:48 +0000841 sk->sk_protocol, csum);
842 if (uh->check == 0)
843 uh->check = CSUM_MANGLED_0;
844
845send:
Eric Dumazetb5ec8ee2012-08-10 02:22:47 +0000846 err = ip_send_skb(sock_net(sk), skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000847 if (err) {
848 if (err == -ENOBUFS && !inet->recverr) {
849 UDP_INC_STATS_USER(sock_net(sk),
850 UDP_MIB_SNDBUFERRORS, is_udplite);
851 err = 0;
852 }
853 } else
854 UDP_INC_STATS_USER(sock_net(sk),
855 UDP_MIB_OUTDATAGRAMS, is_udplite);
856 return err;
857}
858
David S. Millerdb8dac22008-03-06 16:22:02 -0800859/*
860 * Push out all pending data as one UDP datagram. Socket is locked.
861 */
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200862int udp_push_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800863{
864 struct udp_sock *up = udp_sk(sk);
865 struct inet_sock *inet = inet_sk(sk);
David S. Millerb6f21b22011-03-12 02:09:18 -0500866 struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800867 struct sk_buff *skb;
David S. Millerdb8dac22008-03-06 16:22:02 -0800868 int err = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800869
David S. Miller77968b72011-05-08 17:12:19 -0700870 skb = ip_finish_skb(sk, fl4);
Herbert Xuf6b96642011-03-01 02:36:48 +0000871 if (!skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800872 goto out;
873
David S. Miller79ab0532011-05-09 13:31:04 -0700874 err = udp_send_skb(skb, fl4);
David S. Millerdb8dac22008-03-06 16:22:02 -0800875
David S. Millerdb8dac22008-03-06 16:22:02 -0800876out:
877 up->len = 0;
878 up->pending = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800879 return err;
880}
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200881EXPORT_SYMBOL(udp_push_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800882
883int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
884 size_t len)
885{
886 struct inet_sock *inet = inet_sk(sk);
887 struct udp_sock *up = udp_sk(sk);
David S. Millere4749952011-05-08 16:38:45 -0700888 struct flowi4 fl4_stack;
David S. Millerb6f21b22011-03-12 02:09:18 -0500889 struct flowi4 *fl4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800890 int ulen = len;
891 struct ipcm_cookie ipc;
892 struct rtable *rt = NULL;
893 int free = 0;
894 int connected = 0;
895 __be32 daddr, faddr, saddr;
896 __be16 dport;
897 u8 tos;
898 int err, is_udplite = IS_UDPLITE(sk);
899 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
900 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
Herbert Xu903ab862011-03-01 02:36:48 +0000901 struct sk_buff *skb;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000902 struct ip_options_data opt_copy;
David S. Millerdb8dac22008-03-06 16:22:02 -0800903
904 if (len > 0xFFFF)
905 return -EMSGSIZE;
906
907 /*
908 * Check the flags.
909 */
910
Eric Dumazetc482c562009-07-17 00:26:32 +0000911 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
David S. Millerdb8dac22008-03-06 16:22:02 -0800912 return -EOPNOTSUPP;
913
914 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000915 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200916 ipc.ttl = 0;
917 ipc.tos = -1;
David S. Millerdb8dac22008-03-06 16:22:02 -0800918
Herbert Xu903ab862011-03-01 02:36:48 +0000919 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
920
David S. Millerf5fca602011-05-08 17:24:10 -0700921 fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800922 if (up->pending) {
923 /*
924 * There are pending frames.
925 * The socket lock must be held while it's corked.
926 */
927 lock_sock(sk);
928 if (likely(up->pending)) {
929 if (unlikely(up->pending != AF_INET)) {
930 release_sock(sk);
931 return -EINVAL;
932 }
933 goto do_append_data;
934 }
935 release_sock(sk);
936 }
937 ulen += sizeof(struct udphdr);
938
939 /*
940 * Get and verify the address.
941 */
942 if (msg->msg_name) {
Steffen Hurrle342dfc32014-01-17 22:53:15 +0100943 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -0800944 if (msg->msg_namelen < sizeof(*usin))
945 return -EINVAL;
946 if (usin->sin_family != AF_INET) {
947 if (usin->sin_family != AF_UNSPEC)
948 return -EAFNOSUPPORT;
949 }
950
951 daddr = usin->sin_addr.s_addr;
952 dport = usin->sin_port;
953 if (dport == 0)
954 return -EINVAL;
955 } else {
956 if (sk->sk_state != TCP_ESTABLISHED)
957 return -EDESTADDRREQ;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000958 daddr = inet->inet_daddr;
959 dport = inet->inet_dport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800960 /* Open fast path for connected socket.
961 Route will not be used, if at least one option is set.
962 */
963 connected = 1;
964 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000965 ipc.addr = inet->inet_saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800966
967 ipc.oif = sk->sk_bound_dev_if;
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000968
969 sock_tx_timestamp(sk, &ipc.tx_flags);
970
David S. Millerdb8dac22008-03-06 16:22:02 -0800971 if (msg->msg_controllen) {
Hannes Frederic Sowac8e6ad02014-02-18 21:38:08 +0100972 err = ip_cmsg_send(sock_net(sk), msg, &ipc,
973 sk->sk_family == AF_INET6);
David S. Millerdb8dac22008-03-06 16:22:02 -0800974 if (err)
975 return err;
976 if (ipc.opt)
977 free = 1;
978 connected = 0;
979 }
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000980 if (!ipc.opt) {
981 struct ip_options_rcu *inet_opt;
982
983 rcu_read_lock();
984 inet_opt = rcu_dereference(inet->inet_opt);
985 if (inet_opt) {
986 memcpy(&opt_copy, inet_opt,
987 sizeof(*inet_opt) + inet_opt->opt.optlen);
988 ipc.opt = &opt_copy.opt;
989 }
990 rcu_read_unlock();
991 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800992
993 saddr = ipc.addr;
994 ipc.addr = faddr = daddr;
995
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000996 if (ipc.opt && ipc.opt->opt.srr) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800997 if (!daddr)
998 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000999 faddr = ipc.opt->opt.faddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001000 connected = 0;
1001 }
Francesco Fuscoaa661582013-09-24 15:43:09 +02001002 tos = get_rttos(&ipc, inet);
David S. Millerdb8dac22008-03-06 16:22:02 -08001003 if (sock_flag(sk, SOCK_LOCALROUTE) ||
1004 (msg->msg_flags & MSG_DONTROUTE) ||
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001005 (ipc.opt && ipc.opt->opt.is_strictroute)) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001006 tos |= RTO_ONLINK;
1007 connected = 0;
1008 }
1009
1010 if (ipv4_is_multicast(daddr)) {
1011 if (!ipc.oif)
1012 ipc.oif = inet->mc_index;
1013 if (!saddr)
1014 saddr = inet->mc_addr;
1015 connected = 0;
Erich E. Hoover76e21052012-02-08 09:11:07 +00001016 } else if (!ipc.oif)
1017 ipc.oif = inet->uc_index;
David S. Millerdb8dac22008-03-06 16:22:02 -08001018
1019 if (connected)
Eric Dumazetc482c562009-07-17 00:26:32 +00001020 rt = (struct rtable *)sk_dst_check(sk, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001021
1022 if (rt == NULL) {
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001023 struct net *net = sock_net(sk);
1024
David S. Millere4749952011-05-08 16:38:45 -07001025 fl4 = &fl4_stack;
1026 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
David S. Millerc0951cb2011-03-31 04:54:27 -07001027 RT_SCOPE_UNIVERSE, sk->sk_protocol,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001028 inet_sk_flowi_flags(sk),
David S. Millerc0951cb2011-03-31 04:54:27 -07001029 faddr, saddr, dport, inet->inet_sport);
1030
David S. Millere4749952011-05-08 16:38:45 -07001031 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
1032 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001033 if (IS_ERR(rt)) {
1034 err = PTR_ERR(rt);
David S. Miller06dc94b2011-03-03 10:38:01 -08001035 rt = NULL;
David S. Millerdb8dac22008-03-06 16:22:02 -08001036 if (err == -ENETUNREACH)
Eric Dumazetf1d8cba2013-11-28 09:51:22 -08001037 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -08001038 goto out;
1039 }
1040
1041 err = -EACCES;
1042 if ((rt->rt_flags & RTCF_BROADCAST) &&
1043 !sock_flag(sk, SOCK_BROADCAST))
1044 goto out;
1045 if (connected)
Changli Gaod8d1f302010-06-10 23:31:35 -07001046 sk_dst_set(sk, dst_clone(&rt->dst));
David S. Millerdb8dac22008-03-06 16:22:02 -08001047 }
1048
1049 if (msg->msg_flags&MSG_CONFIRM)
1050 goto do_confirm;
1051back_from_confirm:
1052
David S. Millere4749952011-05-08 16:38:45 -07001053 saddr = fl4->saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001054 if (!ipc.addr)
David S. Millere4749952011-05-08 16:38:45 -07001055 daddr = ipc.addr = fl4->daddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001056
Herbert Xu903ab862011-03-01 02:36:48 +00001057 /* Lockless fast path for the non-corking case. */
1058 if (!corkreq) {
David S. Miller77968b72011-05-08 17:12:19 -07001059 skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen,
Herbert Xu903ab862011-03-01 02:36:48 +00001060 sizeof(struct udphdr), &ipc, &rt,
1061 msg->msg_flags);
1062 err = PTR_ERR(skb);
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001063 if (!IS_ERR_OR_NULL(skb))
David S. Miller79ab0532011-05-09 13:31:04 -07001064 err = udp_send_skb(skb, fl4);
Herbert Xu903ab862011-03-01 02:36:48 +00001065 goto out;
1066 }
1067
David S. Millerdb8dac22008-03-06 16:22:02 -08001068 lock_sock(sk);
1069 if (unlikely(up->pending)) {
1070 /* The socket is already corked while preparing it. */
1071 /* ... which is an evident application bug. --ANK */
1072 release_sock(sk);
1073
Joe Perchesafd465032012-03-12 07:03:32 +00001074 LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("cork app bug 2\n"));
David S. Millerdb8dac22008-03-06 16:22:02 -08001075 err = -EINVAL;
1076 goto out;
1077 }
1078 /*
1079 * Now cork the socket to pend data.
1080 */
David S. Millerb6f21b22011-03-12 02:09:18 -05001081 fl4 = &inet->cork.fl.u.ip4;
1082 fl4->daddr = daddr;
1083 fl4->saddr = saddr;
David S. Miller9cce96d2011-03-12 03:00:33 -05001084 fl4->fl4_dport = dport;
1085 fl4->fl4_sport = inet->inet_sport;
David S. Millerdb8dac22008-03-06 16:22:02 -08001086 up->pending = AF_INET;
1087
1088do_append_data:
1089 up->len += ulen;
David S. Millerf5fca602011-05-08 17:24:10 -07001090 err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen,
1091 sizeof(struct udphdr), &ipc, &rt,
1092 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001093 if (err)
1094 udp_flush_pending_frames(sk);
1095 else if (!corkreq)
1096 err = udp_push_pending_frames(sk);
1097 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1098 up->pending = 0;
1099 release_sock(sk);
1100
1101out:
1102 ip_rt_put(rt);
1103 if (free)
1104 kfree(ipc.opt);
1105 if (!err)
1106 return len;
1107 /*
1108 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1109 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1110 * we don't have a good statistic (IpOutDiscards but it can be too many
1111 * things). We could add another new stat but at least for now that
1112 * seems like overkill.
1113 */
1114 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001115 UDP_INC_STATS_USER(sock_net(sk),
1116 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001117 }
1118 return err;
1119
1120do_confirm:
Changli Gaod8d1f302010-06-10 23:31:35 -07001121 dst_confirm(&rt->dst);
David S. Millerdb8dac22008-03-06 16:22:02 -08001122 if (!(msg->msg_flags&MSG_PROBE) || len)
1123 goto back_from_confirm;
1124 err = 0;
1125 goto out;
1126}
Eric Dumazetc482c562009-07-17 00:26:32 +00001127EXPORT_SYMBOL(udp_sendmsg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001128
1129int udp_sendpage(struct sock *sk, struct page *page, int offset,
1130 size_t size, int flags)
1131{
David S. Millerf5fca602011-05-08 17:24:10 -07001132 struct inet_sock *inet = inet_sk(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001133 struct udp_sock *up = udp_sk(sk);
1134 int ret;
1135
Shawn Landdend3f7d562013-11-24 22:36:28 -08001136 if (flags & MSG_SENDPAGE_NOTLAST)
1137 flags |= MSG_MORE;
1138
David S. Millerdb8dac22008-03-06 16:22:02 -08001139 if (!up->pending) {
1140 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
1141
1142 /* Call udp_sendmsg to specify destination address which
1143 * sendpage interface can't pass.
1144 * This will succeed only when the socket is connected.
1145 */
1146 ret = udp_sendmsg(NULL, sk, &msg, 0);
1147 if (ret < 0)
1148 return ret;
1149 }
1150
1151 lock_sock(sk);
1152
1153 if (unlikely(!up->pending)) {
1154 release_sock(sk);
1155
Joe Perchesafd465032012-03-12 07:03:32 +00001156 LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("udp cork app bug 3\n"));
David S. Millerdb8dac22008-03-06 16:22:02 -08001157 return -EINVAL;
1158 }
1159
David S. Millerf5fca602011-05-08 17:24:10 -07001160 ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1161 page, offset, size, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001162 if (ret == -EOPNOTSUPP) {
1163 release_sock(sk);
1164 return sock_no_sendpage(sk->sk_socket, page, offset,
1165 size, flags);
1166 }
1167 if (ret < 0) {
1168 udp_flush_pending_frames(sk);
1169 goto out;
1170 }
1171
1172 up->len += size;
1173 if (!(up->corkflag || (flags&MSG_MORE)))
1174 ret = udp_push_pending_frames(sk);
1175 if (!ret)
1176 ret = size;
1177out:
1178 release_sock(sk);
1179 return ret;
1180}
1181
Eric Dumazet85584672009-10-09 04:43:40 +00001182
1183/**
1184 * first_packet_length - return length of first packet in receive queue
1185 * @sk: socket
1186 *
1187 * Drops all bad checksum frames, until a valid one is found.
1188 * Returns the length of found skb, or 0 if none is found.
1189 */
1190static unsigned int first_packet_length(struct sock *sk)
1191{
1192 struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1193 struct sk_buff *skb;
1194 unsigned int res;
1195
1196 __skb_queue_head_init(&list_kill);
1197
1198 spin_lock_bh(&rcvq->lock);
1199 while ((skb = skb_peek(rcvq)) != NULL &&
1200 udp_lib_checksum_complete(skb)) {
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001201 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS,
1202 IS_UDPLITE(sk));
Eric Dumazet85584672009-10-09 04:43:40 +00001203 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1204 IS_UDPLITE(sk));
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001205 atomic_inc(&sk->sk_drops);
Eric Dumazet85584672009-10-09 04:43:40 +00001206 __skb_unlink(skb, rcvq);
1207 __skb_queue_tail(&list_kill, skb);
1208 }
1209 res = skb ? skb->len : 0;
1210 spin_unlock_bh(&rcvq->lock);
1211
1212 if (!skb_queue_empty(&list_kill)) {
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001213 bool slow = lock_sock_fast(sk);
1214
Eric Dumazet85584672009-10-09 04:43:40 +00001215 __skb_queue_purge(&list_kill);
1216 sk_mem_reclaim_partial(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001217 unlock_sock_fast(sk, slow);
Eric Dumazet85584672009-10-09 04:43:40 +00001218 }
1219 return res;
1220}
1221
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222/*
1223 * IOCTL requests applicable to the UDP protocol
1224 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001225
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1227{
Stephen Hemminger6516c652007-03-08 20:41:55 -08001228 switch (cmd) {
1229 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 {
Eric Dumazet31e6d362009-06-17 19:05:41 -07001231 int amount = sk_wmem_alloc_get(sk);
1232
Stephen Hemminger6516c652007-03-08 20:41:55 -08001233 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 }
Stephen Hemminger6516c652007-03-08 20:41:55 -08001235
1236 case SIOCINQ:
1237 {
Eric Dumazet85584672009-10-09 04:43:40 +00001238 unsigned int amount = first_packet_length(sk);
Stephen Hemminger6516c652007-03-08 20:41:55 -08001239
Eric Dumazet85584672009-10-09 04:43:40 +00001240 if (amount)
Stephen Hemminger6516c652007-03-08 20:41:55 -08001241 /*
1242 * We will only return the amount
1243 * of this packet since that is all
1244 * that will be read.
1245 */
Eric Dumazet85584672009-10-09 04:43:40 +00001246 amount -= sizeof(struct udphdr);
1247
Stephen Hemminger6516c652007-03-08 20:41:55 -08001248 return put_user(amount, (int __user *)arg);
1249 }
1250
1251 default:
1252 return -ENOIOCTLCMD;
1253 }
1254
1255 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256}
Eric Dumazetc482c562009-07-17 00:26:32 +00001257EXPORT_SYMBOL(udp_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258
David S. Millerdb8dac22008-03-06 16:22:02 -08001259/*
1260 * This should be easy, if there is something there we
1261 * return it, otherwise we block.
1262 */
1263
1264int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
1265 size_t len, int noblock, int flags, int *addr_len)
1266{
1267 struct inet_sock *inet = inet_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001268 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -08001269 struct sk_buff *skb;
David S. Miller59c2cda2011-12-01 14:12:55 -05001270 unsigned int ulen, copied;
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001271 int peeked, off = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001272 int err;
1273 int is_udplite = IS_UDPLITE(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001274 bool slow;
David S. Millerdb8dac22008-03-06 16:22:02 -08001275
David S. Millerdb8dac22008-03-06 16:22:02 -08001276 if (flags & MSG_ERRQUEUE)
Hannes Frederic Sowa85fbaa72013-11-23 00:46:12 +01001277 return ip_recv_error(sk, msg, len, addr_len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001278
1279try_again:
1280 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001281 &peeked, &off, &err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001282 if (!skb)
1283 goto out;
1284
1285 ulen = skb->len - sizeof(struct udphdr);
David S. Miller59c2cda2011-12-01 14:12:55 -05001286 copied = len;
1287 if (copied > ulen)
1288 copied = ulen;
1289 else if (copied < ulen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001290 msg->msg_flags |= MSG_TRUNC;
1291
1292 /*
1293 * If checksum is needed at all, try to do it while copying the
1294 * data. If the data is truncated, or if we only want a partial
1295 * coverage checksum (UDP-Lite), do it before the copy.
1296 */
1297
David S. Miller59c2cda2011-12-01 14:12:55 -05001298 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001299 if (udp_lib_checksum_complete(skb))
1300 goto csum_copy_err;
1301 }
1302
1303 if (skb_csum_unnecessary(skb))
1304 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
David S. Miller59c2cda2011-12-01 14:12:55 -05001305 msg->msg_iov, copied);
David S. Millerdb8dac22008-03-06 16:22:02 -08001306 else {
Eric Dumazetc482c562009-07-17 00:26:32 +00001307 err = skb_copy_and_csum_datagram_iovec(skb,
1308 sizeof(struct udphdr),
1309 msg->msg_iov);
David S. Millerdb8dac22008-03-06 16:22:02 -08001310
1311 if (err == -EINVAL)
1312 goto csum_copy_err;
1313 }
1314
Eric Dumazet22911fc2012-06-27 00:23:44 +00001315 if (unlikely(err)) {
1316 trace_kfree_skb(skb, udp_recvmsg);
Eric Dumazet979402b2012-09-05 23:34:44 +00001317 if (!peeked) {
1318 atomic_inc(&sk->sk_drops);
1319 UDP_INC_STATS_USER(sock_net(sk),
1320 UDP_MIB_INERRORS, is_udplite);
1321 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001322 goto out_free;
Eric Dumazet22911fc2012-06-27 00:23:44 +00001323 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001324
1325 if (!peeked)
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001326 UDP_INC_STATS_USER(sock_net(sk),
1327 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001328
Neil Horman3b885782009-10-12 13:26:31 -07001329 sock_recv_ts_and_drops(msg, sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001330
1331 /* Copy the address. */
Eric Dumazetc482c562009-07-17 00:26:32 +00001332 if (sin) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001333 sin->sin_family = AF_INET;
1334 sin->sin_port = udp_hdr(skb)->source;
1335 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1336 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
Hannes Frederic Sowabceaa902013-11-18 04:20:45 +01001337 *addr_len = sizeof(*sin);
David S. Millerdb8dac22008-03-06 16:22:02 -08001338 }
1339 if (inet->cmsg_flags)
1340 ip_cmsg_recv(msg, skb);
1341
David S. Miller59c2cda2011-12-01 14:12:55 -05001342 err = copied;
David S. Millerdb8dac22008-03-06 16:22:02 -08001343 if (flags & MSG_TRUNC)
1344 err = ulen;
1345
1346out_free:
Eric Dumazet9d410c72009-10-30 05:03:53 +00001347 skb_free_datagram_locked(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001348out:
1349 return err;
1350
1351csum_copy_err:
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001352 slow = lock_sock_fast(sk);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001353 if (!skb_kill_datagram(sk, skb, flags)) {
1354 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001355 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001356 }
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001357 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001358
1359 if (noblock)
1360 return -EAGAIN;
Xufeng Zhang9cfaa8d2011-06-21 10:43:40 +00001361
1362 /* starting over for a new packet */
1363 msg->msg_flags &= ~MSG_TRUNC;
David S. Millerdb8dac22008-03-06 16:22:02 -08001364 goto try_again;
1365}
1366
1367
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368int udp_disconnect(struct sock *sk, int flags)
1369{
1370 struct inet_sock *inet = inet_sk(sk);
1371 /*
1372 * 1003.1g - break association.
1373 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001374
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 sk->sk_state = TCP_CLOSE;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001376 inet->inet_daddr = 0;
1377 inet->inet_dport = 0;
Tom Herbertbdeab992011-08-14 19:45:55 +00001378 sock_rps_reset_rxhash(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 sk->sk_bound_dev_if = 0;
1380 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1381 inet_reset_saddr(sk);
1382
1383 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1384 sk->sk_prot->unhash(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001385 inet->inet_sport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 }
1387 sk_dst_reset(sk);
1388 return 0;
1389}
Eric Dumazetc482c562009-07-17 00:26:32 +00001390EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391
Eric Dumazet645ca702008-10-29 01:41:45 -07001392void udp_lib_unhash(struct sock *sk)
1393{
Eric Dumazet723b4612008-11-25 13:55:15 -08001394 if (sk_hashed(sk)) {
1395 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Eric Dumazet512615b2009-11-08 10:17:58 +00001396 struct udp_hslot *hslot, *hslot2;
1397
1398 hslot = udp_hashslot(udptable, sock_net(sk),
1399 udp_sk(sk)->udp_port_hash);
1400 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
Eric Dumazet645ca702008-10-29 01:41:45 -07001401
Eric Dumazet723b4612008-11-25 13:55:15 -08001402 spin_lock_bh(&hslot->lock);
1403 if (sk_nulls_del_node_init_rcu(sk)) {
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00001404 hslot->count--;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001405 inet_sk(sk)->inet_num = 0;
Eric Dumazet723b4612008-11-25 13:55:15 -08001406 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet512615b2009-11-08 10:17:58 +00001407
1408 spin_lock(&hslot2->lock);
1409 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1410 hslot2->count--;
1411 spin_unlock(&hslot2->lock);
Eric Dumazet723b4612008-11-25 13:55:15 -08001412 }
1413 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001414 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001415}
1416EXPORT_SYMBOL(udp_lib_unhash);
1417
Eric Dumazet719f8352010-09-08 05:08:44 +00001418/*
1419 * inet_rcv_saddr was changed, we must rehash secondary hash
1420 */
1421void udp_lib_rehash(struct sock *sk, u16 newhash)
1422{
1423 if (sk_hashed(sk)) {
1424 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1425 struct udp_hslot *hslot, *hslot2, *nhslot2;
1426
1427 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1428 nhslot2 = udp_hashslot2(udptable, newhash);
1429 udp_sk(sk)->udp_portaddr_hash = newhash;
1430 if (hslot2 != nhslot2) {
1431 hslot = udp_hashslot(udptable, sock_net(sk),
1432 udp_sk(sk)->udp_port_hash);
1433 /* we must lock primary chain too */
1434 spin_lock_bh(&hslot->lock);
1435
1436 spin_lock(&hslot2->lock);
1437 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1438 hslot2->count--;
1439 spin_unlock(&hslot2->lock);
1440
1441 spin_lock(&nhslot2->lock);
1442 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
1443 &nhslot2->head);
1444 nhslot2->count++;
1445 spin_unlock(&nhslot2->lock);
1446
1447 spin_unlock_bh(&hslot->lock);
1448 }
1449 }
1450}
1451EXPORT_SYMBOL(udp_lib_rehash);
1452
1453static void udp_v4_rehash(struct sock *sk)
1454{
1455 u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1456 inet_sk(sk)->inet_rcv_saddr,
1457 inet_sk(sk)->inet_num);
1458 udp_lib_rehash(sk, new_hash);
1459}
1460
Herbert Xu93821772008-09-15 11:48:46 -07001461static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1462{
Tom Herbertfec5e652010-04-16 16:01:27 -07001463 int rc;
Herbert Xu93821772008-09-15 11:48:46 -07001464
Shawn Bohrer005ec972013-10-07 11:01:38 -05001465 if (inet_sk(sk)->inet_daddr) {
Tom Herbertbdeab992011-08-14 19:45:55 +00001466 sock_rps_save_rxhash(sk, skb);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001467 sk_mark_napi_id(sk, skb);
1468 }
Tom Herbertfec5e652010-04-16 16:01:27 -07001469
Eric Dumazetd826eb12011-11-09 07:24:35 +00001470 rc = sock_queue_rcv_skb(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -07001471 if (rc < 0) {
1472 int is_udplite = IS_UDPLITE(sk);
1473
Herbert Xu93821772008-09-15 11:48:46 -07001474 /* Note that an ENOMEM error is charged twice */
Eric Dumazet766e90372009-10-14 20:40:11 -07001475 if (rc == -ENOMEM)
Herbert Xu93821772008-09-15 11:48:46 -07001476 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1477 is_udplite);
Eric Dumazet766e90372009-10-14 20:40:11 -07001478 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1479 kfree_skb(skb);
Satoru Moriya296f7ea2011-06-17 11:58:39 +00001480 trace_udp_fail_queue_rcv_skb(rc, sk);
Eric Dumazet766e90372009-10-14 20:40:11 -07001481 return -1;
Herbert Xu93821772008-09-15 11:48:46 -07001482 }
1483
1484 return 0;
1485
Herbert Xu93821772008-09-15 11:48:46 -07001486}
1487
Eric Dumazet447167b2012-04-11 23:05:28 +00001488static struct static_key udp_encap_needed __read_mostly;
1489void udp_encap_enable(void)
1490{
1491 if (!static_key_enabled(&udp_encap_needed))
1492 static_key_slow_inc(&udp_encap_needed);
1493}
1494EXPORT_SYMBOL(udp_encap_enable);
1495
David S. Millerdb8dac22008-03-06 16:22:02 -08001496/* returns:
1497 * -1: error
1498 * 0: success
1499 * >0: "udp encap" protocol resubmission
1500 *
1501 * Note that in the success and error cases, the skb is assumed to
1502 * have either been requeued or freed.
1503 */
Eric Dumazetc482c562009-07-17 00:26:32 +00001504int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -08001505{
1506 struct udp_sock *up = udp_sk(sk);
1507 int rc;
1508 int is_udplite = IS_UDPLITE(sk);
1509
1510 /*
1511 * Charge it to the socket, dropping if the queue is full.
1512 */
1513 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1514 goto drop;
1515 nf_reset(skb);
1516
Eric Dumazet447167b2012-04-11 23:05:28 +00001517 if (static_key_false(&udp_encap_needed) && up->encap_type) {
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001518 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1519
David S. Millerdb8dac22008-03-06 16:22:02 -08001520 /*
1521 * This is an encapsulation socket so pass the skb to
1522 * the socket's udp_encap_rcv() hook. Otherwise, just
1523 * fall through and pass this up the UDP socket.
1524 * up->encap_rcv() returns the following value:
1525 * =0 if skb was successfully passed to the encap
1526 * handler or was discarded by it.
1527 * >0 if skb should be passed on to UDP.
1528 * <0 if skb should be resubmitted as proto -N
1529 */
1530
1531 /* if we're overly short, let UDP handle it */
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001532 encap_rcv = ACCESS_ONCE(up->encap_rcv);
1533 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001534 int ret;
1535
Tom Herbert0a809662014-05-07 16:52:39 -07001536 /* Verify checksum before giving to encap */
1537 if (udp_lib_checksum_complete(skb))
1538 goto csum_error;
1539
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001540 ret = encap_rcv(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001541 if (ret <= 0) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001542 UDP_INC_STATS_BH(sock_net(sk),
1543 UDP_MIB_INDATAGRAMS,
David S. Millerdb8dac22008-03-06 16:22:02 -08001544 is_udplite);
1545 return -ret;
1546 }
1547 }
1548
1549 /* FALLTHROUGH -- it's a UDP Packet */
1550 }
1551
1552 /*
1553 * UDP-Lite specific tests, ignored on UDP sockets
1554 */
1555 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1556
1557 /*
1558 * MIB statistics other than incrementing the error count are
1559 * disabled for the following two types of errors: these depend
1560 * on the application settings, not on the functioning of the
1561 * protocol stack as such.
1562 *
1563 * RFC 3828 here recommends (sec 3.3): "There should also be a
1564 * way ... to ... at least let the receiving application block
1565 * delivery of packets with coverage values less than a value
1566 * provided by the application."
1567 */
1568 if (up->pcrlen == 0) { /* full coverage was set */
Joe Perchesafd465032012-03-12 07:03:32 +00001569 LIMIT_NETDEBUG(KERN_WARNING "UDPLite: partial coverage %d while full coverage %d requested\n",
1570 UDP_SKB_CB(skb)->cscov, skb->len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001571 goto drop;
1572 }
1573 /* The next case involves violating the min. coverage requested
1574 * by the receiver. This is subtle: if receiver wants x and x is
1575 * greater than the buffersize/MTU then receiver will complain
1576 * that it wants x while sender emits packets of smaller size y.
1577 * Therefore the above ...()->partial_cov statement is essential.
1578 */
1579 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
Joe Perchesafd465032012-03-12 07:03:32 +00001580 LIMIT_NETDEBUG(KERN_WARNING "UDPLite: coverage %d too small, need min %d\n",
1581 UDP_SKB_CB(skb)->cscov, up->pcrlen);
David S. Millerdb8dac22008-03-06 16:22:02 -08001582 goto drop;
1583 }
1584 }
1585
Eric Dumazet33d480c2011-08-11 19:30:52 +00001586 if (rcu_access_pointer(sk->sk_filter) &&
1587 udp_lib_checksum_complete(skb))
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001588 goto csum_error;
David S. Millerdb8dac22008-03-06 16:22:02 -08001589
Eric Dumazetc3774112010-04-27 15:13:20 -07001590
Eric Dumazetf545a382012-04-22 23:34:26 +00001591 if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf))
Eric Dumazetc3774112010-04-27 15:13:20 -07001592 goto drop;
1593
Herbert Xu93821772008-09-15 11:48:46 -07001594 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001595
Shawn Bohrerfbf88662013-10-07 11:01:40 -05001596 ipv4_pktinfo_prepare(sk, skb);
Herbert Xu93821772008-09-15 11:48:46 -07001597 bh_lock_sock(sk);
1598 if (!sock_owned_by_user(sk))
1599 rc = __udp_queue_rcv_skb(sk, skb);
Eric Dumazetf545a382012-04-22 23:34:26 +00001600 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
Zhu Yi55349792010-03-04 18:01:42 +00001601 bh_unlock_sock(sk);
1602 goto drop;
1603 }
Herbert Xu93821772008-09-15 11:48:46 -07001604 bh_unlock_sock(sk);
1605
1606 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001607
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001608csum_error:
1609 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001610drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001611 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001612 atomic_inc(&sk->sk_drops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001613 kfree_skb(skb);
1614 return -1;
1615}
1616
Eric Dumazet1240d132009-11-08 10:18:44 +00001617
1618static void flush_stack(struct sock **stack, unsigned int count,
1619 struct sk_buff *skb, unsigned int final)
1620{
1621 unsigned int i;
1622 struct sk_buff *skb1 = NULL;
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001623 struct sock *sk;
Eric Dumazet1240d132009-11-08 10:18:44 +00001624
1625 for (i = 0; i < count; i++) {
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001626 sk = stack[i];
Eric Dumazet1240d132009-11-08 10:18:44 +00001627 if (likely(skb1 == NULL))
1628 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
1629
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001630 if (!skb1) {
1631 atomic_inc(&sk->sk_drops);
1632 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1633 IS_UDPLITE(sk));
1634 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1635 IS_UDPLITE(sk));
1636 }
1637
1638 if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0)
Eric Dumazet1240d132009-11-08 10:18:44 +00001639 skb1 = NULL;
1640 }
1641 if (unlikely(skb1))
1642 kfree_skb(skb1);
1643}
1644
Eric Dumazet97502232013-12-11 14:46:51 -08001645/* For TCP sockets, sk_rx_dst is protected by socket lock
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001646 * For UDP, we use xchg() to guard against concurrent changes.
Eric Dumazet97502232013-12-11 14:46:51 -08001647 */
1648static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
Shawn Bohrer421b3882013-10-07 11:01:39 -05001649{
Eric Dumazet97502232013-12-11 14:46:51 -08001650 struct dst_entry *old;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001651
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001652 dst_hold(dst);
1653 old = xchg(&sk->sk_rx_dst, dst);
1654 dst_release(old);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001655}
1656
David S. Millerdb8dac22008-03-06 16:22:02 -08001657/*
1658 * Multicasts and broadcasts go to each listener.
1659 *
Eric Dumazet1240d132009-11-08 10:18:44 +00001660 * Note: called only from the BH handler context.
David S. Millerdb8dac22008-03-06 16:22:02 -08001661 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001662static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001663 struct udphdr *uh,
1664 __be32 saddr, __be32 daddr,
Eric Dumazet645ca702008-10-29 01:41:45 -07001665 struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -08001666{
Eric Dumazet1240d132009-11-08 10:18:44 +00001667 struct sock *sk, *stack[256 / sizeof(struct sock *)];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00001668 struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001669 int dif;
Eric Dumazet1240d132009-11-08 10:18:44 +00001670 unsigned int i, count = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001671
Eric Dumazet645ca702008-10-29 01:41:45 -07001672 spin_lock(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08001673 sk = sk_nulls_head(&hslot->head);
David S. Millerdb8dac22008-03-06 16:22:02 -08001674 dif = skb->dev->ifindex;
Eric Dumazet920a4612008-11-01 21:22:23 -07001675 sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
Eric Dumazet1240d132009-11-08 10:18:44 +00001676 while (sk) {
1677 stack[count++] = sk;
1678 sk = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest,
1679 daddr, uh->source, saddr, dif);
1680 if (unlikely(count == ARRAY_SIZE(stack))) {
1681 if (!sk)
1682 break;
1683 flush_stack(stack, count, skb, ~0);
1684 count = 0;
1685 }
1686 }
1687 /*
1688 * before releasing chain lock, we must take a reference on sockets
1689 */
1690 for (i = 0; i < count; i++)
1691 sock_hold(stack[i]);
David S. Millerdb8dac22008-03-06 16:22:02 -08001692
Eric Dumazet645ca702008-10-29 01:41:45 -07001693 spin_unlock(&hslot->lock);
Eric Dumazet1240d132009-11-08 10:18:44 +00001694
1695 /*
1696 * do the slow work with no lock held
1697 */
1698 if (count) {
1699 flush_stack(stack, count, skb, count - 1);
1700
1701 for (i = 0; i < count; i++)
1702 sock_put(stack[i]);
1703 } else {
1704 kfree_skb(skb);
1705 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001706 return 0;
1707}
1708
1709/* Initialize UDP checksum. If exited with zero value (success),
1710 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1711 * Otherwise, csum completion requires chacksumming packet body,
1712 * including udp header and folding it to skb->csum.
1713 */
1714static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1715 int proto)
1716{
David S. Millerdb8dac22008-03-06 16:22:02 -08001717 int err;
1718
1719 UDP_SKB_CB(skb)->partial_cov = 0;
1720 UDP_SKB_CB(skb)->cscov = skb->len;
1721
1722 if (proto == IPPROTO_UDPLITE) {
1723 err = udplite_checksum_init(skb, uh);
1724 if (err)
1725 return err;
1726 }
1727
Tom Herberted70fcf2014-05-02 16:29:38 -07001728 return skb_checksum_init_zero_check(skb, proto, uh->check,
1729 inet_compute_pseudo);
David S. Millerdb8dac22008-03-06 16:22:02 -08001730}
1731
1732/*
1733 * All we need to do is get the socket, and then do a checksum.
1734 */
1735
Eric Dumazet645ca702008-10-29 01:41:45 -07001736int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001737 int proto)
1738{
1739 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001740 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001741 unsigned short ulen;
Eric Dumazetadf30902009-06-02 05:19:30 +00001742 struct rtable *rt = skb_rtable(skb);
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001743 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001744 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001745
1746 /*
1747 * Validate the packet.
1748 */
1749 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1750 goto drop; /* No space for header. */
1751
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001752 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001753 ulen = ntohs(uh->len);
Bjørn Morkccc2d972010-05-06 03:44:34 +00001754 saddr = ip_hdr(skb)->saddr;
1755 daddr = ip_hdr(skb)->daddr;
1756
David S. Millerdb8dac22008-03-06 16:22:02 -08001757 if (ulen > skb->len)
1758 goto short_packet;
1759
1760 if (proto == IPPROTO_UDP) {
1761 /* UDP validates ulen. */
1762 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1763 goto short_packet;
1764 uh = udp_hdr(skb);
1765 }
1766
1767 if (udp4_csum_init(skb, uh, proto))
1768 goto csum_error;
1769
Eric Dumazet8afdd992013-12-10 18:07:23 -08001770 sk = skb_steal_sock(skb);
1771 if (sk) {
Eric Dumazet97502232013-12-11 14:46:51 -08001772 struct dst_entry *dst = skb_dst(skb);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001773 int ret;
David S. Millerdb8dac22008-03-06 16:22:02 -08001774
Eric Dumazet97502232013-12-11 14:46:51 -08001775 if (unlikely(sk->sk_rx_dst != dst))
1776 udp_sk_rx_dst_set(sk, dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001777
1778 ret = udp_queue_rcv_skb(sk, skb);
Eric Dumazet8afdd992013-12-10 18:07:23 -08001779 sock_put(sk);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001780 /* a return value > 0 means to resubmit the input, but
1781 * it wants the return to be -protocol, or 0
1782 */
1783 if (ret > 0)
1784 return -ret;
1785 return 0;
1786 } else {
1787 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1788 return __udp4_lib_mcast_deliver(net, skb, uh,
1789 saddr, daddr, udptable);
1790
1791 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
1792 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001793
1794 if (sk != NULL) {
Eliezer Tamira5b50472013-06-10 11:40:00 +03001795 int ret;
1796
Eliezer Tamira5b50472013-06-10 11:40:00 +03001797 ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001798 sock_put(sk);
1799
1800 /* a return value > 0 means to resubmit the input, but
1801 * it wants the return to be -protocol, or 0
1802 */
1803 if (ret > 0)
1804 return -ret;
1805 return 0;
1806 }
1807
1808 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1809 goto drop;
1810 nf_reset(skb);
1811
1812 /* No socket. Drop packet silently, if checksum is wrong */
1813 if (udp_lib_checksum_complete(skb))
1814 goto csum_error;
1815
Pavel Emelyanov02833282008-07-05 21:18:48 -07001816 UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001817 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1818
1819 /*
1820 * Hmm. We got an UDP packet to a port to which we
1821 * don't wanna listen. Ignore it.
1822 */
1823 kfree_skb(skb);
1824 return 0;
1825
1826short_packet:
Harvey Harrison673d57e2008-10-31 00:53:57 -07001827 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
Joe Perchesafd465032012-03-12 07:03:32 +00001828 proto == IPPROTO_UDPLITE ? "Lite" : "",
1829 &saddr, ntohs(uh->source),
1830 ulen, skb->len,
1831 &daddr, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001832 goto drop;
1833
1834csum_error:
1835 /*
1836 * RFC1122: OK. Discards the bad packet silently (as far as
1837 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1838 */
Harvey Harrison673d57e2008-10-31 00:53:57 -07001839 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
Joe Perchesafd465032012-03-12 07:03:32 +00001840 proto == IPPROTO_UDPLITE ? "Lite" : "",
1841 &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
David S. Millerdb8dac22008-03-06 16:22:02 -08001842 ulen);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001843 UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001844drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001845 UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001846 kfree_skb(skb);
1847 return 0;
1848}
1849
Shawn Bohrer421b3882013-10-07 11:01:39 -05001850/* We can only early demux multicast if there is a single matching socket.
1851 * If more than one socket found returns NULL
1852 */
1853static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net,
1854 __be16 loc_port, __be32 loc_addr,
1855 __be16 rmt_port, __be32 rmt_addr,
1856 int dif)
1857{
1858 struct sock *sk, *result;
1859 struct hlist_nulls_node *node;
1860 unsigned short hnum = ntohs(loc_port);
1861 unsigned int count, slot = udp_hashfn(net, hnum, udp_table.mask);
1862 struct udp_hslot *hslot = &udp_table.hash[slot];
1863
Eric Dumazet63c6f812014-06-12 16:13:06 -07001864 /* Do not bother scanning a too big list */
1865 if (hslot->count > 10)
1866 return NULL;
1867
Shawn Bohrer421b3882013-10-07 11:01:39 -05001868 rcu_read_lock();
1869begin:
1870 count = 0;
1871 result = NULL;
1872 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
1873 if (__udp_is_mcast_sock(net, sk,
1874 loc_port, loc_addr,
1875 rmt_port, rmt_addr,
1876 dif, hnum)) {
1877 result = sk;
1878 ++count;
1879 }
1880 }
1881 /*
1882 * if the nulls value we got at the end of this lookup is
1883 * not the expected one, we must restart lookup.
1884 * We probably met an item that was moved to another chain.
1885 */
1886 if (get_nulls_value(node) != slot)
1887 goto begin;
1888
1889 if (result) {
1890 if (count != 1 ||
1891 unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
1892 result = NULL;
Eric Dumazetf69b9232013-10-08 21:47:29 -07001893 else if (unlikely(!__udp_is_mcast_sock(net, result,
Shawn Bohrer421b3882013-10-07 11:01:39 -05001894 loc_port, loc_addr,
1895 rmt_port, rmt_addr,
1896 dif, hnum))) {
1897 sock_put(result);
1898 result = NULL;
1899 }
1900 }
1901 rcu_read_unlock();
1902 return result;
1903}
1904
1905/* For unicast we should only early demux connected sockets or we can
1906 * break forwarding setups. The chains here can be long so only check
1907 * if the first socket is an exact match and if not move on.
1908 */
1909static struct sock *__udp4_lib_demux_lookup(struct net *net,
1910 __be16 loc_port, __be32 loc_addr,
1911 __be16 rmt_port, __be32 rmt_addr,
1912 int dif)
1913{
1914 struct sock *sk, *result;
1915 struct hlist_nulls_node *node;
1916 unsigned short hnum = ntohs(loc_port);
1917 unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum);
1918 unsigned int slot2 = hash2 & udp_table.mask;
1919 struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
Joe Perchesc7228312014-05-13 20:30:07 -07001920 INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001921 const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
1922
1923 rcu_read_lock();
1924 result = NULL;
1925 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
1926 if (INET_MATCH(sk, net, acookie,
1927 rmt_addr, loc_addr, ports, dif))
1928 result = sk;
1929 /* Only check first socket in chain */
1930 break;
1931 }
1932
1933 if (result) {
1934 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
1935 result = NULL;
1936 else if (unlikely(!INET_MATCH(sk, net, acookie,
1937 rmt_addr, loc_addr,
1938 ports, dif))) {
1939 sock_put(result);
1940 result = NULL;
1941 }
1942 }
1943 rcu_read_unlock();
1944 return result;
1945}
1946
1947void udp_v4_early_demux(struct sk_buff *skb)
1948{
Eric Dumazet610438b2013-12-11 08:10:05 -08001949 struct net *net = dev_net(skb->dev);
1950 const struct iphdr *iph;
1951 const struct udphdr *uh;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001952 struct sock *sk;
1953 struct dst_entry *dst;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001954 int dif = skb->dev->ifindex;
1955
1956 /* validate the packet */
1957 if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr)))
1958 return;
1959
Eric Dumazet610438b2013-12-11 08:10:05 -08001960 iph = ip_hdr(skb);
1961 uh = udp_hdr(skb);
1962
Shawn Bohrer421b3882013-10-07 11:01:39 -05001963 if (skb->pkt_type == PACKET_BROADCAST ||
1964 skb->pkt_type == PACKET_MULTICAST)
1965 sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr,
1966 uh->source, iph->saddr, dif);
1967 else if (skb->pkt_type == PACKET_HOST)
1968 sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr,
1969 uh->source, iph->saddr, dif);
1970 else
1971 return;
1972
1973 if (!sk)
1974 return;
1975
1976 skb->sk = sk;
1977 skb->destructor = sock_edemux;
1978 dst = sk->sk_rx_dst;
1979
1980 if (dst)
1981 dst = dst_check(dst, 0);
1982 if (dst)
1983 skb_dst_set_noref(skb, dst);
1984}
1985
David S. Millerdb8dac22008-03-06 16:22:02 -08001986int udp_rcv(struct sk_buff *skb)
1987{
Eric Dumazet645ca702008-10-29 01:41:45 -07001988 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08001989}
1990
Brian Haley7d06b2e2008-06-14 17:04:49 -07001991void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08001992{
Tom Parkin44046a52013-03-19 06:11:12 +00001993 struct udp_sock *up = udp_sk(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001994 bool slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001995 udp_flush_pending_frames(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001996 unlock_sock_fast(sk, slow);
Tom Parkin44046a52013-03-19 06:11:12 +00001997 if (static_key_false(&udp_encap_needed) && up->encap_type) {
1998 void (*encap_destroy)(struct sock *sk);
1999 encap_destroy = ACCESS_ONCE(up->encap_destroy);
2000 if (encap_destroy)
2001 encap_destroy(sk);
2002 }
David S. Millerdb8dac22008-03-06 16:22:02 -08002003}
2004
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005/*
2006 * Socket option code for UDP
2007 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002008int udp_lib_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002009 char __user *optval, unsigned int optlen,
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002010 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011{
2012 struct udp_sock *up = udp_sk(sk);
Tom Herbert1c194482014-05-23 08:47:32 -07002013 int val, valbool;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11002015 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016
Eric Dumazetc482c562009-07-17 00:26:32 +00002017 if (optlen < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 return -EINVAL;
2019
2020 if (get_user(val, (int __user *)optval))
2021 return -EFAULT;
2022
Tom Herbert1c194482014-05-23 08:47:32 -07002023 valbool = val ? 1 : 0;
2024
Stephen Hemminger6516c652007-03-08 20:41:55 -08002025 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026 case UDP_CORK:
2027 if (val != 0) {
2028 up->corkflag = 1;
2029 } else {
2030 up->corkflag = 0;
2031 lock_sock(sk);
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002032 (*push_pending_frames)(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033 release_sock(sk);
2034 }
2035 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002036
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 case UDP_ENCAP:
2038 switch (val) {
2039 case 0:
2040 case UDP_ENCAP_ESPINUDP:
2041 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07002042 up->encap_rcv = xfrm4_udp_encap_rcv;
2043 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07002044 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 up->encap_type = val;
Eric Dumazet447167b2012-04-11 23:05:28 +00002046 udp_encap_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047 break;
2048 default:
2049 err = -ENOPROTOOPT;
2050 break;
2051 }
2052 break;
2053
Tom Herbert1c194482014-05-23 08:47:32 -07002054 case UDP_NO_CHECK6_TX:
2055 up->no_check6_tx = valbool;
2056 break;
2057
2058 case UDP_NO_CHECK6_RX:
2059 up->no_check6_rx = valbool;
2060 break;
2061
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002062 /*
2063 * UDP-Lite's partial checksum coverage (RFC 3828).
2064 */
2065 /* The sender sets actual checksum coverage length via this option.
2066 * The case coverage > packet length is handled by send module. */
2067 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002068 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002069 return -ENOPROTOOPT;
2070 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
2071 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002072 else if (val > USHRT_MAX)
2073 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002074 up->pcslen = val;
2075 up->pcflag |= UDPLITE_SEND_CC;
2076 break;
2077
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002078 /* The receiver specifies a minimum checksum coverage value. To make
2079 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002080 * used, this again means full checksum coverage. */
2081 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002082 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002083 return -ENOPROTOOPT;
2084 if (val != 0 && val < 8) /* Avoid silly minimal values. */
2085 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002086 else if (val > USHRT_MAX)
2087 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002088 up->pcrlen = val;
2089 up->pcflag |= UDPLITE_RECV_CC;
2090 break;
2091
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 default:
2093 err = -ENOPROTOOPT;
2094 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002095 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096
2097 return err;
2098}
Eric Dumazetc482c562009-07-17 00:26:32 +00002099EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100
David S. Millerdb8dac22008-03-06 16:22:02 -08002101int udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002102 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002103{
2104 if (level == SOL_UDP || level == SOL_UDPLITE)
2105 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2106 udp_push_pending_frames);
2107 return ip_setsockopt(sk, level, optname, optval, optlen);
2108}
2109
2110#ifdef CONFIG_COMPAT
2111int compat_udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002112 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002113{
2114 if (level == SOL_UDP || level == SOL_UDPLITE)
2115 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2116 udp_push_pending_frames);
2117 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
2118}
2119#endif
2120
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002121int udp_lib_getsockopt(struct sock *sk, int level, int optname,
2122 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123{
2124 struct udp_sock *up = udp_sk(sk);
2125 int val, len;
2126
Eric Dumazetc482c562009-07-17 00:26:32 +00002127 if (get_user(len, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128 return -EFAULT;
2129
2130 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002131
Stephen Hemminger6516c652007-03-08 20:41:55 -08002132 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133 return -EINVAL;
2134
Stephen Hemminger6516c652007-03-08 20:41:55 -08002135 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136 case UDP_CORK:
2137 val = up->corkflag;
2138 break;
2139
2140 case UDP_ENCAP:
2141 val = up->encap_type;
2142 break;
2143
Tom Herbert1c194482014-05-23 08:47:32 -07002144 case UDP_NO_CHECK6_TX:
2145 val = up->no_check6_tx;
2146 break;
2147
2148 case UDP_NO_CHECK6_RX:
2149 val = up->no_check6_rx;
2150 break;
2151
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002152 /* The following two cannot be changed on UDP sockets, the return is
2153 * always 0 (which corresponds to the full checksum coverage of UDP). */
2154 case UDPLITE_SEND_CSCOV:
2155 val = up->pcslen;
2156 break;
2157
2158 case UDPLITE_RECV_CSCOV:
2159 val = up->pcrlen;
2160 break;
2161
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162 default:
2163 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002164 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165
Stephen Hemminger6516c652007-03-08 20:41:55 -08002166 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002167 return -EFAULT;
Eric Dumazetc482c562009-07-17 00:26:32 +00002168 if (copy_to_user(optval, &val, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002169 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002170 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171}
Eric Dumazetc482c562009-07-17 00:26:32 +00002172EXPORT_SYMBOL(udp_lib_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173
David S. Millerdb8dac22008-03-06 16:22:02 -08002174int udp_getsockopt(struct sock *sk, int level, int optname,
2175 char __user *optval, int __user *optlen)
2176{
2177 if (level == SOL_UDP || level == SOL_UDPLITE)
2178 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2179 return ip_getsockopt(sk, level, optname, optval, optlen);
2180}
2181
2182#ifdef CONFIG_COMPAT
2183int compat_udp_getsockopt(struct sock *sk, int level, int optname,
2184 char __user *optval, int __user *optlen)
2185{
2186 if (level == SOL_UDP || level == SOL_UDPLITE)
2187 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2188 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
2189}
2190#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191/**
2192 * udp_poll - wait for a UDP event.
2193 * @file - file struct
2194 * @sock - socket
2195 * @wait - poll table
2196 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002197 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198 * blocking sockets. If application is using a blocking fd
2199 * and a packet with checksum error is in the queue;
2200 * then it could get return from select indicating data available
2201 * but then block when reading it. Add special case code
2202 * to work around these arguably broken applications.
2203 */
2204unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
2205{
2206 unsigned int mask = datagram_poll(file, sock, wait);
2207 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002208
David Majnemerc3f1dba2013-05-31 13:15:38 +00002209 sock_rps_record_flow(sk);
2210
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 /* Check for false positives due to checksum errors */
Eric Dumazet85584672009-10-09 04:43:40 +00002212 if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
2213 !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
2214 mask &= ~(POLLIN | POLLRDNORM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215
2216 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002217
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218}
Eric Dumazetc482c562009-07-17 00:26:32 +00002219EXPORT_SYMBOL(udp_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220
David S. Millerdb8dac22008-03-06 16:22:02 -08002221struct proto udp_prot = {
2222 .name = "UDP",
2223 .owner = THIS_MODULE,
2224 .close = udp_lib_close,
2225 .connect = ip4_datagram_connect,
2226 .disconnect = udp_disconnect,
2227 .ioctl = udp_ioctl,
2228 .destroy = udp_destroy_sock,
2229 .setsockopt = udp_setsockopt,
2230 .getsockopt = udp_getsockopt,
2231 .sendmsg = udp_sendmsg,
2232 .recvmsg = udp_recvmsg,
2233 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07002234 .backlog_rcv = __udp_queue_rcv_skb,
Steffen Klassert8141ed92013-01-21 02:00:03 +00002235 .release_cb = ip4_datagram_release_cb,
David S. Millerdb8dac22008-03-06 16:22:02 -08002236 .hash = udp_lib_hash,
2237 .unhash = udp_lib_unhash,
Eric Dumazet719f8352010-09-08 05:08:44 +00002238 .rehash = udp_v4_rehash,
David S. Millerdb8dac22008-03-06 16:22:02 -08002239 .get_port = udp_v4_get_port,
2240 .memory_allocated = &udp_memory_allocated,
2241 .sysctl_mem = sysctl_udp_mem,
2242 .sysctl_wmem = &sysctl_udp_wmem_min,
2243 .sysctl_rmem = &sysctl_udp_rmem_min,
2244 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07002245 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07002246 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08002247#ifdef CONFIG_COMPAT
2248 .compat_setsockopt = compat_udp_setsockopt,
2249 .compat_getsockopt = compat_udp_getsockopt,
2250#endif
Octavian Purdilafcbdf092010-12-16 14:26:56 -08002251 .clear_sk = sk_prot_clear_portaddr_nulls,
David S. Millerdb8dac22008-03-06 16:22:02 -08002252};
Eric Dumazetc482c562009-07-17 00:26:32 +00002253EXPORT_SYMBOL(udp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254
2255/* ------------------------------------------------------------------------ */
2256#ifdef CONFIG_PROC_FS
2257
Eric Dumazet645ca702008-10-29 01:41:45 -07002258static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259{
2260 struct sock *sk;
2261 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002262 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002264 for (state->bucket = start; state->bucket <= state->udp_table->mask;
2265 ++state->bucket) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002266 struct hlist_nulls_node *node;
Eric Dumazet645ca702008-10-29 01:41:45 -07002267 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002268
2269 if (hlist_nulls_empty(&hslot->head))
2270 continue;
2271
Eric Dumazet645ca702008-10-29 01:41:45 -07002272 spin_lock_bh(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08002273 sk_nulls_for_each(sk, node, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002274 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002275 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276 if (sk->sk_family == state->family)
2277 goto found;
2278 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002279 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002280 }
2281 sk = NULL;
2282found:
2283 return sk;
2284}
2285
2286static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2287{
2288 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002289 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290
2291 do {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002292 sk = sk_nulls_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002293 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294
Eric Dumazet645ca702008-10-29 01:41:45 -07002295 if (!sk) {
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002296 if (state->bucket <= state->udp_table->mask)
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002297 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07002298 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299 }
2300 return sk;
2301}
2302
2303static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2304{
Eric Dumazet645ca702008-10-29 01:41:45 -07002305 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306
2307 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08002308 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309 --pos;
2310 return pos ? NULL : sk;
2311}
2312
2313static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2314{
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002315 struct udp_iter_state *state = seq->private;
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002316 state->bucket = MAX_UDP_PORTS;
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002317
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002318 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319}
2320
2321static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2322{
2323 struct sock *sk;
2324
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002325 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326 sk = udp_get_idx(seq, 0);
2327 else
2328 sk = udp_get_next(seq, v);
2329
2330 ++*pos;
2331 return sk;
2332}
2333
2334static void udp_seq_stop(struct seq_file *seq, void *v)
2335{
Eric Dumazet645ca702008-10-29 01:41:45 -07002336 struct udp_iter_state *state = seq->private;
2337
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002338 if (state->bucket <= state->udp_table->mask)
Eric Dumazet645ca702008-10-29 01:41:45 -07002339 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340}
2341
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002342int udp_seq_open(struct inode *inode, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343{
Al Virod9dda782013-03-31 18:16:14 -04002344 struct udp_seq_afinfo *afinfo = PDE_DATA(inode);
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002345 struct udp_iter_state *s;
2346 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002348 err = seq_open_net(inode, file, &afinfo->seq_ops,
2349 sizeof(struct udp_iter_state));
2350 if (err < 0)
2351 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002352
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002353 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002354 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07002355 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002356 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002357}
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002358EXPORT_SYMBOL(udp_seq_open);
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002359
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002361int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362{
2363 struct proc_dir_entry *p;
2364 int rc = 0;
2365
Denis V. Lunevdda61922008-03-28 18:24:26 -07002366 afinfo->seq_ops.start = udp_seq_start;
2367 afinfo->seq_ops.next = udp_seq_next;
2368 afinfo->seq_ops.stop = udp_seq_stop;
2369
Denis V. Lunev84841c32008-05-02 04:10:08 -07002370 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002371 afinfo->seq_fops, afinfo);
Denis V. Lunev84841c32008-05-02 04:10:08 -07002372 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002373 rc = -ENOMEM;
2374 return rc;
2375}
Eric Dumazetc482c562009-07-17 00:26:32 +00002376EXPORT_SYMBOL(udp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002378void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002379{
Gao fengece31ff2013-02-18 01:34:56 +00002380 remove_proc_entry(afinfo->name, net->proc_net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381}
Eric Dumazetc482c562009-07-17 00:26:32 +00002382EXPORT_SYMBOL(udp_proc_unregister);
David S. Millerdb8dac22008-03-06 16:22:02 -08002383
2384/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002385static void udp4_format_sock(struct sock *sp, struct seq_file *f,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002386 int bucket)
David S. Millerdb8dac22008-03-06 16:22:02 -08002387{
2388 struct inet_sock *inet = inet_sk(sp);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002389 __be32 dest = inet->inet_daddr;
2390 __be32 src = inet->inet_rcv_saddr;
2391 __u16 destp = ntohs(inet->inet_dport);
2392 __u16 srcp = ntohs(inet->inet_sport);
David S. Millerdb8dac22008-03-06 16:22:02 -08002393
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002394 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
Tetsuo Handa652586d2013-11-14 14:31:57 -08002395 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
David S. Millerdb8dac22008-03-06 16:22:02 -08002396 bucket, src, srcp, dest, destp, sp->sk_state,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002397 sk_wmem_alloc_get(sp),
2398 sk_rmem_alloc_get(sp),
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06002399 0, 0L, 0,
2400 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2401 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002402 atomic_read(&sp->sk_refcnt), sp,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002403 atomic_read(&sp->sk_drops));
David S. Millerdb8dac22008-03-06 16:22:02 -08002404}
2405
2406int udp4_seq_show(struct seq_file *seq, void *v)
2407{
Tetsuo Handa652586d2013-11-14 14:31:57 -08002408 seq_setwidth(seq, 127);
David S. Millerdb8dac22008-03-06 16:22:02 -08002409 if (v == SEQ_START_TOKEN)
Tetsuo Handa652586d2013-11-14 14:31:57 -08002410 seq_puts(seq, " sl local_address rem_address st tx_queue "
David S. Millerdb8dac22008-03-06 16:22:02 -08002411 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002412 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08002413 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08002414 struct udp_iter_state *state = seq->private;
2415
Tetsuo Handa652586d2013-11-14 14:31:57 -08002416 udp4_format_sock(v, seq, state->bucket);
David S. Millerdb8dac22008-03-06 16:22:02 -08002417 }
Tetsuo Handa652586d2013-11-14 14:31:57 -08002418 seq_pad(seq, '\n');
David S. Millerdb8dac22008-03-06 16:22:02 -08002419 return 0;
2420}
2421
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002422static const struct file_operations udp_afinfo_seq_fops = {
2423 .owner = THIS_MODULE,
2424 .open = udp_seq_open,
2425 .read = seq_read,
2426 .llseek = seq_lseek,
2427 .release = seq_release_net
2428};
2429
David S. Millerdb8dac22008-03-06 16:22:02 -08002430/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08002431static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08002432 .name = "udp",
2433 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07002434 .udp_table = &udp_table,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002435 .seq_fops = &udp_afinfo_seq_fops,
Denis V. Lunevdda61922008-03-28 18:24:26 -07002436 .seq_ops = {
2437 .show = udp4_seq_show,
2438 },
David S. Millerdb8dac22008-03-06 16:22:02 -08002439};
2440
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002441static int __net_init udp4_proc_init_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002442{
2443 return udp_proc_register(net, &udp4_seq_afinfo);
2444}
2445
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002446static void __net_exit udp4_proc_exit_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002447{
2448 udp_proc_unregister(net, &udp4_seq_afinfo);
2449}
2450
2451static struct pernet_operations udp4_net_ops = {
2452 .init = udp4_proc_init_net,
2453 .exit = udp4_proc_exit_net,
2454};
2455
David S. Millerdb8dac22008-03-06 16:22:02 -08002456int __init udp4_proc_init(void)
2457{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002458 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002459}
2460
2461void udp4_proc_exit(void)
2462{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002463 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002464}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465#endif /* CONFIG_PROC_FS */
2466
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002467static __initdata unsigned long uhash_entries;
2468static int __init set_uhash_entries(char *str)
Eric Dumazet645ca702008-10-29 01:41:45 -07002469{
Eldad Zack413c27d2012-05-19 14:13:18 +00002470 ssize_t ret;
2471
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002472 if (!str)
2473 return 0;
Eldad Zack413c27d2012-05-19 14:13:18 +00002474
2475 ret = kstrtoul(str, 0, &uhash_entries);
2476 if (ret)
2477 return 0;
2478
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002479 if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2480 uhash_entries = UDP_HTABLE_SIZE_MIN;
2481 return 1;
2482}
2483__setup("uhash_entries=", set_uhash_entries);
Eric Dumazet645ca702008-10-29 01:41:45 -07002484
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002485void __init udp_table_init(struct udp_table *table, const char *name)
2486{
2487 unsigned int i;
2488
Tim Bird31fe62b2012-05-23 13:33:35 +00002489 table->hash = alloc_large_system_hash(name,
2490 2 * sizeof(struct udp_hslot),
2491 uhash_entries,
2492 21, /* one slot per 2 MB */
2493 0,
2494 &table->log,
2495 &table->mask,
2496 UDP_HTABLE_SIZE_MIN,
2497 64 * 1024);
2498
Eric Dumazet512615b2009-11-08 10:17:58 +00002499 table->hash2 = table->hash + (table->mask + 1);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002500 for (i = 0; i <= table->mask; i++) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002501 INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00002502 table->hash[i].count = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -07002503 spin_lock_init(&table->hash[i].lock);
2504 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002505 for (i = 0; i <= table->mask; i++) {
2506 INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i);
2507 table->hash2[i].count = 0;
2508 spin_lock_init(&table->hash2[i].lock);
2509 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002510}
2511
Hideo Aoki95766ff2007-12-31 00:29:24 -08002512void __init udp_init(void)
2513{
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002514 unsigned long limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002515
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002516 udp_table_init(&udp_table, "UDP");
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002517 limit = nr_free_buffer_pages() / 8;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002518 limit = max(limit, 128UL);
2519 sysctl_udp_mem[0] = limit / 4 * 3;
2520 sysctl_udp_mem[1] = limit;
2521 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2522
2523 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2524 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2525}
2526
Daniel Borkmannda5bab02013-06-08 12:56:03 +02002527struct sk_buff *skb_udp_tunnel_segment(struct sk_buff *skb,
2528 netdev_features_t features)
Pravin B Shelar73136262013-03-07 13:21:51 +00002529{
2530 struct sk_buff *segs = ERR_PTR(-EINVAL);
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08002531 u16 mac_offset = skb->mac_header;
Pravin B Shelar73136262013-03-07 13:21:51 +00002532 int mac_len = skb->mac_len;
2533 int tnl_hlen = skb_inner_mac_header(skb) - skb_transport_header(skb);
Pravin B Shelar0d055352013-05-02 16:17:02 +00002534 __be16 protocol = skb->protocol;
Pravin B Shelar73136262013-03-07 13:21:51 +00002535 netdev_features_t enc_features;
Tom Herbert0f4f4ff2014-06-04 17:20:16 -07002536 int udp_offset, outer_hlen;
2537 unsigned int oldlen;
2538 bool need_csum;
2539
2540 oldlen = (u16)~skb->len;
Pravin B Shelar73136262013-03-07 13:21:51 +00002541
2542 if (unlikely(!pskb_may_pull(skb, tnl_hlen)))
2543 goto out;
2544
2545 skb->encapsulation = 0;
2546 __skb_pull(skb, tnl_hlen);
2547 skb_reset_mac_header(skb);
2548 skb_set_network_header(skb, skb_inner_network_offset(skb));
2549 skb->mac_len = skb_inner_network_offset(skb);
Pravin B Shelar19acc322013-05-07 20:41:07 +00002550 skb->protocol = htons(ETH_P_TEB);
Pravin B Shelar73136262013-03-07 13:21:51 +00002551
Tom Herbert0f4f4ff2014-06-04 17:20:16 -07002552 need_csum = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM);
2553 if (need_csum)
2554 skb->encap_hdr_csum = 1;
2555
Pravin B Shelar73136262013-03-07 13:21:51 +00002556 /* segment inner packet. */
2557 enc_features = skb->dev->hw_enc_features & netif_skb_features(skb);
2558 segs = skb_mac_gso_segment(skb, enc_features);
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08002559 if (!segs || IS_ERR(segs)) {
2560 skb_gso_error_unwind(skb, protocol, tnl_hlen, mac_offset,
2561 mac_len);
Pravin B Shelar73136262013-03-07 13:21:51 +00002562 goto out;
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08002563 }
Pravin B Shelar73136262013-03-07 13:21:51 +00002564
2565 outer_hlen = skb_tnl_header_len(skb);
Tom Herbert0f4f4ff2014-06-04 17:20:16 -07002566 udp_offset = outer_hlen - tnl_hlen;
Pravin B Shelar73136262013-03-07 13:21:51 +00002567 skb = segs;
2568 do {
2569 struct udphdr *uh;
Tom Herbert0f4f4ff2014-06-04 17:20:16 -07002570 int len;
Pravin B Shelar73136262013-03-07 13:21:51 +00002571
Alexander Duyckcdbaa0b2013-07-10 17:05:06 -07002572 skb_reset_inner_headers(skb);
2573 skb->encapsulation = 1;
2574
Pravin B Shelar73136262013-03-07 13:21:51 +00002575 skb->mac_len = mac_len;
2576
2577 skb_push(skb, outer_hlen);
2578 skb_reset_mac_header(skb);
2579 skb_set_network_header(skb, mac_len);
2580 skb_set_transport_header(skb, udp_offset);
Tom Herbert0f4f4ff2014-06-04 17:20:16 -07002581 len = skb->len - udp_offset;
Pravin B Shelar73136262013-03-07 13:21:51 +00002582 uh = udp_hdr(skb);
Tom Herbert0f4f4ff2014-06-04 17:20:16 -07002583 uh->len = htons(len);
Pravin B Shelar73136262013-03-07 13:21:51 +00002584
Tom Herbert0f4f4ff2014-06-04 17:20:16 -07002585 if (need_csum) {
2586 __be32 delta = htonl(oldlen + len);
Pravin B Shelar73136262013-03-07 13:21:51 +00002587
Tom Herbert0f4f4ff2014-06-04 17:20:16 -07002588 uh->check = ~csum_fold((__force __wsum)
2589 ((__force u32)uh->check +
2590 (__force u32)delta));
2591 uh->check = gso_make_checksum(skb, ~uh->check);
2592
Pravin B Shelar73136262013-03-07 13:21:51 +00002593 if (uh->check == 0)
2594 uh->check = CSUM_MANGLED_0;
Pravin B Shelar73136262013-03-07 13:21:51 +00002595 }
Cong Wangeb3c0d82013-08-31 13:44:38 +08002596
Pravin B Shelar0d055352013-05-02 16:17:02 +00002597 skb->protocol = protocol;
Pravin B Shelar73136262013-03-07 13:21:51 +00002598 } while ((skb = skb->next));
2599out:
2600 return segs;
2601}