blob: f7d1d5e19e955563178fcddb84c4ffe9abb66036 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * The User Datagram Protocol (UDP).
7 *
Jesper Juhl02c30a82005-05-05 16:16:16 -07008 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
Alan Cox113aa832008-10-13 19:01:08 -070011 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012 * Hirokazu Takahashi, <taka@valinux.co.jp>
13 *
14 * Fixes:
15 * Alan Cox : verify_area() calls
16 * Alan Cox : stopped close while in use off icmp
17 * messages. Not a fix but a botch that
18 * for udp at least is 'valid'.
19 * Alan Cox : Fixed icmp handling properly
20 * Alan Cox : Correct error for oversized datagrams
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090021 * Alan Cox : Tidied select() semantics.
22 * Alan Cox : udp_err() fixed properly, also now
Linus Torvalds1da177e2005-04-16 15:20:36 -070023 * select and read wake correctly on errors
24 * Alan Cox : udp_send verify_area moved to avoid mem leak
25 * Alan Cox : UDP can count its memory
26 * Alan Cox : send to an unknown connection causes
27 * an ECONNREFUSED off the icmp, but
28 * does NOT close.
29 * Alan Cox : Switched to new sk_buff handlers. No more backlog!
30 * Alan Cox : Using generic datagram code. Even smaller and the PEEK
31 * bug no longer crashes it.
32 * Fred Van Kempen : Net2e support for sk->broadcast.
33 * Alan Cox : Uses skb_free_datagram
34 * Alan Cox : Added get/set sockopt support.
35 * Alan Cox : Broadcasting without option set returns EACCES.
36 * Alan Cox : No wakeup calls. Instead we now use the callbacks.
37 * Alan Cox : Use ip_tos and ip_ttl
38 * Alan Cox : SNMP Mibs
39 * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support.
40 * Matt Dillon : UDP length checks.
41 * Alan Cox : Smarter af_inet used properly.
42 * Alan Cox : Use new kernel side addressing.
43 * Alan Cox : Incorrect return on truncated datagram receive.
44 * Arnt Gulbrandsen : New udp_send and stuff
45 * Alan Cox : Cache last socket
46 * Alan Cox : Route cache
47 * Jon Peatfield : Minor efficiency fix to sendto().
48 * Mike Shaver : RFC1122 checks.
49 * Alan Cox : Nonblocking error fix.
50 * Willy Konynenberg : Transparent proxying support.
51 * Mike McLagan : Routing by source
52 * David S. Miller : New socket lookup architecture.
53 * Last socket cache retained as it
54 * does have a high hit rate.
55 * Olaf Kirch : Don't linearise iovec on sendmsg.
56 * Andi Kleen : Some cleanups, cache destination entry
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090057 * for connect.
Linus Torvalds1da177e2005-04-16 15:20:36 -070058 * Vitaly E. Lavrov : Transparent proxy revived after year coma.
59 * Melvin Smith : Check msg_name not msg_namelen in sendto(),
60 * return ENOTCONN for unconnected sockets (POSIX)
61 * Janos Farkas : don't deliver multi/broadcasts to a different
62 * bound-to-device socket
63 * Hirokazu Takahashi : HW checksumming for outgoing UDP
64 * datagrams.
65 * Hirokazu Takahashi : sendfile() on UDP works now.
66 * Arnaldo C. Melo : convert /proc/net/udp to seq_file
67 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
68 * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind
69 * a single port at the same time.
70 * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support
James Chapman342f0232007-06-27 15:37:46 -070071 * James Chapman : Add L2TP encapsulation type.
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 *
73 *
74 * This program is free software; you can redistribute it and/or
75 * modify it under the terms of the GNU General Public License
76 * as published by the Free Software Foundation; either version
77 * 2 of the License, or (at your option) any later version.
78 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090079
Joe Perchesafd465032012-03-12 07:03:32 +000080#define pr_fmt(fmt) "UDP: " fmt
81
Linus Torvalds1da177e2005-04-16 15:20:36 -070082#include <asm/uaccess.h>
83#include <asm/ioctls.h>
Hideo Aoki95766ff2007-12-31 00:29:24 -080084#include <linux/bootmem.h>
Eric Dumazet8203efb2008-10-29 02:32:32 -070085#include <linux/highmem.h>
86#include <linux/swap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070087#include <linux/types.h>
88#include <linux/fcntl.h>
89#include <linux/module.h>
90#include <linux/socket.h>
91#include <linux/sockios.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020092#include <linux/igmp.h>
Shawn Bohrer6e540302015-06-03 16:27:38 -050093#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#include <linux/in.h>
95#include <linux/errno.h>
96#include <linux/timer.h>
97#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <linux/inet.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#include <linux/netdevice.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +0900100#include <linux/slab.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -0700101#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102#include <linux/skbuff.h>
Tom Herbert57c67ff2014-08-22 13:34:44 -0700103#include <linux/netdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104#include <linux/proc_fs.h>
105#include <linux/seq_file.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200106#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#include <net/icmp.h>
Shawn Bohrer421b3882013-10-07 11:01:39 -0500108#include <net/inet_hashtables.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109#include <net/route.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110#include <net/checksum.h>
111#include <net/xfrm.h>
Satoru Moriya296f7ea2011-06-17 11:58:39 +0000112#include <trace/events/udp.h>
Eric Dumazet447167b2012-04-11 23:05:28 +0000113#include <linux/static_key.h>
Eric Dumazet22911fc2012-06-27 00:23:44 +0000114#include <trace/events/skb.h>
Eliezer Tamir076bb0c2013-07-10 17:13:17 +0300115#include <net/busy_poll.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800116#include "udp_impl.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000118struct udp_table udp_table __read_mostly;
Eric Dumazet645ca702008-10-29 01:41:45 -0700119EXPORT_SYMBOL(udp_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
Eric Dumazet8d987e52010-11-09 23:24:26 +0000121long sysctl_udp_mem[3] __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800122EXPORT_SYMBOL(sysctl_udp_mem);
Eric Dumazetc482c562009-07-17 00:26:32 +0000123
124int sysctl_udp_rmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800125EXPORT_SYMBOL(sysctl_udp_rmem_min);
Eric Dumazetc482c562009-07-17 00:26:32 +0000126
127int sysctl_udp_wmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800128EXPORT_SYMBOL(sysctl_udp_wmem_min);
129
Eric Dumazet8d987e52010-11-09 23:24:26 +0000130atomic_long_t udp_memory_allocated;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800131EXPORT_SYMBOL(udp_memory_allocated);
132
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000133#define MAX_UDP_PORTS 65536
134#define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN)
Eric Dumazet98322f22009-01-26 21:35:35 -0800135
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700136static int udp_lib_lport_inuse(struct net *net, __u16 num,
Eric Dumazet645ca702008-10-29 01:41:45 -0700137 const struct udp_hslot *hslot,
Eric Dumazet98322f22009-01-26 21:35:35 -0800138 unsigned long *bitmap,
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700139 struct sock *sk,
140 int (*saddr_comp)(const struct sock *sk1,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000141 const struct sock *sk2),
142 unsigned int log)
Gerrit Renker25030a72006-08-26 20:06:05 -0700143{
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700144 struct sock *sk2;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800145 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000146 kuid_t uid = sock_i_uid(sk);
Gerrit Renker25030a72006-08-26 20:06:05 -0700147
Joe Perches4243cdc2014-11-11 21:59:20 -0800148 sk_nulls_for_each(sk2, node, &hslot->head) {
Joe Perches9d4fb272009-11-23 10:41:23 -0800149 if (net_eq(sock_net(sk2), net) &&
150 sk2 != sk &&
Eric Dumazetd4cada42009-11-08 10:17:30 +0000151 (bitmap || udp_sk(sk2)->udp_port_hash == num) &&
Joe Perches9d4fb272009-11-23 10:41:23 -0800152 (!sk2->sk_reuse || !sk->sk_reuse) &&
153 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
154 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000155 (!sk2->sk_reuseport || !sk->sk_reuseport ||
Joe Perches4243cdc2014-11-11 21:59:20 -0800156 !uid_eq(uid, sock_i_uid(sk2))) &&
157 saddr_comp(sk, sk2)) {
158 if (!bitmap)
Eric Dumazet98322f22009-01-26 21:35:35 -0800159 return 1;
Joe Perches4243cdc2014-11-11 21:59:20 -0800160 __set_bit(udp_sk(sk2)->udp_port_hash >> log, bitmap);
Eric Dumazet98322f22009-01-26 21:35:35 -0800161 }
Joe Perches4243cdc2014-11-11 21:59:20 -0800162 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700163 return 0;
164}
165
Eric Dumazet30fff922009-11-09 05:26:33 +0000166/*
167 * Note: we still hold spinlock of primary hash chain, so no other writer
168 * can insert/delete a socket with local_port == num
169 */
170static int udp_lib_lport_inuse2(struct net *net, __u16 num,
Joe Perches4243cdc2014-11-11 21:59:20 -0800171 struct udp_hslot *hslot2,
172 struct sock *sk,
173 int (*saddr_comp)(const struct sock *sk1,
174 const struct sock *sk2))
Eric Dumazet30fff922009-11-09 05:26:33 +0000175{
176 struct sock *sk2;
177 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000178 kuid_t uid = sock_i_uid(sk);
Eric Dumazet30fff922009-11-09 05:26:33 +0000179 int res = 0;
180
181 spin_lock(&hslot2->lock);
Joe Perches4243cdc2014-11-11 21:59:20 -0800182 udp_portaddr_for_each_entry(sk2, node, &hslot2->head) {
Joe Perches9d4fb272009-11-23 10:41:23 -0800183 if (net_eq(sock_net(sk2), net) &&
184 sk2 != sk &&
185 (udp_sk(sk2)->udp_port_hash == num) &&
186 (!sk2->sk_reuse || !sk->sk_reuse) &&
187 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
188 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000189 (!sk2->sk_reuseport || !sk->sk_reuseport ||
Joe Perches4243cdc2014-11-11 21:59:20 -0800190 !uid_eq(uid, sock_i_uid(sk2))) &&
191 saddr_comp(sk, sk2)) {
Eric Dumazet30fff922009-11-09 05:26:33 +0000192 res = 1;
193 break;
194 }
Joe Perches4243cdc2014-11-11 21:59:20 -0800195 }
Eric Dumazet30fff922009-11-09 05:26:33 +0000196 spin_unlock(&hslot2->lock);
197 return res;
198}
199
Gerrit Renker25030a72006-08-26 20:06:05 -0700200/**
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700201 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
Gerrit Renker25030a72006-08-26 20:06:05 -0700202 *
203 * @sk: socket struct in question
204 * @snum: port number to look up
David S. Millerdf2bc452007-06-05 15:18:43 -0700205 * @saddr_comp: AF-dependent comparison of bound local IP addresses
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300206 * @hash2_nulladdr: AF-dependent hash value in secondary hash chains,
Eric Dumazet30fff922009-11-09 05:26:33 +0000207 * with NULL address
Gerrit Renker25030a72006-08-26 20:06:05 -0700208 */
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700209int udp_lib_get_port(struct sock *sk, unsigned short snum,
Joe Perches4243cdc2014-11-11 21:59:20 -0800210 int (*saddr_comp)(const struct sock *sk1,
211 const struct sock *sk2),
Eric Dumazet30fff922009-11-09 05:26:33 +0000212 unsigned int hash2_nulladdr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213{
Eric Dumazet512615b2009-11-08 10:17:58 +0000214 struct udp_hslot *hslot, *hslot2;
Eric Dumazet645ca702008-10-29 01:41:45 -0700215 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Gerrit Renker25030a72006-08-26 20:06:05 -0700216 int error = 1;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900217 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700219 if (!snum) {
Eric Dumazet9088c562008-10-08 11:44:17 -0700220 int low, high, remaining;
Eric Dumazet95c96172012-04-15 05:58:06 +0000221 unsigned int rand;
Eric Dumazet98322f22009-01-26 21:35:35 -0800222 unsigned short first, last;
223 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -0700225 inet_get_local_port_range(net, &low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700226 remaining = (high - low) + 1;
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700227
Aruna-Hewapathirane63862b52014-01-11 07:15:59 -0500228 rand = prandom_u32();
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200229 first = reciprocal_scale(rand, remaining) + low;
Eric Dumazet98322f22009-01-26 21:35:35 -0800230 /*
231 * force rand to be an odd multiple of UDP_HTABLE_SIZE
232 */
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000233 rand = (rand | 1) * (udptable->mask + 1);
Eric Dumazet5781b232009-12-13 19:32:39 -0800234 last = first + udptable->mask + 1;
235 do {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000236 hslot = udp_hashslot(udptable, net, first);
Eric Dumazet98322f22009-01-26 21:35:35 -0800237 bitmap_zero(bitmap, PORTS_PER_CHAIN);
Eric Dumazet645ca702008-10-29 01:41:45 -0700238 spin_lock_bh(&hslot->lock);
Eric Dumazet98322f22009-01-26 21:35:35 -0800239 udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000240 saddr_comp, udptable->log);
Eric Dumazet98322f22009-01-26 21:35:35 -0800241
242 snum = first;
243 /*
244 * Iterate on all possible values of snum for this hash.
245 * Using steps of an odd multiple of UDP_HTABLE_SIZE
246 * give us randomization and full range coverage.
247 */
Eric Dumazet9088c562008-10-08 11:44:17 -0700248 do {
Eric Dumazet98322f22009-01-26 21:35:35 -0800249 if (low <= snum && snum <= high &&
Amerigo Wange3826f12010-05-05 00:27:06 +0000250 !test_bit(snum >> udptable->log, bitmap) &&
WANG Cong122ff242014-05-12 16:04:53 -0700251 !inet_is_local_reserved_port(net, snum))
Eric Dumazet98322f22009-01-26 21:35:35 -0800252 goto found;
253 snum += rand;
254 } while (snum != first);
255 spin_unlock_bh(&hslot->lock);
Eric Dumazet5781b232009-12-13 19:32:39 -0800256 } while (++first != last);
Eric Dumazet98322f22009-01-26 21:35:35 -0800257 goto fail;
Eric Dumazet645ca702008-10-29 01:41:45 -0700258 } else {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000259 hslot = udp_hashslot(udptable, net, snum);
Eric Dumazet645ca702008-10-29 01:41:45 -0700260 spin_lock_bh(&hslot->lock);
Eric Dumazet30fff922009-11-09 05:26:33 +0000261 if (hslot->count > 10) {
262 int exist;
263 unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum;
264
265 slot2 &= udptable->mask;
266 hash2_nulladdr &= udptable->mask;
267
268 hslot2 = udp_hashslot2(udptable, slot2);
269 if (hslot->count < hslot2->count)
270 goto scan_primary_hash;
271
272 exist = udp_lib_lport_inuse2(net, snum, hslot2,
273 sk, saddr_comp);
274 if (!exist && (hash2_nulladdr != slot2)) {
275 hslot2 = udp_hashslot2(udptable, hash2_nulladdr);
276 exist = udp_lib_lport_inuse2(net, snum, hslot2,
277 sk, saddr_comp);
278 }
279 if (exist)
280 goto fail_unlock;
281 else
282 goto found;
283 }
284scan_primary_hash:
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000285 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk,
286 saddr_comp, 0))
Eric Dumazet645ca702008-10-29 01:41:45 -0700287 goto fail_unlock;
288 }
Eric Dumazet98322f22009-01-26 21:35:35 -0800289found:
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000290 inet_sk(sk)->inet_num = snum;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000291 udp_sk(sk)->udp_port_hash = snum;
292 udp_sk(sk)->udp_portaddr_hash ^= snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 if (sk_unhashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800294 sk_nulls_add_node_rcu(sk, &hslot->head);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +0000295 hslot->count++;
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700296 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet512615b2009-11-08 10:17:58 +0000297
298 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
299 spin_lock(&hslot2->lock);
300 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
301 &hslot2->head);
302 hslot2->count++;
303 spin_unlock(&hslot2->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700305 error = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -0700306fail_unlock:
307 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308fail:
Gerrit Renker25030a72006-08-26 20:06:05 -0700309 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310}
Eric Dumazetc482c562009-07-17 00:26:32 +0000311EXPORT_SYMBOL(udp_lib_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
Vlad Yasevich499923c2009-04-09 17:37:33 +0000313static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
David S. Millerdb8dac22008-03-06 16:22:02 -0800314{
315 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
316
Eric Dumazetc482c562009-07-17 00:26:32 +0000317 return (!ipv6_only_sock(sk2) &&
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000318 (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr ||
319 inet1->inet_rcv_saddr == inet2->inet_rcv_saddr));
David S. Millerdb8dac22008-03-06 16:22:02 -0800320}
321
Eric Dumazet6eada012015-03-18 14:05:33 -0700322static u32 udp4_portaddr_hash(const struct net *net, __be32 saddr,
323 unsigned int port)
Eric Dumazetd4cada42009-11-08 10:17:30 +0000324{
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700325 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000326}
327
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700328int udp_v4_get_port(struct sock *sk, unsigned short snum)
David S. Millerdb8dac22008-03-06 16:22:02 -0800329{
Eric Dumazet30fff922009-11-09 05:26:33 +0000330 unsigned int hash2_nulladdr =
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700331 udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum);
Eric Dumazet30fff922009-11-09 05:26:33 +0000332 unsigned int hash2_partial =
333 udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0);
334
Eric Dumazetd4cada42009-11-08 10:17:30 +0000335 /* precompute partial secondary hash */
Eric Dumazet30fff922009-11-09 05:26:33 +0000336 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
337 return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr);
David S. Millerdb8dac22008-03-06 16:22:02 -0800338}
339
Joe Perches60c04ae2014-12-01 20:29:06 -0800340static inline int compute_score(struct sock *sk, struct net *net,
341 __be32 saddr, unsigned short hnum, __be16 sport,
342 __be32 daddr, __be16 dport, int dif)
Eric Dumazet645ca702008-10-29 01:41:45 -0700343{
Joe Perches60c04ae2014-12-01 20:29:06 -0800344 int score;
345 struct inet_sock *inet;
Eric Dumazet645ca702008-10-29 01:41:45 -0700346
Joe Perches60c04ae2014-12-01 20:29:06 -0800347 if (!net_eq(sock_net(sk), net) ||
348 udp_sk(sk)->udp_port_hash != hnum ||
349 ipv6_only_sock(sk))
350 return -1;
Eric Dumazet645ca702008-10-29 01:41:45 -0700351
Joe Perches60c04ae2014-12-01 20:29:06 -0800352 score = (sk->sk_family == PF_INET) ? 2 : 1;
353 inet = inet_sk(sk);
354
355 if (inet->inet_rcv_saddr) {
356 if (inet->inet_rcv_saddr != daddr)
357 return -1;
358 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700359 }
Joe Perches60c04ae2014-12-01 20:29:06 -0800360
361 if (inet->inet_daddr) {
362 if (inet->inet_daddr != saddr)
363 return -1;
364 score += 4;
365 }
366
367 if (inet->inet_dport) {
368 if (inet->inet_dport != sport)
369 return -1;
370 score += 4;
371 }
372
373 if (sk->sk_bound_dev_if) {
374 if (sk->sk_bound_dev_if != dif)
375 return -1;
376 score += 4;
377 }
378
Eric Dumazet645ca702008-10-29 01:41:45 -0700379 return score;
380}
381
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000382/*
383 * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num)
384 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000385static inline int compute_score2(struct sock *sk, struct net *net,
386 __be32 saddr, __be16 sport,
387 __be32 daddr, unsigned int hnum, int dif)
388{
Joe Perches60c04ae2014-12-01 20:29:06 -0800389 int score;
390 struct inet_sock *inet;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000391
Joe Perches60c04ae2014-12-01 20:29:06 -0800392 if (!net_eq(sock_net(sk), net) ||
393 ipv6_only_sock(sk))
394 return -1;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000395
Joe Perches60c04ae2014-12-01 20:29:06 -0800396 inet = inet_sk(sk);
397
398 if (inet->inet_rcv_saddr != daddr ||
399 inet->inet_num != hnum)
400 return -1;
401
402 score = (sk->sk_family == PF_INET) ? 2 : 1;
403
404 if (inet->inet_daddr) {
405 if (inet->inet_daddr != saddr)
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000406 return -1;
Joe Perches60c04ae2014-12-01 20:29:06 -0800407 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000408 }
Joe Perches60c04ae2014-12-01 20:29:06 -0800409
410 if (inet->inet_dport) {
411 if (inet->inet_dport != sport)
412 return -1;
413 score += 4;
414 }
415
416 if (sk->sk_bound_dev_if) {
417 if (sk->sk_bound_dev_if != dif)
418 return -1;
419 score += 4;
420 }
421
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000422 return score;
423}
424
Eric Dumazet6eada012015-03-18 14:05:33 -0700425static u32 udp_ehashfn(const struct net *net, const __be32 laddr,
426 const __u16 lport, const __be32 faddr,
427 const __be16 fport)
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200428{
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200429 static u32 udp_ehash_secret __read_mostly;
430
431 net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret));
432
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200433 return __inet_ehashfn(laddr, lport, faddr, fport,
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200434 udp_ehash_secret + net_hash_mix(net));
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200435}
436
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000437/* called with read_rcu_lock() */
438static struct sock *udp4_lib_lookup2(struct net *net,
439 __be32 saddr, __be16 sport,
440 __be32 daddr, unsigned int hnum, int dif,
441 struct udp_hslot *hslot2, unsigned int slot2)
442{
443 struct sock *sk, *result;
444 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000445 int score, badness, matches = 0, reuseport = 0;
446 u32 hash = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000447
448begin:
449 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000450 badness = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000451 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
452 score = compute_score2(sk, net, saddr, sport,
453 daddr, hnum, dif);
454 if (score > badness) {
455 result = sk;
456 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000457 reuseport = sk->sk_reuseport;
458 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200459 hash = udp_ehashfn(net, daddr, hnum,
460 saddr, sport);
Tom Herbertba418fa2013-01-22 09:50:32 +0000461 matches = 1;
462 }
463 } else if (score == badness && reuseport) {
464 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200465 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000466 result = sk;
467 hash = next_pseudo_random32(hash);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000468 }
469 }
470 /*
471 * if the nulls value we got at the end of this lookup is
472 * not the expected one, we must restart lookup.
473 * We probably met an item that was moved to another chain.
474 */
475 if (get_nulls_value(node) != slot2)
476 goto begin;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000477 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000478 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000479 result = NULL;
480 else if (unlikely(compute_score2(result, net, saddr, sport,
481 daddr, hnum, dif) < badness)) {
482 sock_put(result);
483 goto begin;
484 }
485 }
486 return result;
487}
488
David S. Millerdb8dac22008-03-06 16:22:02 -0800489/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
490 * harder than this. -DaveM
491 */
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000492struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -0800493 __be16 sport, __be32 daddr, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700494 int dif, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800495{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700496 struct sock *sk, *result;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800497 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800498 unsigned short hnum = ntohs(dport);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000499 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
500 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
Tom Herbertba418fa2013-01-22 09:50:32 +0000501 int score, badness, matches = 0, reuseport = 0;
502 u32 hash = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800503
Eric Dumazet271b72c2008-10-29 02:11:14 -0700504 rcu_read_lock();
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000505 if (hslot->count > 10) {
506 hash2 = udp4_portaddr_hash(net, daddr, hnum);
507 slot2 = hash2 & udptable->mask;
508 hslot2 = &udptable->hash2[slot2];
509 if (hslot->count < hslot2->count)
510 goto begin;
511
512 result = udp4_lib_lookup2(net, saddr, sport,
513 daddr, hnum, dif,
514 hslot2, slot2);
515 if (!result) {
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700516 hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000517 slot2 = hash2 & udptable->mask;
518 hslot2 = &udptable->hash2[slot2];
519 if (hslot->count < hslot2->count)
520 goto begin;
521
Jorge Boncompte [DTI2]1223c672010-04-08 04:56:48 +0000522 result = udp4_lib_lookup2(net, saddr, sport,
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700523 htonl(INADDR_ANY), hnum, dif,
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000524 hslot2, slot2);
525 }
526 rcu_read_unlock();
527 return result;
528 }
Eric Dumazet271b72c2008-10-29 02:11:14 -0700529begin:
530 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000531 badness = 0;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800532 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
Eric Dumazet645ca702008-10-29 01:41:45 -0700533 score = compute_score(sk, net, saddr, hnum, sport,
534 daddr, dport, dif);
535 if (score > badness) {
536 result = sk;
537 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000538 reuseport = sk->sk_reuseport;
539 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200540 hash = udp_ehashfn(net, daddr, hnum,
541 saddr, sport);
Tom Herbertba418fa2013-01-22 09:50:32 +0000542 matches = 1;
543 }
544 } else if (score == badness && reuseport) {
545 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200546 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000547 result = sk;
548 hash = next_pseudo_random32(hash);
David S. Millerdb8dac22008-03-06 16:22:02 -0800549 }
550 }
Eric Dumazet88ab1932008-11-16 19:39:21 -0800551 /*
552 * if the nulls value we got at the end of this lookup is
553 * not the expected one, we must restart lookup.
554 * We probably met an item that was moved to another chain.
555 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000556 if (get_nulls_value(node) != slot)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800557 goto begin;
558
Eric Dumazet271b72c2008-10-29 02:11:14 -0700559 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000560 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet271b72c2008-10-29 02:11:14 -0700561 result = NULL;
562 else if (unlikely(compute_score(result, net, saddr, hnum, sport,
563 daddr, dport, dif) < badness)) {
564 sock_put(result);
565 goto begin;
566 }
567 }
568 rcu_read_unlock();
David S. Millerdb8dac22008-03-06 16:22:02 -0800569 return result;
570}
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000571EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
David S. Millerdb8dac22008-03-06 16:22:02 -0800572
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700573static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
574 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700575 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700576{
577 const struct iphdr *iph = ip_hdr(skb);
578
Eric Dumazet8afdd992013-12-10 18:07:23 -0800579 return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport,
580 iph->daddr, dport, inet_iif(skb),
581 udptable);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700582}
583
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700584struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
585 __be32 daddr, __be16 dport, int dif)
586{
Eric Dumazet645ca702008-10-29 01:41:45 -0700587 return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700588}
589EXPORT_SYMBOL_GPL(udp4_lib_lookup);
590
Shawn Bohrer421b3882013-10-07 11:01:39 -0500591static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk,
592 __be16 loc_port, __be32 loc_addr,
593 __be16 rmt_port, __be32 rmt_addr,
594 int dif, unsigned short hnum)
595{
596 struct inet_sock *inet = inet_sk(sk);
597
598 if (!net_eq(sock_net(sk), net) ||
599 udp_sk(sk)->udp_port_hash != hnum ||
600 (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
601 (inet->inet_dport != rmt_port && inet->inet_dport) ||
602 (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) ||
603 ipv6_only_sock(sk) ||
604 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
605 return false;
606 if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif))
607 return false;
608 return true;
609}
610
David S. Millerdb8dac22008-03-06 16:22:02 -0800611/*
612 * This routine is called by the ICMP module when it gets some
613 * sort of error condition. If err < 0 then the socket should
614 * be closed and the error returned to the user. If err > 0
615 * it's just the icmp type << 8 | icmp code.
616 * Header points to the ip header of the error packet. We move
617 * on past this. Then (as it used to claim before adjustment)
618 * header points to the first 8 bytes of the udp header. We need
619 * to find the appropriate port.
620 */
621
Eric Dumazet645ca702008-10-29 01:41:45 -0700622void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800623{
624 struct inet_sock *inet;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000625 const struct iphdr *iph = (const struct iphdr *)skb->data;
Eric Dumazetc482c562009-07-17 00:26:32 +0000626 struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
David S. Millerdb8dac22008-03-06 16:22:02 -0800627 const int type = icmp_hdr(skb)->type;
628 const int code = icmp_hdr(skb)->code;
629 struct sock *sk;
630 int harderr;
631 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700632 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800633
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700634 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
David S. Millerdb8dac22008-03-06 16:22:02 -0800635 iph->saddr, uh->source, skb->dev->ifindex, udptable);
Ian Morris51456b22015-04-03 09:17:26 +0100636 if (!sk) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700637 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800638 return; /* No socket for error */
639 }
640
641 err = 0;
642 harderr = 0;
643 inet = inet_sk(sk);
644
645 switch (type) {
646 default:
647 case ICMP_TIME_EXCEEDED:
648 err = EHOSTUNREACH;
649 break;
650 case ICMP_SOURCE_QUENCH:
651 goto out;
652 case ICMP_PARAMETERPROB:
653 err = EPROTO;
654 harderr = 1;
655 break;
656 case ICMP_DEST_UNREACH:
657 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
David S. Miller36393392012-06-14 22:21:46 -0700658 ipv4_sk_update_pmtu(skb, sk, info);
David S. Millerdb8dac22008-03-06 16:22:02 -0800659 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
660 err = EMSGSIZE;
661 harderr = 1;
662 break;
663 }
664 goto out;
665 }
666 err = EHOSTUNREACH;
667 if (code <= NR_ICMP_UNREACH) {
668 harderr = icmp_err_convert[code].fatal;
669 err = icmp_err_convert[code].errno;
670 }
671 break;
David S. Miller55be7a92012-07-11 21:27:49 -0700672 case ICMP_REDIRECT:
673 ipv4_sk_redirect(skb, sk);
Duan Jiong1a462d12013-09-20 18:20:28 +0800674 goto out;
David S. Millerdb8dac22008-03-06 16:22:02 -0800675 }
676
677 /*
678 * RFC1122: OK. Passes ICMP errors back to application, as per
679 * 4.1.3.3.
680 */
681 if (!inet->recverr) {
682 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
683 goto out;
Eric Dumazetb1faf562010-05-31 23:44:05 -0700684 } else
Eric Dumazetc482c562009-07-17 00:26:32 +0000685 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
Eric Dumazetb1faf562010-05-31 23:44:05 -0700686
David S. Millerdb8dac22008-03-06 16:22:02 -0800687 sk->sk_err = err;
688 sk->sk_error_report(sk);
689out:
690 sock_put(sk);
691}
692
693void udp_err(struct sk_buff *skb, u32 info)
694{
Eric Dumazet645ca702008-10-29 01:41:45 -0700695 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800696}
697
698/*
699 * Throw away all pending data and cancel the corking. Socket is locked.
700 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400701void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800702{
703 struct udp_sock *up = udp_sk(sk);
704
705 if (up->pending) {
706 up->len = 0;
707 up->pending = 0;
708 ip_flush_pending_frames(sk);
709 }
710}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400711EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800712
713/**
Herbert Xuf6b96642011-03-01 02:36:48 +0000714 * udp4_hwcsum - handle outgoing HW checksumming
David S. Millerdb8dac22008-03-06 16:22:02 -0800715 * @skb: sk_buff containing the filled-in UDP header
716 * (checksum field must be zeroed out)
Herbert Xuf6b96642011-03-01 02:36:48 +0000717 * @src: source IP address
718 * @dst: destination IP address
David S. Millerdb8dac22008-03-06 16:22:02 -0800719 */
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200720void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
David S. Millerdb8dac22008-03-06 16:22:02 -0800721{
David S. Millerdb8dac22008-03-06 16:22:02 -0800722 struct udphdr *uh = udp_hdr(skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000723 int offset = skb_transport_offset(skb);
724 int len = skb->len - offset;
725 int hlen = len;
David S. Millerdb8dac22008-03-06 16:22:02 -0800726 __wsum csum = 0;
727
WANG Congebbe4952014-06-02 16:12:02 -0700728 if (!skb_has_frag_list(skb)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800729 /*
730 * Only one fragment on the socket.
731 */
732 skb->csum_start = skb_transport_header(skb) - skb->head;
733 skb->csum_offset = offsetof(struct udphdr, check);
Herbert Xuf6b96642011-03-01 02:36:48 +0000734 uh->check = ~csum_tcpudp_magic(src, dst, len,
735 IPPROTO_UDP, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800736 } else {
WANG Congebbe4952014-06-02 16:12:02 -0700737 struct sk_buff *frags;
738
David S. Millerdb8dac22008-03-06 16:22:02 -0800739 /*
740 * HW-checksum won't work as there are two or more
741 * fragments on the socket so that all csums of sk_buffs
742 * should be together
743 */
WANG Congebbe4952014-06-02 16:12:02 -0700744 skb_walk_frags(skb, frags) {
Herbert Xuf6b96642011-03-01 02:36:48 +0000745 csum = csum_add(csum, frags->csum);
746 hlen -= frags->len;
WANG Congebbe4952014-06-02 16:12:02 -0700747 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800748
Herbert Xuf6b96642011-03-01 02:36:48 +0000749 csum = skb_checksum(skb, offset, hlen, csum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800750 skb->ip_summed = CHECKSUM_NONE;
751
David S. Millerdb8dac22008-03-06 16:22:02 -0800752 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
753 if (uh->check == 0)
754 uh->check = CSUM_MANGLED_0;
755 }
756}
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200757EXPORT_SYMBOL_GPL(udp4_hwcsum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800758
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700759/* Function to set UDP checksum for an IPv4 UDP packet. This is intended
760 * for the simple case like when setting the checksum for a UDP tunnel.
761 */
762void udp_set_csum(bool nocheck, struct sk_buff *skb,
763 __be32 saddr, __be32 daddr, int len)
764{
765 struct udphdr *uh = udp_hdr(skb);
766
767 if (nocheck)
768 uh->check = 0;
769 else if (skb_is_gso(skb))
770 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
771 else if (skb_dst(skb) && skb_dst(skb)->dev &&
772 (skb_dst(skb)->dev->features & NETIF_F_V4_CSUM)) {
773
774 BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL);
775
776 skb->ip_summed = CHECKSUM_PARTIAL;
777 skb->csum_start = skb_transport_header(skb) - skb->head;
778 skb->csum_offset = offsetof(struct udphdr, check);
779 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
780 } else {
781 __wsum csum;
782
783 BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL);
784
785 uh->check = 0;
786 csum = skb_checksum(skb, 0, len, 0);
787 uh->check = udp_v4_check(len, saddr, daddr, csum);
788 if (uh->check == 0)
789 uh->check = CSUM_MANGLED_0;
790
791 skb->ip_summed = CHECKSUM_UNNECESSARY;
792 }
793}
794EXPORT_SYMBOL(udp_set_csum);
795
David S. Miller79ab0532011-05-09 13:31:04 -0700796static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
Herbert Xuf6b96642011-03-01 02:36:48 +0000797{
798 struct sock *sk = skb->sk;
799 struct inet_sock *inet = inet_sk(sk);
800 struct udphdr *uh;
Herbert Xuf6b96642011-03-01 02:36:48 +0000801 int err = 0;
802 int is_udplite = IS_UDPLITE(sk);
803 int offset = skb_transport_offset(skb);
804 int len = skb->len - offset;
805 __wsum csum = 0;
806
807 /*
808 * Create a UDP header
809 */
810 uh = udp_hdr(skb);
811 uh->source = inet->inet_sport;
David S. Miller79ab0532011-05-09 13:31:04 -0700812 uh->dest = fl4->fl4_dport;
Herbert Xuf6b96642011-03-01 02:36:48 +0000813 uh->len = htons(len);
814 uh->check = 0;
815
816 if (is_udplite) /* UDP-Lite */
817 csum = udplite_csum(skb);
818
Tom Herbert28448b82014-05-23 08:47:19 -0700819 else if (sk->sk_no_check_tx) { /* UDP csum disabled */
Herbert Xuf6b96642011-03-01 02:36:48 +0000820
821 skb->ip_summed = CHECKSUM_NONE;
822 goto send;
823
824 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
825
David S. Miller79ab0532011-05-09 13:31:04 -0700826 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
Herbert Xuf6b96642011-03-01 02:36:48 +0000827 goto send;
828
829 } else
830 csum = udp_csum(skb);
831
832 /* add protocol-dependent pseudo-header */
David S. Miller79ab0532011-05-09 13:31:04 -0700833 uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
Herbert Xuf6b96642011-03-01 02:36:48 +0000834 sk->sk_protocol, csum);
835 if (uh->check == 0)
836 uh->check = CSUM_MANGLED_0;
837
838send:
Eric Dumazetb5ec8ee2012-08-10 02:22:47 +0000839 err = ip_send_skb(sock_net(sk), skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000840 if (err) {
841 if (err == -ENOBUFS && !inet->recverr) {
842 UDP_INC_STATS_USER(sock_net(sk),
843 UDP_MIB_SNDBUFERRORS, is_udplite);
844 err = 0;
845 }
846 } else
847 UDP_INC_STATS_USER(sock_net(sk),
848 UDP_MIB_OUTDATAGRAMS, is_udplite);
849 return err;
850}
851
David S. Millerdb8dac22008-03-06 16:22:02 -0800852/*
853 * Push out all pending data as one UDP datagram. Socket is locked.
854 */
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200855int udp_push_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800856{
857 struct udp_sock *up = udp_sk(sk);
858 struct inet_sock *inet = inet_sk(sk);
David S. Millerb6f21b22011-03-12 02:09:18 -0500859 struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800860 struct sk_buff *skb;
David S. Millerdb8dac22008-03-06 16:22:02 -0800861 int err = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800862
David S. Miller77968b72011-05-08 17:12:19 -0700863 skb = ip_finish_skb(sk, fl4);
Herbert Xuf6b96642011-03-01 02:36:48 +0000864 if (!skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800865 goto out;
866
David S. Miller79ab0532011-05-09 13:31:04 -0700867 err = udp_send_skb(skb, fl4);
David S. Millerdb8dac22008-03-06 16:22:02 -0800868
David S. Millerdb8dac22008-03-06 16:22:02 -0800869out:
870 up->len = 0;
871 up->pending = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800872 return err;
873}
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200874EXPORT_SYMBOL(udp_push_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800875
Ying Xue1b784142015-03-02 15:37:48 +0800876int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
David S. Millerdb8dac22008-03-06 16:22:02 -0800877{
878 struct inet_sock *inet = inet_sk(sk);
879 struct udp_sock *up = udp_sk(sk);
David S. Millere4749952011-05-08 16:38:45 -0700880 struct flowi4 fl4_stack;
David S. Millerb6f21b22011-03-12 02:09:18 -0500881 struct flowi4 *fl4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800882 int ulen = len;
883 struct ipcm_cookie ipc;
884 struct rtable *rt = NULL;
885 int free = 0;
886 int connected = 0;
887 __be32 daddr, faddr, saddr;
888 __be16 dport;
889 u8 tos;
890 int err, is_udplite = IS_UDPLITE(sk);
891 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
892 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
Herbert Xu903ab862011-03-01 02:36:48 +0000893 struct sk_buff *skb;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000894 struct ip_options_data opt_copy;
David S. Millerdb8dac22008-03-06 16:22:02 -0800895
896 if (len > 0xFFFF)
897 return -EMSGSIZE;
898
899 /*
900 * Check the flags.
901 */
902
Eric Dumazetc482c562009-07-17 00:26:32 +0000903 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
David S. Millerdb8dac22008-03-06 16:22:02 -0800904 return -EOPNOTSUPP;
905
906 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000907 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200908 ipc.ttl = 0;
909 ipc.tos = -1;
David S. Millerdb8dac22008-03-06 16:22:02 -0800910
Herbert Xu903ab862011-03-01 02:36:48 +0000911 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
912
David S. Millerf5fca602011-05-08 17:24:10 -0700913 fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800914 if (up->pending) {
915 /*
916 * There are pending frames.
917 * The socket lock must be held while it's corked.
918 */
919 lock_sock(sk);
920 if (likely(up->pending)) {
921 if (unlikely(up->pending != AF_INET)) {
922 release_sock(sk);
923 return -EINVAL;
924 }
925 goto do_append_data;
926 }
927 release_sock(sk);
928 }
929 ulen += sizeof(struct udphdr);
930
931 /*
932 * Get and verify the address.
933 */
934 if (msg->msg_name) {
Steffen Hurrle342dfc32014-01-17 22:53:15 +0100935 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -0800936 if (msg->msg_namelen < sizeof(*usin))
937 return -EINVAL;
938 if (usin->sin_family != AF_INET) {
939 if (usin->sin_family != AF_UNSPEC)
940 return -EAFNOSUPPORT;
941 }
942
943 daddr = usin->sin_addr.s_addr;
944 dport = usin->sin_port;
945 if (dport == 0)
946 return -EINVAL;
947 } else {
948 if (sk->sk_state != TCP_ESTABLISHED)
949 return -EDESTADDRREQ;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000950 daddr = inet->inet_daddr;
951 dport = inet->inet_dport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800952 /* Open fast path for connected socket.
953 Route will not be used, if at least one option is set.
954 */
955 connected = 1;
956 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000957 ipc.addr = inet->inet_saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800958
959 ipc.oif = sk->sk_bound_dev_if;
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000960
961 sock_tx_timestamp(sk, &ipc.tx_flags);
962
David S. Millerdb8dac22008-03-06 16:22:02 -0800963 if (msg->msg_controllen) {
Hannes Frederic Sowac8e6ad02014-02-18 21:38:08 +0100964 err = ip_cmsg_send(sock_net(sk), msg, &ipc,
965 sk->sk_family == AF_INET6);
David S. Millerdb8dac22008-03-06 16:22:02 -0800966 if (err)
967 return err;
968 if (ipc.opt)
969 free = 1;
970 connected = 0;
971 }
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000972 if (!ipc.opt) {
973 struct ip_options_rcu *inet_opt;
974
975 rcu_read_lock();
976 inet_opt = rcu_dereference(inet->inet_opt);
977 if (inet_opt) {
978 memcpy(&opt_copy, inet_opt,
979 sizeof(*inet_opt) + inet_opt->opt.optlen);
980 ipc.opt = &opt_copy.opt;
981 }
982 rcu_read_unlock();
983 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800984
985 saddr = ipc.addr;
986 ipc.addr = faddr = daddr;
987
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000988 if (ipc.opt && ipc.opt->opt.srr) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800989 if (!daddr)
990 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000991 faddr = ipc.opt->opt.faddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800992 connected = 0;
993 }
Francesco Fuscoaa661582013-09-24 15:43:09 +0200994 tos = get_rttos(&ipc, inet);
David S. Millerdb8dac22008-03-06 16:22:02 -0800995 if (sock_flag(sk, SOCK_LOCALROUTE) ||
996 (msg->msg_flags & MSG_DONTROUTE) ||
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000997 (ipc.opt && ipc.opt->opt.is_strictroute)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800998 tos |= RTO_ONLINK;
999 connected = 0;
1000 }
1001
1002 if (ipv4_is_multicast(daddr)) {
1003 if (!ipc.oif)
1004 ipc.oif = inet->mc_index;
1005 if (!saddr)
1006 saddr = inet->mc_addr;
1007 connected = 0;
Erich E. Hoover76e21052012-02-08 09:11:07 +00001008 } else if (!ipc.oif)
1009 ipc.oif = inet->uc_index;
David S. Millerdb8dac22008-03-06 16:22:02 -08001010
1011 if (connected)
Eric Dumazetc482c562009-07-17 00:26:32 +00001012 rt = (struct rtable *)sk_dst_check(sk, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001013
Ian Morris51456b22015-04-03 09:17:26 +01001014 if (!rt) {
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001015 struct net *net = sock_net(sk);
David Ahern9a24abf2015-08-13 14:59:03 -06001016 __u8 flow_flags = inet_sk_flowi_flags(sk);
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001017
David S. Millere4749952011-05-08 16:38:45 -07001018 fl4 = &fl4_stack;
David Ahern9a24abf2015-08-13 14:59:03 -06001019
1020 /* unconnected socket. If output device is enslaved to a VRF
1021 * device lookup source address from VRF table. This mimics
1022 * behavior of ip_route_connect{_init}.
1023 */
1024 if (netif_index_is_vrf(net, ipc.oif)) {
1025 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
1026 RT_SCOPE_UNIVERSE, sk->sk_protocol,
David Ahern58189ca2015-09-15 15:10:50 -07001027 (flow_flags | FLOWI_FLAG_VRFSRC |
1028 FLOWI_FLAG_SKIP_NH_OIF),
David Ahern9a24abf2015-08-13 14:59:03 -06001029 faddr, saddr, dport,
1030 inet->inet_sport);
1031
1032 rt = ip_route_output_flow(net, fl4, sk);
1033 if (!IS_ERR(rt)) {
1034 saddr = fl4->saddr;
1035 ip_rt_put(rt);
1036 }
1037 }
1038
David S. Millere4749952011-05-08 16:38:45 -07001039 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
David S. Millerc0951cb2011-03-31 04:54:27 -07001040 RT_SCOPE_UNIVERSE, sk->sk_protocol,
David Ahern9a24abf2015-08-13 14:59:03 -06001041 flow_flags,
David S. Millerc0951cb2011-03-31 04:54:27 -07001042 faddr, saddr, dport, inet->inet_sport);
1043
David S. Millere4749952011-05-08 16:38:45 -07001044 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
1045 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001046 if (IS_ERR(rt)) {
1047 err = PTR_ERR(rt);
David S. Miller06dc94b2011-03-03 10:38:01 -08001048 rt = NULL;
David S. Millerdb8dac22008-03-06 16:22:02 -08001049 if (err == -ENETUNREACH)
Eric Dumazetf1d8cba2013-11-28 09:51:22 -08001050 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -08001051 goto out;
1052 }
1053
1054 err = -EACCES;
1055 if ((rt->rt_flags & RTCF_BROADCAST) &&
1056 !sock_flag(sk, SOCK_BROADCAST))
1057 goto out;
1058 if (connected)
Changli Gaod8d1f302010-06-10 23:31:35 -07001059 sk_dst_set(sk, dst_clone(&rt->dst));
David S. Millerdb8dac22008-03-06 16:22:02 -08001060 }
1061
1062 if (msg->msg_flags&MSG_CONFIRM)
1063 goto do_confirm;
1064back_from_confirm:
1065
David S. Millere4749952011-05-08 16:38:45 -07001066 saddr = fl4->saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001067 if (!ipc.addr)
David S. Millere4749952011-05-08 16:38:45 -07001068 daddr = ipc.addr = fl4->daddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001069
Herbert Xu903ab862011-03-01 02:36:48 +00001070 /* Lockless fast path for the non-corking case. */
1071 if (!corkreq) {
Al Virof69e6d12014-11-24 13:23:40 -05001072 skb = ip_make_skb(sk, fl4, getfrag, msg, ulen,
Herbert Xu903ab862011-03-01 02:36:48 +00001073 sizeof(struct udphdr), &ipc, &rt,
1074 msg->msg_flags);
1075 err = PTR_ERR(skb);
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001076 if (!IS_ERR_OR_NULL(skb))
David S. Miller79ab0532011-05-09 13:31:04 -07001077 err = udp_send_skb(skb, fl4);
Herbert Xu903ab862011-03-01 02:36:48 +00001078 goto out;
1079 }
1080
David S. Millerdb8dac22008-03-06 16:22:02 -08001081 lock_sock(sk);
1082 if (unlikely(up->pending)) {
1083 /* The socket is already corked while preparing it. */
1084 /* ... which is an evident application bug. --ANK */
1085 release_sock(sk);
1086
Joe Perchesba7a46f2014-11-11 10:59:17 -08001087 net_dbg_ratelimited("cork app bug 2\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001088 err = -EINVAL;
1089 goto out;
1090 }
1091 /*
1092 * Now cork the socket to pend data.
1093 */
David S. Millerb6f21b22011-03-12 02:09:18 -05001094 fl4 = &inet->cork.fl.u.ip4;
1095 fl4->daddr = daddr;
1096 fl4->saddr = saddr;
David S. Miller9cce96d2011-03-12 03:00:33 -05001097 fl4->fl4_dport = dport;
1098 fl4->fl4_sport = inet->inet_sport;
David S. Millerdb8dac22008-03-06 16:22:02 -08001099 up->pending = AF_INET;
1100
1101do_append_data:
1102 up->len += ulen;
Al Virof69e6d12014-11-24 13:23:40 -05001103 err = ip_append_data(sk, fl4, getfrag, msg, ulen,
David S. Millerf5fca602011-05-08 17:24:10 -07001104 sizeof(struct udphdr), &ipc, &rt,
1105 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001106 if (err)
1107 udp_flush_pending_frames(sk);
1108 else if (!corkreq)
1109 err = udp_push_pending_frames(sk);
1110 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1111 up->pending = 0;
1112 release_sock(sk);
1113
1114out:
1115 ip_rt_put(rt);
1116 if (free)
1117 kfree(ipc.opt);
1118 if (!err)
1119 return len;
1120 /*
1121 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1122 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1123 * we don't have a good statistic (IpOutDiscards but it can be too many
1124 * things). We could add another new stat but at least for now that
1125 * seems like overkill.
1126 */
1127 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001128 UDP_INC_STATS_USER(sock_net(sk),
1129 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001130 }
1131 return err;
1132
1133do_confirm:
Changli Gaod8d1f302010-06-10 23:31:35 -07001134 dst_confirm(&rt->dst);
David S. Millerdb8dac22008-03-06 16:22:02 -08001135 if (!(msg->msg_flags&MSG_PROBE) || len)
1136 goto back_from_confirm;
1137 err = 0;
1138 goto out;
1139}
Eric Dumazetc482c562009-07-17 00:26:32 +00001140EXPORT_SYMBOL(udp_sendmsg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001141
1142int udp_sendpage(struct sock *sk, struct page *page, int offset,
1143 size_t size, int flags)
1144{
David S. Millerf5fca602011-05-08 17:24:10 -07001145 struct inet_sock *inet = inet_sk(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001146 struct udp_sock *up = udp_sk(sk);
1147 int ret;
1148
Shawn Landdend3f7d562013-11-24 22:36:28 -08001149 if (flags & MSG_SENDPAGE_NOTLAST)
1150 flags |= MSG_MORE;
1151
David S. Millerdb8dac22008-03-06 16:22:02 -08001152 if (!up->pending) {
1153 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
1154
1155 /* Call udp_sendmsg to specify destination address which
1156 * sendpage interface can't pass.
1157 * This will succeed only when the socket is connected.
1158 */
Ying Xue1b784142015-03-02 15:37:48 +08001159 ret = udp_sendmsg(sk, &msg, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001160 if (ret < 0)
1161 return ret;
1162 }
1163
1164 lock_sock(sk);
1165
1166 if (unlikely(!up->pending)) {
1167 release_sock(sk);
1168
Joe Perchesba7a46f2014-11-11 10:59:17 -08001169 net_dbg_ratelimited("udp cork app bug 3\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001170 return -EINVAL;
1171 }
1172
David S. Millerf5fca602011-05-08 17:24:10 -07001173 ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1174 page, offset, size, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001175 if (ret == -EOPNOTSUPP) {
1176 release_sock(sk);
1177 return sock_no_sendpage(sk->sk_socket, page, offset,
1178 size, flags);
1179 }
1180 if (ret < 0) {
1181 udp_flush_pending_frames(sk);
1182 goto out;
1183 }
1184
1185 up->len += size;
1186 if (!(up->corkflag || (flags&MSG_MORE)))
1187 ret = udp_push_pending_frames(sk);
1188 if (!ret)
1189 ret = size;
1190out:
1191 release_sock(sk);
1192 return ret;
1193}
1194
Eric Dumazet85584672009-10-09 04:43:40 +00001195/**
1196 * first_packet_length - return length of first packet in receive queue
1197 * @sk: socket
1198 *
1199 * Drops all bad checksum frames, until a valid one is found.
1200 * Returns the length of found skb, or 0 if none is found.
1201 */
1202static unsigned int first_packet_length(struct sock *sk)
1203{
1204 struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1205 struct sk_buff *skb;
1206 unsigned int res;
1207
1208 __skb_queue_head_init(&list_kill);
1209
1210 spin_lock_bh(&rcvq->lock);
1211 while ((skb = skb_peek(rcvq)) != NULL &&
1212 udp_lib_checksum_complete(skb)) {
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001213 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS,
1214 IS_UDPLITE(sk));
Eric Dumazet85584672009-10-09 04:43:40 +00001215 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1216 IS_UDPLITE(sk));
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001217 atomic_inc(&sk->sk_drops);
Eric Dumazet85584672009-10-09 04:43:40 +00001218 __skb_unlink(skb, rcvq);
1219 __skb_queue_tail(&list_kill, skb);
1220 }
1221 res = skb ? skb->len : 0;
1222 spin_unlock_bh(&rcvq->lock);
1223
1224 if (!skb_queue_empty(&list_kill)) {
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001225 bool slow = lock_sock_fast(sk);
1226
Eric Dumazet85584672009-10-09 04:43:40 +00001227 __skb_queue_purge(&list_kill);
1228 sk_mem_reclaim_partial(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001229 unlock_sock_fast(sk, slow);
Eric Dumazet85584672009-10-09 04:43:40 +00001230 }
1231 return res;
1232}
1233
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234/*
1235 * IOCTL requests applicable to the UDP protocol
1236 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001237
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1239{
Stephen Hemminger6516c652007-03-08 20:41:55 -08001240 switch (cmd) {
1241 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 {
Eric Dumazet31e6d362009-06-17 19:05:41 -07001243 int amount = sk_wmem_alloc_get(sk);
1244
Stephen Hemminger6516c652007-03-08 20:41:55 -08001245 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 }
Stephen Hemminger6516c652007-03-08 20:41:55 -08001247
1248 case SIOCINQ:
1249 {
Eric Dumazet85584672009-10-09 04:43:40 +00001250 unsigned int amount = first_packet_length(sk);
Stephen Hemminger6516c652007-03-08 20:41:55 -08001251
Eric Dumazet85584672009-10-09 04:43:40 +00001252 if (amount)
Stephen Hemminger6516c652007-03-08 20:41:55 -08001253 /*
1254 * We will only return the amount
1255 * of this packet since that is all
1256 * that will be read.
1257 */
Eric Dumazet85584672009-10-09 04:43:40 +00001258 amount -= sizeof(struct udphdr);
1259
Stephen Hemminger6516c652007-03-08 20:41:55 -08001260 return put_user(amount, (int __user *)arg);
1261 }
1262
1263 default:
1264 return -ENOIOCTLCMD;
1265 }
1266
1267 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268}
Eric Dumazetc482c562009-07-17 00:26:32 +00001269EXPORT_SYMBOL(udp_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270
David S. Millerdb8dac22008-03-06 16:22:02 -08001271/*
1272 * This should be easy, if there is something there we
1273 * return it, otherwise we block.
1274 */
1275
Ying Xue1b784142015-03-02 15:37:48 +08001276int udp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
1277 int flags, int *addr_len)
David S. Millerdb8dac22008-03-06 16:22:02 -08001278{
1279 struct inet_sock *inet = inet_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001280 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -08001281 struct sk_buff *skb;
David S. Miller59c2cda2011-12-01 14:12:55 -05001282 unsigned int ulen, copied;
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001283 int peeked, off = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001284 int err;
1285 int is_udplite = IS_UDPLITE(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001286 bool slow;
David S. Millerdb8dac22008-03-06 16:22:02 -08001287
David S. Millerdb8dac22008-03-06 16:22:02 -08001288 if (flags & MSG_ERRQUEUE)
Hannes Frederic Sowa85fbaa72013-11-23 00:46:12 +01001289 return ip_recv_error(sk, msg, len, addr_len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001290
1291try_again:
1292 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001293 &peeked, &off, &err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001294 if (!skb)
1295 goto out;
1296
1297 ulen = skb->len - sizeof(struct udphdr);
David S. Miller59c2cda2011-12-01 14:12:55 -05001298 copied = len;
1299 if (copied > ulen)
1300 copied = ulen;
1301 else if (copied < ulen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001302 msg->msg_flags |= MSG_TRUNC;
1303
1304 /*
1305 * If checksum is needed at all, try to do it while copying the
1306 * data. If the data is truncated, or if we only want a partial
1307 * coverage checksum (UDP-Lite), do it before the copy.
1308 */
1309
David S. Miller59c2cda2011-12-01 14:12:55 -05001310 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001311 if (udp_lib_checksum_complete(skb))
1312 goto csum_copy_err;
1313 }
1314
1315 if (skb_csum_unnecessary(skb))
David S. Miller51f3d022014-11-05 16:46:40 -05001316 err = skb_copy_datagram_msg(skb, sizeof(struct udphdr),
1317 msg, copied);
David S. Millerdb8dac22008-03-06 16:22:02 -08001318 else {
Al Viro227158d2014-04-06 18:47:38 -04001319 err = skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr),
1320 msg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001321
1322 if (err == -EINVAL)
1323 goto csum_copy_err;
1324 }
1325
Eric Dumazet22911fc2012-06-27 00:23:44 +00001326 if (unlikely(err)) {
1327 trace_kfree_skb(skb, udp_recvmsg);
Eric Dumazet979402b2012-09-05 23:34:44 +00001328 if (!peeked) {
1329 atomic_inc(&sk->sk_drops);
1330 UDP_INC_STATS_USER(sock_net(sk),
1331 UDP_MIB_INERRORS, is_udplite);
1332 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001333 goto out_free;
Eric Dumazet22911fc2012-06-27 00:23:44 +00001334 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001335
1336 if (!peeked)
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001337 UDP_INC_STATS_USER(sock_net(sk),
1338 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001339
Neil Horman3b885782009-10-12 13:26:31 -07001340 sock_recv_ts_and_drops(msg, sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001341
1342 /* Copy the address. */
Eric Dumazetc482c562009-07-17 00:26:32 +00001343 if (sin) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001344 sin->sin_family = AF_INET;
1345 sin->sin_port = udp_hdr(skb)->source;
1346 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1347 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
Hannes Frederic Sowabceaa902013-11-18 04:20:45 +01001348 *addr_len = sizeof(*sin);
David S. Millerdb8dac22008-03-06 16:22:02 -08001349 }
1350 if (inet->cmsg_flags)
Tom Herbertad6f9392015-01-05 13:56:17 -08001351 ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr));
David S. Millerdb8dac22008-03-06 16:22:02 -08001352
David S. Miller59c2cda2011-12-01 14:12:55 -05001353 err = copied;
David S. Millerdb8dac22008-03-06 16:22:02 -08001354 if (flags & MSG_TRUNC)
1355 err = ulen;
1356
1357out_free:
Eric Dumazet9d410c72009-10-30 05:03:53 +00001358 skb_free_datagram_locked(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001359out:
1360 return err;
1361
1362csum_copy_err:
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001363 slow = lock_sock_fast(sk);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001364 if (!skb_kill_datagram(sk, skb, flags)) {
1365 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001366 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001367 }
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001368 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001369
Eric Dumazetbeb39db2015-05-30 09:16:53 -07001370 /* starting over for a new packet, but check if we need to yield */
1371 cond_resched();
Xufeng Zhang9cfaa8d2011-06-21 10:43:40 +00001372 msg->msg_flags &= ~MSG_TRUNC;
David S. Millerdb8dac22008-03-06 16:22:02 -08001373 goto try_again;
1374}
1375
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376int udp_disconnect(struct sock *sk, int flags)
1377{
1378 struct inet_sock *inet = inet_sk(sk);
1379 /*
1380 * 1003.1g - break association.
1381 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001382
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 sk->sk_state = TCP_CLOSE;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001384 inet->inet_daddr = 0;
1385 inet->inet_dport = 0;
Tom Herbertbdeab992011-08-14 19:45:55 +00001386 sock_rps_reset_rxhash(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 sk->sk_bound_dev_if = 0;
1388 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1389 inet_reset_saddr(sk);
1390
1391 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1392 sk->sk_prot->unhash(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001393 inet->inet_sport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 }
1395 sk_dst_reset(sk);
1396 return 0;
1397}
Eric Dumazetc482c562009-07-17 00:26:32 +00001398EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399
Eric Dumazet645ca702008-10-29 01:41:45 -07001400void udp_lib_unhash(struct sock *sk)
1401{
Eric Dumazet723b4612008-11-25 13:55:15 -08001402 if (sk_hashed(sk)) {
1403 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Eric Dumazet512615b2009-11-08 10:17:58 +00001404 struct udp_hslot *hslot, *hslot2;
1405
1406 hslot = udp_hashslot(udptable, sock_net(sk),
1407 udp_sk(sk)->udp_port_hash);
1408 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
Eric Dumazet645ca702008-10-29 01:41:45 -07001409
Eric Dumazet723b4612008-11-25 13:55:15 -08001410 spin_lock_bh(&hslot->lock);
1411 if (sk_nulls_del_node_init_rcu(sk)) {
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00001412 hslot->count--;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001413 inet_sk(sk)->inet_num = 0;
Eric Dumazet723b4612008-11-25 13:55:15 -08001414 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet512615b2009-11-08 10:17:58 +00001415
1416 spin_lock(&hslot2->lock);
1417 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1418 hslot2->count--;
1419 spin_unlock(&hslot2->lock);
Eric Dumazet723b4612008-11-25 13:55:15 -08001420 }
1421 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001422 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001423}
1424EXPORT_SYMBOL(udp_lib_unhash);
1425
Eric Dumazet719f8352010-09-08 05:08:44 +00001426/*
1427 * inet_rcv_saddr was changed, we must rehash secondary hash
1428 */
1429void udp_lib_rehash(struct sock *sk, u16 newhash)
1430{
1431 if (sk_hashed(sk)) {
1432 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1433 struct udp_hslot *hslot, *hslot2, *nhslot2;
1434
1435 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1436 nhslot2 = udp_hashslot2(udptable, newhash);
1437 udp_sk(sk)->udp_portaddr_hash = newhash;
1438 if (hslot2 != nhslot2) {
1439 hslot = udp_hashslot(udptable, sock_net(sk),
1440 udp_sk(sk)->udp_port_hash);
1441 /* we must lock primary chain too */
1442 spin_lock_bh(&hslot->lock);
1443
1444 spin_lock(&hslot2->lock);
1445 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1446 hslot2->count--;
1447 spin_unlock(&hslot2->lock);
1448
1449 spin_lock(&nhslot2->lock);
1450 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
1451 &nhslot2->head);
1452 nhslot2->count++;
1453 spin_unlock(&nhslot2->lock);
1454
1455 spin_unlock_bh(&hslot->lock);
1456 }
1457 }
1458}
1459EXPORT_SYMBOL(udp_lib_rehash);
1460
1461static void udp_v4_rehash(struct sock *sk)
1462{
1463 u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1464 inet_sk(sk)->inet_rcv_saddr,
1465 inet_sk(sk)->inet_num);
1466 udp_lib_rehash(sk, new_hash);
1467}
1468
Herbert Xu93821772008-09-15 11:48:46 -07001469static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1470{
Tom Herbertfec5e652010-04-16 16:01:27 -07001471 int rc;
Herbert Xu93821772008-09-15 11:48:46 -07001472
Shawn Bohrer005ec972013-10-07 11:01:38 -05001473 if (inet_sk(sk)->inet_daddr) {
Tom Herbertbdeab992011-08-14 19:45:55 +00001474 sock_rps_save_rxhash(sk, skb);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001475 sk_mark_napi_id(sk, skb);
Eric Dumazet2c8c56e2014-11-11 05:54:28 -08001476 sk_incoming_cpu_update(sk);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001477 }
Tom Herbertfec5e652010-04-16 16:01:27 -07001478
Eric Dumazetd826eb12011-11-09 07:24:35 +00001479 rc = sock_queue_rcv_skb(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -07001480 if (rc < 0) {
1481 int is_udplite = IS_UDPLITE(sk);
1482
Herbert Xu93821772008-09-15 11:48:46 -07001483 /* Note that an ENOMEM error is charged twice */
Eric Dumazet766e90372009-10-14 20:40:11 -07001484 if (rc == -ENOMEM)
Herbert Xu93821772008-09-15 11:48:46 -07001485 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1486 is_udplite);
Eric Dumazet766e90372009-10-14 20:40:11 -07001487 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1488 kfree_skb(skb);
Satoru Moriya296f7ea2011-06-17 11:58:39 +00001489 trace_udp_fail_queue_rcv_skb(rc, sk);
Eric Dumazet766e90372009-10-14 20:40:11 -07001490 return -1;
Herbert Xu93821772008-09-15 11:48:46 -07001491 }
1492
1493 return 0;
1494
Herbert Xu93821772008-09-15 11:48:46 -07001495}
1496
Eric Dumazet447167b2012-04-11 23:05:28 +00001497static struct static_key udp_encap_needed __read_mostly;
1498void udp_encap_enable(void)
1499{
1500 if (!static_key_enabled(&udp_encap_needed))
1501 static_key_slow_inc(&udp_encap_needed);
1502}
1503EXPORT_SYMBOL(udp_encap_enable);
1504
David S. Millerdb8dac22008-03-06 16:22:02 -08001505/* returns:
1506 * -1: error
1507 * 0: success
1508 * >0: "udp encap" protocol resubmission
1509 *
1510 * Note that in the success and error cases, the skb is assumed to
1511 * have either been requeued or freed.
1512 */
Eric Dumazetc482c562009-07-17 00:26:32 +00001513int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -08001514{
1515 struct udp_sock *up = udp_sk(sk);
1516 int rc;
1517 int is_udplite = IS_UDPLITE(sk);
1518
1519 /*
1520 * Charge it to the socket, dropping if the queue is full.
1521 */
1522 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1523 goto drop;
1524 nf_reset(skb);
1525
Eric Dumazet447167b2012-04-11 23:05:28 +00001526 if (static_key_false(&udp_encap_needed) && up->encap_type) {
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001527 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1528
David S. Millerdb8dac22008-03-06 16:22:02 -08001529 /*
1530 * This is an encapsulation socket so pass the skb to
1531 * the socket's udp_encap_rcv() hook. Otherwise, just
1532 * fall through and pass this up the UDP socket.
1533 * up->encap_rcv() returns the following value:
1534 * =0 if skb was successfully passed to the encap
1535 * handler or was discarded by it.
1536 * >0 if skb should be passed on to UDP.
1537 * <0 if skb should be resubmitted as proto -N
1538 */
1539
1540 /* if we're overly short, let UDP handle it */
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001541 encap_rcv = ACCESS_ONCE(up->encap_rcv);
Ian Morris00db4122015-04-03 09:17:27 +01001542 if (skb->len > sizeof(struct udphdr) && encap_rcv) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001543 int ret;
1544
Tom Herbert0a809662014-05-07 16:52:39 -07001545 /* Verify checksum before giving to encap */
1546 if (udp_lib_checksum_complete(skb))
1547 goto csum_error;
1548
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001549 ret = encap_rcv(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001550 if (ret <= 0) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001551 UDP_INC_STATS_BH(sock_net(sk),
1552 UDP_MIB_INDATAGRAMS,
David S. Millerdb8dac22008-03-06 16:22:02 -08001553 is_udplite);
1554 return -ret;
1555 }
1556 }
1557
1558 /* FALLTHROUGH -- it's a UDP Packet */
1559 }
1560
1561 /*
1562 * UDP-Lite specific tests, ignored on UDP sockets
1563 */
1564 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1565
1566 /*
1567 * MIB statistics other than incrementing the error count are
1568 * disabled for the following two types of errors: these depend
1569 * on the application settings, not on the functioning of the
1570 * protocol stack as such.
1571 *
1572 * RFC 3828 here recommends (sec 3.3): "There should also be a
1573 * way ... to ... at least let the receiving application block
1574 * delivery of packets with coverage values less than a value
1575 * provided by the application."
1576 */
1577 if (up->pcrlen == 0) { /* full coverage was set */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001578 net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n",
1579 UDP_SKB_CB(skb)->cscov, skb->len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001580 goto drop;
1581 }
1582 /* The next case involves violating the min. coverage requested
1583 * by the receiver. This is subtle: if receiver wants x and x is
1584 * greater than the buffersize/MTU then receiver will complain
1585 * that it wants x while sender emits packets of smaller size y.
1586 * Therefore the above ...()->partial_cov statement is essential.
1587 */
1588 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
Joe Perchesba7a46f2014-11-11 10:59:17 -08001589 net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n",
1590 UDP_SKB_CB(skb)->cscov, up->pcrlen);
David S. Millerdb8dac22008-03-06 16:22:02 -08001591 goto drop;
1592 }
1593 }
1594
Eric Dumazet33d480c2011-08-11 19:30:52 +00001595 if (rcu_access_pointer(sk->sk_filter) &&
1596 udp_lib_checksum_complete(skb))
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001597 goto csum_error;
David S. Millerdb8dac22008-03-06 16:22:02 -08001598
Sorin Dumitru274f4822014-07-22 21:16:51 +03001599 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
James M Leddy3e215c82014-06-25 17:38:13 -04001600 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1601 is_udplite);
Eric Dumazetc3774112010-04-27 15:13:20 -07001602 goto drop;
James M Leddy3e215c82014-06-25 17:38:13 -04001603 }
Eric Dumazetc3774112010-04-27 15:13:20 -07001604
Herbert Xu93821772008-09-15 11:48:46 -07001605 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001606
Shawn Bohrerfbf88662013-10-07 11:01:40 -05001607 ipv4_pktinfo_prepare(sk, skb);
Herbert Xu93821772008-09-15 11:48:46 -07001608 bh_lock_sock(sk);
1609 if (!sock_owned_by_user(sk))
1610 rc = __udp_queue_rcv_skb(sk, skb);
Eric Dumazetf545a382012-04-22 23:34:26 +00001611 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
Zhu Yi55349792010-03-04 18:01:42 +00001612 bh_unlock_sock(sk);
1613 goto drop;
1614 }
Herbert Xu93821772008-09-15 11:48:46 -07001615 bh_unlock_sock(sk);
1616
1617 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001618
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001619csum_error:
1620 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001621drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001622 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001623 atomic_inc(&sk->sk_drops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001624 kfree_skb(skb);
1625 return -1;
1626}
1627
Eric Dumazet1240d132009-11-08 10:18:44 +00001628static void flush_stack(struct sock **stack, unsigned int count,
1629 struct sk_buff *skb, unsigned int final)
1630{
1631 unsigned int i;
1632 struct sk_buff *skb1 = NULL;
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001633 struct sock *sk;
Eric Dumazet1240d132009-11-08 10:18:44 +00001634
1635 for (i = 0; i < count; i++) {
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001636 sk = stack[i];
Ian Morris51456b22015-04-03 09:17:26 +01001637 if (likely(!skb1))
Eric Dumazet1240d132009-11-08 10:18:44 +00001638 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
1639
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001640 if (!skb1) {
1641 atomic_inc(&sk->sk_drops);
1642 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1643 IS_UDPLITE(sk));
1644 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1645 IS_UDPLITE(sk));
1646 }
1647
1648 if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0)
Eric Dumazet1240d132009-11-08 10:18:44 +00001649 skb1 = NULL;
David Held2dc41cf2014-07-15 23:28:32 -04001650
1651 sock_put(sk);
Eric Dumazet1240d132009-11-08 10:18:44 +00001652 }
1653 if (unlikely(skb1))
1654 kfree_skb(skb1);
1655}
1656
Eric Dumazet97502232013-12-11 14:46:51 -08001657/* For TCP sockets, sk_rx_dst is protected by socket lock
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001658 * For UDP, we use xchg() to guard against concurrent changes.
Eric Dumazet97502232013-12-11 14:46:51 -08001659 */
1660static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
Shawn Bohrer421b3882013-10-07 11:01:39 -05001661{
Eric Dumazet97502232013-12-11 14:46:51 -08001662 struct dst_entry *old;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001663
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001664 dst_hold(dst);
1665 old = xchg(&sk->sk_rx_dst, dst);
1666 dst_release(old);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001667}
1668
David S. Millerdb8dac22008-03-06 16:22:02 -08001669/*
1670 * Multicasts and broadcasts go to each listener.
1671 *
Eric Dumazet1240d132009-11-08 10:18:44 +00001672 * Note: called only from the BH handler context.
David S. Millerdb8dac22008-03-06 16:22:02 -08001673 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001674static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001675 struct udphdr *uh,
1676 __be32 saddr, __be32 daddr,
Rick Jones36cbb242014-11-06 10:37:54 -08001677 struct udp_table *udptable,
1678 int proto)
David S. Millerdb8dac22008-03-06 16:22:02 -08001679{
Eric Dumazet1240d132009-11-08 10:18:44 +00001680 struct sock *sk, *stack[256 / sizeof(struct sock *)];
David Held5cf3d462014-07-15 23:28:31 -04001681 struct hlist_nulls_node *node;
1682 unsigned short hnum = ntohs(uh->dest);
1683 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
1684 int dif = skb->dev->ifindex;
David Held2dc41cf2014-07-15 23:28:32 -04001685 unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node);
1686 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
Rick Jones36cbb242014-11-06 10:37:54 -08001687 bool inner_flushed = false;
David Held2dc41cf2014-07-15 23:28:32 -04001688
1689 if (use_hash2) {
1690 hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) &
1691 udp_table.mask;
1692 hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask;
1693start_lookup:
1694 hslot = &udp_table.hash2[hash2];
1695 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
1696 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001697
Eric Dumazet645ca702008-10-29 01:41:45 -07001698 spin_lock(&hslot->lock);
David Held2dc41cf2014-07-15 23:28:32 -04001699 sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) {
David Held5cf3d462014-07-15 23:28:31 -04001700 if (__udp_is_mcast_sock(net, sk,
1701 uh->dest, daddr,
1702 uh->source, saddr,
1703 dif, hnum)) {
1704 if (unlikely(count == ARRAY_SIZE(stack))) {
1705 flush_stack(stack, count, skb, ~0);
Rick Jones36cbb242014-11-06 10:37:54 -08001706 inner_flushed = true;
David Held5cf3d462014-07-15 23:28:31 -04001707 count = 0;
1708 }
1709 stack[count++] = sk;
David Held2dc41cf2014-07-15 23:28:32 -04001710 sock_hold(sk);
Eric Dumazet1240d132009-11-08 10:18:44 +00001711 }
1712 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001713
Eric Dumazet645ca702008-10-29 01:41:45 -07001714 spin_unlock(&hslot->lock);
Eric Dumazet1240d132009-11-08 10:18:44 +00001715
David Held2dc41cf2014-07-15 23:28:32 -04001716 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
1717 if (use_hash2 && hash2 != hash2_any) {
1718 hash2 = hash2_any;
1719 goto start_lookup;
1720 }
1721
Eric Dumazet1240d132009-11-08 10:18:44 +00001722 /*
1723 * do the slow work with no lock held
1724 */
1725 if (count) {
1726 flush_stack(stack, count, skb, count - 1);
Eric Dumazet1240d132009-11-08 10:18:44 +00001727 } else {
Rick Jones36cbb242014-11-06 10:37:54 -08001728 if (!inner_flushed)
1729 UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
1730 proto == IPPROTO_UDPLITE);
1731 consume_skb(skb);
Eric Dumazet1240d132009-11-08 10:18:44 +00001732 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001733 return 0;
1734}
1735
1736/* Initialize UDP checksum. If exited with zero value (success),
1737 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1738 * Otherwise, csum completion requires chacksumming packet body,
1739 * including udp header and folding it to skb->csum.
1740 */
1741static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1742 int proto)
1743{
David S. Millerdb8dac22008-03-06 16:22:02 -08001744 int err;
1745
1746 UDP_SKB_CB(skb)->partial_cov = 0;
1747 UDP_SKB_CB(skb)->cscov = skb->len;
1748
1749 if (proto == IPPROTO_UDPLITE) {
1750 err = udplite_checksum_init(skb, uh);
1751 if (err)
1752 return err;
1753 }
1754
Tom Herberted70fcf2014-05-02 16:29:38 -07001755 return skb_checksum_init_zero_check(skb, proto, uh->check,
1756 inet_compute_pseudo);
David S. Millerdb8dac22008-03-06 16:22:02 -08001757}
1758
1759/*
1760 * All we need to do is get the socket, and then do a checksum.
1761 */
1762
Eric Dumazet645ca702008-10-29 01:41:45 -07001763int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001764 int proto)
1765{
1766 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001767 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001768 unsigned short ulen;
Eric Dumazetadf30902009-06-02 05:19:30 +00001769 struct rtable *rt = skb_rtable(skb);
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001770 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001771 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001772
1773 /*
1774 * Validate the packet.
1775 */
1776 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1777 goto drop; /* No space for header. */
1778
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001779 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001780 ulen = ntohs(uh->len);
Bjørn Morkccc2d972010-05-06 03:44:34 +00001781 saddr = ip_hdr(skb)->saddr;
1782 daddr = ip_hdr(skb)->daddr;
1783
David S. Millerdb8dac22008-03-06 16:22:02 -08001784 if (ulen > skb->len)
1785 goto short_packet;
1786
1787 if (proto == IPPROTO_UDP) {
1788 /* UDP validates ulen. */
1789 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1790 goto short_packet;
1791 uh = udp_hdr(skb);
1792 }
1793
1794 if (udp4_csum_init(skb, uh, proto))
1795 goto csum_error;
1796
Eric Dumazet8afdd992013-12-10 18:07:23 -08001797 sk = skb_steal_sock(skb);
1798 if (sk) {
Eric Dumazet97502232013-12-11 14:46:51 -08001799 struct dst_entry *dst = skb_dst(skb);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001800 int ret;
David S. Millerdb8dac22008-03-06 16:22:02 -08001801
Eric Dumazet97502232013-12-11 14:46:51 -08001802 if (unlikely(sk->sk_rx_dst != dst))
1803 udp_sk_rx_dst_set(sk, dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001804
1805 ret = udp_queue_rcv_skb(sk, skb);
Eric Dumazet8afdd992013-12-10 18:07:23 -08001806 sock_put(sk);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001807 /* a return value > 0 means to resubmit the input, but
1808 * it wants the return to be -protocol, or 0
1809 */
1810 if (ret > 0)
1811 return -ret;
1812 return 0;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001813 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001814
Fabian Frederickc18450a2014-11-04 20:48:41 +01001815 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1816 return __udp4_lib_mcast_deliver(net, skb, uh,
Rick Jones36cbb242014-11-06 10:37:54 -08001817 saddr, daddr, udptable, proto);
Fabian Frederickc18450a2014-11-04 20:48:41 +01001818
1819 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
Ian Morris00db4122015-04-03 09:17:27 +01001820 if (sk) {
Eliezer Tamira5b50472013-06-10 11:40:00 +03001821 int ret;
1822
Tom Herbert224d0192015-01-05 13:56:14 -08001823 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
Tom Herbert2abb7cd2014-08-31 15:12:43 -07001824 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
1825 inet_compute_pseudo);
1826
Eliezer Tamira5b50472013-06-10 11:40:00 +03001827 ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001828 sock_put(sk);
1829
1830 /* a return value > 0 means to resubmit the input, but
1831 * it wants the return to be -protocol, or 0
1832 */
1833 if (ret > 0)
1834 return -ret;
1835 return 0;
1836 }
1837
1838 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1839 goto drop;
1840 nf_reset(skb);
1841
1842 /* No socket. Drop packet silently, if checksum is wrong */
1843 if (udp_lib_checksum_complete(skb))
1844 goto csum_error;
1845
Pavel Emelyanov02833282008-07-05 21:18:48 -07001846 UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001847 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1848
1849 /*
1850 * Hmm. We got an UDP packet to a port to which we
1851 * don't wanna listen. Ignore it.
1852 */
1853 kfree_skb(skb);
1854 return 0;
1855
1856short_packet:
Joe Perchesba7a46f2014-11-11 10:59:17 -08001857 net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
1858 proto == IPPROTO_UDPLITE ? "Lite" : "",
1859 &saddr, ntohs(uh->source),
1860 ulen, skb->len,
1861 &daddr, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001862 goto drop;
1863
1864csum_error:
1865 /*
1866 * RFC1122: OK. Discards the bad packet silently (as far as
1867 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1868 */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001869 net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
1870 proto == IPPROTO_UDPLITE ? "Lite" : "",
1871 &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
1872 ulen);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001873 UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001874drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001875 UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001876 kfree_skb(skb);
1877 return 0;
1878}
1879
Shawn Bohrer421b3882013-10-07 11:01:39 -05001880/* We can only early demux multicast if there is a single matching socket.
1881 * If more than one socket found returns NULL
1882 */
1883static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net,
1884 __be16 loc_port, __be32 loc_addr,
1885 __be16 rmt_port, __be32 rmt_addr,
1886 int dif)
1887{
1888 struct sock *sk, *result;
1889 struct hlist_nulls_node *node;
1890 unsigned short hnum = ntohs(loc_port);
1891 unsigned int count, slot = udp_hashfn(net, hnum, udp_table.mask);
1892 struct udp_hslot *hslot = &udp_table.hash[slot];
1893
Eric Dumazet63c6f812014-06-12 16:13:06 -07001894 /* Do not bother scanning a too big list */
1895 if (hslot->count > 10)
1896 return NULL;
1897
Shawn Bohrer421b3882013-10-07 11:01:39 -05001898 rcu_read_lock();
1899begin:
1900 count = 0;
1901 result = NULL;
1902 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
1903 if (__udp_is_mcast_sock(net, sk,
1904 loc_port, loc_addr,
1905 rmt_port, rmt_addr,
1906 dif, hnum)) {
1907 result = sk;
1908 ++count;
1909 }
1910 }
1911 /*
1912 * if the nulls value we got at the end of this lookup is
1913 * not the expected one, we must restart lookup.
1914 * We probably met an item that was moved to another chain.
1915 */
1916 if (get_nulls_value(node) != slot)
1917 goto begin;
1918
1919 if (result) {
1920 if (count != 1 ||
1921 unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
1922 result = NULL;
Eric Dumazetf69b9232013-10-08 21:47:29 -07001923 else if (unlikely(!__udp_is_mcast_sock(net, result,
Shawn Bohrer421b3882013-10-07 11:01:39 -05001924 loc_port, loc_addr,
1925 rmt_port, rmt_addr,
1926 dif, hnum))) {
1927 sock_put(result);
1928 result = NULL;
1929 }
1930 }
1931 rcu_read_unlock();
1932 return result;
1933}
1934
1935/* For unicast we should only early demux connected sockets or we can
1936 * break forwarding setups. The chains here can be long so only check
1937 * if the first socket is an exact match and if not move on.
1938 */
1939static struct sock *__udp4_lib_demux_lookup(struct net *net,
1940 __be16 loc_port, __be32 loc_addr,
1941 __be16 rmt_port, __be32 rmt_addr,
1942 int dif)
1943{
1944 struct sock *sk, *result;
1945 struct hlist_nulls_node *node;
1946 unsigned short hnum = ntohs(loc_port);
1947 unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum);
1948 unsigned int slot2 = hash2 & udp_table.mask;
1949 struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
Joe Perchesc7228312014-05-13 20:30:07 -07001950 INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001951 const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
1952
1953 rcu_read_lock();
1954 result = NULL;
1955 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
1956 if (INET_MATCH(sk, net, acookie,
1957 rmt_addr, loc_addr, ports, dif))
1958 result = sk;
1959 /* Only check first socket in chain */
1960 break;
1961 }
1962
1963 if (result) {
1964 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
1965 result = NULL;
1966 else if (unlikely(!INET_MATCH(sk, net, acookie,
1967 rmt_addr, loc_addr,
1968 ports, dif))) {
1969 sock_put(result);
1970 result = NULL;
1971 }
1972 }
1973 rcu_read_unlock();
1974 return result;
1975}
1976
1977void udp_v4_early_demux(struct sk_buff *skb)
1978{
Eric Dumazet610438b2013-12-11 08:10:05 -08001979 struct net *net = dev_net(skb->dev);
1980 const struct iphdr *iph;
1981 const struct udphdr *uh;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001982 struct sock *sk;
1983 struct dst_entry *dst;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001984 int dif = skb->dev->ifindex;
Shawn Bohrer6e540302015-06-03 16:27:38 -05001985 int ours;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001986
1987 /* validate the packet */
1988 if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr)))
1989 return;
1990
Eric Dumazet610438b2013-12-11 08:10:05 -08001991 iph = ip_hdr(skb);
1992 uh = udp_hdr(skb);
1993
Shawn Bohrer421b3882013-10-07 11:01:39 -05001994 if (skb->pkt_type == PACKET_BROADCAST ||
Shawn Bohrer6e540302015-06-03 16:27:38 -05001995 skb->pkt_type == PACKET_MULTICAST) {
1996 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
1997
1998 if (!in_dev)
1999 return;
2000
2001 ours = ip_check_mc_rcu(in_dev, iph->daddr, iph->saddr,
2002 iph->protocol);
2003 if (!ours)
2004 return;
Shawn Bohrer421b3882013-10-07 11:01:39 -05002005 sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr,
2006 uh->source, iph->saddr, dif);
Shawn Bohrer6e540302015-06-03 16:27:38 -05002007 } else if (skb->pkt_type == PACKET_HOST) {
Shawn Bohrer421b3882013-10-07 11:01:39 -05002008 sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr,
2009 uh->source, iph->saddr, dif);
Shawn Bohrer6e540302015-06-03 16:27:38 -05002010 } else {
Shawn Bohrer421b3882013-10-07 11:01:39 -05002011 return;
Shawn Bohrer6e540302015-06-03 16:27:38 -05002012 }
Shawn Bohrer421b3882013-10-07 11:01:39 -05002013
2014 if (!sk)
2015 return;
2016
2017 skb->sk = sk;
Alexander Duyck82eabd92014-09-04 13:32:11 -04002018 skb->destructor = sock_efree;
Eric Dumazet10e2eb82015-08-01 12:14:33 +02002019 dst = READ_ONCE(sk->sk_rx_dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05002020
2021 if (dst)
2022 dst = dst_check(dst, 0);
Eric Dumazet10e2eb82015-08-01 12:14:33 +02002023 if (dst) {
2024 /* DST_NOCACHE can not be used without taking a reference */
2025 if (dst->flags & DST_NOCACHE) {
2026 if (likely(atomic_inc_not_zero(&dst->__refcnt)))
2027 skb_dst_set(skb, dst);
2028 } else {
2029 skb_dst_set_noref(skb, dst);
2030 }
2031 }
Shawn Bohrer421b3882013-10-07 11:01:39 -05002032}
2033
David S. Millerdb8dac22008-03-06 16:22:02 -08002034int udp_rcv(struct sk_buff *skb)
2035{
Eric Dumazet645ca702008-10-29 01:41:45 -07002036 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08002037}
2038
Brian Haley7d06b2e2008-06-14 17:04:49 -07002039void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08002040{
Tom Parkin44046a52013-03-19 06:11:12 +00002041 struct udp_sock *up = udp_sk(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00002042 bool slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08002043 udp_flush_pending_frames(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00002044 unlock_sock_fast(sk, slow);
Tom Parkin44046a52013-03-19 06:11:12 +00002045 if (static_key_false(&udp_encap_needed) && up->encap_type) {
2046 void (*encap_destroy)(struct sock *sk);
2047 encap_destroy = ACCESS_ONCE(up->encap_destroy);
2048 if (encap_destroy)
2049 encap_destroy(sk);
2050 }
David S. Millerdb8dac22008-03-06 16:22:02 -08002051}
2052
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053/*
2054 * Socket option code for UDP
2055 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002056int udp_lib_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002057 char __user *optval, unsigned int optlen,
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002058 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059{
2060 struct udp_sock *up = udp_sk(sk);
Tom Herbert1c194482014-05-23 08:47:32 -07002061 int val, valbool;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11002063 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064
Eric Dumazetc482c562009-07-17 00:26:32 +00002065 if (optlen < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066 return -EINVAL;
2067
2068 if (get_user(val, (int __user *)optval))
2069 return -EFAULT;
2070
Tom Herbert1c194482014-05-23 08:47:32 -07002071 valbool = val ? 1 : 0;
2072
Stephen Hemminger6516c652007-03-08 20:41:55 -08002073 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074 case UDP_CORK:
2075 if (val != 0) {
2076 up->corkflag = 1;
2077 } else {
2078 up->corkflag = 0;
2079 lock_sock(sk);
Joe Perches4243cdc2014-11-11 21:59:20 -08002080 push_pending_frames(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081 release_sock(sk);
2082 }
2083 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002084
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085 case UDP_ENCAP:
2086 switch (val) {
2087 case 0:
2088 case UDP_ENCAP_ESPINUDP:
2089 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07002090 up->encap_rcv = xfrm4_udp_encap_rcv;
2091 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07002092 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093 up->encap_type = val;
Eric Dumazet447167b2012-04-11 23:05:28 +00002094 udp_encap_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 break;
2096 default:
2097 err = -ENOPROTOOPT;
2098 break;
2099 }
2100 break;
2101
Tom Herbert1c194482014-05-23 08:47:32 -07002102 case UDP_NO_CHECK6_TX:
2103 up->no_check6_tx = valbool;
2104 break;
2105
2106 case UDP_NO_CHECK6_RX:
2107 up->no_check6_rx = valbool;
2108 break;
2109
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002110 /*
2111 * UDP-Lite's partial checksum coverage (RFC 3828).
2112 */
2113 /* The sender sets actual checksum coverage length via this option.
2114 * The case coverage > packet length is handled by send module. */
2115 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002116 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002117 return -ENOPROTOOPT;
2118 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
2119 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002120 else if (val > USHRT_MAX)
2121 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002122 up->pcslen = val;
2123 up->pcflag |= UDPLITE_SEND_CC;
2124 break;
2125
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002126 /* The receiver specifies a minimum checksum coverage value. To make
2127 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002128 * used, this again means full checksum coverage. */
2129 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002130 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002131 return -ENOPROTOOPT;
2132 if (val != 0 && val < 8) /* Avoid silly minimal values. */
2133 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002134 else if (val > USHRT_MAX)
2135 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002136 up->pcrlen = val;
2137 up->pcflag |= UDPLITE_RECV_CC;
2138 break;
2139
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140 default:
2141 err = -ENOPROTOOPT;
2142 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002143 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144
2145 return err;
2146}
Eric Dumazetc482c562009-07-17 00:26:32 +00002147EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148
David S. Millerdb8dac22008-03-06 16:22:02 -08002149int udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002150 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002151{
2152 if (level == SOL_UDP || level == SOL_UDPLITE)
2153 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2154 udp_push_pending_frames);
2155 return ip_setsockopt(sk, level, optname, optval, optlen);
2156}
2157
2158#ifdef CONFIG_COMPAT
2159int compat_udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002160 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002161{
2162 if (level == SOL_UDP || level == SOL_UDPLITE)
2163 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2164 udp_push_pending_frames);
2165 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
2166}
2167#endif
2168
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002169int udp_lib_getsockopt(struct sock *sk, int level, int optname,
2170 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171{
2172 struct udp_sock *up = udp_sk(sk);
2173 int val, len;
2174
Eric Dumazetc482c562009-07-17 00:26:32 +00002175 if (get_user(len, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176 return -EFAULT;
2177
2178 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002179
Stephen Hemminger6516c652007-03-08 20:41:55 -08002180 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181 return -EINVAL;
2182
Stephen Hemminger6516c652007-03-08 20:41:55 -08002183 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184 case UDP_CORK:
2185 val = up->corkflag;
2186 break;
2187
2188 case UDP_ENCAP:
2189 val = up->encap_type;
2190 break;
2191
Tom Herbert1c194482014-05-23 08:47:32 -07002192 case UDP_NO_CHECK6_TX:
2193 val = up->no_check6_tx;
2194 break;
2195
2196 case UDP_NO_CHECK6_RX:
2197 val = up->no_check6_rx;
2198 break;
2199
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002200 /* The following two cannot be changed on UDP sockets, the return is
2201 * always 0 (which corresponds to the full checksum coverage of UDP). */
2202 case UDPLITE_SEND_CSCOV:
2203 val = up->pcslen;
2204 break;
2205
2206 case UDPLITE_RECV_CSCOV:
2207 val = up->pcrlen;
2208 break;
2209
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210 default:
2211 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002212 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213
Stephen Hemminger6516c652007-03-08 20:41:55 -08002214 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002215 return -EFAULT;
Eric Dumazetc482c562009-07-17 00:26:32 +00002216 if (copy_to_user(optval, &val, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002218 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219}
Eric Dumazetc482c562009-07-17 00:26:32 +00002220EXPORT_SYMBOL(udp_lib_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221
David S. Millerdb8dac22008-03-06 16:22:02 -08002222int udp_getsockopt(struct sock *sk, int level, int optname,
2223 char __user *optval, int __user *optlen)
2224{
2225 if (level == SOL_UDP || level == SOL_UDPLITE)
2226 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2227 return ip_getsockopt(sk, level, optname, optval, optlen);
2228}
2229
2230#ifdef CONFIG_COMPAT
2231int compat_udp_getsockopt(struct sock *sk, int level, int optname,
2232 char __user *optval, int __user *optlen)
2233{
2234 if (level == SOL_UDP || level == SOL_UDPLITE)
2235 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2236 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
2237}
2238#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002239/**
2240 * udp_poll - wait for a UDP event.
2241 * @file - file struct
2242 * @sock - socket
2243 * @wait - poll table
2244 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002245 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07002246 * blocking sockets. If application is using a blocking fd
2247 * and a packet with checksum error is in the queue;
2248 * then it could get return from select indicating data available
2249 * but then block when reading it. Add special case code
2250 * to work around these arguably broken applications.
2251 */
2252unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
2253{
2254 unsigned int mask = datagram_poll(file, sock, wait);
2255 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002256
David Majnemerc3f1dba2013-05-31 13:15:38 +00002257 sock_rps_record_flow(sk);
2258
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259 /* Check for false positives due to checksum errors */
Eric Dumazet85584672009-10-09 04:43:40 +00002260 if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
2261 !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
2262 mask &= ~(POLLIN | POLLRDNORM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263
2264 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002265
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266}
Eric Dumazetc482c562009-07-17 00:26:32 +00002267EXPORT_SYMBOL(udp_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268
David S. Millerdb8dac22008-03-06 16:22:02 -08002269struct proto udp_prot = {
2270 .name = "UDP",
2271 .owner = THIS_MODULE,
2272 .close = udp_lib_close,
2273 .connect = ip4_datagram_connect,
2274 .disconnect = udp_disconnect,
2275 .ioctl = udp_ioctl,
2276 .destroy = udp_destroy_sock,
2277 .setsockopt = udp_setsockopt,
2278 .getsockopt = udp_getsockopt,
2279 .sendmsg = udp_sendmsg,
2280 .recvmsg = udp_recvmsg,
2281 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07002282 .backlog_rcv = __udp_queue_rcv_skb,
Steffen Klassert8141ed92013-01-21 02:00:03 +00002283 .release_cb = ip4_datagram_release_cb,
David S. Millerdb8dac22008-03-06 16:22:02 -08002284 .hash = udp_lib_hash,
2285 .unhash = udp_lib_unhash,
Eric Dumazet719f8352010-09-08 05:08:44 +00002286 .rehash = udp_v4_rehash,
David S. Millerdb8dac22008-03-06 16:22:02 -08002287 .get_port = udp_v4_get_port,
2288 .memory_allocated = &udp_memory_allocated,
2289 .sysctl_mem = sysctl_udp_mem,
2290 .sysctl_wmem = &sysctl_udp_wmem_min,
2291 .sysctl_rmem = &sysctl_udp_rmem_min,
2292 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07002293 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07002294 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08002295#ifdef CONFIG_COMPAT
2296 .compat_setsockopt = compat_udp_setsockopt,
2297 .compat_getsockopt = compat_udp_getsockopt,
2298#endif
Octavian Purdilafcbdf092010-12-16 14:26:56 -08002299 .clear_sk = sk_prot_clear_portaddr_nulls,
David S. Millerdb8dac22008-03-06 16:22:02 -08002300};
Eric Dumazetc482c562009-07-17 00:26:32 +00002301EXPORT_SYMBOL(udp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002302
2303/* ------------------------------------------------------------------------ */
2304#ifdef CONFIG_PROC_FS
2305
Eric Dumazet645ca702008-10-29 01:41:45 -07002306static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002307{
2308 struct sock *sk;
2309 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002310 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002312 for (state->bucket = start; state->bucket <= state->udp_table->mask;
2313 ++state->bucket) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002314 struct hlist_nulls_node *node;
Eric Dumazet645ca702008-10-29 01:41:45 -07002315 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002316
2317 if (hlist_nulls_empty(&hslot->head))
2318 continue;
2319
Eric Dumazet645ca702008-10-29 01:41:45 -07002320 spin_lock_bh(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08002321 sk_nulls_for_each(sk, node, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002322 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002323 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002324 if (sk->sk_family == state->family)
2325 goto found;
2326 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002327 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002328 }
2329 sk = NULL;
2330found:
2331 return sk;
2332}
2333
2334static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2335{
2336 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002337 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338
2339 do {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002340 sk = sk_nulls_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002341 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002342
Eric Dumazet645ca702008-10-29 01:41:45 -07002343 if (!sk) {
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002344 if (state->bucket <= state->udp_table->mask)
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002345 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07002346 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347 }
2348 return sk;
2349}
2350
2351static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2352{
Eric Dumazet645ca702008-10-29 01:41:45 -07002353 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002354
2355 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08002356 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002357 --pos;
2358 return pos ? NULL : sk;
2359}
2360
2361static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2362{
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002363 struct udp_iter_state *state = seq->private;
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002364 state->bucket = MAX_UDP_PORTS;
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002365
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002366 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367}
2368
2369static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2370{
2371 struct sock *sk;
2372
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002373 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002374 sk = udp_get_idx(seq, 0);
2375 else
2376 sk = udp_get_next(seq, v);
2377
2378 ++*pos;
2379 return sk;
2380}
2381
2382static void udp_seq_stop(struct seq_file *seq, void *v)
2383{
Eric Dumazet645ca702008-10-29 01:41:45 -07002384 struct udp_iter_state *state = seq->private;
2385
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002386 if (state->bucket <= state->udp_table->mask)
Eric Dumazet645ca702008-10-29 01:41:45 -07002387 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388}
2389
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002390int udp_seq_open(struct inode *inode, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391{
Al Virod9dda782013-03-31 18:16:14 -04002392 struct udp_seq_afinfo *afinfo = PDE_DATA(inode);
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002393 struct udp_iter_state *s;
2394 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002396 err = seq_open_net(inode, file, &afinfo->seq_ops,
2397 sizeof(struct udp_iter_state));
2398 if (err < 0)
2399 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002400
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002401 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002402 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07002403 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002404 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002405}
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002406EXPORT_SYMBOL(udp_seq_open);
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002407
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002409int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002410{
2411 struct proc_dir_entry *p;
2412 int rc = 0;
2413
Denis V. Lunevdda61922008-03-28 18:24:26 -07002414 afinfo->seq_ops.start = udp_seq_start;
2415 afinfo->seq_ops.next = udp_seq_next;
2416 afinfo->seq_ops.stop = udp_seq_stop;
2417
Denis V. Lunev84841c32008-05-02 04:10:08 -07002418 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002419 afinfo->seq_fops, afinfo);
Denis V. Lunev84841c32008-05-02 04:10:08 -07002420 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002421 rc = -ENOMEM;
2422 return rc;
2423}
Eric Dumazetc482c562009-07-17 00:26:32 +00002424EXPORT_SYMBOL(udp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002425
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002426void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002427{
Gao fengece31ff2013-02-18 01:34:56 +00002428 remove_proc_entry(afinfo->name, net->proc_net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002429}
Eric Dumazetc482c562009-07-17 00:26:32 +00002430EXPORT_SYMBOL(udp_proc_unregister);
David S. Millerdb8dac22008-03-06 16:22:02 -08002431
2432/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002433static void udp4_format_sock(struct sock *sp, struct seq_file *f,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002434 int bucket)
David S. Millerdb8dac22008-03-06 16:22:02 -08002435{
2436 struct inet_sock *inet = inet_sk(sp);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002437 __be32 dest = inet->inet_daddr;
2438 __be32 src = inet->inet_rcv_saddr;
2439 __u16 destp = ntohs(inet->inet_dport);
2440 __u16 srcp = ntohs(inet->inet_sport);
David S. Millerdb8dac22008-03-06 16:22:02 -08002441
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002442 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
Tetsuo Handa652586d2013-11-14 14:31:57 -08002443 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
David S. Millerdb8dac22008-03-06 16:22:02 -08002444 bucket, src, srcp, dest, destp, sp->sk_state,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002445 sk_wmem_alloc_get(sp),
2446 sk_rmem_alloc_get(sp),
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06002447 0, 0L, 0,
2448 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2449 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002450 atomic_read(&sp->sk_refcnt), sp,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002451 atomic_read(&sp->sk_drops));
David S. Millerdb8dac22008-03-06 16:22:02 -08002452}
2453
2454int udp4_seq_show(struct seq_file *seq, void *v)
2455{
Tetsuo Handa652586d2013-11-14 14:31:57 -08002456 seq_setwidth(seq, 127);
David S. Millerdb8dac22008-03-06 16:22:02 -08002457 if (v == SEQ_START_TOKEN)
Tetsuo Handa652586d2013-11-14 14:31:57 -08002458 seq_puts(seq, " sl local_address rem_address st tx_queue "
David S. Millerdb8dac22008-03-06 16:22:02 -08002459 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002460 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08002461 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08002462 struct udp_iter_state *state = seq->private;
2463
Tetsuo Handa652586d2013-11-14 14:31:57 -08002464 udp4_format_sock(v, seq, state->bucket);
David S. Millerdb8dac22008-03-06 16:22:02 -08002465 }
Tetsuo Handa652586d2013-11-14 14:31:57 -08002466 seq_pad(seq, '\n');
David S. Millerdb8dac22008-03-06 16:22:02 -08002467 return 0;
2468}
2469
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002470static const struct file_operations udp_afinfo_seq_fops = {
2471 .owner = THIS_MODULE,
2472 .open = udp_seq_open,
2473 .read = seq_read,
2474 .llseek = seq_lseek,
2475 .release = seq_release_net
2476};
2477
David S. Millerdb8dac22008-03-06 16:22:02 -08002478/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08002479static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08002480 .name = "udp",
2481 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07002482 .udp_table = &udp_table,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002483 .seq_fops = &udp_afinfo_seq_fops,
Denis V. Lunevdda61922008-03-28 18:24:26 -07002484 .seq_ops = {
2485 .show = udp4_seq_show,
2486 },
David S. Millerdb8dac22008-03-06 16:22:02 -08002487};
2488
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002489static int __net_init udp4_proc_init_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002490{
2491 return udp_proc_register(net, &udp4_seq_afinfo);
2492}
2493
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002494static void __net_exit udp4_proc_exit_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002495{
2496 udp_proc_unregister(net, &udp4_seq_afinfo);
2497}
2498
2499static struct pernet_operations udp4_net_ops = {
2500 .init = udp4_proc_init_net,
2501 .exit = udp4_proc_exit_net,
2502};
2503
David S. Millerdb8dac22008-03-06 16:22:02 -08002504int __init udp4_proc_init(void)
2505{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002506 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002507}
2508
2509void udp4_proc_exit(void)
2510{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002511 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002512}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002513#endif /* CONFIG_PROC_FS */
2514
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002515static __initdata unsigned long uhash_entries;
2516static int __init set_uhash_entries(char *str)
Eric Dumazet645ca702008-10-29 01:41:45 -07002517{
Eldad Zack413c27d2012-05-19 14:13:18 +00002518 ssize_t ret;
2519
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002520 if (!str)
2521 return 0;
Eldad Zack413c27d2012-05-19 14:13:18 +00002522
2523 ret = kstrtoul(str, 0, &uhash_entries);
2524 if (ret)
2525 return 0;
2526
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002527 if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2528 uhash_entries = UDP_HTABLE_SIZE_MIN;
2529 return 1;
2530}
2531__setup("uhash_entries=", set_uhash_entries);
Eric Dumazet645ca702008-10-29 01:41:45 -07002532
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002533void __init udp_table_init(struct udp_table *table, const char *name)
2534{
2535 unsigned int i;
2536
Tim Bird31fe62b2012-05-23 13:33:35 +00002537 table->hash = alloc_large_system_hash(name,
2538 2 * sizeof(struct udp_hslot),
2539 uhash_entries,
2540 21, /* one slot per 2 MB */
2541 0,
2542 &table->log,
2543 &table->mask,
2544 UDP_HTABLE_SIZE_MIN,
2545 64 * 1024);
2546
Eric Dumazet512615b2009-11-08 10:17:58 +00002547 table->hash2 = table->hash + (table->mask + 1);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002548 for (i = 0; i <= table->mask; i++) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002549 INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00002550 table->hash[i].count = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -07002551 spin_lock_init(&table->hash[i].lock);
2552 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002553 for (i = 0; i <= table->mask; i++) {
2554 INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i);
2555 table->hash2[i].count = 0;
2556 spin_lock_init(&table->hash2[i].lock);
2557 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002558}
2559
Tom Herbert723b8e42015-02-24 09:17:31 -08002560u32 udp_flow_hashrnd(void)
2561{
2562 static u32 hashrnd __read_mostly;
2563
2564 net_get_random_once(&hashrnd, sizeof(hashrnd));
2565
2566 return hashrnd;
2567}
2568EXPORT_SYMBOL(udp_flow_hashrnd);
2569
Hideo Aoki95766ff2007-12-31 00:29:24 -08002570void __init udp_init(void)
2571{
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002572 unsigned long limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002573
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002574 udp_table_init(&udp_table, "UDP");
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002575 limit = nr_free_buffer_pages() / 8;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002576 limit = max(limit, 128UL);
2577 sysctl_udp_mem[0] = limit / 4 * 3;
2578 sysctl_udp_mem[1] = limit;
2579 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2580
2581 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2582 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2583}