blob: bc1cf7d88ccb922bd0093f6284cd3f03fbe00336 [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
Eric Dumazetc31504d2010-11-15 19:58:26 +000060#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#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 Melo8feaf0c2005-08-09 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 Dumazet68835ab2010-11-30 19:04:07 +0000108 * @skc_daddr: Foreign IPv4 addr
109 * @skc_rcv_saddr: Bound local IPv4 addr
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000110 * @skc_hash: hash value used with various protocol lookup tables
Eric Dumazetd4cada42009-11-08 10:17:30 +0000111 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700112 * @skc_family: network address family
113 * @skc_state: Connection state
114 * @skc_reuse: %SO_REUSEADDR setting
115 * @skc_bound_dev_if: bound device index if != 0
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700116 * @skc_bind_node: bind hash linkage for various protocol lookup tables
Eric Dumazet512615b2009-11-08 10:17:58 +0000117 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700118 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200119 * @skc_net: reference to the network namespace of this socket
Eric Dumazet68835ab2010-11-30 19:04:07 +0000120 * @skc_node: main hash linkage for various protocol lookup tables
121 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
122 * @skc_tx_queue_mapping: tx queue number for this connection
123 * @skc_refcnt: reference count
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700124 *
125 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700126 * for struct sock and struct inet_timewait_sock.
127 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128struct sock_common {
Eric Dumazet68835ab2010-11-30 19:04:07 +0000129 /* skc_daddr and skc_rcv_saddr must be grouped :
130 * cf INET_MATCH() and INET_TW_MATCH()
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000131 */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000132 __be32 skc_daddr;
133 __be32 skc_rcv_saddr;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000134
Eric Dumazetd4cada42009-11-08 10:17:30 +0000135 union {
136 unsigned int skc_hash;
137 __u16 skc_u16hashes[2];
138 };
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000139 unsigned short skc_family;
140 volatile unsigned char skc_state;
141 unsigned char skc_reuse;
142 int skc_bound_dev_if;
Eric Dumazet512615b2009-11-08 10:17:58 +0000143 union {
144 struct hlist_node skc_bind_node;
145 struct hlist_nulls_node skc_portaddr_node;
146 };
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700147 struct proto *skc_prot;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900148#ifdef CONFIG_NET_NS
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200149 struct net *skc_net;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900150#endif
Eric Dumazet68835ab2010-11-30 19:04:07 +0000151 /*
152 * fields between dontcopy_begin/dontcopy_end
153 * are not copied in sock_copy()
154 */
Randy Dunlap928c41e2011-01-08 17:39:21 +0000155 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000156 int skc_dontcopy_begin[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000157 /* public: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000158 union {
159 struct hlist_node skc_node;
160 struct hlist_nulls_node skc_nulls_node;
161 };
162 int skc_tx_queue_mapping;
163 atomic_t skc_refcnt;
Randy Dunlap928c41e2011-01-08 17:39:21 +0000164 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000165 int skc_dontcopy_end[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000166 /* public: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167};
168
169/**
170 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700171 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700172 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
173 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
174 * @sk_lock: synchronizer
175 * @sk_rcvbuf: size of receive buffer in bytes
Eric Dumazet43815482010-04-29 11:01:49 +0000176 * @sk_wq: sock wait queue and async head
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700177 * @sk_dst_cache: destination cache
178 * @sk_dst_lock: destination cache lock
179 * @sk_policy: flow policy
180 * @sk_rmem_alloc: receive queue bytes committed
181 * @sk_receive_queue: incoming packets
182 * @sk_wmem_alloc: transmit queue bytes committed
183 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700184 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700185 * @sk_omem_alloc: "o" is "option" or "other"
186 * @sk_wmem_queued: persistent queue size
187 * @sk_forward_alloc: space allocated forward
188 * @sk_allocation: allocation mode
189 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800190 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
Patrick Ohly20d49472009-02-12 05:03:38 +0000191 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700192 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
193 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700194 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700195 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700196 * @sk_gso_max_size: Maximum GSO segment size to build
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700197 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700198 * @sk_backlog: always used with the per-socket spinlock held
199 * @sk_callback_lock: used with the callbacks in the end of this struct
200 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800201 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
202 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700203 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800204 * @sk_err_soft: errors that don't cause failure but are the cause of a
205 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700206 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700207 * @sk_ack_backlog: current listen backlog
208 * @sk_max_ack_backlog: listen backlog set in listen()
209 * @sk_priority: %SO_PRIORITY setting
210 * @sk_type: socket type (%SOCK_STREAM, etc)
211 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000212 * @sk_peer_pid: &struct pid for this socket's peer
213 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700214 * @sk_rcvlowat: %SO_RCVLOWAT setting
215 * @sk_rcvtimeo: %SO_RCVTIMEO setting
216 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700217 * @sk_rxhash: flow hash received from netif layer
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700218 * @sk_filter: socket filtering instructions
219 * @sk_protinfo: private area, net family specific, when not using slab
220 * @sk_timer: sock cleanup timer
221 * @sk_stamp: time stamp of last packet received
222 * @sk_socket: Identd and reporting IO signals
223 * @sk_user_data: RPC layer private data
224 * @sk_sndmsg_page: cached page for sendmsg
225 * @sk_sndmsg_off: cached offset for sendmsg
226 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700227 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800228 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000229 * @sk_classid: this socket's cgroup classid
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700230 * @sk_write_pending: a write to stream socket waits to start
231 * @sk_state_change: callback to indicate change in the state of the sock
232 * @sk_data_ready: callback to indicate there is data to be processed
233 * @sk_write_space: callback to indicate there is bf sending space available
234 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
235 * @sk_backlog_rcv: callback to process the backlog
236 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 */
238struct sock {
239 /*
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700240 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 * don't add nothing before this first member (__sk_common) --acme
242 */
243 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000244#define sk_node __sk_common.skc_node
245#define sk_nulls_node __sk_common.skc_nulls_node
246#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000247#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000248
Eric Dumazet68835ab2010-11-30 19:04:07 +0000249#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
250#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000251#define sk_hash __sk_common.skc_hash
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252#define sk_family __sk_common.skc_family
253#define sk_state __sk_common.skc_state
254#define sk_reuse __sk_common.skc_reuse
255#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700257#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200258#define sk_net __sk_common.skc_net
Eric Dumazetb178bb32010-11-16 05:56:04 +0000259 socket_lock_t sk_lock;
260 struct sk_buff_head sk_receive_queue;
261 /*
262 * The backlog queue is special, it is always used with
263 * the per-socket spinlock held and requires low latency
264 * access. Therefore we special case it's implementation.
265 * Note : rmem_alloc is in this structure to fill a hole
266 * on 64bit arches, not because its logically part of
267 * backlog.
268 */
269 struct {
270 atomic_t rmem_alloc;
271 int len;
272 struct sk_buff *head;
273 struct sk_buff *tail;
274 } sk_backlog;
275#define sk_rmem_alloc sk_backlog.rmem_alloc
276 int sk_forward_alloc;
277#ifdef CONFIG_RPS
278 __u32 sk_rxhash;
279#endif
280 atomic_t sk_drops;
281 int sk_rcvbuf;
282
283 struct sk_filter __rcu *sk_filter;
284 struct socket_wq *sk_wq;
285
286#ifdef CONFIG_NET_DMA
287 struct sk_buff_head sk_async_wait_queue;
288#endif
289
290#ifdef CONFIG_XFRM
291 struct xfrm_policy *sk_policy[2];
292#endif
293 unsigned long sk_flags;
294 struct dst_entry *sk_dst_cache;
295 spinlock_t sk_dst_lock;
296 atomic_t sk_wmem_alloc;
297 atomic_t sk_omem_alloc;
298 int sk_sndbuf;
299 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100300 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000301 unsigned int sk_shutdown : 2,
302 sk_no_check : 2,
303 sk_userlocks : 4,
304 sk_protocol : 8,
305 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100306 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400308 gfp_t sk_allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 int sk_route_caps;
Eric Dumazeta4654192010-05-16 00:36:33 -0700310 int sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700311 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700312 unsigned int sk_gso_max_size;
David S. Miller9932cf92006-03-24 15:12:37 -0800313 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700316 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 rwlock_t sk_callback_lock;
318 int sk_err,
319 sk_err_soft;
320 unsigned short sk_ack_backlog;
321 unsigned short sk_max_ack_backlog;
322 __u32 sk_priority;
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000323 struct pid *sk_peer_pid;
324 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 long sk_rcvtimeo;
326 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 void *sk_protinfo;
328 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700329 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 struct socket *sk_socket;
331 void *sk_user_data;
332 struct page *sk_sndmsg_page;
333 struct sk_buff *sk_send_head;
334 __u32 sk_sndmsg_off;
335 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800336#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800338#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800339 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700340 u32 sk_classid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 void (*sk_state_change)(struct sock *sk);
342 void (*sk_data_ready)(struct sock *sk, int bytes);
343 void (*sk_write_space)(struct sock *sk);
344 void (*sk_error_report)(struct sock *sk);
345 int (*sk_backlog_rcv)(struct sock *sk,
346 struct sk_buff *skb);
347 void (*sk_destruct)(struct sock *sk);
348};
349
350/*
351 * Hashed lists helper routines
352 */
Li Zefanc4146642010-02-08 23:18:45 +0000353static inline struct sock *sk_entry(const struct hlist_node *node)
354{
355 return hlist_entry(node, struct sock, sk_node);
356}
357
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700358static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359{
360 return hlist_entry(head->first, struct sock, sk_node);
361}
362
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700363static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364{
365 return hlist_empty(head) ? NULL : __sk_head(head);
366}
367
Eric Dumazet88ab1932008-11-16 19:39:21 -0800368static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
369{
370 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
371}
372
373static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
374{
375 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
376}
377
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700378static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379{
380 return sk->sk_node.next ?
381 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
382}
383
Eric Dumazet88ab1932008-11-16 19:39:21 -0800384static inline struct sock *sk_nulls_next(const struct sock *sk)
385{
386 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
387 hlist_nulls_entry(sk->sk_nulls_node.next,
388 struct sock, sk_nulls_node) :
389 NULL;
390}
391
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700392static inline int sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393{
394 return hlist_unhashed(&sk->sk_node);
395}
396
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700397static inline int sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398{
Akinobu Mitada753be2006-04-28 15:21:23 -0700399 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400}
401
402static __inline__ void sk_node_init(struct hlist_node *node)
403{
404 node->pprev = NULL;
405}
406
Eric Dumazet88ab1932008-11-16 19:39:21 -0800407static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
408{
409 node->pprev = NULL;
410}
411
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412static __inline__ void __sk_del_node(struct sock *sk)
413{
414 __hlist_del(&sk->sk_node);
415}
416
stephen hemminger808f5112010-02-22 07:57:18 +0000417/* NB: equivalent to hlist_del_init_rcu */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418static __inline__ int __sk_del_node_init(struct sock *sk)
419{
420 if (sk_hashed(sk)) {
421 __sk_del_node(sk);
422 sk_node_init(&sk->sk_node);
423 return 1;
424 }
425 return 0;
426}
427
428/* Grab socket reference count. This operation is valid only
429 when sk is ALREADY grabbed f.e. it is found in hash table
430 or a list and the lookup is made under lock preventing hash table
431 modifications.
432 */
433
434static inline void sock_hold(struct sock *sk)
435{
436 atomic_inc(&sk->sk_refcnt);
437}
438
439/* Ungrab socket in the context, which assumes that socket refcnt
440 cannot hit zero, f.e. it is true in context of any socketcall.
441 */
442static inline void __sock_put(struct sock *sk)
443{
444 atomic_dec(&sk->sk_refcnt);
445}
446
447static __inline__ int sk_del_node_init(struct sock *sk)
448{
449 int rc = __sk_del_node_init(sk);
450
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}
stephen hemminger808f5112010-02-22 07:57:18 +0000458#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459
Eric Dumazet88ab1932008-11-16 19:39:21 -0800460static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700461{
462 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800463 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700464 return 1;
465 }
466 return 0;
467}
468
Eric Dumazet88ab1932008-11-16 19:39:21 -0800469static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700470{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800471 int rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700472
473 if (rc) {
474 /* paranoid for a while -acme */
475 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
476 __sock_put(sk);
477 }
478 return rc;
479}
480
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
482{
483 hlist_add_head(&sk->sk_node, list);
484}
485
486static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
487{
488 sock_hold(sk);
489 __sk_add_node(sk, list);
490}
491
stephen hemminger808f5112010-02-22 07:57:18 +0000492static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
493{
494 sock_hold(sk);
495 hlist_add_head_rcu(&sk->sk_node, list);
496}
497
Eric Dumazet88ab1932008-11-16 19:39:21 -0800498static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700499{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800500 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700501}
502
Eric Dumazet88ab1932008-11-16 19:39:21 -0800503static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700504{
505 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800506 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700507}
508
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509static __inline__ void __sk_del_bind_node(struct sock *sk)
510{
511 __hlist_del(&sk->sk_bind_node);
512}
513
514static __inline__ void sk_add_bind_node(struct sock *sk,
515 struct hlist_head *list)
516{
517 hlist_add_head(&sk->sk_bind_node, list);
518}
519
520#define sk_for_each(__sk, node, list) \
521 hlist_for_each_entry(__sk, node, list, sk_node)
stephen hemminger808f5112010-02-22 07:57:18 +0000522#define sk_for_each_rcu(__sk, node, list) \
523 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800524#define sk_nulls_for_each(__sk, node, list) \
525 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
526#define sk_nulls_for_each_rcu(__sk, node, list) \
527 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528#define sk_for_each_from(__sk, node) \
529 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
530 hlist_for_each_entry_from(__sk, node, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800531#define sk_nulls_for_each_from(__sk, node) \
532 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
533 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534#define sk_for_each_safe(__sk, node, tmp, list) \
535 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
536#define sk_for_each_bound(__sk, node, list) \
537 hlist_for_each_entry(__sk, node, list, sk_bind_node)
538
539/* Sock flags */
540enum sock_flags {
541 SOCK_DEAD,
542 SOCK_DONE,
543 SOCK_URGINLINE,
544 SOCK_KEEPOPEN,
545 SOCK_LINGER,
546 SOCK_DESTROY,
547 SOCK_BROADCAST,
548 SOCK_TIMESTAMP,
549 SOCK_ZAPPED,
550 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
551 SOCK_DBG, /* %SO_DEBUG setting */
552 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700553 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
555 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Patrick Ohly20d49472009-02-12 05:03:38 +0000556 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
557 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
558 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
559 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
560 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
561 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
562 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700563 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700564 SOCK_RXQ_OVFL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565};
566
Ralf Baechle53b924b2005-08-23 10:11:30 -0700567static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
568{
569 nsk->sk_flags = osk->sk_flags;
570}
571
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
573{
574 __set_bit(flag, &sk->sk_flags);
575}
576
577static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
578{
579 __clear_bit(flag, &sk->sk_flags);
580}
581
582static inline int sock_flag(struct sock *sk, enum sock_flags flag)
583{
584 return test_bit(flag, &sk->sk_flags);
585}
586
587static inline void sk_acceptq_removed(struct sock *sk)
588{
589 sk->sk_ack_backlog--;
590}
591
592static inline void sk_acceptq_added(struct sock *sk)
593{
594 sk->sk_ack_backlog++;
595}
596
597static inline int sk_acceptq_is_full(struct sock *sk)
598{
David S. Miller64a14652007-03-06 11:21:05 -0800599 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600}
601
602/*
603 * Compute minimal free write space needed to queue new packets.
604 */
605static inline int sk_stream_min_wspace(struct sock *sk)
606{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800607 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608}
609
610static inline int sk_stream_wspace(struct sock *sk)
611{
612 return sk->sk_sndbuf - sk->sk_wmem_queued;
613}
614
615extern void sk_stream_write_space(struct sock *sk);
616
617static inline int sk_stream_memory_free(struct sock *sk)
618{
619 return sk->sk_wmem_queued < sk->sk_sndbuf;
620}
621
Zhu Yi8eae9392010-03-04 18:01:40 +0000622/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000623static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800624{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000625 /* dont let skb dst not refcounted, we are going to leave rcu lock */
626 skb_dst_force(skb);
627
628 if (!sk->sk_backlog.tail)
629 sk->sk_backlog.head = skb;
630 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800631 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000632
633 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800634 skb->next = NULL;
635}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636
Eric Dumazetc3774112010-04-27 15:13:20 -0700637/*
638 * Take into account size of receive queue and backlog queue
639 */
640static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
641{
642 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
643
644 return qsize + skb->truesize > sk->sk_rcvbuf;
645}
646
Zhu Yi8eae9392010-03-04 18:01:40 +0000647/* The per-socket spinlock must be held here. */
Zhu Yi40456352010-03-07 16:21:39 +0000648static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Zhu Yi8eae9392010-03-04 18:01:40 +0000649{
Eric Dumazetc3774112010-04-27 15:13:20 -0700650 if (sk_rcvqueues_full(sk, skb))
Zhu Yi8eae9392010-03-04 18:01:40 +0000651 return -ENOBUFS;
652
Zhu Yia3a858f2010-03-04 18:01:47 +0000653 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000654 sk->sk_backlog.len += skb->truesize;
655 return 0;
656}
657
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700658static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
659{
660 return sk->sk_backlog_rcv(sk, skb);
661}
662
David S. Millerc58dc012010-04-27 15:05:31 -0700663static inline void sock_rps_record_flow(const struct sock *sk)
664{
665#ifdef CONFIG_RPS
666 struct rps_sock_flow_table *sock_flow_table;
667
668 rcu_read_lock();
669 sock_flow_table = rcu_dereference(rps_sock_flow_table);
670 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
671 rcu_read_unlock();
672#endif
673}
674
675static inline void sock_rps_reset_flow(const struct sock *sk)
676{
677#ifdef CONFIG_RPS
678 struct rps_sock_flow_table *sock_flow_table;
679
680 rcu_read_lock();
681 sock_flow_table = rcu_dereference(rps_sock_flow_table);
682 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
683 rcu_read_unlock();
684#endif
685}
686
687static inline void sock_rps_save_rxhash(struct sock *sk, u32 rxhash)
688{
689#ifdef CONFIG_RPS
690 if (unlikely(sk->sk_rxhash != rxhash)) {
691 sock_rps_reset_flow(sk);
692 sk->sk_rxhash = rxhash;
693 }
694#endif
695}
696
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700697#define sk_wait_event(__sk, __timeo, __condition) \
698 ({ int __rc; \
699 release_sock(__sk); \
700 __rc = __condition; \
701 if (!__rc) { \
702 *(__timeo) = schedule_timeout(*(__timeo)); \
703 } \
704 lock_sock(__sk); \
705 __rc = __condition; \
706 __rc; \
707 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708
709extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
710extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
711extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
712extern int sk_stream_error(struct sock *sk, int flags, int err);
713extern void sk_stream_kill_queues(struct sock *sk);
714
715extern int sk_wait_data(struct sock *sk, long *timeo);
716
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700717struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800718struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800719struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700720struct raw_hashinfo;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700721
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722/* Networking protocol blocks we attach to sockets.
723 * socket layer -> transport layer interface
724 * transport -> network interface is defined by struct inet_proto
725 */
726struct proto {
727 void (*close)(struct sock *sk,
728 long timeout);
729 int (*connect)(struct sock *sk,
730 struct sockaddr *uaddr,
731 int addr_len);
732 int (*disconnect)(struct sock *sk, int flags);
733
734 struct sock * (*accept) (struct sock *sk, int flags, int *err);
735
736 int (*ioctl)(struct sock *sk, int cmd,
737 unsigned long arg);
738 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700739 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 void (*shutdown)(struct sock *sk, int how);
741 int (*setsockopt)(struct sock *sk, int level,
742 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700743 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 int (*getsockopt)(struct sock *sk, int level,
745 int optname, char __user *optval,
746 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700747#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800748 int (*compat_setsockopt)(struct sock *sk,
749 int level,
750 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700751 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800752 int (*compat_getsockopt)(struct sock *sk,
753 int level,
754 int optname, char __user *optval,
755 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000756 int (*compat_ioctl)(struct sock *sk,
757 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700758#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
760 struct msghdr *msg, size_t len);
761 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
762 struct msghdr *msg,
763 size_t len, int noblock, int flags,
764 int *addr_len);
765 int (*sendpage)(struct sock *sk, struct page *page,
766 int offset, size_t size, int flags);
767 int (*bind)(struct sock *sk,
768 struct sockaddr *uaddr, int addr_len);
769
770 int (*backlog_rcv) (struct sock *sk,
771 struct sk_buff *skb);
772
773 /* Keeping track of sk's, looking them up, and port selection methods. */
774 void (*hash)(struct sock *sk);
775 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000776 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800778 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800780 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800781#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700782 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800783#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800784
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700786 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +0000787 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -0800788 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 /*
790 * Pressure flag: try to collapse.
791 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800792 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 * is strict, actions are advisory and have some latency.
794 */
795 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000796 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 int *sysctl_wmem;
798 int *sysctl_rmem;
799 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +0000800 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801
Eric Dumazet271b72c2008-10-29 02:11:14 -0700802 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700804 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805
Eric Dumazetdd24c002008-11-25 21:17:14 -0800806 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700807
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700808 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800809 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700810
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700811 union {
812 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -0700813 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700814 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700815 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800816
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 struct module *owner;
818
819 char name[32];
820
821 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700822#ifdef SOCK_REFCNT_DEBUG
823 atomic_t socks;
824#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825};
826
827extern int proto_register(struct proto *prot, int alloc_slab);
828extern void proto_unregister(struct proto *prot);
829
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700830#ifdef SOCK_REFCNT_DEBUG
831static inline void sk_refcnt_debug_inc(struct sock *sk)
832{
833 atomic_inc(&sk->sk_prot->socks);
834}
835
836static inline void sk_refcnt_debug_dec(struct sock *sk)
837{
838 atomic_dec(&sk->sk_prot->socks);
839 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
840 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
841}
842
843static inline void sk_refcnt_debug_release(const struct sock *sk)
844{
845 if (atomic_read(&sk->sk_refcnt) != 1)
846 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
847 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
848}
849#else /* SOCK_REFCNT_DEBUG */
850#define sk_refcnt_debug_inc(sk) do { } while (0)
851#define sk_refcnt_debug_dec(sk) do { } while (0)
852#define sk_refcnt_debug_release(sk) do { } while (0)
853#endif /* SOCK_REFCNT_DEBUG */
854
Eric Dumazet65f76512008-01-03 20:46:48 -0800855
856#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857/* Called with local bh disabled */
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700858extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
859extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -0800860#else
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700861static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
862 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -0800863{
864}
Eric Dumazet65f76512008-01-03 20:46:48 -0800865#endif
866
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -0700868/* With per-bucket locks this operation is not-atomic, so that
869 * this version is not worse.
870 */
871static inline void __sk_prot_rehash(struct sock *sk)
872{
873 sk->sk_prot->unhash(sk);
874 sk->sk_prot->hash(sk);
875}
876
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800877void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
878
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879/* About 10 seconds */
880#define SOCK_DESTROY_TIME (10*HZ)
881
882/* Sockets 0-1023 can't be bound to unless you are superuser */
883#define PROT_SOCK 1024
884
885#define SHUTDOWN_MASK 3
886#define RCV_SHUTDOWN 1
887#define SEND_SHUTDOWN 2
888
889#define SOCK_SNDBUF_LOCK 1
890#define SOCK_RCVBUF_LOCK 2
891#define SOCK_BINDADDR_LOCK 4
892#define SOCK_BINDPORT_LOCK 8
893
894/* sock_iocb: used to kick off async processing of socket ios */
895struct sock_iocb {
896 struct list_head list;
897
898 int flags;
899 int size;
900 struct socket *sock;
901 struct sock *sk;
902 struct scm_cookie *scm;
903 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 struct kiocb *kiocb;
905};
906
907static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
908{
909 return (struct sock_iocb *)iocb->private;
910}
911
912static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
913{
914 return si->kiocb;
915}
916
917struct socket_alloc {
918 struct socket socket;
919 struct inode vfs_inode;
920};
921
922static inline struct socket *SOCKET_I(struct inode *inode)
923{
924 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
925}
926
927static inline struct inode *SOCK_INODE(struct socket *socket)
928{
929 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
930}
931
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800932/*
933 * Functions for memory accounting
934 */
935extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
936extern void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800938#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
939#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
940#define SK_MEM_SEND 0
941#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800943static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800945 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946}
947
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800948static inline int sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800950 /* return true if protocol supports memory accounting */
951 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952}
953
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800954static inline int sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800956 if (!sk_has_account(sk))
957 return 1;
Herbert Xud80d99d2005-09-01 17:48:23 -0700958 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800959 __sk_mem_schedule(sk, size, SK_MEM_SEND);
960}
961
962static inline int sk_rmem_schedule(struct sock *sk, int size)
963{
964 if (!sk_has_account(sk))
965 return 1;
966 return size <= sk->sk_forward_alloc ||
967 __sk_mem_schedule(sk, size, SK_MEM_RECV);
968}
969
970static inline void sk_mem_reclaim(struct sock *sk)
971{
972 if (!sk_has_account(sk))
973 return;
974 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
975 __sk_mem_reclaim(sk);
976}
977
David S. Miller9993e7d2008-01-10 21:56:38 -0800978static inline void sk_mem_reclaim_partial(struct sock *sk)
979{
980 if (!sk_has_account(sk))
981 return;
982 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
983 __sk_mem_reclaim(sk);
984}
985
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800986static inline void sk_mem_charge(struct sock *sk, int size)
987{
988 if (!sk_has_account(sk))
989 return;
990 sk->sk_forward_alloc -= size;
991}
992
993static inline void sk_mem_uncharge(struct sock *sk, int size)
994{
995 if (!sk_has_account(sk))
996 return;
997 sk->sk_forward_alloc += size;
998}
999
1000static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1001{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001002 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1003 sk->sk_wmem_queued -= skb->truesize;
1004 sk_mem_uncharge(sk, skb->truesize);
1005 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001006}
1007
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008/* Used by processes to "lock" a socket state, so that
1009 * interrupts and bottom half handlers won't change it
1010 * from under us. It essentially blocks any incoming
1011 * packets, so that we won't get any new data or any
1012 * packets that change the state of the socket.
1013 *
1014 * While locked, BH processing will add new packets to
1015 * the backlog queue. This queue is processed by the
1016 * owner of the socket lock right before it is released.
1017 *
1018 * Since ~2.3.5 it is also exclusive sleep lock serializing
1019 * accesses from user process context.
1020 */
John Heffnerd2e91172007-09-12 10:44:19 +02001021#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022
Peter Zijlstraed075362006-12-06 20:35:24 -08001023/*
1024 * Macro so as to not evaluate some arguments when
1025 * lockdep is not enabled.
1026 *
1027 * Mark both the sk_lock and the sk_lock.slock as a
1028 * per-address-family lock class.
1029 */
1030#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1031do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001032 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001033 init_waitqueue_head(&sk->sk_lock.wq); \
1034 spin_lock_init(&(sk)->sk_lock.slock); \
1035 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1036 sizeof((sk)->sk_lock)); \
1037 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1038 (skey), (sname)); \
1039 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1040} while (0)
1041
Harvey Harrison41380932007-12-12 10:46:51 -08001042extern void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001043
1044static inline void lock_sock(struct sock *sk)
1045{
1046 lock_sock_nested(sk, 0);
1047}
1048
Harvey Harrison41380932007-12-12 10:46:51 -08001049extern void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050
1051/* BH context may only use the following locking interface. */
1052#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001053#define bh_lock_sock_nested(__sk) \
1054 spin_lock_nested(&((__sk)->sk_lock.slock), \
1055 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1057
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001058extern bool lock_sock_fast(struct sock *sk);
1059/**
1060 * unlock_sock_fast - complement of lock_sock_fast
1061 * @sk: socket
1062 * @slow: slow mode
1063 *
1064 * fast unlock socket for user context.
1065 * If slow mode is on, we call regular release_sock()
1066 */
1067static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001068{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001069 if (slow)
1070 release_sock(sk);
1071 else
1072 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001073}
1074
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001075
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001076extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +01001077 gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001078 struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079extern void sk_free(struct sock *sk);
Denis V. Lunevedf02082008-02-29 11:18:32 -08001080extern void sk_release_kernel(struct sock *sk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001081extern struct sock *sk_clone(const struct sock *sk,
Al Virodd0fc662005-10-07 07:46:04 +01001082 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
1084extern struct sk_buff *sock_wmalloc(struct sock *sk,
1085 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001086 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087extern struct sk_buff *sock_rmalloc(struct sock *sk,
1088 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001089 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090extern void sock_wfree(struct sk_buff *skb);
1091extern void sock_rfree(struct sk_buff *skb);
1092
1093extern int sock_setsockopt(struct socket *sock, int level,
1094 int op, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07001095 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096
1097extern int sock_getsockopt(struct socket *sock, int level,
1098 int op, char __user *optval,
1099 int __user *optlen);
1100extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1101 unsigned long size,
1102 int noblock,
1103 int *errcode);
Herbert Xu4cc7f682009-02-04 16:55:54 -08001104extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1105 unsigned long header_len,
1106 unsigned long data_len,
1107 int noblock,
1108 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -07001109extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +01001110 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1112extern void sk_send_sigurg(struct sock *sk);
1113
Herbert Xuf8451722010-05-24 00:12:34 -07001114#ifdef CONFIG_CGROUPS
1115extern void sock_update_classid(struct sock *sk);
1116#else
1117static inline void sock_update_classid(struct sock *sk)
1118{
1119}
1120#endif
1121
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122/*
1123 * Functions to fill in entries in struct proto_ops when a protocol
1124 * does not implement a particular function.
1125 */
1126extern int sock_no_bind(struct socket *,
1127 struct sockaddr *, int);
1128extern int sock_no_connect(struct socket *,
1129 struct sockaddr *, int, int);
1130extern int sock_no_socketpair(struct socket *,
1131 struct socket *);
1132extern int sock_no_accept(struct socket *,
1133 struct socket *, int);
1134extern int sock_no_getname(struct socket *,
1135 struct sockaddr *, int *, int);
1136extern unsigned int sock_no_poll(struct file *, struct socket *,
1137 struct poll_table_struct *);
1138extern int sock_no_ioctl(struct socket *, unsigned int,
1139 unsigned long);
1140extern int sock_no_listen(struct socket *, int);
1141extern int sock_no_shutdown(struct socket *, int);
1142extern int sock_no_getsockopt(struct socket *, int , int,
1143 char __user *, int __user *);
1144extern int sock_no_setsockopt(struct socket *, int, int,
David S. Millerb7058842009-09-30 16:12:20 -07001145 char __user *, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1147 struct msghdr *, size_t);
1148extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1149 struct msghdr *, size_t, int);
1150extern int sock_no_mmap(struct file *file,
1151 struct socket *sock,
1152 struct vm_area_struct *vma);
1153extern ssize_t sock_no_sendpage(struct socket *sock,
1154 struct page *page,
1155 int offset, size_t size,
1156 int flags);
1157
1158/*
1159 * Functions to fill in entries in struct proto_ops when a protocol
1160 * uses the inet style.
1161 */
1162extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1163 char __user *optval, int __user *optlen);
1164extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1165 struct msghdr *msg, size_t size, int flags);
1166extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001167 char __user *optval, unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001168extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1169 int optname, char __user *optval, int __user *optlen);
1170extern int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001171 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172
1173extern void sk_common_release(struct sock *sk);
1174
1175/*
1176 * Default socket callbacks and setup code
1177 */
1178
1179/* Initialise core socket variables */
1180extern void sock_init_data(struct socket *sock, struct sock *sk);
1181
Eric Dumazet46bcf142010-12-06 09:29:43 -08001182extern void sk_filter_release_rcu(struct rcu_head *rcu);
1183
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001185 * sk_filter_release - release a socket filter
Paul Bonserdc9b3342006-11-23 17:56:13 -08001186 * @fp: filter to remove
1187 *
1188 * Remove a filter from a socket and release its resources.
1189 */
1190
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001191static inline void sk_filter_release(struct sk_filter *fp)
1192{
1193 if (atomic_dec_and_test(&fp->refcnt))
Eric Dumazet46bcf142010-12-06 09:29:43 -08001194 call_rcu_bh(&fp->rcu, sk_filter_release_rcu);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001195}
1196
1197static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198{
1199 unsigned int size = sk_filter_len(fp);
1200
1201 atomic_sub(size, &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001202 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203}
1204
1205static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1206{
1207 atomic_inc(&fp->refcnt);
1208 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1209}
1210
1211/*
1212 * Socket reference counting postulates.
1213 *
1214 * * Each user of socket SHOULD hold a reference count.
1215 * * Each access point to socket (an hash table bucket, reference from a list,
1216 * running timer, skb in flight MUST hold a reference count.
1217 * * When reference count hits 0, it means it will never increase back.
1218 * * When reference count hits 0, it means that no references from
1219 * outside exist to this socket and current process on current CPU
1220 * is last user and may/should destroy this socket.
1221 * * sk_free is called from any context: process, BH, IRQ. When
1222 * it is called, socket has no references from outside -> sk_free
1223 * may release descendant resources allocated by the socket, but
1224 * to the time when it is called, socket is NOT referenced by any
1225 * hash tables, lists etc.
1226 * * Packets, delivered from outside (from network or from another process)
1227 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1228 * when they sit in queue. Otherwise, packets will leak to hole, when
1229 * socket is looked up by one cpu and unhasing is made by another CPU.
1230 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1231 * (leak to backlog). Packet socket does all the processing inside
1232 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1233 * use separate SMP lock, so that they are prone too.
1234 */
1235
1236/* Ungrab socket and destroy it, if it was the last reference. */
1237static inline void sock_put(struct sock *sk)
1238{
1239 if (atomic_dec_and_test(&sk->sk_refcnt))
1240 sk_free(sk);
1241}
1242
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -02001243extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1244 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001245
Krishna Kumare022f0b2009-10-19 23:46:20 +00001246static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1247{
1248 sk->sk_tx_queue_mapping = tx_queue;
1249}
1250
1251static inline void sk_tx_queue_clear(struct sock *sk)
1252{
1253 sk->sk_tx_queue_mapping = -1;
1254}
1255
1256static inline int sk_tx_queue_get(const struct sock *sk)
1257{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001258 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001259}
1260
David S. Miller972692e2008-06-17 22:41:38 -07001261static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1262{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001263 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001264 sk->sk_socket = sock;
1265}
1266
Eric Dumazetaa395142010-04-20 13:03:51 +00001267static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1268{
Eric Dumazet43815482010-04-29 11:01:49 +00001269 return &sk->sk_wq->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001270}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271/* Detach socket from process context.
1272 * Announce socket dead, detach it from wait queue and inode.
1273 * Note that parent inode held reference count on this struct sock,
1274 * we do not release it in this function, because protocol
1275 * probably wants some additional cleanups or even continuing
1276 * to work with this socket (TCP).
1277 */
1278static inline void sock_orphan(struct sock *sk)
1279{
1280 write_lock_bh(&sk->sk_callback_lock);
1281 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001282 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001283 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 write_unlock_bh(&sk->sk_callback_lock);
1285}
1286
1287static inline void sock_graft(struct sock *sk, struct socket *parent)
1288{
1289 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazet43815482010-04-29 11:01:49 +00001290 rcu_assign_pointer(sk->sk_wq, parent->wq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001292 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001293 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 write_unlock_bh(&sk->sk_callback_lock);
1295}
1296
1297extern int sock_i_uid(struct sock *sk);
1298extern unsigned long sock_i_ino(struct sock *sk);
1299
1300static inline struct dst_entry *
1301__sk_dst_get(struct sock *sk)
1302{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001303 return rcu_dereference_check(sk->sk_dst_cache, rcu_read_lock_held() ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001304 sock_owned_by_user(sk) ||
1305 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306}
1307
1308static inline struct dst_entry *
1309sk_dst_get(struct sock *sk)
1310{
1311 struct dst_entry *dst;
1312
Eric Dumazetb6c67122010-04-08 23:03:29 +00001313 rcu_read_lock();
1314 dst = rcu_dereference(sk->sk_dst_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 if (dst)
1316 dst_hold(dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001317 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 return dst;
1319}
1320
Eric Dumazetb6c67122010-04-08 23:03:29 +00001321extern void sk_reset_txq(struct sock *sk);
1322
1323static inline void dst_negative_advice(struct sock *sk)
1324{
1325 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1326
1327 if (dst && dst->ops->negative_advice) {
1328 ndst = dst->ops->negative_advice(dst);
1329
1330 if (ndst != dst) {
1331 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1332 sk_reset_txq(sk);
1333 }
1334 }
1335}
1336
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337static inline void
1338__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1339{
1340 struct dst_entry *old_dst;
1341
Krishna Kumare022f0b2009-10-19 23:46:20 +00001342 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001343 /*
1344 * This can be called while sk is owned by the caller only,
1345 * with no state that can be checked in a rcu_dereference_check() cond
1346 */
1347 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001348 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 dst_release(old_dst);
1350}
1351
1352static inline void
1353sk_dst_set(struct sock *sk, struct dst_entry *dst)
1354{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001355 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 __sk_dst_set(sk, dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001357 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358}
1359
1360static inline void
1361__sk_dst_reset(struct sock *sk)
1362{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001363 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364}
1365
1366static inline void
1367sk_dst_reset(struct sock *sk)
1368{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001369 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 __sk_dst_reset(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001371 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372}
1373
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001374extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001376extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377
Herbert Xubcd76112006-06-30 13:36:35 -07001378static inline int sk_can_gso(const struct sock *sk)
1379{
1380 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1381}
1382
Andi Kleen99580892007-04-20 17:12:43 -07001383extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001384
Eric Dumazeta4654192010-05-16 00:36:33 -07001385static inline void sk_nocaps_add(struct sock *sk, int flags)
1386{
1387 sk->sk_route_nocaps |= flags;
1388 sk->sk_route_caps &= ~flags;
1389}
1390
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1392 struct sk_buff *skb, struct page *page,
1393 int off, int copy)
1394{
1395 if (skb->ip_summed == CHECKSUM_NONE) {
1396 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001397 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 page_address(page) + off,
1399 copy, 0, &err);
1400 if (err)
1401 return err;
1402 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1403 } else if (copy_from_user(page_address(page) + off, from, copy))
1404 return -EFAULT;
1405
1406 skb->len += copy;
1407 skb->data_len += copy;
1408 skb->truesize += copy;
1409 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001410 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 return 0;
1412}
1413
Eric Dumazetc5640392009-06-16 10:12:03 +00001414/**
1415 * sk_wmem_alloc_get - returns write allocations
1416 * @sk: socket
1417 *
1418 * Returns sk_wmem_alloc minus initial offset of one
1419 */
1420static inline int sk_wmem_alloc_get(const struct sock *sk)
1421{
1422 return atomic_read(&sk->sk_wmem_alloc) - 1;
1423}
1424
1425/**
1426 * sk_rmem_alloc_get - returns read allocations
1427 * @sk: socket
1428 *
1429 * Returns sk_rmem_alloc
1430 */
1431static inline int sk_rmem_alloc_get(const struct sock *sk)
1432{
1433 return atomic_read(&sk->sk_rmem_alloc);
1434}
1435
1436/**
1437 * sk_has_allocations - check if allocations are outstanding
1438 * @sk: socket
1439 *
1440 * Returns true if socket has write or read allocations
1441 */
1442static inline int sk_has_allocations(const struct sock *sk)
1443{
1444 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1445}
1446
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001447/**
Eric Dumazet43815482010-04-29 11:01:49 +00001448 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001449 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001450 *
Eric Dumazet43815482010-04-29 11:01:49 +00001451 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001452 *
Eric Dumazet43815482010-04-29 11:01:49 +00001453 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001454 * barrier call. They were added due to the race found within the tcp code.
1455 *
1456 * Consider following tcp code paths:
1457 *
1458 * CPU1 CPU2
1459 *
1460 * sys_select receive packet
1461 * ... ...
1462 * __add_wait_queue update tp->rcv_nxt
1463 * ... ...
1464 * tp->rcv_nxt check sock_def_readable
1465 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001466 * schedule rcu_read_lock();
1467 * wq = rcu_dereference(sk->sk_wq);
1468 * if (wq && waitqueue_active(&wq->wait))
1469 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001470 * ...
1471 * }
1472 *
1473 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1474 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1475 * could then endup calling schedule and sleep forever if there are no more
1476 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001477 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001478 */
Eric Dumazet43815482010-04-29 11:01:49 +00001479static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001480{
Eric Dumazet43815482010-04-29 11:01:49 +00001481
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001482 /*
1483 * We need to be sure we are in sync with the
1484 * add_wait_queue modifications to the wait queue.
1485 *
1486 * This memory barrier is paired in the sock_poll_wait.
1487 */
Eric Dumazet43815482010-04-29 11:01:49 +00001488 smp_mb();
1489 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001490}
1491
1492/**
1493 * sock_poll_wait - place memory barrier behind the poll_wait call.
1494 * @filp: file
1495 * @wait_address: socket wait queue
1496 * @p: poll_table
1497 *
Eric Dumazet43815482010-04-29 11:01:49 +00001498 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001499 */
1500static inline void sock_poll_wait(struct file *filp,
1501 wait_queue_head_t *wait_address, poll_table *p)
1502{
1503 if (p && wait_address) {
1504 poll_wait(filp, wait_address, p);
1505 /*
1506 * We need to be sure we are in sync with the
1507 * socket flags modification.
1508 *
Eric Dumazet43815482010-04-29 11:01:49 +00001509 * This memory barrier is paired in the wq_has_sleeper.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001510 */
1511 smp_mb();
1512 }
1513}
1514
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515/*
1516 * Queue a received datagram if it will fit. Stream and sequenced
1517 * protocols can't normally use this as they need to fit buffers in
1518 * and play with them.
1519 *
1520 * Inlined as it's very short and called for pretty much every
1521 * packet ever received.
1522 */
1523
1524static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1525{
Herbert Xud55d87f2009-06-22 02:25:25 +00001526 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 skb->sk = sk;
1528 skb->destructor = sock_wfree;
Eric Dumazet2b85a342009-06-11 02:55:43 -07001529 /*
1530 * We used to take a refcount on sk, but following operation
1531 * is enough to guarantee sk_free() wont free this sock until
1532 * all in-flight packets are completed
1533 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1535}
1536
1537static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1538{
Herbert Xud55d87f2009-06-22 02:25:25 +00001539 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 skb->sk = sk;
1541 skb->destructor = sock_rfree;
1542 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001543 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544}
1545
1546extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1547 unsigned long expires);
1548
1549extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1550
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001551extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552
Eric Dumazetb1faf562010-05-31 23:44:05 -07001553extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554
1555/*
1556 * Recover an error report and clear atomically
1557 */
1558
1559static inline int sock_error(struct sock *sk)
1560{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001561 int err;
1562 if (likely(!sk->sk_err))
1563 return 0;
1564 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 return -err;
1566}
1567
1568static inline unsigned long sock_wspace(struct sock *sk)
1569{
1570 int amt = 0;
1571
1572 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1573 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1574 if (amt < 0)
1575 amt = 0;
1576 }
1577 return amt;
1578}
1579
1580static inline void sk_wake_async(struct sock *sk, int how, int band)
1581{
Eric Dumazetbcdce712009-10-06 17:28:29 -07001582 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 sock_wake_async(sk->sk_socket, how, band);
1584}
1585
1586#define SOCK_MIN_SNDBUF 2048
Eric Dumazet7a91b4342010-09-26 18:53:07 -07001587/*
1588 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1589 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1590 */
1591#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592
1593static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1594{
1595 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001596 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1598 }
1599}
1600
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11001601struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602
1603static inline struct page *sk_stream_alloc_page(struct sock *sk)
1604{
1605 struct page *page = NULL;
1606
Herbert Xuef015782005-09-01 17:48:59 -07001607 page = alloc_pages(sk->sk_allocation, 0);
1608 if (!page) {
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001609 sk->sk_prot->enter_memory_pressure(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610 sk_stream_moderate_sndbuf(sk);
1611 }
1612 return page;
1613}
1614
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615/*
1616 * Default write policy as shown to user space via poll/select/SIGIO
1617 */
1618static inline int sock_writeable(const struct sock *sk)
1619{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001620 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621}
1622
Al Virodd0fc662005-10-07 07:46:04 +01001623static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624{
Andrew Morton99709372009-02-12 16:43:17 -08001625 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626}
1627
1628static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1629{
1630 return noblock ? 0 : sk->sk_rcvtimeo;
1631}
1632
1633static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1634{
1635 return noblock ? 0 : sk->sk_sndtimeo;
1636}
1637
1638static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1639{
1640 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1641}
1642
1643/* Alas, with timeout socket operations are not restartable.
1644 * Compare this to poll().
1645 */
1646static inline int sock_intr_errno(long timeo)
1647{
1648 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1649}
1650
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001651extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1652 struct sk_buff *skb);
1653
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654static __inline__ void
1655sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1656{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001657 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00001658 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001659
Patrick Ohly20d49472009-02-12 05:03:38 +00001660 /*
1661 * generate control messages if
1662 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1663 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1664 * - software time stamp available and wanted
1665 * (SOCK_TIMESTAMPING_SOFTWARE)
1666 * - hardware time stamps available and wanted
1667 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1668 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1669 */
1670 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1671 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1672 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1673 (hwtstamps->hwtstamp.tv64 &&
1674 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1675 (hwtstamps->syststamp.tv64 &&
1676 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001677 __sock_recv_timestamp(msg, sk, skb);
1678 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001679 sk->sk_stamp = kt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680}
1681
Eric Dumazet767dd032010-04-28 19:14:43 +00001682extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1683 struct sk_buff *skb);
1684
1685static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1686 struct sk_buff *skb)
1687{
1688#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
1689 (1UL << SOCK_RCVTSTAMP) | \
1690 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
1691 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
1692 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
1693 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
1694
1695 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
1696 __sock_recv_ts_and_drops(msg, sk, skb);
1697 else
1698 sk->sk_stamp = skb->tstamp;
1699}
Neil Horman3b885782009-10-12 13:26:31 -07001700
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701/**
Patrick Ohly20d49472009-02-12 05:03:38 +00001702 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00001703 * @sk: socket sending this packet
Oliver Hartkopp2244d072010-08-17 08:59:14 +00001704 * @tx_flags: filled with instructions for time stamping
Patrick Ohly20d49472009-02-12 05:03:38 +00001705 *
1706 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1707 * parameters are invalid.
1708 */
Oliver Hartkopp2244d072010-08-17 08:59:14 +00001709extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
Patrick Ohly20d49472009-02-12 05:03:38 +00001710
1711/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001713 * @sk: socket to eat this skb from
1714 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07001715 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 *
1717 * This routine must be called with interrupts disabled or with the socket
1718 * locked so that the sk_buff queue operation is ok.
1719*/
Chris Leech624d1162006-05-23 18:01:28 -07001720#ifdef CONFIG_NET_DMA
1721static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1722{
1723 __skb_unlink(skb, &sk->sk_receive_queue);
1724 if (!copied_early)
1725 __kfree_skb(skb);
1726 else
1727 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1728}
1729#else
1730static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731{
1732 __skb_unlink(skb, &sk->sk_receive_queue);
1733 __kfree_skb(skb);
1734}
Chris Leech624d1162006-05-23 18:01:28 -07001735#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001737static inline
1738struct net *sock_net(const struct sock *sk)
1739{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001740 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001741}
1742
1743static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07001744void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001745{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001746 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001747}
1748
Denis V. Lunevedf02082008-02-29 11:18:32 -08001749/*
1750 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1751 * They should not hold a referrence to a namespace in order to allow
1752 * to stop it.
1753 * Sockets after sk_change_net should be released using sk_release_kernel
1754 */
1755static inline void sk_change_net(struct sock *sk, struct net *net)
1756{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001757 put_net(sock_net(sk));
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07001758 sock_net_set(sk, hold_net(net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08001759}
1760
KOVACS Krisztian23542612008-10-07 12:41:01 -07001761static inline struct sock *skb_steal_sock(struct sk_buff *skb)
1762{
1763 if (unlikely(skb->sk)) {
1764 struct sock *sk = skb->sk;
1765
1766 skb->destructor = NULL;
1767 skb->sk = NULL;
1768 return sk;
1769 }
1770 return NULL;
1771}
1772
Patrick Ohly20d49472009-02-12 05:03:38 +00001773extern void sock_enable_timestamp(struct sock *sk, int flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774extern int sock_get_timestamp(struct sock *, struct timeval __user *);
Eric Dumazetae40eb12007-03-18 17:33:16 -07001775extern int sock_get_timestampns(struct sock *, struct timespec __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776
1777/*
1778 * Enable debug/info messages
1779 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08001780extern int net_msg_warn;
1781#define NETDEBUG(fmt, args...) \
1782 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08001784#define LIMIT_NETDEBUG(fmt, args...) \
1785 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787extern __u32 sysctl_wmem_max;
1788extern __u32 sysctl_rmem_max;
1789
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001790extern void sk_init(void);
1791
David S. Miller6baf1f42005-09-05 18:14:11 -07001792extern int sysctl_optmem_max;
1793
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001794extern __u32 sysctl_wmem_default;
1795extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001796
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797#endif /* _SOCK_H */