blob: adab9dc5818355c603a699101c4787c99434eb12 [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 * Definitions for the AF_INET socket handler.
7 *
8 * Version: @(#)sock.h 1.0.4 05/13/93
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
14 *
15 * Fixes:
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
19 * than the reverse.
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
33 *
34 *
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
39 */
40#ifndef _SOCK_H
41#define _SOCK_H
42
Ilpo Järvinen172589c2007-08-28 15:50:33 -070043#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/list.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080045#include <linux/list_nulls.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070046#include <linux/timer.h>
47#include <linux/cache.h>
48#include <linux/module.h>
Ingo Molnara5b5bb92006-07-03 00:25:35 -070049#include <linux/lockdep.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050#include <linux/netdevice.h>
51#include <linux/skbuff.h> /* struct sk_buff */
Al Virod7fe0f22006-12-03 23:15:30 -050052#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070053#include <linux/security.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090054#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070055
56#include <linux/filter.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080057#include <linux/rculist_nulls.h>
Jiri Olsaa57de0b2009-07-08 12:09:13 +000058#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
60#include <asm/atomic.h>
61#include <net/dst.h>
62#include <net/checksum.h>
63
64/*
65 * This structure really needs to be cleaned up.
66 * Most of it is for TCP, and not used by any of
67 * the other protocols.
68 */
69
70/* Define this to get the SOCK_DBG debugging facility. */
71#define SOCK_DEBUGGING
72#ifdef SOCK_DEBUGGING
73#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
74 printk(KERN_DEBUG msg); } while (0)
75#else
Stephen Hemminger4cd90292008-03-21 15:54:53 -070076/* Validate arguments and do nothing */
Jan Engelhardt1183f382010-05-02 13:42:39 -070077static inline void __attribute__ ((format (printf, 2, 3)))
Stephen Hemminger4cd90292008-03-21 15:54:53 -070078SOCK_DEBUG(struct sock *sk, const char *msg, ...)
79{
80}
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#endif
82
83/* This is the per-socket lock. The spinlock provides a synchronization
84 * between user contexts and software interrupt processing, whereas the
85 * mini-semaphore synchronizes multiple users amongst themselves.
86 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070087typedef struct {
88 spinlock_t slock;
John Heffnerd2e91172007-09-12 10:44:19 +020089 int owned;
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 wait_queue_head_t wq;
Ingo Molnara5b5bb92006-07-03 00:25:35 -070091 /*
92 * We express the mutex-alike socket_lock semantics
93 * to the lock validator by explicitly managing
94 * the slock as a lock variant (in addition to
95 * the slock itself):
96 */
97#ifdef CONFIG_DEBUG_LOCK_ALLOC
98 struct lockdep_map dep_map;
99#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100} socket_lock_t;
101
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102struct sock;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700103struct proto;
Denis V. Lunev0eeb8ff2007-12-04 01:15:45 -0800104struct net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105
106/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700107 * struct sock_common - minimal network layer representation of sockets
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000108 * @skc_node: main hash linkage for various protocol lookup tables
Eric Dumazet512615b2009-11-08 10:17:58 +0000109 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000110 * @skc_refcnt: reference count
Krishna Kumare022f0b2009-10-19 23:46:20 +0000111 * @skc_tx_queue_mapping: tx queue number for this connection
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000112 * @skc_hash: hash value used with various protocol lookup tables
Eric Dumazetd4cada42009-11-08 10:17:30 +0000113 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700114 * @skc_family: network address family
115 * @skc_state: Connection state
116 * @skc_reuse: %SO_REUSEADDR setting
117 * @skc_bound_dev_if: bound device index if != 0
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700118 * @skc_bind_node: bind hash linkage for various protocol lookup tables
Eric Dumazet512615b2009-11-08 10:17:58 +0000119 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700120 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200121 * @skc_net: reference to the network namespace of this socket
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700122 *
123 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700124 * for struct sock and struct inet_timewait_sock.
125 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126struct sock_common {
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000127 /*
128 * first fields are not copied in sock_copy()
129 */
Eric Dumazet88ab1932008-11-16 19:39:21 -0800130 union {
131 struct hlist_node skc_node;
132 struct hlist_nulls_node skc_nulls_node;
133 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 atomic_t skc_refcnt;
Krishna Kumare022f0b2009-10-19 23:46:20 +0000135 int skc_tx_queue_mapping;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000136
Eric Dumazetd4cada42009-11-08 10:17:30 +0000137 union {
138 unsigned int skc_hash;
139 __u16 skc_u16hashes[2];
140 };
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000141 unsigned short skc_family;
142 volatile unsigned char skc_state;
143 unsigned char skc_reuse;
144 int skc_bound_dev_if;
Eric Dumazet512615b2009-11-08 10:17:58 +0000145 union {
146 struct hlist_node skc_bind_node;
147 struct hlist_nulls_node skc_portaddr_node;
148 };
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700149 struct proto *skc_prot;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900150#ifdef CONFIG_NET_NS
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200151 struct net *skc_net;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900152#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153};
154
155/**
156 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700157 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700158 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
159 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
160 * @sk_lock: synchronizer
161 * @sk_rcvbuf: size of receive buffer in bytes
Eric Dumazet43815482010-04-29 11:01:49 +0000162 * @sk_wq: sock wait queue and async head
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700163 * @sk_dst_cache: destination cache
164 * @sk_dst_lock: destination cache lock
165 * @sk_policy: flow policy
166 * @sk_rmem_alloc: receive queue bytes committed
167 * @sk_receive_queue: incoming packets
168 * @sk_wmem_alloc: transmit queue bytes committed
169 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700170 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700171 * @sk_omem_alloc: "o" is "option" or "other"
172 * @sk_wmem_queued: persistent queue size
173 * @sk_forward_alloc: space allocated forward
174 * @sk_allocation: allocation mode
175 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800176 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
Patrick Ohly20d49472009-02-12 05:03:38 +0000177 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700178 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
179 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700180 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700181 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700182 * @sk_gso_max_size: Maximum GSO segment size to build
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700183 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700184 * @sk_backlog: always used with the per-socket spinlock held
185 * @sk_callback_lock: used with the callbacks in the end of this struct
186 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800187 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
188 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700189 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800190 * @sk_err_soft: errors that don't cause failure but are the cause of a
191 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700192 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700193 * @sk_ack_backlog: current listen backlog
194 * @sk_max_ack_backlog: listen backlog set in listen()
195 * @sk_priority: %SO_PRIORITY setting
196 * @sk_type: socket type (%SOCK_STREAM, etc)
197 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000198 * @sk_peer_pid: &struct pid for this socket's peer
199 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700200 * @sk_rcvlowat: %SO_RCVLOWAT setting
201 * @sk_rcvtimeo: %SO_RCVTIMEO setting
202 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700203 * @sk_rxhash: flow hash received from netif layer
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700204 * @sk_filter: socket filtering instructions
205 * @sk_protinfo: private area, net family specific, when not using slab
206 * @sk_timer: sock cleanup timer
207 * @sk_stamp: time stamp of last packet received
208 * @sk_socket: Identd and reporting IO signals
209 * @sk_user_data: RPC layer private data
210 * @sk_sndmsg_page: cached page for sendmsg
211 * @sk_sndmsg_off: cached offset for sendmsg
212 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700213 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800214 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000215 * @sk_classid: this socket's cgroup classid
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700216 * @sk_write_pending: a write to stream socket waits to start
217 * @sk_state_change: callback to indicate change in the state of the sock
218 * @sk_data_ready: callback to indicate there is data to be processed
219 * @sk_write_space: callback to indicate there is bf sending space available
220 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
221 * @sk_backlog_rcv: callback to process the backlog
222 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 */
224struct sock {
225 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700226 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 * don't add nothing before this first member (__sk_common) --acme
228 */
229 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000230#define sk_node __sk_common.skc_node
231#define sk_nulls_node __sk_common.skc_nulls_node
232#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000233#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000234
235#define sk_copy_start __sk_common.skc_hash
236#define sk_hash __sk_common.skc_hash
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237#define sk_family __sk_common.skc_family
238#define sk_state __sk_common.skc_state
239#define sk_reuse __sk_common.skc_reuse
240#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700242#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200243#define sk_net __sk_common.skc_net
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100244 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000245 unsigned int sk_shutdown : 2,
246 sk_no_check : 2,
247 sk_userlocks : 4,
248 sk_protocol : 8,
249 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100250 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 int sk_rcvbuf;
252 socket_lock_t sk_lock;
Eric Dumazetfa438cc2007-03-04 16:05:44 -0800253 /*
254 * The backlog queue is special, it is always used with
255 * the per-socket spinlock held and requires low latency
256 * access. Therefore we special case it's implementation.
257 */
258 struct {
259 struct sk_buff *head;
260 struct sk_buff *tail;
Zhu Yi8eae9392010-03-04 18:01:40 +0000261 int len;
Eric Dumazetfa438cc2007-03-04 16:05:44 -0800262 } sk_backlog;
Eric Dumazet43815482010-04-29 11:01:49 +0000263 struct socket_wq *sk_wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 struct dst_entry *sk_dst_cache;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700265#ifdef CONFIG_XFRM
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 struct xfrm_policy *sk_policy[2];
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700267#endif
Eric Dumazetb6c67122010-04-08 23:03:29 +0000268 spinlock_t sk_dst_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 atomic_t sk_rmem_alloc;
270 atomic_t sk_wmem_alloc;
271 atomic_t sk_omem_alloc;
Arnaldo Carvalho de Melo4e07a912007-05-29 13:17:47 -0700272 int sk_sndbuf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 struct sk_buff_head sk_receive_queue;
274 struct sk_buff_head sk_write_queue;
Alexey Dobriyan23789822008-11-12 23:25:32 -0800275#ifdef CONFIG_NET_DMA
Chris Leech97fc2f02006-05-23 17:55:33 -0700276 struct sk_buff_head sk_async_wait_queue;
Alexey Dobriyan23789822008-11-12 23:25:32 -0800277#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 int sk_wmem_queued;
279 int sk_forward_alloc;
Al Viro7d877f32005-10-21 03:20:43 -0400280 gfp_t sk_allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 int sk_route_caps;
Eric Dumazeta4654192010-05-16 00:36:33 -0700282 int sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700283 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700284 unsigned int sk_gso_max_size;
David S. Miller9932cf92006-03-24 15:12:37 -0800285 int sk_rcvlowat;
David S. Millerc58dc012010-04-27 15:05:31 -0700286#ifdef CONFIG_RPS
287 __u32 sk_rxhash;
288#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 unsigned long sk_flags;
290 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700292 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 rwlock_t sk_callback_lock;
294 int sk_err,
295 sk_err_soft;
Wang Chen33c732c2007-11-13 20:30:01 -0800296 atomic_t sk_drops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 unsigned short sk_ack_backlog;
298 unsigned short sk_max_ack_backlog;
299 __u32 sk_priority;
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000300 struct pid *sk_peer_pid;
301 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 long sk_rcvtimeo;
303 long sk_sndtimeo;
304 struct sk_filter *sk_filter;
305 void *sk_protinfo;
306 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700307 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 struct socket *sk_socket;
309 void *sk_user_data;
310 struct page *sk_sndmsg_page;
311 struct sk_buff *sk_send_head;
312 __u32 sk_sndmsg_off;
313 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800314#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800316#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800317 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700318 u32 sk_classid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 void (*sk_state_change)(struct sock *sk);
320 void (*sk_data_ready)(struct sock *sk, int bytes);
321 void (*sk_write_space)(struct sock *sk);
322 void (*sk_error_report)(struct sock *sk);
323 int (*sk_backlog_rcv)(struct sock *sk,
324 struct sk_buff *skb);
325 void (*sk_destruct)(struct sock *sk);
326};
327
328/*
329 * Hashed lists helper routines
330 */
Li Zefanc4146642010-02-08 23:18:45 +0000331static inline struct sock *sk_entry(const struct hlist_node *node)
332{
333 return hlist_entry(node, struct sock, sk_node);
334}
335
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700336static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337{
338 return hlist_entry(head->first, struct sock, sk_node);
339}
340
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700341static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342{
343 return hlist_empty(head) ? NULL : __sk_head(head);
344}
345
Eric Dumazet88ab1932008-11-16 19:39:21 -0800346static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
347{
348 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
349}
350
351static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
352{
353 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
354}
355
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700356static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357{
358 return sk->sk_node.next ?
359 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
360}
361
Eric Dumazet88ab1932008-11-16 19:39:21 -0800362static inline struct sock *sk_nulls_next(const struct sock *sk)
363{
364 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
365 hlist_nulls_entry(sk->sk_nulls_node.next,
366 struct sock, sk_nulls_node) :
367 NULL;
368}
369
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700370static inline int sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371{
372 return hlist_unhashed(&sk->sk_node);
373}
374
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700375static inline int sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376{
Akinobu Mitada753be2006-04-28 15:21:23 -0700377 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378}
379
380static __inline__ void sk_node_init(struct hlist_node *node)
381{
382 node->pprev = NULL;
383}
384
Eric Dumazet88ab1932008-11-16 19:39:21 -0800385static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
386{
387 node->pprev = NULL;
388}
389
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390static __inline__ void __sk_del_node(struct sock *sk)
391{
392 __hlist_del(&sk->sk_node);
393}
394
stephen hemminger808f5112010-02-22 07:57:18 +0000395/* NB: equivalent to hlist_del_init_rcu */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396static __inline__ int __sk_del_node_init(struct sock *sk)
397{
398 if (sk_hashed(sk)) {
399 __sk_del_node(sk);
400 sk_node_init(&sk->sk_node);
401 return 1;
402 }
403 return 0;
404}
405
406/* Grab socket reference count. This operation is valid only
407 when sk is ALREADY grabbed f.e. it is found in hash table
408 or a list and the lookup is made under lock preventing hash table
409 modifications.
410 */
411
412static inline void sock_hold(struct sock *sk)
413{
414 atomic_inc(&sk->sk_refcnt);
415}
416
417/* Ungrab socket in the context, which assumes that socket refcnt
418 cannot hit zero, f.e. it is true in context of any socketcall.
419 */
420static inline void __sock_put(struct sock *sk)
421{
422 atomic_dec(&sk->sk_refcnt);
423}
424
425static __inline__ int sk_del_node_init(struct sock *sk)
426{
427 int rc = __sk_del_node_init(sk);
428
429 if (rc) {
430 /* paranoid for a while -acme */
431 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
432 __sock_put(sk);
433 }
434 return rc;
435}
stephen hemminger808f5112010-02-22 07:57:18 +0000436#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
Eric Dumazet88ab1932008-11-16 19:39:21 -0800438static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700439{
440 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800441 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700442 return 1;
443 }
444 return 0;
445}
446
Eric Dumazet88ab1932008-11-16 19:39:21 -0800447static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700448{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800449 int rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700450
451 if (rc) {
452 /* paranoid for a while -acme */
453 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
454 __sock_put(sk);
455 }
456 return rc;
457}
458
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
460{
461 hlist_add_head(&sk->sk_node, list);
462}
463
464static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
465{
466 sock_hold(sk);
467 __sk_add_node(sk, list);
468}
469
stephen hemminger808f5112010-02-22 07:57:18 +0000470static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
471{
472 sock_hold(sk);
473 hlist_add_head_rcu(&sk->sk_node, list);
474}
475
Eric Dumazet88ab1932008-11-16 19:39:21 -0800476static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700477{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800478 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700479}
480
Eric Dumazet88ab1932008-11-16 19:39:21 -0800481static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700482{
483 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800484 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700485}
486
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487static __inline__ void __sk_del_bind_node(struct sock *sk)
488{
489 __hlist_del(&sk->sk_bind_node);
490}
491
492static __inline__ void sk_add_bind_node(struct sock *sk,
493 struct hlist_head *list)
494{
495 hlist_add_head(&sk->sk_bind_node, list);
496}
497
498#define sk_for_each(__sk, node, list) \
499 hlist_for_each_entry(__sk, node, list, sk_node)
stephen hemminger808f5112010-02-22 07:57:18 +0000500#define sk_for_each_rcu(__sk, node, list) \
501 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800502#define sk_nulls_for_each(__sk, node, list) \
503 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
504#define sk_nulls_for_each_rcu(__sk, node, list) \
505 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506#define sk_for_each_from(__sk, node) \
507 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
508 hlist_for_each_entry_from(__sk, node, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800509#define sk_nulls_for_each_from(__sk, node) \
510 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
511 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512#define sk_for_each_continue(__sk, node) \
513 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
514 hlist_for_each_entry_continue(__sk, node, sk_node)
515#define sk_for_each_safe(__sk, node, tmp, list) \
516 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
517#define sk_for_each_bound(__sk, node, list) \
518 hlist_for_each_entry(__sk, node, list, sk_bind_node)
519
520/* Sock flags */
521enum sock_flags {
522 SOCK_DEAD,
523 SOCK_DONE,
524 SOCK_URGINLINE,
525 SOCK_KEEPOPEN,
526 SOCK_LINGER,
527 SOCK_DESTROY,
528 SOCK_BROADCAST,
529 SOCK_TIMESTAMP,
530 SOCK_ZAPPED,
531 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
532 SOCK_DBG, /* %SO_DEBUG setting */
533 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700534 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
536 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Patrick Ohly20d49472009-02-12 05:03:38 +0000537 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
538 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
539 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
540 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
541 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
542 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
543 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700544 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700545 SOCK_RXQ_OVFL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546};
547
Ralf Baechle53b924b2005-08-23 10:11:30 -0700548static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
549{
550 nsk->sk_flags = osk->sk_flags;
551}
552
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
554{
555 __set_bit(flag, &sk->sk_flags);
556}
557
558static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
559{
560 __clear_bit(flag, &sk->sk_flags);
561}
562
563static inline int sock_flag(struct sock *sk, enum sock_flags flag)
564{
565 return test_bit(flag, &sk->sk_flags);
566}
567
568static inline void sk_acceptq_removed(struct sock *sk)
569{
570 sk->sk_ack_backlog--;
571}
572
573static inline void sk_acceptq_added(struct sock *sk)
574{
575 sk->sk_ack_backlog++;
576}
577
578static inline int sk_acceptq_is_full(struct sock *sk)
579{
David S. Miller64a14652007-03-06 11:21:05 -0800580 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581}
582
583/*
584 * Compute minimal free write space needed to queue new packets.
585 */
586static inline int sk_stream_min_wspace(struct sock *sk)
587{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800588 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589}
590
591static inline int sk_stream_wspace(struct sock *sk)
592{
593 return sk->sk_sndbuf - sk->sk_wmem_queued;
594}
595
596extern void sk_stream_write_space(struct sock *sk);
597
598static inline int sk_stream_memory_free(struct sock *sk)
599{
600 return sk->sk_wmem_queued < sk->sk_sndbuf;
601}
602
Zhu Yi8eae9392010-03-04 18:01:40 +0000603/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000604static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800605{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000606 /* dont let skb dst not refcounted, we are going to leave rcu lock */
607 skb_dst_force(skb);
608
609 if (!sk->sk_backlog.tail)
610 sk->sk_backlog.head = skb;
611 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800612 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000613
614 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800615 skb->next = NULL;
616}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617
Eric Dumazetc3774112010-04-27 15:13:20 -0700618/*
619 * Take into account size of receive queue and backlog queue
620 */
621static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
622{
623 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
624
625 return qsize + skb->truesize > sk->sk_rcvbuf;
626}
627
Zhu Yi8eae9392010-03-04 18:01:40 +0000628/* The per-socket spinlock must be held here. */
Zhu Yi40456352010-03-07 16:21:39 +0000629static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Zhu Yi8eae9392010-03-04 18:01:40 +0000630{
Eric Dumazetc3774112010-04-27 15:13:20 -0700631 if (sk_rcvqueues_full(sk, skb))
Zhu Yi8eae9392010-03-04 18:01:40 +0000632 return -ENOBUFS;
633
Zhu Yia3a858f2010-03-04 18:01:47 +0000634 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000635 sk->sk_backlog.len += skb->truesize;
636 return 0;
637}
638
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700639static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
640{
641 return sk->sk_backlog_rcv(sk, skb);
642}
643
David S. Millerc58dc012010-04-27 15:05:31 -0700644static inline void sock_rps_record_flow(const struct sock *sk)
645{
646#ifdef CONFIG_RPS
647 struct rps_sock_flow_table *sock_flow_table;
648
649 rcu_read_lock();
650 sock_flow_table = rcu_dereference(rps_sock_flow_table);
651 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
652 rcu_read_unlock();
653#endif
654}
655
656static inline void sock_rps_reset_flow(const struct sock *sk)
657{
658#ifdef CONFIG_RPS
659 struct rps_sock_flow_table *sock_flow_table;
660
661 rcu_read_lock();
662 sock_flow_table = rcu_dereference(rps_sock_flow_table);
663 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
664 rcu_read_unlock();
665#endif
666}
667
668static inline void sock_rps_save_rxhash(struct sock *sk, u32 rxhash)
669{
670#ifdef CONFIG_RPS
671 if (unlikely(sk->sk_rxhash != rxhash)) {
672 sock_rps_reset_flow(sk);
673 sk->sk_rxhash = rxhash;
674 }
675#endif
676}
677
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700678#define sk_wait_event(__sk, __timeo, __condition) \
679 ({ int __rc; \
680 release_sock(__sk); \
681 __rc = __condition; \
682 if (!__rc) { \
683 *(__timeo) = schedule_timeout(*(__timeo)); \
684 } \
685 lock_sock(__sk); \
686 __rc = __condition; \
687 __rc; \
688 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689
690extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
691extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
692extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
693extern int sk_stream_error(struct sock *sk, int flags, int err);
694extern void sk_stream_kill_queues(struct sock *sk);
695
696extern int sk_wait_data(struct sock *sk, long *timeo);
697
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700698struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800699struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800700struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700701struct raw_hashinfo;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700702
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703/* Networking protocol blocks we attach to sockets.
704 * socket layer -> transport layer interface
705 * transport -> network interface is defined by struct inet_proto
706 */
707struct proto {
708 void (*close)(struct sock *sk,
709 long timeout);
710 int (*connect)(struct sock *sk,
711 struct sockaddr *uaddr,
712 int addr_len);
713 int (*disconnect)(struct sock *sk, int flags);
714
715 struct sock * (*accept) (struct sock *sk, int flags, int *err);
716
717 int (*ioctl)(struct sock *sk, int cmd,
718 unsigned long arg);
719 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700720 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 void (*shutdown)(struct sock *sk, int how);
722 int (*setsockopt)(struct sock *sk, int level,
723 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700724 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 int (*getsockopt)(struct sock *sk, int level,
726 int optname, char __user *optval,
727 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700728#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800729 int (*compat_setsockopt)(struct sock *sk,
730 int level,
731 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700732 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800733 int (*compat_getsockopt)(struct sock *sk,
734 int level,
735 int optname, char __user *optval,
736 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700737#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
739 struct msghdr *msg, size_t len);
740 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
741 struct msghdr *msg,
742 size_t len, int noblock, int flags,
743 int *addr_len);
744 int (*sendpage)(struct sock *sk, struct page *page,
745 int offset, size_t size, int flags);
746 int (*bind)(struct sock *sk,
747 struct sockaddr *uaddr, int addr_len);
748
749 int (*backlog_rcv) (struct sock *sk,
750 struct sk_buff *skb);
751
752 /* Keeping track of sk's, looking them up, and port selection methods. */
753 void (*hash)(struct sock *sk);
754 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000755 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 int (*get_port)(struct sock *sk, unsigned short snum);
757
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800758 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800759#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700760 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800761#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800762
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700764 void (*enter_memory_pressure)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 atomic_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -0800766 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 /*
768 * Pressure flag: try to collapse.
769 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800770 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 * is strict, actions are advisory and have some latency.
772 */
773 int *memory_pressure;
774 int *sysctl_mem;
775 int *sysctl_wmem;
776 int *sysctl_rmem;
777 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +0000778 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779
Eric Dumazet271b72c2008-10-29 02:11:14 -0700780 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700782 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783
Eric Dumazetdd24c002008-11-25 21:17:14 -0800784 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700785
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700786 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800787 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700788
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700789 union {
790 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -0700791 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700792 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700793 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800794
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 struct module *owner;
796
797 char name[32];
798
799 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700800#ifdef SOCK_REFCNT_DEBUG
801 atomic_t socks;
802#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803};
804
805extern int proto_register(struct proto *prot, int alloc_slab);
806extern void proto_unregister(struct proto *prot);
807
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700808#ifdef SOCK_REFCNT_DEBUG
809static inline void sk_refcnt_debug_inc(struct sock *sk)
810{
811 atomic_inc(&sk->sk_prot->socks);
812}
813
814static inline void sk_refcnt_debug_dec(struct sock *sk)
815{
816 atomic_dec(&sk->sk_prot->socks);
817 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
818 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
819}
820
821static inline void sk_refcnt_debug_release(const struct sock *sk)
822{
823 if (atomic_read(&sk->sk_refcnt) != 1)
824 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
825 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
826}
827#else /* SOCK_REFCNT_DEBUG */
828#define sk_refcnt_debug_inc(sk) do { } while (0)
829#define sk_refcnt_debug_dec(sk) do { } while (0)
830#define sk_refcnt_debug_release(sk) do { } while (0)
831#endif /* SOCK_REFCNT_DEBUG */
832
Eric Dumazet65f76512008-01-03 20:46:48 -0800833
834#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835/* Called with local bh disabled */
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700836extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
837extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -0800838#else
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700839static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
840 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -0800841{
842}
Eric Dumazet65f76512008-01-03 20:46:48 -0800843#endif
844
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -0700846/* With per-bucket locks this operation is not-atomic, so that
847 * this version is not worse.
848 */
849static inline void __sk_prot_rehash(struct sock *sk)
850{
851 sk->sk_prot->unhash(sk);
852 sk->sk_prot->hash(sk);
853}
854
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855/* About 10 seconds */
856#define SOCK_DESTROY_TIME (10*HZ)
857
858/* Sockets 0-1023 can't be bound to unless you are superuser */
859#define PROT_SOCK 1024
860
861#define SHUTDOWN_MASK 3
862#define RCV_SHUTDOWN 1
863#define SEND_SHUTDOWN 2
864
865#define SOCK_SNDBUF_LOCK 1
866#define SOCK_RCVBUF_LOCK 2
867#define SOCK_BINDADDR_LOCK 4
868#define SOCK_BINDPORT_LOCK 8
869
870/* sock_iocb: used to kick off async processing of socket ios */
871struct sock_iocb {
872 struct list_head list;
873
874 int flags;
875 int size;
876 struct socket *sock;
877 struct sock *sk;
878 struct scm_cookie *scm;
879 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 struct kiocb *kiocb;
881};
882
883static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
884{
885 return (struct sock_iocb *)iocb->private;
886}
887
888static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
889{
890 return si->kiocb;
891}
892
893struct socket_alloc {
894 struct socket socket;
895 struct inode vfs_inode;
896};
897
898static inline struct socket *SOCKET_I(struct inode *inode)
899{
900 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
901}
902
903static inline struct inode *SOCK_INODE(struct socket *socket)
904{
905 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
906}
907
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800908/*
909 * Functions for memory accounting
910 */
911extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
912extern void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800914#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
915#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
916#define SK_MEM_SEND 0
917#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800919static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800921 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922}
923
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800924static inline int sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800926 /* return true if protocol supports memory accounting */
927 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928}
929
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800930static inline int sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800932 if (!sk_has_account(sk))
933 return 1;
Herbert Xud80d99d2005-09-01 17:48:23 -0700934 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800935 __sk_mem_schedule(sk, size, SK_MEM_SEND);
936}
937
938static inline int sk_rmem_schedule(struct sock *sk, int size)
939{
940 if (!sk_has_account(sk))
941 return 1;
942 return size <= sk->sk_forward_alloc ||
943 __sk_mem_schedule(sk, size, SK_MEM_RECV);
944}
945
946static inline void sk_mem_reclaim(struct sock *sk)
947{
948 if (!sk_has_account(sk))
949 return;
950 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
951 __sk_mem_reclaim(sk);
952}
953
David S. Miller9993e7d2008-01-10 21:56:38 -0800954static inline void sk_mem_reclaim_partial(struct sock *sk)
955{
956 if (!sk_has_account(sk))
957 return;
958 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
959 __sk_mem_reclaim(sk);
960}
961
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800962static inline void sk_mem_charge(struct sock *sk, int size)
963{
964 if (!sk_has_account(sk))
965 return;
966 sk->sk_forward_alloc -= size;
967}
968
969static inline void sk_mem_uncharge(struct sock *sk, int size)
970{
971 if (!sk_has_account(sk))
972 return;
973 sk->sk_forward_alloc += size;
974}
975
976static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
977{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800978 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
979 sk->sk_wmem_queued -= skb->truesize;
980 sk_mem_uncharge(sk, skb->truesize);
981 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -0700982}
983
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984/* Used by processes to "lock" a socket state, so that
985 * interrupts and bottom half handlers won't change it
986 * from under us. It essentially blocks any incoming
987 * packets, so that we won't get any new data or any
988 * packets that change the state of the socket.
989 *
990 * While locked, BH processing will add new packets to
991 * the backlog queue. This queue is processed by the
992 * owner of the socket lock right before it is released.
993 *
994 * Since ~2.3.5 it is also exclusive sleep lock serializing
995 * accesses from user process context.
996 */
John Heffnerd2e91172007-09-12 10:44:19 +0200997#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998
Peter Zijlstraed075362006-12-06 20:35:24 -0800999/*
1000 * Macro so as to not evaluate some arguments when
1001 * lockdep is not enabled.
1002 *
1003 * Mark both the sk_lock and the sk_lock.slock as a
1004 * per-address-family lock class.
1005 */
1006#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1007do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001008 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001009 init_waitqueue_head(&sk->sk_lock.wq); \
1010 spin_lock_init(&(sk)->sk_lock.slock); \
1011 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1012 sizeof((sk)->sk_lock)); \
1013 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1014 (skey), (sname)); \
1015 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1016} while (0)
1017
Harvey Harrison41380932007-12-12 10:46:51 -08001018extern void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001019
1020static inline void lock_sock(struct sock *sk)
1021{
1022 lock_sock_nested(sk, 0);
1023}
1024
Harvey Harrison41380932007-12-12 10:46:51 -08001025extern void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026
1027/* BH context may only use the following locking interface. */
1028#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001029#define bh_lock_sock_nested(__sk) \
1030 spin_lock_nested(&((__sk)->sk_lock.slock), \
1031 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1033
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001034extern bool lock_sock_fast(struct sock *sk);
1035/**
1036 * unlock_sock_fast - complement of lock_sock_fast
1037 * @sk: socket
1038 * @slow: slow mode
1039 *
1040 * fast unlock socket for user context.
1041 * If slow mode is on, we call regular release_sock()
1042 */
1043static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001044{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001045 if (slow)
1046 release_sock(sk);
1047 else
1048 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001049}
1050
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001051
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001052extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +01001053 gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001054 struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055extern void sk_free(struct sock *sk);
Denis V. Lunevedf02082008-02-29 11:18:32 -08001056extern void sk_release_kernel(struct sock *sk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001057extern struct sock *sk_clone(const struct sock *sk,
Al Virodd0fc662005-10-07 07:46:04 +01001058 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059
1060extern struct sk_buff *sock_wmalloc(struct sock *sk,
1061 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001062 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063extern struct sk_buff *sock_rmalloc(struct sock *sk,
1064 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001065 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066extern void sock_wfree(struct sk_buff *skb);
1067extern void sock_rfree(struct sk_buff *skb);
1068
1069extern int sock_setsockopt(struct socket *sock, int level,
1070 int op, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07001071 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072
1073extern int sock_getsockopt(struct socket *sock, int level,
1074 int op, char __user *optval,
1075 int __user *optlen);
1076extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1077 unsigned long size,
1078 int noblock,
1079 int *errcode);
Herbert Xu4cc7f682009-02-04 16:55:54 -08001080extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1081 unsigned long header_len,
1082 unsigned long data_len,
1083 int noblock,
1084 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -07001085extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +01001086 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1088extern void sk_send_sigurg(struct sock *sk);
1089
Herbert Xuf8451722010-05-24 00:12:34 -07001090#ifdef CONFIG_CGROUPS
1091extern void sock_update_classid(struct sock *sk);
1092#else
1093static inline void sock_update_classid(struct sock *sk)
1094{
1095}
1096#endif
1097
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098/*
1099 * Functions to fill in entries in struct proto_ops when a protocol
1100 * does not implement a particular function.
1101 */
1102extern int sock_no_bind(struct socket *,
1103 struct sockaddr *, int);
1104extern int sock_no_connect(struct socket *,
1105 struct sockaddr *, int, int);
1106extern int sock_no_socketpair(struct socket *,
1107 struct socket *);
1108extern int sock_no_accept(struct socket *,
1109 struct socket *, int);
1110extern int sock_no_getname(struct socket *,
1111 struct sockaddr *, int *, int);
1112extern unsigned int sock_no_poll(struct file *, struct socket *,
1113 struct poll_table_struct *);
1114extern int sock_no_ioctl(struct socket *, unsigned int,
1115 unsigned long);
1116extern int sock_no_listen(struct socket *, int);
1117extern int sock_no_shutdown(struct socket *, int);
1118extern int sock_no_getsockopt(struct socket *, int , int,
1119 char __user *, int __user *);
1120extern int sock_no_setsockopt(struct socket *, int, int,
David S. Millerb7058842009-09-30 16:12:20 -07001121 char __user *, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1123 struct msghdr *, size_t);
1124extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1125 struct msghdr *, size_t, int);
1126extern int sock_no_mmap(struct file *file,
1127 struct socket *sock,
1128 struct vm_area_struct *vma);
1129extern ssize_t sock_no_sendpage(struct socket *sock,
1130 struct page *page,
1131 int offset, size_t size,
1132 int flags);
1133
1134/*
1135 * Functions to fill in entries in struct proto_ops when a protocol
1136 * uses the inet style.
1137 */
1138extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1139 char __user *optval, int __user *optlen);
1140extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1141 struct msghdr *msg, size_t size, int flags);
1142extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001143 char __user *optval, unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001144extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1145 int optname, char __user *optval, int __user *optlen);
1146extern int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001147 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148
1149extern void sk_common_release(struct sock *sk);
1150
1151/*
1152 * Default socket callbacks and setup code
1153 */
1154
1155/* Initialise core socket variables */
1156extern void sock_init_data(struct socket *sock, struct sock *sk);
1157
1158/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001159 * sk_filter_release - release a socket filter
Paul Bonserdc9b3342006-11-23 17:56:13 -08001160 * @fp: filter to remove
1161 *
1162 * Remove a filter from a socket and release its resources.
1163 */
1164
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001165static inline void sk_filter_release(struct sk_filter *fp)
1166{
1167 if (atomic_dec_and_test(&fp->refcnt))
Pavel Emelyanov47e958e2007-10-17 21:22:42 -07001168 kfree(fp);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001169}
1170
1171static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172{
1173 unsigned int size = sk_filter_len(fp);
1174
1175 atomic_sub(size, &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001176 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177}
1178
1179static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1180{
1181 atomic_inc(&fp->refcnt);
1182 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1183}
1184
1185/*
1186 * Socket reference counting postulates.
1187 *
1188 * * Each user of socket SHOULD hold a reference count.
1189 * * Each access point to socket (an hash table bucket, reference from a list,
1190 * running timer, skb in flight MUST hold a reference count.
1191 * * When reference count hits 0, it means it will never increase back.
1192 * * When reference count hits 0, it means that no references from
1193 * outside exist to this socket and current process on current CPU
1194 * is last user and may/should destroy this socket.
1195 * * sk_free is called from any context: process, BH, IRQ. When
1196 * it is called, socket has no references from outside -> sk_free
1197 * may release descendant resources allocated by the socket, but
1198 * to the time when it is called, socket is NOT referenced by any
1199 * hash tables, lists etc.
1200 * * Packets, delivered from outside (from network or from another process)
1201 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1202 * when they sit in queue. Otherwise, packets will leak to hole, when
1203 * socket is looked up by one cpu and unhasing is made by another CPU.
1204 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1205 * (leak to backlog). Packet socket does all the processing inside
1206 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1207 * use separate SMP lock, so that they are prone too.
1208 */
1209
1210/* Ungrab socket and destroy it, if it was the last reference. */
1211static inline void sock_put(struct sock *sk)
1212{
1213 if (atomic_dec_and_test(&sk->sk_refcnt))
1214 sk_free(sk);
1215}
1216
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -02001217extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1218 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001219
Krishna Kumare022f0b2009-10-19 23:46:20 +00001220static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1221{
1222 sk->sk_tx_queue_mapping = tx_queue;
1223}
1224
1225static inline void sk_tx_queue_clear(struct sock *sk)
1226{
1227 sk->sk_tx_queue_mapping = -1;
1228}
1229
1230static inline int sk_tx_queue_get(const struct sock *sk)
1231{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001232 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001233}
1234
David S. Miller972692e2008-06-17 22:41:38 -07001235static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1236{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001237 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001238 sk->sk_socket = sock;
1239}
1240
Eric Dumazetaa395142010-04-20 13:03:51 +00001241static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1242{
Eric Dumazet43815482010-04-29 11:01:49 +00001243 return &sk->sk_wq->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001244}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245/* Detach socket from process context.
1246 * Announce socket dead, detach it from wait queue and inode.
1247 * Note that parent inode held reference count on this struct sock,
1248 * we do not release it in this function, because protocol
1249 * probably wants some additional cleanups or even continuing
1250 * to work with this socket (TCP).
1251 */
1252static inline void sock_orphan(struct sock *sk)
1253{
1254 write_lock_bh(&sk->sk_callback_lock);
1255 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001256 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001257 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 write_unlock_bh(&sk->sk_callback_lock);
1259}
1260
1261static inline void sock_graft(struct sock *sk, struct socket *parent)
1262{
1263 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazet43815482010-04-29 11:01:49 +00001264 rcu_assign_pointer(sk->sk_wq, parent->wq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001266 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001267 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 write_unlock_bh(&sk->sk_callback_lock);
1269}
1270
1271extern int sock_i_uid(struct sock *sk);
1272extern unsigned long sock_i_ino(struct sock *sk);
1273
1274static inline struct dst_entry *
1275__sk_dst_get(struct sock *sk)
1276{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001277 return rcu_dereference_check(sk->sk_dst_cache, rcu_read_lock_held() ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001278 sock_owned_by_user(sk) ||
1279 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280}
1281
1282static inline struct dst_entry *
1283sk_dst_get(struct sock *sk)
1284{
1285 struct dst_entry *dst;
1286
Eric Dumazetb6c67122010-04-08 23:03:29 +00001287 rcu_read_lock();
1288 dst = rcu_dereference(sk->sk_dst_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 if (dst)
1290 dst_hold(dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001291 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 return dst;
1293}
1294
Eric Dumazetb6c67122010-04-08 23:03:29 +00001295extern void sk_reset_txq(struct sock *sk);
1296
1297static inline void dst_negative_advice(struct sock *sk)
1298{
1299 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1300
1301 if (dst && dst->ops->negative_advice) {
1302 ndst = dst->ops->negative_advice(dst);
1303
1304 if (ndst != dst) {
1305 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1306 sk_reset_txq(sk);
1307 }
1308 }
1309}
1310
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311static inline void
1312__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1313{
1314 struct dst_entry *old_dst;
1315
Krishna Kumare022f0b2009-10-19 23:46:20 +00001316 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001317 /*
1318 * This can be called while sk is owned by the caller only,
1319 * with no state that can be checked in a rcu_dereference_check() cond
1320 */
1321 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001322 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 dst_release(old_dst);
1324}
1325
1326static inline void
1327sk_dst_set(struct sock *sk, struct dst_entry *dst)
1328{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001329 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 __sk_dst_set(sk, dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001331 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332}
1333
1334static inline void
1335__sk_dst_reset(struct sock *sk)
1336{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001337 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338}
1339
1340static inline void
1341sk_dst_reset(struct sock *sk)
1342{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001343 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 __sk_dst_reset(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001345 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346}
1347
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001348extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001350extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351
Herbert Xubcd76112006-06-30 13:36:35 -07001352static inline int sk_can_gso(const struct sock *sk)
1353{
1354 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1355}
1356
Andi Kleen99580892007-04-20 17:12:43 -07001357extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001358
Eric Dumazeta4654192010-05-16 00:36:33 -07001359static inline void sk_nocaps_add(struct sock *sk, int flags)
1360{
1361 sk->sk_route_nocaps |= flags;
1362 sk->sk_route_caps &= ~flags;
1363}
1364
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1366 struct sk_buff *skb, struct page *page,
1367 int off, int copy)
1368{
1369 if (skb->ip_summed == CHECKSUM_NONE) {
1370 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001371 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 page_address(page) + off,
1373 copy, 0, &err);
1374 if (err)
1375 return err;
1376 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1377 } else if (copy_from_user(page_address(page) + off, from, copy))
1378 return -EFAULT;
1379
1380 skb->len += copy;
1381 skb->data_len += copy;
1382 skb->truesize += copy;
1383 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001384 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 return 0;
1386}
1387
Eric Dumazetc5640392009-06-16 10:12:03 +00001388/**
1389 * sk_wmem_alloc_get - returns write allocations
1390 * @sk: socket
1391 *
1392 * Returns sk_wmem_alloc minus initial offset of one
1393 */
1394static inline int sk_wmem_alloc_get(const struct sock *sk)
1395{
1396 return atomic_read(&sk->sk_wmem_alloc) - 1;
1397}
1398
1399/**
1400 * sk_rmem_alloc_get - returns read allocations
1401 * @sk: socket
1402 *
1403 * Returns sk_rmem_alloc
1404 */
1405static inline int sk_rmem_alloc_get(const struct sock *sk)
1406{
1407 return atomic_read(&sk->sk_rmem_alloc);
1408}
1409
1410/**
1411 * sk_has_allocations - check if allocations are outstanding
1412 * @sk: socket
1413 *
1414 * Returns true if socket has write or read allocations
1415 */
1416static inline int sk_has_allocations(const struct sock *sk)
1417{
1418 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1419}
1420
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001421/**
Eric Dumazet43815482010-04-29 11:01:49 +00001422 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001423 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001424 *
Eric Dumazet43815482010-04-29 11:01:49 +00001425 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001426 *
Eric Dumazet43815482010-04-29 11:01:49 +00001427 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001428 * barrier call. They were added due to the race found within the tcp code.
1429 *
1430 * Consider following tcp code paths:
1431 *
1432 * CPU1 CPU2
1433 *
1434 * sys_select receive packet
1435 * ... ...
1436 * __add_wait_queue update tp->rcv_nxt
1437 * ... ...
1438 * tp->rcv_nxt check sock_def_readable
1439 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001440 * schedule rcu_read_lock();
1441 * wq = rcu_dereference(sk->sk_wq);
1442 * if (wq && waitqueue_active(&wq->wait))
1443 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001444 * ...
1445 * }
1446 *
1447 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1448 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1449 * could then endup calling schedule and sleep forever if there are no more
1450 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001451 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001452 */
Eric Dumazet43815482010-04-29 11:01:49 +00001453static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001454{
Eric Dumazet43815482010-04-29 11:01:49 +00001455
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001456 /*
1457 * We need to be sure we are in sync with the
1458 * add_wait_queue modifications to the wait queue.
1459 *
1460 * This memory barrier is paired in the sock_poll_wait.
1461 */
Eric Dumazet43815482010-04-29 11:01:49 +00001462 smp_mb();
1463 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001464}
1465
1466/**
1467 * sock_poll_wait - place memory barrier behind the poll_wait call.
1468 * @filp: file
1469 * @wait_address: socket wait queue
1470 * @p: poll_table
1471 *
Eric Dumazet43815482010-04-29 11:01:49 +00001472 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001473 */
1474static inline void sock_poll_wait(struct file *filp,
1475 wait_queue_head_t *wait_address, poll_table *p)
1476{
1477 if (p && wait_address) {
1478 poll_wait(filp, wait_address, p);
1479 /*
1480 * We need to be sure we are in sync with the
1481 * socket flags modification.
1482 *
Eric Dumazet43815482010-04-29 11:01:49 +00001483 * This memory barrier is paired in the wq_has_sleeper.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001484 */
1485 smp_mb();
1486 }
1487}
1488
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489/*
1490 * Queue a received datagram if it will fit. Stream and sequenced
1491 * protocols can't normally use this as they need to fit buffers in
1492 * and play with them.
1493 *
1494 * Inlined as it's very short and called for pretty much every
1495 * packet ever received.
1496 */
1497
1498static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1499{
Herbert Xud55d87f2009-06-22 02:25:25 +00001500 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 skb->sk = sk;
1502 skb->destructor = sock_wfree;
Eric Dumazet2b85a342009-06-11 02:55:43 -07001503 /*
1504 * We used to take a refcount on sk, but following operation
1505 * is enough to guarantee sk_free() wont free this sock until
1506 * all in-flight packets are completed
1507 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1509}
1510
1511static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1512{
Herbert Xud55d87f2009-06-22 02:25:25 +00001513 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 skb->sk = sk;
1515 skb->destructor = sock_rfree;
1516 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001517 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518}
1519
1520extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1521 unsigned long expires);
1522
1523extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1524
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001525extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526
Eric Dumazetb1faf562010-05-31 23:44:05 -07001527extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528
1529/*
1530 * Recover an error report and clear atomically
1531 */
1532
1533static inline int sock_error(struct sock *sk)
1534{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001535 int err;
1536 if (likely(!sk->sk_err))
1537 return 0;
1538 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 return -err;
1540}
1541
1542static inline unsigned long sock_wspace(struct sock *sk)
1543{
1544 int amt = 0;
1545
1546 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1547 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1548 if (amt < 0)
1549 amt = 0;
1550 }
1551 return amt;
1552}
1553
1554static inline void sk_wake_async(struct sock *sk, int how, int band)
1555{
Eric Dumazetbcdce712009-10-06 17:28:29 -07001556 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 sock_wake_async(sk->sk_socket, how, band);
1558}
1559
1560#define SOCK_MIN_SNDBUF 2048
1561#define SOCK_MIN_RCVBUF 256
1562
1563static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1564{
1565 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001566 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1568 }
1569}
1570
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11001571struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572
1573static inline struct page *sk_stream_alloc_page(struct sock *sk)
1574{
1575 struct page *page = NULL;
1576
Herbert Xuef015782005-09-01 17:48:59 -07001577 page = alloc_pages(sk->sk_allocation, 0);
1578 if (!page) {
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001579 sk->sk_prot->enter_memory_pressure(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 sk_stream_moderate_sndbuf(sk);
1581 }
1582 return page;
1583}
1584
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585/*
1586 * Default write policy as shown to user space via poll/select/SIGIO
1587 */
1588static inline int sock_writeable(const struct sock *sk)
1589{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001590 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591}
1592
Al Virodd0fc662005-10-07 07:46:04 +01001593static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594{
Andrew Morton99709372009-02-12 16:43:17 -08001595 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596}
1597
1598static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1599{
1600 return noblock ? 0 : sk->sk_rcvtimeo;
1601}
1602
1603static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1604{
1605 return noblock ? 0 : sk->sk_sndtimeo;
1606}
1607
1608static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1609{
1610 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1611}
1612
1613/* Alas, with timeout socket operations are not restartable.
1614 * Compare this to poll().
1615 */
1616static inline int sock_intr_errno(long timeo)
1617{
1618 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1619}
1620
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001621extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1622 struct sk_buff *skb);
1623
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624static __inline__ void
1625sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1626{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001627 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00001628 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001629
Patrick Ohly20d49472009-02-12 05:03:38 +00001630 /*
1631 * generate control messages if
1632 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1633 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1634 * - software time stamp available and wanted
1635 * (SOCK_TIMESTAMPING_SOFTWARE)
1636 * - hardware time stamps available and wanted
1637 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1638 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1639 */
1640 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1641 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1642 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1643 (hwtstamps->hwtstamp.tv64 &&
1644 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1645 (hwtstamps->syststamp.tv64 &&
1646 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001647 __sock_recv_timestamp(msg, sk, skb);
1648 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001649 sk->sk_stamp = kt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650}
1651
Eric Dumazet767dd032010-04-28 19:14:43 +00001652extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1653 struct sk_buff *skb);
1654
1655static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1656 struct sk_buff *skb)
1657{
1658#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
1659 (1UL << SOCK_RCVTSTAMP) | \
1660 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
1661 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
1662 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
1663 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
1664
1665 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
1666 __sock_recv_ts_and_drops(msg, sk, skb);
1667 else
1668 sk->sk_stamp = skb->tstamp;
1669}
Neil Horman3b885782009-10-12 13:26:31 -07001670
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671/**
Patrick Ohly20d49472009-02-12 05:03:38 +00001672 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1673 * @msg: outgoing packet
1674 * @sk: socket sending this packet
1675 * @shtx: filled with instructions for time stamping
1676 *
1677 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1678 * parameters are invalid.
1679 */
1680extern int sock_tx_timestamp(struct msghdr *msg,
1681 struct sock *sk,
1682 union skb_shared_tx *shtx);
1683
1684
1685/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001687 * @sk: socket to eat this skb from
1688 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07001689 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 *
1691 * This routine must be called with interrupts disabled or with the socket
1692 * locked so that the sk_buff queue operation is ok.
1693*/
Chris Leech624d1162006-05-23 18:01:28 -07001694#ifdef CONFIG_NET_DMA
1695static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1696{
1697 __skb_unlink(skb, &sk->sk_receive_queue);
1698 if (!copied_early)
1699 __kfree_skb(skb);
1700 else
1701 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1702}
1703#else
1704static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705{
1706 __skb_unlink(skb, &sk->sk_receive_queue);
1707 __kfree_skb(skb);
1708}
Chris Leech624d1162006-05-23 18:01:28 -07001709#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001711static inline
1712struct net *sock_net(const struct sock *sk)
1713{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001714 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001715}
1716
1717static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07001718void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001719{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001720 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001721}
1722
Denis V. Lunevedf02082008-02-29 11:18:32 -08001723/*
1724 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1725 * They should not hold a referrence to a namespace in order to allow
1726 * to stop it.
1727 * Sockets after sk_change_net should be released using sk_release_kernel
1728 */
1729static inline void sk_change_net(struct sock *sk, struct net *net)
1730{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001731 put_net(sock_net(sk));
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07001732 sock_net_set(sk, hold_net(net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08001733}
1734
KOVACS Krisztian23542612008-10-07 12:41:01 -07001735static inline struct sock *skb_steal_sock(struct sk_buff *skb)
1736{
1737 if (unlikely(skb->sk)) {
1738 struct sock *sk = skb->sk;
1739
1740 skb->destructor = NULL;
1741 skb->sk = NULL;
1742 return sk;
1743 }
1744 return NULL;
1745}
1746
Patrick Ohly20d49472009-02-12 05:03:38 +00001747extern void sock_enable_timestamp(struct sock *sk, int flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748extern int sock_get_timestamp(struct sock *, struct timeval __user *);
Eric Dumazetae40eb12007-03-18 17:33:16 -07001749extern int sock_get_timestampns(struct sock *, struct timespec __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750
1751/*
1752 * Enable debug/info messages
1753 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08001754extern int net_msg_warn;
1755#define NETDEBUG(fmt, args...) \
1756 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08001758#define LIMIT_NETDEBUG(fmt, args...) \
1759 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761extern __u32 sysctl_wmem_max;
1762extern __u32 sysctl_rmem_max;
1763
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001764extern void sk_init(void);
1765
David S. Miller6baf1f42005-09-05 18:14:11 -07001766extern int sysctl_optmem_max;
1767
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001768extern __u32 sysctl_wmem_default;
1769extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001770
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771#endif /* _SOCK_H */