blob: f673ba5b6b1a1066f413c389e6e1b818aa8c33c2 [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
Alexey Dobriyana6b7a402011-06-06 10:43:46 +000043#include <linux/hardirq.h>
Ilpo Järvinen172589c2007-08-28 15:50:33 -070044#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#include <linux/list.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080046#include <linux/list_nulls.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070047#include <linux/timer.h>
48#include <linux/cache.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>
Tom Herbertc6e1a0d2011-04-04 22:30:30 -070055#include <linux/uaccess.h>
Glauber Costa180d8cd2011-12-11 21:47:02 +000056#include <linux/memcontrol.h>
Glauber Costae1aab162011-12-11 21:47:03 +000057#include <linux/res_counter.h>
Ingo Molnarc5905af2012-02-24 08:31:31 +010058#include <linux/static_key.h>
Al Viro40401532012-02-13 03:58:52 +000059#include <linux/aio.h>
60#include <linux/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
62#include <linux/filter.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080063#include <linux/rculist_nulls.h>
Jiri Olsaa57de0b2009-07-08 12:09:13 +000064#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070065
Eric Dumazetc31504d2010-11-15 19:58:26 +000066#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <net/dst.h>
68#include <net/checksum.h>
69
Eric Dumazet9f048bf2011-12-13 03:59:08 +000070struct cgroup;
71struct cgroup_subsys;
Glauber Costac607b2e2011-12-16 00:52:00 +000072#ifdef CONFIG_NET
Glauber Costad1a4c0b2011-12-11 21:47:04 +000073int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss);
Li Zefan761b3ef2012-01-31 13:47:36 +080074void mem_cgroup_sockets_destroy(struct cgroup *cgrp);
Glauber Costac607b2e2011-12-16 00:52:00 +000075#else
76static inline
77int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss)
78{
79 return 0;
80}
81static inline
Li Zefan761b3ef2012-01-31 13:47:36 +080082void mem_cgroup_sockets_destroy(struct cgroup *cgrp)
Glauber Costac607b2e2011-12-16 00:52:00 +000083{
84}
85#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070086/*
87 * This structure really needs to be cleaned up.
88 * Most of it is for TCP, and not used by any of
89 * the other protocols.
90 */
91
92/* Define this to get the SOCK_DBG debugging facility. */
93#define SOCK_DEBUGGING
94#ifdef SOCK_DEBUGGING
95#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
96 printk(KERN_DEBUG msg); } while (0)
97#else
Stephen Hemminger4cd90292008-03-21 15:54:53 -070098/* Validate arguments and do nothing */
Joe Perchesb9075fa2011-10-31 17:11:33 -070099static inline __printf(2, 3)
100void SOCK_DEBUG(struct sock *sk, const char *msg, ...)
Stephen Hemminger4cd90292008-03-21 15:54:53 -0700101{
102}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103#endif
104
105/* This is the per-socket lock. The spinlock provides a synchronization
106 * between user contexts and software interrupt processing, whereas the
107 * mini-semaphore synchronizes multiple users amongst themselves.
108 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109typedef struct {
110 spinlock_t slock;
John Heffnerd2e91172007-09-12 10:44:19 +0200111 int owned;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 wait_queue_head_t wq;
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700113 /*
114 * We express the mutex-alike socket_lock semantics
115 * to the lock validator by explicitly managing
116 * the slock as a lock variant (in addition to
117 * the slock itself):
118 */
119#ifdef CONFIG_DEBUG_LOCK_ALLOC
120 struct lockdep_map dep_map;
121#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122} socket_lock_t;
123
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124struct sock;
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700125struct proto;
Denis V. Lunev0eeb8ff2007-12-04 01:15:45 -0800126struct net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127
128/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700129 * struct sock_common - minimal network layer representation of sockets
Eric Dumazet68835ab2010-11-30 19:04:07 +0000130 * @skc_daddr: Foreign IPv4 addr
131 * @skc_rcv_saddr: Bound local IPv4 addr
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000132 * @skc_hash: hash value used with various protocol lookup tables
Eric Dumazetd4cada42009-11-08 10:17:30 +0000133 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700134 * @skc_family: network address family
135 * @skc_state: Connection state
136 * @skc_reuse: %SO_REUSEADDR setting
137 * @skc_bound_dev_if: bound device index if != 0
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700138 * @skc_bind_node: bind hash linkage for various protocol lookup tables
Eric Dumazet512615b2009-11-08 10:17:58 +0000139 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700140 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200141 * @skc_net: reference to the network namespace of this socket
Eric Dumazet68835ab2010-11-30 19:04:07 +0000142 * @skc_node: main hash linkage for various protocol lookup tables
143 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
144 * @skc_tx_queue_mapping: tx queue number for this connection
145 * @skc_refcnt: reference count
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700146 *
147 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700148 * for struct sock and struct inet_timewait_sock.
149 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150struct sock_common {
Eric Dumazet68835ab2010-11-30 19:04:07 +0000151 /* skc_daddr and skc_rcv_saddr must be grouped :
152 * cf INET_MATCH() and INET_TW_MATCH()
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000153 */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000154 __be32 skc_daddr;
155 __be32 skc_rcv_saddr;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000156
Eric Dumazetd4cada42009-11-08 10:17:30 +0000157 union {
158 unsigned int skc_hash;
159 __u16 skc_u16hashes[2];
160 };
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000161 unsigned short skc_family;
162 volatile unsigned char skc_state;
163 unsigned char skc_reuse;
164 int skc_bound_dev_if;
Eric Dumazet512615b2009-11-08 10:17:58 +0000165 union {
166 struct hlist_node skc_bind_node;
167 struct hlist_nulls_node skc_portaddr_node;
168 };
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700169 struct proto *skc_prot;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900170#ifdef CONFIG_NET_NS
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200171 struct net *skc_net;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900172#endif
Eric Dumazet68835ab2010-11-30 19:04:07 +0000173 /*
174 * fields between dontcopy_begin/dontcopy_end
175 * are not copied in sock_copy()
176 */
Randy Dunlap928c41e2011-01-08 17:39:21 +0000177 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000178 int skc_dontcopy_begin[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000179 /* public: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000180 union {
181 struct hlist_node skc_node;
182 struct hlist_nulls_node skc_nulls_node;
183 };
184 int skc_tx_queue_mapping;
185 atomic_t skc_refcnt;
Randy Dunlap928c41e2011-01-08 17:39:21 +0000186 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000187 int skc_dontcopy_end[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000188 /* public: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189};
190
Glauber Costae1aab162011-12-11 21:47:03 +0000191struct cg_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192/**
193 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700194 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700195 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
196 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
197 * @sk_lock: synchronizer
198 * @sk_rcvbuf: size of receive buffer in bytes
Eric Dumazet43815482010-04-29 11:01:49 +0000199 * @sk_wq: sock wait queue and async head
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700200 * @sk_dst_cache: destination cache
201 * @sk_dst_lock: destination cache lock
202 * @sk_policy: flow policy
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700203 * @sk_receive_queue: incoming packets
204 * @sk_wmem_alloc: transmit queue bytes committed
205 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700206 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700207 * @sk_omem_alloc: "o" is "option" or "other"
208 * @sk_wmem_queued: persistent queue size
209 * @sk_forward_alloc: space allocated forward
210 * @sk_allocation: allocation mode
211 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800212 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
Patrick Ohly20d49472009-02-12 05:03:38 +0000213 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700214 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
215 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700216 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700217 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700218 * @sk_gso_max_size: Maximum GSO segment size to build
Ben Hutchings4658b242012-07-30 16:11:42 +0000219 * @sk_gso_max_segs: Maximum number of GSO segments
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700220 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700221 * @sk_backlog: always used with the per-socket spinlock held
222 * @sk_callback_lock: used with the callbacks in the end of this struct
223 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800224 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
225 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700226 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800227 * @sk_err_soft: errors that don't cause failure but are the cause of a
228 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700229 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700230 * @sk_ack_backlog: current listen backlog
231 * @sk_max_ack_backlog: listen backlog set in listen()
232 * @sk_priority: %SO_PRIORITY setting
Randy Dunlap1a3bc362012-01-21 09:03:10 +0000233 * @sk_cgrp_prioidx: socket group's priority map index
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700234 * @sk_type: socket type (%SOCK_STREAM, etc)
235 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000236 * @sk_peer_pid: &struct pid for this socket's peer
237 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700238 * @sk_rcvlowat: %SO_RCVLOWAT setting
239 * @sk_rcvtimeo: %SO_RCVTIMEO setting
240 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700241 * @sk_rxhash: flow hash received from netif layer
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700242 * @sk_filter: socket filtering instructions
243 * @sk_protinfo: private area, net family specific, when not using slab
244 * @sk_timer: sock cleanup timer
245 * @sk_stamp: time stamp of last packet received
246 * @sk_socket: Identd and reporting IO signals
247 * @sk_user_data: RPC layer private data
248 * @sk_sndmsg_page: cached page for sendmsg
249 * @sk_sndmsg_off: cached offset for sendmsg
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000250 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700251 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700252 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800253 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000254 * @sk_classid: this socket's cgroup classid
Glauber Costae1aab162011-12-11 21:47:03 +0000255 * @sk_cgrp: this socket's cgroup-specific proto data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700256 * @sk_write_pending: a write to stream socket waits to start
257 * @sk_state_change: callback to indicate change in the state of the sock
258 * @sk_data_ready: callback to indicate there is data to be processed
259 * @sk_write_space: callback to indicate there is bf sending space available
260 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
261 * @sk_backlog_rcv: callback to process the backlog
262 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 */
264struct sock {
265 /*
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700266 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 * don't add nothing before this first member (__sk_common) --acme
268 */
269 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000270#define sk_node __sk_common.skc_node
271#define sk_nulls_node __sk_common.skc_nulls_node
272#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000273#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000274
Eric Dumazet68835ab2010-11-30 19:04:07 +0000275#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
276#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000277#define sk_hash __sk_common.skc_hash
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278#define sk_family __sk_common.skc_family
279#define sk_state __sk_common.skc_state
280#define sk_reuse __sk_common.skc_reuse
281#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700283#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200284#define sk_net __sk_common.skc_net
Eric Dumazetb178bb32010-11-16 05:56:04 +0000285 socket_lock_t sk_lock;
286 struct sk_buff_head sk_receive_queue;
287 /*
288 * The backlog queue is special, it is always used with
289 * the per-socket spinlock held and requires low latency
290 * access. Therefore we special case it's implementation.
291 * Note : rmem_alloc is in this structure to fill a hole
292 * on 64bit arches, not because its logically part of
293 * backlog.
294 */
295 struct {
296 atomic_t rmem_alloc;
297 int len;
298 struct sk_buff *head;
299 struct sk_buff *tail;
300 } sk_backlog;
301#define sk_rmem_alloc sk_backlog.rmem_alloc
302 int sk_forward_alloc;
303#ifdef CONFIG_RPS
304 __u32 sk_rxhash;
305#endif
306 atomic_t sk_drops;
307 int sk_rcvbuf;
308
309 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000310 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000311
312#ifdef CONFIG_NET_DMA
313 struct sk_buff_head sk_async_wait_queue;
314#endif
315
316#ifdef CONFIG_XFRM
317 struct xfrm_policy *sk_policy[2];
318#endif
319 unsigned long sk_flags;
320 struct dst_entry *sk_dst_cache;
321 spinlock_t sk_dst_lock;
322 atomic_t sk_wmem_alloc;
323 atomic_t sk_omem_alloc;
324 int sk_sndbuf;
325 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100326 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000327 unsigned int sk_shutdown : 2,
328 sk_no_check : 2,
329 sk_userlocks : 4,
330 sk_protocol : 8,
331 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100332 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400334 gfp_t sk_allocation;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000335 netdev_features_t sk_route_caps;
336 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700337 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700338 unsigned int sk_gso_max_size;
Ben Hutchings4658b242012-07-30 16:11:42 +0000339 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800340 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700343 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 rwlock_t sk_callback_lock;
345 int sk_err,
346 sk_err_soft;
347 unsigned short sk_ack_backlog;
348 unsigned short sk_max_ack_backlog;
349 __u32 sk_priority;
Neil Horman5bc14212011-11-22 05:10:51 +0000350#ifdef CONFIG_CGROUPS
351 __u32 sk_cgrp_prioidx;
352#endif
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000353 struct pid *sk_peer_pid;
354 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 long sk_rcvtimeo;
356 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 void *sk_protinfo;
358 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700359 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 struct socket *sk_socket;
361 void *sk_user_data;
362 struct page *sk_sndmsg_page;
363 struct sk_buff *sk_send_head;
364 __u32 sk_sndmsg_off;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000365 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800367#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800369#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800370 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700371 u32 sk_classid;
Glauber Costae1aab162011-12-11 21:47:03 +0000372 struct cg_proto *sk_cgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 void (*sk_state_change)(struct sock *sk);
374 void (*sk_data_ready)(struct sock *sk, int bytes);
375 void (*sk_write_space)(struct sock *sk);
376 void (*sk_error_report)(struct sock *sk);
377 int (*sk_backlog_rcv)(struct sock *sk,
378 struct sk_buff *skb);
379 void (*sk_destruct)(struct sock *sk);
380};
381
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000382static inline int sk_peek_offset(struct sock *sk, int flags)
383{
384 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
385 return sk->sk_peek_off;
386 else
387 return 0;
388}
389
390static inline void sk_peek_offset_bwd(struct sock *sk, int val)
391{
392 if (sk->sk_peek_off >= 0) {
393 if (sk->sk_peek_off >= val)
394 sk->sk_peek_off -= val;
395 else
396 sk->sk_peek_off = 0;
397 }
398}
399
400static inline void sk_peek_offset_fwd(struct sock *sk, int val)
401{
402 if (sk->sk_peek_off >= 0)
403 sk->sk_peek_off += val;
404}
405
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406/*
407 * Hashed lists helper routines
408 */
Li Zefanc4146642010-02-08 23:18:45 +0000409static inline struct sock *sk_entry(const struct hlist_node *node)
410{
411 return hlist_entry(node, struct sock, sk_node);
412}
413
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700414static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415{
416 return hlist_entry(head->first, struct sock, sk_node);
417}
418
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700419static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420{
421 return hlist_empty(head) ? NULL : __sk_head(head);
422}
423
Eric Dumazet88ab1932008-11-16 19:39:21 -0800424static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
425{
426 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
427}
428
429static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
430{
431 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
432}
433
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700434static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435{
436 return sk->sk_node.next ?
437 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
438}
439
Eric Dumazet88ab1932008-11-16 19:39:21 -0800440static inline struct sock *sk_nulls_next(const struct sock *sk)
441{
442 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
443 hlist_nulls_entry(sk->sk_nulls_node.next,
444 struct sock, sk_nulls_node) :
445 NULL;
446}
447
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700448static inline int sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449{
450 return hlist_unhashed(&sk->sk_node);
451}
452
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700453static inline int sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454{
Akinobu Mitada753be2006-04-28 15:21:23 -0700455 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456}
457
458static __inline__ void sk_node_init(struct hlist_node *node)
459{
460 node->pprev = NULL;
461}
462
Eric Dumazet88ab1932008-11-16 19:39:21 -0800463static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
464{
465 node->pprev = NULL;
466}
467
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468static __inline__ void __sk_del_node(struct sock *sk)
469{
470 __hlist_del(&sk->sk_node);
471}
472
stephen hemminger808f5112010-02-22 07:57:18 +0000473/* NB: equivalent to hlist_del_init_rcu */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474static __inline__ int __sk_del_node_init(struct sock *sk)
475{
476 if (sk_hashed(sk)) {
477 __sk_del_node(sk);
478 sk_node_init(&sk->sk_node);
479 return 1;
480 }
481 return 0;
482}
483
484/* Grab socket reference count. This operation is valid only
485 when sk is ALREADY grabbed f.e. it is found in hash table
486 or a list and the lookup is made under lock preventing hash table
487 modifications.
488 */
489
490static inline void sock_hold(struct sock *sk)
491{
492 atomic_inc(&sk->sk_refcnt);
493}
494
495/* Ungrab socket in the context, which assumes that socket refcnt
496 cannot hit zero, f.e. it is true in context of any socketcall.
497 */
498static inline void __sock_put(struct sock *sk)
499{
500 atomic_dec(&sk->sk_refcnt);
501}
502
503static __inline__ int sk_del_node_init(struct sock *sk)
504{
505 int rc = __sk_del_node_init(sk);
506
507 if (rc) {
508 /* paranoid for a while -acme */
509 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
510 __sock_put(sk);
511 }
512 return rc;
513}
stephen hemminger808f5112010-02-22 07:57:18 +0000514#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Eric Dumazet88ab1932008-11-16 19:39:21 -0800516static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700517{
518 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800519 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700520 return 1;
521 }
522 return 0;
523}
524
Eric Dumazet88ab1932008-11-16 19:39:21 -0800525static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700526{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800527 int rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700528
529 if (rc) {
530 /* paranoid for a while -acme */
531 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
532 __sock_put(sk);
533 }
534 return rc;
535}
536
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
538{
539 hlist_add_head(&sk->sk_node, list);
540}
541
542static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
543{
544 sock_hold(sk);
545 __sk_add_node(sk, list);
546}
547
stephen hemminger808f5112010-02-22 07:57:18 +0000548static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
549{
550 sock_hold(sk);
551 hlist_add_head_rcu(&sk->sk_node, list);
552}
553
Eric Dumazet88ab1932008-11-16 19:39:21 -0800554static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700555{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800556 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700557}
558
Eric Dumazet88ab1932008-11-16 19:39:21 -0800559static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700560{
561 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800562 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700563}
564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565static __inline__ void __sk_del_bind_node(struct sock *sk)
566{
567 __hlist_del(&sk->sk_bind_node);
568}
569
570static __inline__ void sk_add_bind_node(struct sock *sk,
571 struct hlist_head *list)
572{
573 hlist_add_head(&sk->sk_bind_node, list);
574}
575
576#define sk_for_each(__sk, node, list) \
577 hlist_for_each_entry(__sk, node, list, sk_node)
stephen hemminger808f5112010-02-22 07:57:18 +0000578#define sk_for_each_rcu(__sk, node, list) \
579 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800580#define sk_nulls_for_each(__sk, node, list) \
581 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
582#define sk_nulls_for_each_rcu(__sk, node, list) \
583 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584#define sk_for_each_from(__sk, node) \
585 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
586 hlist_for_each_entry_from(__sk, node, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800587#define sk_nulls_for_each_from(__sk, node) \
588 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
589 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590#define sk_for_each_safe(__sk, node, tmp, list) \
591 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
592#define sk_for_each_bound(__sk, node, list) \
593 hlist_for_each_entry(__sk, node, list, sk_bind_node)
594
595/* Sock flags */
596enum sock_flags {
597 SOCK_DEAD,
598 SOCK_DONE,
599 SOCK_URGINLINE,
600 SOCK_KEEPOPEN,
601 SOCK_LINGER,
602 SOCK_DESTROY,
603 SOCK_BROADCAST,
604 SOCK_TIMESTAMP,
605 SOCK_ZAPPED,
606 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
607 SOCK_DBG, /* %SO_DEBUG setting */
608 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700609 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
611 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Patrick Ohly20d49472009-02-12 05:03:38 +0000612 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
613 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
614 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
615 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
616 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
617 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
618 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700619 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700620 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000621 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100622 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000623 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
624 * Will use last 4 bytes of packet sent from
625 * user-space instead.
626 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627};
628
Ralf Baechle53b924b2005-08-23 10:11:30 -0700629static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
630{
631 nsk->sk_flags = osk->sk_flags;
632}
633
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
635{
636 __set_bit(flag, &sk->sk_flags);
637}
638
639static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
640{
641 __clear_bit(flag, &sk->sk_flags);
642}
643
644static inline int sock_flag(struct sock *sk, enum sock_flags flag)
645{
646 return test_bit(flag, &sk->sk_flags);
647}
648
649static inline void sk_acceptq_removed(struct sock *sk)
650{
651 sk->sk_ack_backlog--;
652}
653
654static inline void sk_acceptq_added(struct sock *sk)
655{
656 sk->sk_ack_backlog++;
657}
658
659static inline int sk_acceptq_is_full(struct sock *sk)
660{
David S. Miller64a14652007-03-06 11:21:05 -0800661 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662}
663
664/*
665 * Compute minimal free write space needed to queue new packets.
666 */
667static inline int sk_stream_min_wspace(struct sock *sk)
668{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800669 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670}
671
672static inline int sk_stream_wspace(struct sock *sk)
673{
674 return sk->sk_sndbuf - sk->sk_wmem_queued;
675}
676
677extern void sk_stream_write_space(struct sock *sk);
678
679static inline int sk_stream_memory_free(struct sock *sk)
680{
681 return sk->sk_wmem_queued < sk->sk_sndbuf;
682}
683
Zhu Yi8eae9392010-03-04 18:01:40 +0000684/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000685static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800686{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000687 /* dont let skb dst not refcounted, we are going to leave rcu lock */
688 skb_dst_force(skb);
689
690 if (!sk->sk_backlog.tail)
691 sk->sk_backlog.head = skb;
692 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800693 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000694
695 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800696 skb->next = NULL;
697}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
Eric Dumazetc3774112010-04-27 15:13:20 -0700699/*
700 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000701 * Do not take into account this skb truesize,
702 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700703 */
704static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
705{
706 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
707
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000708 return qsize > sk->sk_rcvbuf;
Eric Dumazetc3774112010-04-27 15:13:20 -0700709}
710
Zhu Yi8eae9392010-03-04 18:01:40 +0000711/* The per-socket spinlock must be held here. */
Zhu Yi40456352010-03-07 16:21:39 +0000712static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Zhu Yi8eae9392010-03-04 18:01:40 +0000713{
Eric Dumazetc3774112010-04-27 15:13:20 -0700714 if (sk_rcvqueues_full(sk, skb))
Zhu Yi8eae9392010-03-04 18:01:40 +0000715 return -ENOBUFS;
716
Zhu Yia3a858f2010-03-04 18:01:47 +0000717 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000718 sk->sk_backlog.len += skb->truesize;
719 return 0;
720}
721
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700722static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
723{
724 return sk->sk_backlog_rcv(sk, skb);
725}
726
David S. Millerc58dc012010-04-27 15:05:31 -0700727static inline void sock_rps_record_flow(const struct sock *sk)
728{
729#ifdef CONFIG_RPS
730 struct rps_sock_flow_table *sock_flow_table;
731
732 rcu_read_lock();
733 sock_flow_table = rcu_dereference(rps_sock_flow_table);
734 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
735 rcu_read_unlock();
736#endif
737}
738
739static inline void sock_rps_reset_flow(const struct sock *sk)
740{
741#ifdef CONFIG_RPS
742 struct rps_sock_flow_table *sock_flow_table;
743
744 rcu_read_lock();
745 sock_flow_table = rcu_dereference(rps_sock_flow_table);
746 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
747 rcu_read_unlock();
748#endif
749}
750
Tom Herbertbdeab992011-08-14 19:45:55 +0000751static inline void sock_rps_save_rxhash(struct sock *sk,
752 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700753{
754#ifdef CONFIG_RPS
Tom Herbertbdeab992011-08-14 19:45:55 +0000755 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
David S. Millerc58dc012010-04-27 15:05:31 -0700756 sock_rps_reset_flow(sk);
Tom Herbertbdeab992011-08-14 19:45:55 +0000757 sk->sk_rxhash = skb->rxhash;
David S. Millerc58dc012010-04-27 15:05:31 -0700758 }
759#endif
760}
761
Tom Herbertbdeab992011-08-14 19:45:55 +0000762static inline void sock_rps_reset_rxhash(struct sock *sk)
763{
764#ifdef CONFIG_RPS
765 sock_rps_reset_flow(sk);
766 sk->sk_rxhash = 0;
767#endif
768}
769
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700770#define sk_wait_event(__sk, __timeo, __condition) \
771 ({ int __rc; \
772 release_sock(__sk); \
773 __rc = __condition; \
774 if (!__rc) { \
775 *(__timeo) = schedule_timeout(*(__timeo)); \
776 } \
777 lock_sock(__sk); \
778 __rc = __condition; \
779 __rc; \
780 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781
782extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
783extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
784extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
785extern int sk_stream_error(struct sock *sk, int flags, int err);
786extern void sk_stream_kill_queues(struct sock *sk);
787
788extern int sk_wait_data(struct sock *sk, long *timeo);
789
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700790struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800791struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800792struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700793struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400794struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700795
Eric Dumazet7d9577d2013-05-09 10:28:16 +0000796/*
797 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
798 * un-modified. Special care is taken when initializing object to zero.
799 */
800static inline void sk_prot_clear_nulls(struct sock *sk, int size)
801{
802 if (offsetof(struct sock, sk_node.next) != 0)
803 memset(sk, 0, offsetof(struct sock, sk_node.next));
804 memset(&sk->sk_node.pprev, 0,
805 size - offsetof(struct sock, sk_node.pprev));
806}
807
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808/* Networking protocol blocks we attach to sockets.
809 * socket layer -> transport layer interface
810 * transport -> network interface is defined by struct inet_proto
811 */
812struct proto {
813 void (*close)(struct sock *sk,
814 long timeout);
815 int (*connect)(struct sock *sk,
816 struct sockaddr *uaddr,
817 int addr_len);
818 int (*disconnect)(struct sock *sk, int flags);
819
820 struct sock * (*accept) (struct sock *sk, int flags, int *err);
821
822 int (*ioctl)(struct sock *sk, int cmd,
823 unsigned long arg);
824 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700825 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 void (*shutdown)(struct sock *sk, int how);
827 int (*setsockopt)(struct sock *sk, int level,
828 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700829 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 int (*getsockopt)(struct sock *sk, int level,
831 int optname, char __user *optval,
832 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700833#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800834 int (*compat_setsockopt)(struct sock *sk,
835 int level,
836 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700837 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800838 int (*compat_getsockopt)(struct sock *sk,
839 int level,
840 int optname, char __user *optval,
841 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000842 int (*compat_ioctl)(struct sock *sk,
843 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700844#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
846 struct msghdr *msg, size_t len);
847 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
848 struct msghdr *msg,
849 size_t len, int noblock, int flags,
850 int *addr_len);
851 int (*sendpage)(struct sock *sk, struct page *page,
852 int offset, size_t size, int flags);
853 int (*bind)(struct sock *sk,
854 struct sockaddr *uaddr, int addr_len);
855
856 int (*backlog_rcv) (struct sock *sk,
857 struct sk_buff *skb);
858
859 /* Keeping track of sk's, looking them up, and port selection methods. */
860 void (*hash)(struct sock *sk);
861 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000862 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800864 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800866 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800867#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700868 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800869#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800870
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700872 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +0000873 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -0800874 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 /*
876 * Pressure flag: try to collapse.
877 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800878 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 * is strict, actions are advisory and have some latency.
880 */
881 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000882 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 int *sysctl_wmem;
884 int *sysctl_rmem;
885 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +0000886 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887
Eric Dumazet271b72c2008-10-29 02:11:14 -0700888 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700890 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891
Eric Dumazetdd24c002008-11-25 21:17:14 -0800892 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700893
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700894 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800895 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700896
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700897 union {
898 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -0700899 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700900 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700901 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800902
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 struct module *owner;
904
905 char name[32];
906
907 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700908#ifdef SOCK_REFCNT_DEBUG
909 atomic_t socks;
910#endif
Glauber Costae1aab162011-12-11 21:47:03 +0000911#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
912 /*
913 * cgroup specific init/deinit functions. Called once for all
914 * protocols that implement it, from cgroups populate function.
915 * This function has to setup any files the protocol want to
916 * appear in the kmem cgroup filesystem.
917 */
918 int (*init_cgroup)(struct cgroup *cgrp,
919 struct cgroup_subsys *ss);
Li Zefan761b3ef2012-01-31 13:47:36 +0800920 void (*destroy_cgroup)(struct cgroup *cgrp);
Glauber Costae1aab162011-12-11 21:47:03 +0000921 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
922#endif
923};
924
925struct cg_proto {
926 void (*enter_memory_pressure)(struct sock *sk);
927 struct res_counter *memory_allocated; /* Current allocated memory. */
928 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
929 int *memory_pressure;
930 long *sysctl_mem;
931 /*
932 * memcg field is used to find which memcg we belong directly
933 * Each memcg struct can hold more than one cg_proto, so container_of
934 * won't really cut.
935 *
936 * The elegant solution would be having an inverse function to
937 * proto_cgroup in struct proto, but that means polluting the structure
938 * for everybody, instead of just for memcg users.
939 */
940 struct mem_cgroup *memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941};
942
943extern int proto_register(struct proto *prot, int alloc_slab);
944extern void proto_unregister(struct proto *prot);
945
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700946#ifdef SOCK_REFCNT_DEBUG
947static inline void sk_refcnt_debug_inc(struct sock *sk)
948{
949 atomic_inc(&sk->sk_prot->socks);
950}
951
952static inline void sk_refcnt_debug_dec(struct sock *sk)
953{
954 atomic_dec(&sk->sk_prot->socks);
955 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
956 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
957}
958
Ying Xue1e6b5fb2013-02-15 22:28:25 +0000959static inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700960{
961 if (atomic_read(&sk->sk_refcnt) != 1)
962 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
963 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
964}
965#else /* SOCK_REFCNT_DEBUG */
966#define sk_refcnt_debug_inc(sk) do { } while (0)
967#define sk_refcnt_debug_dec(sk) do { } while (0)
968#define sk_refcnt_debug_release(sk) do { } while (0)
969#endif /* SOCK_REFCNT_DEBUG */
970
Glauber Costa376be5f2012-01-20 04:57:14 +0000971#if defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) && defined(CONFIG_NET)
Ingo Molnarc5905af2012-02-24 08:31:31 +0100972extern struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +0000973static inline struct cg_proto *parent_cg_proto(struct proto *proto,
974 struct cg_proto *cg_proto)
975{
976 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
977}
Ingo Molnarc5905af2012-02-24 08:31:31 +0100978#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
Glauber Costae1aab162011-12-11 21:47:03 +0000979#else
980#define mem_cgroup_sockets_enabled 0
981static inline struct cg_proto *parent_cg_proto(struct proto *proto,
982 struct cg_proto *cg_proto)
983{
984 return NULL;
985}
986#endif
987
988
Glauber Costa180d8cd2011-12-11 21:47:02 +0000989static inline bool sk_has_memory_pressure(const struct sock *sk)
990{
991 return sk->sk_prot->memory_pressure != NULL;
992}
993
994static inline bool sk_under_memory_pressure(const struct sock *sk)
995{
996 if (!sk->sk_prot->memory_pressure)
997 return false;
Glauber Costae1aab162011-12-11 21:47:03 +0000998
999 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1000 return !!*sk->sk_cgrp->memory_pressure;
1001
Glauber Costa180d8cd2011-12-11 21:47:02 +00001002 return !!*sk->sk_prot->memory_pressure;
1003}
1004
1005static inline void sk_leave_memory_pressure(struct sock *sk)
1006{
1007 int *memory_pressure = sk->sk_prot->memory_pressure;
1008
Glauber Costae1aab162011-12-11 21:47:03 +00001009 if (!memory_pressure)
1010 return;
1011
1012 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001013 *memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001014
1015 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1016 struct cg_proto *cg_proto = sk->sk_cgrp;
1017 struct proto *prot = sk->sk_prot;
1018
1019 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1020 if (*cg_proto->memory_pressure)
1021 *cg_proto->memory_pressure = 0;
1022 }
1023
Glauber Costa180d8cd2011-12-11 21:47:02 +00001024}
1025
1026static inline void sk_enter_memory_pressure(struct sock *sk)
1027{
Glauber Costae1aab162011-12-11 21:47:03 +00001028 if (!sk->sk_prot->enter_memory_pressure)
1029 return;
1030
1031 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1032 struct cg_proto *cg_proto = sk->sk_cgrp;
1033 struct proto *prot = sk->sk_prot;
1034
1035 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1036 cg_proto->enter_memory_pressure(sk);
1037 }
1038
1039 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001040}
1041
1042static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1043{
1044 long *prot = sk->sk_prot->sysctl_mem;
Glauber Costae1aab162011-12-11 21:47:03 +00001045 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1046 prot = sk->sk_cgrp->sysctl_mem;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001047 return prot[index];
1048}
1049
Glauber Costae1aab162011-12-11 21:47:03 +00001050static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1051 unsigned long amt,
1052 int *parent_status)
1053{
1054 struct res_counter *fail;
1055 int ret;
1056
Glauber Costa0e90b312012-01-20 04:57:16 +00001057 ret = res_counter_charge_nofail(prot->memory_allocated,
1058 amt << PAGE_SHIFT, &fail);
Glauber Costae1aab162011-12-11 21:47:03 +00001059 if (ret < 0)
1060 *parent_status = OVER_LIMIT;
1061}
1062
1063static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1064 unsigned long amt)
1065{
1066 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1067}
1068
1069static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1070{
1071 u64 ret;
1072 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1073 return ret >> PAGE_SHIFT;
1074}
1075
Glauber Costa180d8cd2011-12-11 21:47:02 +00001076static inline long
1077sk_memory_allocated(const struct sock *sk)
1078{
1079 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001080 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1081 return memcg_memory_allocated_read(sk->sk_cgrp);
1082
Glauber Costa180d8cd2011-12-11 21:47:02 +00001083 return atomic_long_read(prot->memory_allocated);
1084}
1085
1086static inline long
Glauber Costae1aab162011-12-11 21:47:03 +00001087sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001088{
1089 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001090
1091 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1092 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1093 /* update the root cgroup regardless */
1094 atomic_long_add_return(amt, prot->memory_allocated);
1095 return memcg_memory_allocated_read(sk->sk_cgrp);
1096 }
1097
Glauber Costa180d8cd2011-12-11 21:47:02 +00001098 return atomic_long_add_return(amt, prot->memory_allocated);
1099}
1100
1101static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001102sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001103{
1104 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001105
Glauber Costa0e90b312012-01-20 04:57:16 +00001106 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Glauber Costae1aab162011-12-11 21:47:03 +00001107 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1108
Glauber Costa180d8cd2011-12-11 21:47:02 +00001109 atomic_long_sub(amt, prot->memory_allocated);
1110}
1111
1112static inline void sk_sockets_allocated_dec(struct sock *sk)
1113{
1114 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001115
1116 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1117 struct cg_proto *cg_proto = sk->sk_cgrp;
1118
1119 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1120 percpu_counter_dec(cg_proto->sockets_allocated);
1121 }
1122
Glauber Costa180d8cd2011-12-11 21:47:02 +00001123 percpu_counter_dec(prot->sockets_allocated);
1124}
1125
1126static inline void sk_sockets_allocated_inc(struct sock *sk)
1127{
1128 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001129
1130 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1131 struct cg_proto *cg_proto = sk->sk_cgrp;
1132
1133 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1134 percpu_counter_inc(cg_proto->sockets_allocated);
1135 }
1136
Glauber Costa180d8cd2011-12-11 21:47:02 +00001137 percpu_counter_inc(prot->sockets_allocated);
1138}
1139
1140static inline int
1141sk_sockets_allocated_read_positive(struct sock *sk)
1142{
1143 struct proto *prot = sk->sk_prot;
1144
Glauber Costae1aab162011-12-11 21:47:03 +00001145 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric Dumazet518fbf92012-04-28 23:21:56 +00001146 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001147
Eric Dumazet518fbf92012-04-28 23:21:56 +00001148 return percpu_counter_read_positive(prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001149}
1150
1151static inline int
1152proto_sockets_allocated_sum_positive(struct proto *prot)
1153{
1154 return percpu_counter_sum_positive(prot->sockets_allocated);
1155}
1156
1157static inline long
1158proto_memory_allocated(struct proto *prot)
1159{
1160 return atomic_long_read(prot->memory_allocated);
1161}
1162
1163static inline bool
1164proto_memory_pressure(struct proto *prot)
1165{
1166 if (!prot->memory_pressure)
1167 return false;
1168 return !!*prot->memory_pressure;
1169}
1170
Eric Dumazet65f76512008-01-03 20:46:48 -08001171
1172#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173/* Called with local bh disabled */
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001174extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1175extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001176#else
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001177static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
1178 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001179{
1180}
Eric Dumazet65f76512008-01-03 20:46:48 -08001181#endif
1182
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001184/* With per-bucket locks this operation is not-atomic, so that
1185 * this version is not worse.
1186 */
1187static inline void __sk_prot_rehash(struct sock *sk)
1188{
1189 sk->sk_prot->unhash(sk);
1190 sk->sk_prot->hash(sk);
1191}
1192
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001193void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1194
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195/* About 10 seconds */
1196#define SOCK_DESTROY_TIME (10*HZ)
1197
1198/* Sockets 0-1023 can't be bound to unless you are superuser */
1199#define PROT_SOCK 1024
1200
1201#define SHUTDOWN_MASK 3
1202#define RCV_SHUTDOWN 1
1203#define SEND_SHUTDOWN 2
1204
1205#define SOCK_SNDBUF_LOCK 1
1206#define SOCK_RCVBUF_LOCK 2
1207#define SOCK_BINDADDR_LOCK 4
1208#define SOCK_BINDPORT_LOCK 8
1209
1210/* sock_iocb: used to kick off async processing of socket ios */
1211struct sock_iocb {
1212 struct list_head list;
1213
1214 int flags;
1215 int size;
1216 struct socket *sock;
1217 struct sock *sk;
1218 struct scm_cookie *scm;
1219 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 struct kiocb *kiocb;
1221};
1222
1223static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1224{
1225 return (struct sock_iocb *)iocb->private;
1226}
1227
1228static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1229{
1230 return si->kiocb;
1231}
1232
1233struct socket_alloc {
1234 struct socket socket;
1235 struct inode vfs_inode;
1236};
1237
1238static inline struct socket *SOCKET_I(struct inode *inode)
1239{
1240 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1241}
1242
1243static inline struct inode *SOCK_INODE(struct socket *socket)
1244{
1245 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1246}
1247
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001248/*
1249 * Functions for memory accounting
1250 */
1251extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1252extern void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001254#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1255#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1256#define SK_MEM_SEND 0
1257#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001259static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001261 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262}
1263
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001264static inline int sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001266 /* return true if protocol supports memory accounting */
1267 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268}
1269
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001270static inline int sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001272 if (!sk_has_account(sk))
1273 return 1;
Herbert Xud80d99d2005-09-01 17:48:23 -07001274 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001275 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1276}
1277
1278static inline int sk_rmem_schedule(struct sock *sk, int size)
1279{
1280 if (!sk_has_account(sk))
1281 return 1;
1282 return size <= sk->sk_forward_alloc ||
1283 __sk_mem_schedule(sk, size, SK_MEM_RECV);
1284}
1285
1286static inline void sk_mem_reclaim(struct sock *sk)
1287{
1288 if (!sk_has_account(sk))
1289 return;
1290 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1291 __sk_mem_reclaim(sk);
1292}
1293
David S. Miller9993e7d2008-01-10 21:56:38 -08001294static inline void sk_mem_reclaim_partial(struct sock *sk)
1295{
1296 if (!sk_has_account(sk))
1297 return;
1298 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1299 __sk_mem_reclaim(sk);
1300}
1301
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001302static inline void sk_mem_charge(struct sock *sk, int size)
1303{
1304 if (!sk_has_account(sk))
1305 return;
1306 sk->sk_forward_alloc -= size;
1307}
1308
1309static inline void sk_mem_uncharge(struct sock *sk, int size)
1310{
1311 if (!sk_has_account(sk))
1312 return;
1313 sk->sk_forward_alloc += size;
1314}
1315
1316static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1317{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001318 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1319 sk->sk_wmem_queued -= skb->truesize;
1320 sk_mem_uncharge(sk, skb->truesize);
1321 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001322}
1323
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324/* Used by processes to "lock" a socket state, so that
1325 * interrupts and bottom half handlers won't change it
1326 * from under us. It essentially blocks any incoming
1327 * packets, so that we won't get any new data or any
1328 * packets that change the state of the socket.
1329 *
1330 * While locked, BH processing will add new packets to
1331 * the backlog queue. This queue is processed by the
1332 * owner of the socket lock right before it is released.
1333 *
1334 * Since ~2.3.5 it is also exclusive sleep lock serializing
1335 * accesses from user process context.
1336 */
John Heffnerd2e91172007-09-12 10:44:19 +02001337#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338
Peter Zijlstraed075362006-12-06 20:35:24 -08001339/*
1340 * Macro so as to not evaluate some arguments when
1341 * lockdep is not enabled.
1342 *
1343 * Mark both the sk_lock and the sk_lock.slock as a
1344 * per-address-family lock class.
1345 */
1346#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1347do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001348 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001349 init_waitqueue_head(&sk->sk_lock.wq); \
1350 spin_lock_init(&(sk)->sk_lock.slock); \
1351 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1352 sizeof((sk)->sk_lock)); \
1353 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1354 (skey), (sname)); \
1355 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1356} while (0)
1357
Harvey Harrison41380932007-12-12 10:46:51 -08001358extern void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001359
1360static inline void lock_sock(struct sock *sk)
1361{
1362 lock_sock_nested(sk, 0);
1363}
1364
Harvey Harrison41380932007-12-12 10:46:51 -08001365extern void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366
1367/* BH context may only use the following locking interface. */
1368#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001369#define bh_lock_sock_nested(__sk) \
1370 spin_lock_nested(&((__sk)->sk_lock.slock), \
1371 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1373
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001374extern bool lock_sock_fast(struct sock *sk);
1375/**
1376 * unlock_sock_fast - complement of lock_sock_fast
1377 * @sk: socket
1378 * @slow: slow mode
1379 *
1380 * fast unlock socket for user context.
1381 * If slow mode is on, we call regular release_sock()
1382 */
1383static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001384{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001385 if (slow)
1386 release_sock(sk);
1387 else
1388 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001389}
1390
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001391
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001392extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +01001393 gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001394 struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395extern void sk_free(struct sock *sk);
Denis V. Lunevedf02082008-02-29 11:18:32 -08001396extern void sk_release_kernel(struct sock *sk);
Eric Dumazete56c57d2011-11-08 17:07:07 -05001397extern struct sock *sk_clone_lock(const struct sock *sk,
1398 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399
1400extern struct sk_buff *sock_wmalloc(struct sock *sk,
1401 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001402 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403extern struct sk_buff *sock_rmalloc(struct sock *sk,
1404 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001405 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406extern void sock_wfree(struct sk_buff *skb);
1407extern void sock_rfree(struct sk_buff *skb);
1408
1409extern int sock_setsockopt(struct socket *sock, int level,
1410 int op, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07001411 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412
1413extern int sock_getsockopt(struct socket *sock, int level,
1414 int op, char __user *optval,
1415 int __user *optlen);
1416extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1417 unsigned long size,
1418 int noblock,
1419 int *errcode);
Herbert Xu4cc7f682009-02-04 16:55:54 -08001420extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1421 unsigned long header_len,
1422 unsigned long data_len,
1423 int noblock,
1424 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -07001425extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +01001426 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1428extern void sk_send_sigurg(struct sock *sk);
1429
Herbert Xuf8451722010-05-24 00:12:34 -07001430#ifdef CONFIG_CGROUPS
1431extern void sock_update_classid(struct sock *sk);
1432#else
1433static inline void sock_update_classid(struct sock *sk)
1434{
1435}
1436#endif
1437
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438/*
1439 * Functions to fill in entries in struct proto_ops when a protocol
1440 * does not implement a particular function.
1441 */
1442extern int sock_no_bind(struct socket *,
1443 struct sockaddr *, int);
1444extern int sock_no_connect(struct socket *,
1445 struct sockaddr *, int, int);
1446extern int sock_no_socketpair(struct socket *,
1447 struct socket *);
1448extern int sock_no_accept(struct socket *,
1449 struct socket *, int);
1450extern int sock_no_getname(struct socket *,
1451 struct sockaddr *, int *, int);
1452extern unsigned int sock_no_poll(struct file *, struct socket *,
1453 struct poll_table_struct *);
1454extern int sock_no_ioctl(struct socket *, unsigned int,
1455 unsigned long);
1456extern int sock_no_listen(struct socket *, int);
1457extern int sock_no_shutdown(struct socket *, int);
1458extern int sock_no_getsockopt(struct socket *, int , int,
1459 char __user *, int __user *);
1460extern int sock_no_setsockopt(struct socket *, int, int,
David S. Millerb7058842009-09-30 16:12:20 -07001461 char __user *, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1463 struct msghdr *, size_t);
1464extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1465 struct msghdr *, size_t, int);
1466extern int sock_no_mmap(struct file *file,
1467 struct socket *sock,
1468 struct vm_area_struct *vma);
1469extern ssize_t sock_no_sendpage(struct socket *sock,
1470 struct page *page,
1471 int offset, size_t size,
1472 int flags);
1473
1474/*
1475 * Functions to fill in entries in struct proto_ops when a protocol
1476 * uses the inet style.
1477 */
1478extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1479 char __user *optval, int __user *optlen);
1480extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1481 struct msghdr *msg, size_t size, int flags);
1482extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001483 char __user *optval, unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001484extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1485 int optname, char __user *optval, int __user *optlen);
1486extern int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001487 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488
1489extern void sk_common_release(struct sock *sk);
1490
1491/*
1492 * Default socket callbacks and setup code
1493 */
1494
1495/* Initialise core socket variables */
1496extern void sock_init_data(struct socket *sock, struct sock *sk);
1497
Eric Dumazet46bcf142010-12-06 09:29:43 -08001498extern void sk_filter_release_rcu(struct rcu_head *rcu);
1499
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001501 * sk_filter_release - release a socket filter
Paul Bonserdc9b3342006-11-23 17:56:13 -08001502 * @fp: filter to remove
1503 *
1504 * Remove a filter from a socket and release its resources.
1505 */
1506
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001507static inline void sk_filter_release(struct sk_filter *fp)
1508{
1509 if (atomic_dec_and_test(&fp->refcnt))
Eric Dumazet80f8f102011-01-18 07:46:52 +00001510 call_rcu(&fp->rcu, sk_filter_release_rcu);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001511}
1512
1513static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514{
1515 unsigned int size = sk_filter_len(fp);
1516
1517 atomic_sub(size, &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001518 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519}
1520
1521static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1522{
1523 atomic_inc(&fp->refcnt);
1524 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1525}
1526
1527/*
1528 * Socket reference counting postulates.
1529 *
1530 * * Each user of socket SHOULD hold a reference count.
1531 * * Each access point to socket (an hash table bucket, reference from a list,
1532 * running timer, skb in flight MUST hold a reference count.
1533 * * When reference count hits 0, it means it will never increase back.
1534 * * When reference count hits 0, it means that no references from
1535 * outside exist to this socket and current process on current CPU
1536 * is last user and may/should destroy this socket.
1537 * * sk_free is called from any context: process, BH, IRQ. When
1538 * it is called, socket has no references from outside -> sk_free
1539 * may release descendant resources allocated by the socket, but
1540 * to the time when it is called, socket is NOT referenced by any
1541 * hash tables, lists etc.
1542 * * Packets, delivered from outside (from network or from another process)
1543 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1544 * when they sit in queue. Otherwise, packets will leak to hole, when
1545 * socket is looked up by one cpu and unhasing is made by another CPU.
1546 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1547 * (leak to backlog). Packet socket does all the processing inside
1548 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1549 * use separate SMP lock, so that they are prone too.
1550 */
1551
1552/* Ungrab socket and destroy it, if it was the last reference. */
1553static inline void sock_put(struct sock *sk)
1554{
1555 if (atomic_dec_and_test(&sk->sk_refcnt))
1556 sk_free(sk);
1557}
1558
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -02001559extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1560 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001561
Krishna Kumare022f0b2009-10-19 23:46:20 +00001562static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1563{
1564 sk->sk_tx_queue_mapping = tx_queue;
1565}
1566
1567static inline void sk_tx_queue_clear(struct sock *sk)
1568{
1569 sk->sk_tx_queue_mapping = -1;
1570}
1571
1572static inline int sk_tx_queue_get(const struct sock *sk)
1573{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001574 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001575}
1576
David S. Miller972692e2008-06-17 22:41:38 -07001577static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1578{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001579 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001580 sk->sk_socket = sock;
1581}
1582
Eric Dumazetaa395142010-04-20 13:03:51 +00001583static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1584{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001585 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1586 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001587}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588/* Detach socket from process context.
1589 * Announce socket dead, detach it from wait queue and inode.
1590 * Note that parent inode held reference count on this struct sock,
1591 * we do not release it in this function, because protocol
1592 * probably wants some additional cleanups or even continuing
1593 * to work with this socket (TCP).
1594 */
1595static inline void sock_orphan(struct sock *sk)
1596{
1597 write_lock_bh(&sk->sk_callback_lock);
1598 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001599 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001600 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 write_unlock_bh(&sk->sk_callback_lock);
1602}
1603
1604static inline void sock_graft(struct sock *sk, struct socket *parent)
1605{
1606 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001607 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001609 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001610 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 write_unlock_bh(&sk->sk_callback_lock);
1612}
1613
1614extern int sock_i_uid(struct sock *sk);
1615extern unsigned long sock_i_ino(struct sock *sk);
1616
1617static inline struct dst_entry *
1618__sk_dst_get(struct sock *sk)
1619{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001620 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001621 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622}
1623
1624static inline struct dst_entry *
1625sk_dst_get(struct sock *sk)
1626{
1627 struct dst_entry *dst;
1628
Eric Dumazetb6c67122010-04-08 23:03:29 +00001629 rcu_read_lock();
1630 dst = rcu_dereference(sk->sk_dst_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 if (dst)
1632 dst_hold(dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001633 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634 return dst;
1635}
1636
Eric Dumazetb6c67122010-04-08 23:03:29 +00001637extern void sk_reset_txq(struct sock *sk);
1638
1639static inline void dst_negative_advice(struct sock *sk)
1640{
1641 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1642
1643 if (dst && dst->ops->negative_advice) {
1644 ndst = dst->ops->negative_advice(dst);
1645
1646 if (ndst != dst) {
1647 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1648 sk_reset_txq(sk);
1649 }
1650 }
1651}
1652
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653static inline void
1654__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1655{
1656 struct dst_entry *old_dst;
1657
Krishna Kumare022f0b2009-10-19 23:46:20 +00001658 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001659 /*
1660 * This can be called while sk is owned by the caller only,
1661 * with no state that can be checked in a rcu_dereference_check() cond
1662 */
1663 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001664 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 dst_release(old_dst);
1666}
1667
1668static inline void
1669sk_dst_set(struct sock *sk, struct dst_entry *dst)
1670{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001671 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 __sk_dst_set(sk, dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001673 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674}
1675
1676static inline void
1677__sk_dst_reset(struct sock *sk)
1678{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001679 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680}
1681
1682static inline void
1683sk_dst_reset(struct sock *sk)
1684{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001685 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 __sk_dst_reset(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001687 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688}
1689
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001690extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001692extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693
Herbert Xubcd76112006-06-30 13:36:35 -07001694static inline int sk_can_gso(const struct sock *sk)
1695{
1696 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1697}
1698
Andi Kleen99580892007-04-20 17:12:43 -07001699extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001700
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001701static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001702{
1703 sk->sk_route_nocaps |= flags;
1704 sk->sk_route_caps &= ~flags;
1705}
1706
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001707static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1708 char __user *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001709 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001710{
1711 if (skb->ip_summed == CHECKSUM_NONE) {
1712 int err = 0;
1713 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1714 if (err)
1715 return err;
Wei Yongjun912d3982011-04-06 18:40:12 +00001716 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001717 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1718 if (!access_ok(VERIFY_READ, from, copy) ||
1719 __copy_from_user_nocache(to, from, copy))
1720 return -EFAULT;
1721 } else if (copy_from_user(to, from, copy))
1722 return -EFAULT;
1723
1724 return 0;
1725}
1726
1727static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1728 char __user *from, int copy)
1729{
Wei Yongjun912d3982011-04-06 18:40:12 +00001730 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001731
Wei Yongjun912d3982011-04-06 18:40:12 +00001732 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1733 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001734 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001735 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001736
1737 return err;
1738}
1739
1740static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1741 struct sk_buff *skb,
1742 struct page *page,
1743 int off, int copy)
1744{
1745 int err;
1746
Wei Yongjun912d3982011-04-06 18:40:12 +00001747 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1748 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001749 if (err)
1750 return err;
1751
1752 skb->len += copy;
1753 skb->data_len += copy;
1754 skb->truesize += copy;
1755 sk->sk_wmem_queued += copy;
1756 sk_mem_charge(sk, copy);
1757 return 0;
1758}
1759
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1761 struct sk_buff *skb, struct page *page,
1762 int off, int copy)
1763{
1764 if (skb->ip_summed == CHECKSUM_NONE) {
1765 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001766 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 page_address(page) + off,
1768 copy, 0, &err);
1769 if (err)
1770 return err;
1771 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1772 } else if (copy_from_user(page_address(page) + off, from, copy))
1773 return -EFAULT;
1774
1775 skb->len += copy;
1776 skb->data_len += copy;
1777 skb->truesize += copy;
1778 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001779 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 return 0;
1781}
1782
Eric Dumazetc5640392009-06-16 10:12:03 +00001783/**
1784 * sk_wmem_alloc_get - returns write allocations
1785 * @sk: socket
1786 *
1787 * Returns sk_wmem_alloc minus initial offset of one
1788 */
1789static inline int sk_wmem_alloc_get(const struct sock *sk)
1790{
1791 return atomic_read(&sk->sk_wmem_alloc) - 1;
1792}
1793
1794/**
1795 * sk_rmem_alloc_get - returns read allocations
1796 * @sk: socket
1797 *
1798 * Returns sk_rmem_alloc
1799 */
1800static inline int sk_rmem_alloc_get(const struct sock *sk)
1801{
1802 return atomic_read(&sk->sk_rmem_alloc);
1803}
1804
1805/**
1806 * sk_has_allocations - check if allocations are outstanding
1807 * @sk: socket
1808 *
1809 * Returns true if socket has write or read allocations
1810 */
1811static inline int sk_has_allocations(const struct sock *sk)
1812{
1813 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1814}
1815
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001816/**
Eric Dumazet43815482010-04-29 11:01:49 +00001817 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001818 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001819 *
Eric Dumazet43815482010-04-29 11:01:49 +00001820 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001821 *
Eric Dumazet43815482010-04-29 11:01:49 +00001822 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001823 * barrier call. They were added due to the race found within the tcp code.
1824 *
1825 * Consider following tcp code paths:
1826 *
1827 * CPU1 CPU2
1828 *
1829 * sys_select receive packet
1830 * ... ...
1831 * __add_wait_queue update tp->rcv_nxt
1832 * ... ...
1833 * tp->rcv_nxt check sock_def_readable
1834 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001835 * schedule rcu_read_lock();
1836 * wq = rcu_dereference(sk->sk_wq);
1837 * if (wq && waitqueue_active(&wq->wait))
1838 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001839 * ...
1840 * }
1841 *
1842 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1843 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1844 * could then endup calling schedule and sleep forever if there are no more
1845 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001846 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001847 */
Eric Dumazet43815482010-04-29 11:01:49 +00001848static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001849{
Eric Dumazet43815482010-04-29 11:01:49 +00001850
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001851 /*
1852 * We need to be sure we are in sync with the
1853 * add_wait_queue modifications to the wait queue.
1854 *
1855 * This memory barrier is paired in the sock_poll_wait.
1856 */
Eric Dumazet43815482010-04-29 11:01:49 +00001857 smp_mb();
1858 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001859}
1860
1861/**
1862 * sock_poll_wait - place memory barrier behind the poll_wait call.
1863 * @filp: file
1864 * @wait_address: socket wait queue
1865 * @p: poll_table
1866 *
Eric Dumazet43815482010-04-29 11:01:49 +00001867 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001868 */
1869static inline void sock_poll_wait(struct file *filp,
1870 wait_queue_head_t *wait_address, poll_table *p)
1871{
Hans Verkuil626cf232012-03-23 15:02:27 -07001872 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001873 poll_wait(filp, wait_address, p);
1874 /*
1875 * We need to be sure we are in sync with the
1876 * socket flags modification.
1877 *
Eric Dumazet43815482010-04-29 11:01:49 +00001878 * This memory barrier is paired in the wq_has_sleeper.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001879 */
1880 smp_mb();
1881 }
1882}
1883
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884/*
1885 * Queue a received datagram if it will fit. Stream and sequenced
1886 * protocols can't normally use this as they need to fit buffers in
1887 * and play with them.
1888 *
1889 * Inlined as it's very short and called for pretty much every
1890 * packet ever received.
1891 */
1892
1893static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1894{
Herbert Xud55d87f2009-06-22 02:25:25 +00001895 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 skb->sk = sk;
1897 skb->destructor = sock_wfree;
Eric Dumazet2b85a342009-06-11 02:55:43 -07001898 /*
1899 * We used to take a refcount on sk, but following operation
1900 * is enough to guarantee sk_free() wont free this sock until
1901 * all in-flight packets are completed
1902 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1904}
1905
1906static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1907{
Herbert Xud55d87f2009-06-22 02:25:25 +00001908 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 skb->sk = sk;
1910 skb->destructor = sock_rfree;
1911 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001912 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913}
1914
1915extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1916 unsigned long expires);
1917
1918extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1919
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001920extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921
Eric Dumazetb1faf562010-05-31 23:44:05 -07001922extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923
1924/*
1925 * Recover an error report and clear atomically
1926 */
1927
1928static inline int sock_error(struct sock *sk)
1929{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001930 int err;
1931 if (likely(!sk->sk_err))
1932 return 0;
1933 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 return -err;
1935}
1936
1937static inline unsigned long sock_wspace(struct sock *sk)
1938{
1939 int amt = 0;
1940
1941 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1942 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1943 if (amt < 0)
1944 amt = 0;
1945 }
1946 return amt;
1947}
1948
1949static inline void sk_wake_async(struct sock *sk, int how, int band)
1950{
Eric Dumazetbcdce712009-10-06 17:28:29 -07001951 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 sock_wake_async(sk->sk_socket, how, band);
1953}
1954
1955#define SOCK_MIN_SNDBUF 2048
Eric Dumazet7a91b4342010-09-26 18:53:07 -07001956/*
1957 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1958 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1959 */
1960#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961
1962static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1963{
1964 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001965 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1967 }
1968}
1969
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11001970struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971
1972static inline struct page *sk_stream_alloc_page(struct sock *sk)
1973{
1974 struct page *page = NULL;
1975
Herbert Xuef015782005-09-01 17:48:59 -07001976 page = alloc_pages(sk->sk_allocation, 0);
1977 if (!page) {
Glauber Costa180d8cd2011-12-11 21:47:02 +00001978 sk_enter_memory_pressure(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979 sk_stream_moderate_sndbuf(sk);
1980 }
1981 return page;
1982}
1983
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984/*
1985 * Default write policy as shown to user space via poll/select/SIGIO
1986 */
1987static inline int sock_writeable(const struct sock *sk)
1988{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001989 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990}
1991
Al Virodd0fc662005-10-07 07:46:04 +01001992static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993{
Andrew Morton99709372009-02-12 16:43:17 -08001994 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995}
1996
1997static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1998{
1999 return noblock ? 0 : sk->sk_rcvtimeo;
2000}
2001
2002static inline long sock_sndtimeo(const struct sock *sk, int noblock)
2003{
2004 return noblock ? 0 : sk->sk_sndtimeo;
2005}
2006
2007static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2008{
2009 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2010}
2011
2012/* Alas, with timeout socket operations are not restartable.
2013 * Compare this to poll().
2014 */
2015static inline int sock_intr_errno(long timeo)
2016{
2017 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2018}
2019
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002020extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2021 struct sk_buff *skb);
Johannes Berg6e3e9392011-11-09 10:15:42 +01002022extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2023 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002024
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025static __inline__ void
2026sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2027{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002028 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002029 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002030
Patrick Ohly20d49472009-02-12 05:03:38 +00002031 /*
2032 * generate control messages if
2033 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2034 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2035 * - software time stamp available and wanted
2036 * (SOCK_TIMESTAMPING_SOFTWARE)
2037 * - hardware time stamps available and wanted
2038 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2039 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2040 */
2041 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2042 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2043 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2044 (hwtstamps->hwtstamp.tv64 &&
2045 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2046 (hwtstamps->syststamp.tv64 &&
2047 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002048 __sock_recv_timestamp(msg, sk, skb);
2049 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002050 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002051
2052 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2053 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054}
2055
Eric Dumazet767dd032010-04-28 19:14:43 +00002056extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2057 struct sk_buff *skb);
2058
2059static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2060 struct sk_buff *skb)
2061{
2062#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2063 (1UL << SOCK_RCVTSTAMP) | \
2064 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2065 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2066 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2067 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2068
2069 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2070 __sock_recv_ts_and_drops(msg, sk, skb);
2071 else
2072 sk->sk_stamp = skb->tstamp;
2073}
Neil Horman3b885782009-10-12 13:26:31 -07002074
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002076 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002077 * @sk: socket sending this packet
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002078 * @tx_flags: filled with instructions for time stamping
Patrick Ohly20d49472009-02-12 05:03:38 +00002079 *
2080 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2081 * parameters are invalid.
2082 */
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002083extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
Patrick Ohly20d49472009-02-12 05:03:38 +00002084
2085/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002087 * @sk: socket to eat this skb from
2088 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07002089 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090 *
2091 * This routine must be called with interrupts disabled or with the socket
2092 * locked so that the sk_buff queue operation is ok.
2093*/
Chris Leech624d1162006-05-23 18:01:28 -07002094#ifdef CONFIG_NET_DMA
2095static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2096{
2097 __skb_unlink(skb, &sk->sk_receive_queue);
2098 if (!copied_early)
2099 __kfree_skb(skb);
2100 else
2101 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2102}
2103#else
2104static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105{
2106 __skb_unlink(skb, &sk->sk_receive_queue);
2107 __kfree_skb(skb);
2108}
Chris Leech624d1162006-05-23 18:01:28 -07002109#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002111static inline
2112struct net *sock_net(const struct sock *sk)
2113{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002114 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002115}
2116
2117static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002118void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002119{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002120 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002121}
2122
Denis V. Lunevedf02082008-02-29 11:18:32 -08002123/*
2124 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002125 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08002126 * to stop it.
2127 * Sockets after sk_change_net should be released using sk_release_kernel
2128 */
2129static inline void sk_change_net(struct sock *sk, struct net *net)
2130{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002131 put_net(sock_net(sk));
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07002132 sock_net_set(sk, hold_net(net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08002133}
2134
KOVACS Krisztian23542612008-10-07 12:41:01 -07002135static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2136{
2137 if (unlikely(skb->sk)) {
2138 struct sock *sk = skb->sk;
2139
2140 skb->destructor = NULL;
2141 skb->sk = NULL;
2142 return sk;
2143 }
2144 return NULL;
2145}
2146
Patrick Ohly20d49472009-02-12 05:03:38 +00002147extern void sock_enable_timestamp(struct sock *sk, int flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148extern int sock_get_timestamp(struct sock *, struct timeval __user *);
Eric Dumazetae40eb12007-03-18 17:33:16 -07002149extern int sock_get_timestampns(struct sock *, struct timespec __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150
2151/*
2152 * Enable debug/info messages
2153 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002154extern int net_msg_warn;
2155#define NETDEBUG(fmt, args...) \
2156 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002158#define LIMIT_NETDEBUG(fmt, args...) \
2159 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002160
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161extern __u32 sysctl_wmem_max;
2162extern __u32 sysctl_rmem_max;
2163
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002164extern void sk_init(void);
2165
David S. Miller6baf1f42005-09-05 18:14:11 -07002166extern int sysctl_optmem_max;
2167
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002168extern __u32 sysctl_wmem_default;
2169extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002170
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171#endif /* _SOCK_H */