blob: 5a0a58ac4126c47517ea5a4da0fdb051cc88daba [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 Melo8feaf0c02005-08-09 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 Melo8feaf0c02005-08-09 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 Melo8feaf0c02005-08-09 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 Melo8feaf0c02005-08-09 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 Melo8feaf0c02005-08-09 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
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700219 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700220 * @sk_backlog: always used with the per-socket spinlock held
221 * @sk_callback_lock: used with the callbacks in the end of this struct
222 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800223 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
224 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700225 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800226 * @sk_err_soft: errors that don't cause failure but are the cause of a
227 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700228 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700229 * @sk_ack_backlog: current listen backlog
230 * @sk_max_ack_backlog: listen backlog set in listen()
231 * @sk_priority: %SO_PRIORITY setting
Randy Dunlap1a3bc362012-01-21 09:03:10 +0000232 * @sk_cgrp_prioidx: socket group's priority map index
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700233 * @sk_type: socket type (%SOCK_STREAM, etc)
234 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000235 * @sk_peer_pid: &struct pid for this socket's peer
236 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700237 * @sk_rcvlowat: %SO_RCVLOWAT setting
238 * @sk_rcvtimeo: %SO_RCVTIMEO setting
239 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700240 * @sk_rxhash: flow hash received from netif layer
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700241 * @sk_filter: socket filtering instructions
242 * @sk_protinfo: private area, net family specific, when not using slab
243 * @sk_timer: sock cleanup timer
244 * @sk_stamp: time stamp of last packet received
245 * @sk_socket: Identd and reporting IO signals
246 * @sk_user_data: RPC layer private data
247 * @sk_sndmsg_page: cached page for sendmsg
248 * @sk_sndmsg_off: cached offset for sendmsg
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000249 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700250 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700251 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800252 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000253 * @sk_classid: this socket's cgroup classid
Glauber Costae1aab162011-12-11 21:47:03 +0000254 * @sk_cgrp: this socket's cgroup-specific proto data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700255 * @sk_write_pending: a write to stream socket waits to start
256 * @sk_state_change: callback to indicate change in the state of the sock
257 * @sk_data_ready: callback to indicate there is data to be processed
258 * @sk_write_space: callback to indicate there is bf sending space available
259 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
260 * @sk_backlog_rcv: callback to process the backlog
261 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 */
263struct sock {
264 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700265 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 * don't add nothing before this first member (__sk_common) --acme
267 */
268 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000269#define sk_node __sk_common.skc_node
270#define sk_nulls_node __sk_common.skc_nulls_node
271#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000272#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000273
Eric Dumazet68835ab2010-11-30 19:04:07 +0000274#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
275#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000276#define sk_hash __sk_common.skc_hash
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277#define sk_family __sk_common.skc_family
278#define sk_state __sk_common.skc_state
279#define sk_reuse __sk_common.skc_reuse
280#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700282#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200283#define sk_net __sk_common.skc_net
Eric Dumazetb178bb32010-11-16 05:56:04 +0000284 socket_lock_t sk_lock;
285 struct sk_buff_head sk_receive_queue;
286 /*
287 * The backlog queue is special, it is always used with
288 * the per-socket spinlock held and requires low latency
289 * access. Therefore we special case it's implementation.
290 * Note : rmem_alloc is in this structure to fill a hole
291 * on 64bit arches, not because its logically part of
292 * backlog.
293 */
294 struct {
295 atomic_t rmem_alloc;
296 int len;
297 struct sk_buff *head;
298 struct sk_buff *tail;
299 } sk_backlog;
300#define sk_rmem_alloc sk_backlog.rmem_alloc
301 int sk_forward_alloc;
302#ifdef CONFIG_RPS
303 __u32 sk_rxhash;
304#endif
305 atomic_t sk_drops;
306 int sk_rcvbuf;
307
308 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000309 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000310
311#ifdef CONFIG_NET_DMA
312 struct sk_buff_head sk_async_wait_queue;
313#endif
314
315#ifdef CONFIG_XFRM
316 struct xfrm_policy *sk_policy[2];
317#endif
318 unsigned long sk_flags;
319 struct dst_entry *sk_dst_cache;
320 spinlock_t sk_dst_lock;
321 atomic_t sk_wmem_alloc;
322 atomic_t sk_omem_alloc;
323 int sk_sndbuf;
324 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100325 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000326 unsigned int sk_shutdown : 2,
327 sk_no_check : 2,
328 sk_userlocks : 4,
329 sk_protocol : 8,
330 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100331 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400333 gfp_t sk_allocation;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000334 netdev_features_t sk_route_caps;
335 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700336 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700337 unsigned int sk_gso_max_size;
David S. Miller9932cf92006-03-24 15:12:37 -0800338 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700341 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 rwlock_t sk_callback_lock;
343 int sk_err,
344 sk_err_soft;
345 unsigned short sk_ack_backlog;
346 unsigned short sk_max_ack_backlog;
347 __u32 sk_priority;
Neil Horman5bc14212011-11-22 05:10:51 +0000348#ifdef CONFIG_CGROUPS
349 __u32 sk_cgrp_prioidx;
350#endif
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000351 struct pid *sk_peer_pid;
352 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 long sk_rcvtimeo;
354 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 void *sk_protinfo;
356 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700357 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 struct socket *sk_socket;
359 void *sk_user_data;
360 struct page *sk_sndmsg_page;
361 struct sk_buff *sk_send_head;
362 __u32 sk_sndmsg_off;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000363 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800365#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800367#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800368 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700369 u32 sk_classid;
Glauber Costae1aab162011-12-11 21:47:03 +0000370 struct cg_proto *sk_cgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 void (*sk_state_change)(struct sock *sk);
372 void (*sk_data_ready)(struct sock *sk, int bytes);
373 void (*sk_write_space)(struct sock *sk);
374 void (*sk_error_report)(struct sock *sk);
375 int (*sk_backlog_rcv)(struct sock *sk,
376 struct sk_buff *skb);
377 void (*sk_destruct)(struct sock *sk);
378};
379
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000380static inline int sk_peek_offset(struct sock *sk, int flags)
381{
382 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
383 return sk->sk_peek_off;
384 else
385 return 0;
386}
387
388static inline void sk_peek_offset_bwd(struct sock *sk, int val)
389{
390 if (sk->sk_peek_off >= 0) {
391 if (sk->sk_peek_off >= val)
392 sk->sk_peek_off -= val;
393 else
394 sk->sk_peek_off = 0;
395 }
396}
397
398static inline void sk_peek_offset_fwd(struct sock *sk, int val)
399{
400 if (sk->sk_peek_off >= 0)
401 sk->sk_peek_off += val;
402}
403
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404/*
405 * Hashed lists helper routines
406 */
Li Zefanc4146642010-02-08 23:18:45 +0000407static inline struct sock *sk_entry(const struct hlist_node *node)
408{
409 return hlist_entry(node, struct sock, sk_node);
410}
411
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700412static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413{
414 return hlist_entry(head->first, struct sock, sk_node);
415}
416
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700417static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418{
419 return hlist_empty(head) ? NULL : __sk_head(head);
420}
421
Eric Dumazet88ab1932008-11-16 19:39:21 -0800422static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
423{
424 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
425}
426
427static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
428{
429 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
430}
431
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700432static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433{
434 return sk->sk_node.next ?
435 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
436}
437
Eric Dumazet88ab1932008-11-16 19:39:21 -0800438static inline struct sock *sk_nulls_next(const struct sock *sk)
439{
440 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
441 hlist_nulls_entry(sk->sk_nulls_node.next,
442 struct sock, sk_nulls_node) :
443 NULL;
444}
445
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700446static inline int sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447{
448 return hlist_unhashed(&sk->sk_node);
449}
450
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700451static inline int sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452{
Akinobu Mitada753be2006-04-28 15:21:23 -0700453 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454}
455
456static __inline__ void sk_node_init(struct hlist_node *node)
457{
458 node->pprev = NULL;
459}
460
Eric Dumazet88ab1932008-11-16 19:39:21 -0800461static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
462{
463 node->pprev = NULL;
464}
465
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466static __inline__ void __sk_del_node(struct sock *sk)
467{
468 __hlist_del(&sk->sk_node);
469}
470
stephen hemminger808f5112010-02-22 07:57:18 +0000471/* NB: equivalent to hlist_del_init_rcu */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472static __inline__ int __sk_del_node_init(struct sock *sk)
473{
474 if (sk_hashed(sk)) {
475 __sk_del_node(sk);
476 sk_node_init(&sk->sk_node);
477 return 1;
478 }
479 return 0;
480}
481
482/* Grab socket reference count. This operation is valid only
483 when sk is ALREADY grabbed f.e. it is found in hash table
484 or a list and the lookup is made under lock preventing hash table
485 modifications.
486 */
487
488static inline void sock_hold(struct sock *sk)
489{
490 atomic_inc(&sk->sk_refcnt);
491}
492
493/* Ungrab socket in the context, which assumes that socket refcnt
494 cannot hit zero, f.e. it is true in context of any socketcall.
495 */
496static inline void __sock_put(struct sock *sk)
497{
498 atomic_dec(&sk->sk_refcnt);
499}
500
501static __inline__ int sk_del_node_init(struct sock *sk)
502{
503 int rc = __sk_del_node_init(sk);
504
505 if (rc) {
506 /* paranoid for a while -acme */
507 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
508 __sock_put(sk);
509 }
510 return rc;
511}
stephen hemminger808f5112010-02-22 07:57:18 +0000512#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513
Eric Dumazet88ab1932008-11-16 19:39:21 -0800514static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700515{
516 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800517 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700518 return 1;
519 }
520 return 0;
521}
522
Eric Dumazet88ab1932008-11-16 19:39:21 -0800523static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700524{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800525 int rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700526
527 if (rc) {
528 /* paranoid for a while -acme */
529 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
530 __sock_put(sk);
531 }
532 return rc;
533}
534
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
536{
537 hlist_add_head(&sk->sk_node, list);
538}
539
540static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
541{
542 sock_hold(sk);
543 __sk_add_node(sk, list);
544}
545
stephen hemminger808f5112010-02-22 07:57:18 +0000546static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
547{
548 sock_hold(sk);
549 hlist_add_head_rcu(&sk->sk_node, list);
550}
551
Eric Dumazet88ab1932008-11-16 19:39:21 -0800552static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700553{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800554 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700555}
556
Eric Dumazet88ab1932008-11-16 19:39:21 -0800557static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700558{
559 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800560 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700561}
562
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563static __inline__ void __sk_del_bind_node(struct sock *sk)
564{
565 __hlist_del(&sk->sk_bind_node);
566}
567
568static __inline__ void sk_add_bind_node(struct sock *sk,
569 struct hlist_head *list)
570{
571 hlist_add_head(&sk->sk_bind_node, list);
572}
573
574#define sk_for_each(__sk, node, list) \
575 hlist_for_each_entry(__sk, node, list, sk_node)
stephen hemminger808f5112010-02-22 07:57:18 +0000576#define sk_for_each_rcu(__sk, node, list) \
577 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800578#define sk_nulls_for_each(__sk, node, list) \
579 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
580#define sk_nulls_for_each_rcu(__sk, node, list) \
581 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582#define sk_for_each_from(__sk, node) \
583 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
584 hlist_for_each_entry_from(__sk, node, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800585#define sk_nulls_for_each_from(__sk, node) \
586 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
587 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588#define sk_for_each_safe(__sk, node, tmp, list) \
589 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
590#define sk_for_each_bound(__sk, node, list) \
591 hlist_for_each_entry(__sk, node, list, sk_bind_node)
592
593/* Sock flags */
594enum sock_flags {
595 SOCK_DEAD,
596 SOCK_DONE,
597 SOCK_URGINLINE,
598 SOCK_KEEPOPEN,
599 SOCK_LINGER,
600 SOCK_DESTROY,
601 SOCK_BROADCAST,
602 SOCK_TIMESTAMP,
603 SOCK_ZAPPED,
604 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
605 SOCK_DBG, /* %SO_DEBUG setting */
606 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700607 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
609 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Patrick Ohly20d49472009-02-12 05:03:38 +0000610 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
611 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
612 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
613 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
614 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
615 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
616 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700617 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700618 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000619 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100620 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000621 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
622 * Will use last 4 bytes of packet sent from
623 * user-space instead.
624 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625};
626
Ralf Baechle53b924b2005-08-23 10:11:30 -0700627static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
628{
629 nsk->sk_flags = osk->sk_flags;
630}
631
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
633{
634 __set_bit(flag, &sk->sk_flags);
635}
636
637static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
638{
639 __clear_bit(flag, &sk->sk_flags);
640}
641
642static inline int sock_flag(struct sock *sk, enum sock_flags flag)
643{
644 return test_bit(flag, &sk->sk_flags);
645}
646
647static inline void sk_acceptq_removed(struct sock *sk)
648{
649 sk->sk_ack_backlog--;
650}
651
652static inline void sk_acceptq_added(struct sock *sk)
653{
654 sk->sk_ack_backlog++;
655}
656
657static inline int sk_acceptq_is_full(struct sock *sk)
658{
David S. Miller64a14652007-03-06 11:21:05 -0800659 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660}
661
662/*
663 * Compute minimal free write space needed to queue new packets.
664 */
665static inline int sk_stream_min_wspace(struct sock *sk)
666{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800667 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668}
669
670static inline int sk_stream_wspace(struct sock *sk)
671{
672 return sk->sk_sndbuf - sk->sk_wmem_queued;
673}
674
675extern void sk_stream_write_space(struct sock *sk);
676
677static inline int sk_stream_memory_free(struct sock *sk)
678{
679 return sk->sk_wmem_queued < sk->sk_sndbuf;
680}
681
Zhu Yi8eae9392010-03-04 18:01:40 +0000682/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000683static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800684{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000685 /* dont let skb dst not refcounted, we are going to leave rcu lock */
686 skb_dst_force(skb);
687
688 if (!sk->sk_backlog.tail)
689 sk->sk_backlog.head = skb;
690 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800691 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000692
693 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800694 skb->next = NULL;
695}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696
Eric Dumazetc3774112010-04-27 15:13:20 -0700697/*
698 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000699 * Do not take into account this skb truesize,
700 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700701 */
702static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
703{
704 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
705
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000706 return qsize > sk->sk_rcvbuf;
Eric Dumazetc3774112010-04-27 15:13:20 -0700707}
708
Zhu Yi8eae9392010-03-04 18:01:40 +0000709/* The per-socket spinlock must be held here. */
Zhu Yi40456352010-03-07 16:21:39 +0000710static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Zhu Yi8eae9392010-03-04 18:01:40 +0000711{
Eric Dumazetc3774112010-04-27 15:13:20 -0700712 if (sk_rcvqueues_full(sk, skb))
Zhu Yi8eae9392010-03-04 18:01:40 +0000713 return -ENOBUFS;
714
Zhu Yia3a858f2010-03-04 18:01:47 +0000715 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000716 sk->sk_backlog.len += skb->truesize;
717 return 0;
718}
719
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700720static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
721{
722 return sk->sk_backlog_rcv(sk, skb);
723}
724
David S. Millerc58dc012010-04-27 15:05:31 -0700725static inline void sock_rps_record_flow(const struct sock *sk)
726{
727#ifdef CONFIG_RPS
728 struct rps_sock_flow_table *sock_flow_table;
729
730 rcu_read_lock();
731 sock_flow_table = rcu_dereference(rps_sock_flow_table);
732 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
733 rcu_read_unlock();
734#endif
735}
736
737static inline void sock_rps_reset_flow(const struct sock *sk)
738{
739#ifdef CONFIG_RPS
740 struct rps_sock_flow_table *sock_flow_table;
741
742 rcu_read_lock();
743 sock_flow_table = rcu_dereference(rps_sock_flow_table);
744 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
745 rcu_read_unlock();
746#endif
747}
748
Tom Herbertbdeab992011-08-14 19:45:55 +0000749static inline void sock_rps_save_rxhash(struct sock *sk,
750 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700751{
752#ifdef CONFIG_RPS
Tom Herbertbdeab992011-08-14 19:45:55 +0000753 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
David S. Millerc58dc012010-04-27 15:05:31 -0700754 sock_rps_reset_flow(sk);
Tom Herbertbdeab992011-08-14 19:45:55 +0000755 sk->sk_rxhash = skb->rxhash;
David S. Millerc58dc012010-04-27 15:05:31 -0700756 }
757#endif
758}
759
Tom Herbertbdeab992011-08-14 19:45:55 +0000760static inline void sock_rps_reset_rxhash(struct sock *sk)
761{
762#ifdef CONFIG_RPS
763 sock_rps_reset_flow(sk);
764 sk->sk_rxhash = 0;
765#endif
766}
767
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700768#define sk_wait_event(__sk, __timeo, __condition) \
769 ({ int __rc; \
770 release_sock(__sk); \
771 __rc = __condition; \
772 if (!__rc) { \
773 *(__timeo) = schedule_timeout(*(__timeo)); \
774 } \
775 lock_sock(__sk); \
776 __rc = __condition; \
777 __rc; \
778 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779
780extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
781extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
782extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
783extern int sk_stream_error(struct sock *sk, int flags, int err);
784extern void sk_stream_kill_queues(struct sock *sk);
785
786extern int sk_wait_data(struct sock *sk, long *timeo);
787
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700788struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800789struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800790struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700791struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400792struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700793
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794/* Networking protocol blocks we attach to sockets.
795 * socket layer -> transport layer interface
796 * transport -> network interface is defined by struct inet_proto
797 */
798struct proto {
799 void (*close)(struct sock *sk,
800 long timeout);
801 int (*connect)(struct sock *sk,
802 struct sockaddr *uaddr,
803 int addr_len);
804 int (*disconnect)(struct sock *sk, int flags);
805
806 struct sock * (*accept) (struct sock *sk, int flags, int *err);
807
808 int (*ioctl)(struct sock *sk, int cmd,
809 unsigned long arg);
810 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700811 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 void (*shutdown)(struct sock *sk, int how);
813 int (*setsockopt)(struct sock *sk, int level,
814 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700815 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 int (*getsockopt)(struct sock *sk, int level,
817 int optname, char __user *optval,
818 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700819#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800820 int (*compat_setsockopt)(struct sock *sk,
821 int level,
822 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700823 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800824 int (*compat_getsockopt)(struct sock *sk,
825 int level,
826 int optname, char __user *optval,
827 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000828 int (*compat_ioctl)(struct sock *sk,
829 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700830#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
832 struct msghdr *msg, size_t len);
833 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
834 struct msghdr *msg,
835 size_t len, int noblock, int flags,
836 int *addr_len);
837 int (*sendpage)(struct sock *sk, struct page *page,
838 int offset, size_t size, int flags);
839 int (*bind)(struct sock *sk,
840 struct sockaddr *uaddr, int addr_len);
841
842 int (*backlog_rcv) (struct sock *sk,
843 struct sk_buff *skb);
844
845 /* Keeping track of sk's, looking them up, and port selection methods. */
846 void (*hash)(struct sock *sk);
847 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000848 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800850 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800852 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800853#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700854 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800855#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800856
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700858 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +0000859 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -0800860 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 /*
862 * Pressure flag: try to collapse.
863 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800864 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 * is strict, actions are advisory and have some latency.
866 */
867 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000868 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 int *sysctl_wmem;
870 int *sysctl_rmem;
871 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +0000872 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873
Eric Dumazet271b72c2008-10-29 02:11:14 -0700874 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700876 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877
Eric Dumazetdd24c002008-11-25 21:17:14 -0800878 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700879
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700880 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800881 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700882
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700883 union {
884 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -0700885 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700886 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700887 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800888
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 struct module *owner;
890
891 char name[32];
892
893 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700894#ifdef SOCK_REFCNT_DEBUG
895 atomic_t socks;
896#endif
Glauber Costae1aab162011-12-11 21:47:03 +0000897#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
898 /*
899 * cgroup specific init/deinit functions. Called once for all
900 * protocols that implement it, from cgroups populate function.
901 * This function has to setup any files the protocol want to
902 * appear in the kmem cgroup filesystem.
903 */
904 int (*init_cgroup)(struct cgroup *cgrp,
905 struct cgroup_subsys *ss);
Li Zefan761b3ef2012-01-31 13:47:36 +0800906 void (*destroy_cgroup)(struct cgroup *cgrp);
Glauber Costae1aab162011-12-11 21:47:03 +0000907 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
908#endif
909};
910
911struct cg_proto {
912 void (*enter_memory_pressure)(struct sock *sk);
913 struct res_counter *memory_allocated; /* Current allocated memory. */
914 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
915 int *memory_pressure;
916 long *sysctl_mem;
917 /*
918 * memcg field is used to find which memcg we belong directly
919 * Each memcg struct can hold more than one cg_proto, so container_of
920 * won't really cut.
921 *
922 * The elegant solution would be having an inverse function to
923 * proto_cgroup in struct proto, but that means polluting the structure
924 * for everybody, instead of just for memcg users.
925 */
926 struct mem_cgroup *memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927};
928
929extern int proto_register(struct proto *prot, int alloc_slab);
930extern void proto_unregister(struct proto *prot);
931
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700932#ifdef SOCK_REFCNT_DEBUG
933static inline void sk_refcnt_debug_inc(struct sock *sk)
934{
935 atomic_inc(&sk->sk_prot->socks);
936}
937
938static inline void sk_refcnt_debug_dec(struct sock *sk)
939{
940 atomic_dec(&sk->sk_prot->socks);
941 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
942 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
943}
944
Glauber Costae1aab162011-12-11 21:47:03 +0000945inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700946{
947 if (atomic_read(&sk->sk_refcnt) != 1)
948 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
949 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
950}
951#else /* SOCK_REFCNT_DEBUG */
952#define sk_refcnt_debug_inc(sk) do { } while (0)
953#define sk_refcnt_debug_dec(sk) do { } while (0)
954#define sk_refcnt_debug_release(sk) do { } while (0)
955#endif /* SOCK_REFCNT_DEBUG */
956
Glauber Costa376be5f2012-01-20 04:57:14 +0000957#if defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) && defined(CONFIG_NET)
Ingo Molnarc5905af2012-02-24 08:31:31 +0100958extern struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +0000959static inline struct cg_proto *parent_cg_proto(struct proto *proto,
960 struct cg_proto *cg_proto)
961{
962 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
963}
Ingo Molnarc5905af2012-02-24 08:31:31 +0100964#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
Glauber Costae1aab162011-12-11 21:47:03 +0000965#else
966#define mem_cgroup_sockets_enabled 0
967static inline struct cg_proto *parent_cg_proto(struct proto *proto,
968 struct cg_proto *cg_proto)
969{
970 return NULL;
971}
972#endif
973
974
Glauber Costa180d8cd2011-12-11 21:47:02 +0000975static inline bool sk_has_memory_pressure(const struct sock *sk)
976{
977 return sk->sk_prot->memory_pressure != NULL;
978}
979
980static inline bool sk_under_memory_pressure(const struct sock *sk)
981{
982 if (!sk->sk_prot->memory_pressure)
983 return false;
Glauber Costae1aab162011-12-11 21:47:03 +0000984
985 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
986 return !!*sk->sk_cgrp->memory_pressure;
987
Glauber Costa180d8cd2011-12-11 21:47:02 +0000988 return !!*sk->sk_prot->memory_pressure;
989}
990
991static inline void sk_leave_memory_pressure(struct sock *sk)
992{
993 int *memory_pressure = sk->sk_prot->memory_pressure;
994
Glauber Costae1aab162011-12-11 21:47:03 +0000995 if (!memory_pressure)
996 return;
997
998 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +0000999 *memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001000
1001 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1002 struct cg_proto *cg_proto = sk->sk_cgrp;
1003 struct proto *prot = sk->sk_prot;
1004
1005 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1006 if (*cg_proto->memory_pressure)
1007 *cg_proto->memory_pressure = 0;
1008 }
1009
Glauber Costa180d8cd2011-12-11 21:47:02 +00001010}
1011
1012static inline void sk_enter_memory_pressure(struct sock *sk)
1013{
Glauber Costae1aab162011-12-11 21:47:03 +00001014 if (!sk->sk_prot->enter_memory_pressure)
1015 return;
1016
1017 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1018 struct cg_proto *cg_proto = sk->sk_cgrp;
1019 struct proto *prot = sk->sk_prot;
1020
1021 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1022 cg_proto->enter_memory_pressure(sk);
1023 }
1024
1025 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001026}
1027
1028static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1029{
1030 long *prot = sk->sk_prot->sysctl_mem;
Glauber Costae1aab162011-12-11 21:47:03 +00001031 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1032 prot = sk->sk_cgrp->sysctl_mem;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001033 return prot[index];
1034}
1035
Glauber Costae1aab162011-12-11 21:47:03 +00001036static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1037 unsigned long amt,
1038 int *parent_status)
1039{
1040 struct res_counter *fail;
1041 int ret;
1042
Glauber Costa0e90b312012-01-20 04:57:16 +00001043 ret = res_counter_charge_nofail(prot->memory_allocated,
1044 amt << PAGE_SHIFT, &fail);
Glauber Costae1aab162011-12-11 21:47:03 +00001045 if (ret < 0)
1046 *parent_status = OVER_LIMIT;
1047}
1048
1049static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1050 unsigned long amt)
1051{
1052 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1053}
1054
1055static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1056{
1057 u64 ret;
1058 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1059 return ret >> PAGE_SHIFT;
1060}
1061
Glauber Costa180d8cd2011-12-11 21:47:02 +00001062static inline long
1063sk_memory_allocated(const struct sock *sk)
1064{
1065 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001066 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1067 return memcg_memory_allocated_read(sk->sk_cgrp);
1068
Glauber Costa180d8cd2011-12-11 21:47:02 +00001069 return atomic_long_read(prot->memory_allocated);
1070}
1071
1072static inline long
Glauber Costae1aab162011-12-11 21:47:03 +00001073sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001074{
1075 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001076
1077 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1078 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1079 /* update the root cgroup regardless */
1080 atomic_long_add_return(amt, prot->memory_allocated);
1081 return memcg_memory_allocated_read(sk->sk_cgrp);
1082 }
1083
Glauber Costa180d8cd2011-12-11 21:47:02 +00001084 return atomic_long_add_return(amt, prot->memory_allocated);
1085}
1086
1087static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001088sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001089{
1090 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001091
Glauber Costa0e90b312012-01-20 04:57:16 +00001092 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Glauber Costae1aab162011-12-11 21:47:03 +00001093 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1094
Glauber Costa180d8cd2011-12-11 21:47:02 +00001095 atomic_long_sub(amt, prot->memory_allocated);
1096}
1097
1098static inline void sk_sockets_allocated_dec(struct sock *sk)
1099{
1100 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001101
1102 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1103 struct cg_proto *cg_proto = sk->sk_cgrp;
1104
1105 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1106 percpu_counter_dec(cg_proto->sockets_allocated);
1107 }
1108
Glauber Costa180d8cd2011-12-11 21:47:02 +00001109 percpu_counter_dec(prot->sockets_allocated);
1110}
1111
1112static inline void sk_sockets_allocated_inc(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_inc(cg_proto->sockets_allocated);
1121 }
1122
Glauber Costa180d8cd2011-12-11 21:47:02 +00001123 percpu_counter_inc(prot->sockets_allocated);
1124}
1125
1126static inline int
1127sk_sockets_allocated_read_positive(struct sock *sk)
1128{
1129 struct proto *prot = sk->sk_prot;
1130
Glauber Costae1aab162011-12-11 21:47:03 +00001131 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric Dumazet518fbf92012-04-28 23:21:56 +00001132 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001133
Eric Dumazet518fbf92012-04-28 23:21:56 +00001134 return percpu_counter_read_positive(prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001135}
1136
1137static inline int
1138proto_sockets_allocated_sum_positive(struct proto *prot)
1139{
1140 return percpu_counter_sum_positive(prot->sockets_allocated);
1141}
1142
1143static inline long
1144proto_memory_allocated(struct proto *prot)
1145{
1146 return atomic_long_read(prot->memory_allocated);
1147}
1148
1149static inline bool
1150proto_memory_pressure(struct proto *prot)
1151{
1152 if (!prot->memory_pressure)
1153 return false;
1154 return !!*prot->memory_pressure;
1155}
1156
Eric Dumazet65f76512008-01-03 20:46:48 -08001157
1158#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159/* Called with local bh disabled */
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001160extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1161extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001162#else
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001163static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
1164 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001165{
1166}
Eric Dumazet65f76512008-01-03 20:46:48 -08001167#endif
1168
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001170/* With per-bucket locks this operation is not-atomic, so that
1171 * this version is not worse.
1172 */
1173static inline void __sk_prot_rehash(struct sock *sk)
1174{
1175 sk->sk_prot->unhash(sk);
1176 sk->sk_prot->hash(sk);
1177}
1178
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001179void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1180
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181/* About 10 seconds */
1182#define SOCK_DESTROY_TIME (10*HZ)
1183
1184/* Sockets 0-1023 can't be bound to unless you are superuser */
1185#define PROT_SOCK 1024
1186
1187#define SHUTDOWN_MASK 3
1188#define RCV_SHUTDOWN 1
1189#define SEND_SHUTDOWN 2
1190
1191#define SOCK_SNDBUF_LOCK 1
1192#define SOCK_RCVBUF_LOCK 2
1193#define SOCK_BINDADDR_LOCK 4
1194#define SOCK_BINDPORT_LOCK 8
1195
1196/* sock_iocb: used to kick off async processing of socket ios */
1197struct sock_iocb {
1198 struct list_head list;
1199
1200 int flags;
1201 int size;
1202 struct socket *sock;
1203 struct sock *sk;
1204 struct scm_cookie *scm;
1205 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 struct kiocb *kiocb;
1207};
1208
1209static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1210{
1211 return (struct sock_iocb *)iocb->private;
1212}
1213
1214static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1215{
1216 return si->kiocb;
1217}
1218
1219struct socket_alloc {
1220 struct socket socket;
1221 struct inode vfs_inode;
1222};
1223
1224static inline struct socket *SOCKET_I(struct inode *inode)
1225{
1226 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1227}
1228
1229static inline struct inode *SOCK_INODE(struct socket *socket)
1230{
1231 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1232}
1233
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001234/*
1235 * Functions for memory accounting
1236 */
1237extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1238extern void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001240#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1241#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1242#define SK_MEM_SEND 0
1243#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001245static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001247 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248}
1249
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001250static inline int sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001252 /* return true if protocol supports memory accounting */
1253 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254}
1255
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001256static inline int sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001258 if (!sk_has_account(sk))
1259 return 1;
Herbert Xud80d99d2005-09-01 17:48:23 -07001260 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001261 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1262}
1263
1264static inline int sk_rmem_schedule(struct sock *sk, int size)
1265{
1266 if (!sk_has_account(sk))
1267 return 1;
1268 return size <= sk->sk_forward_alloc ||
1269 __sk_mem_schedule(sk, size, SK_MEM_RECV);
1270}
1271
1272static inline void sk_mem_reclaim(struct sock *sk)
1273{
1274 if (!sk_has_account(sk))
1275 return;
1276 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1277 __sk_mem_reclaim(sk);
1278}
1279
David S. Miller9993e7d2008-01-10 21:56:38 -08001280static inline void sk_mem_reclaim_partial(struct sock *sk)
1281{
1282 if (!sk_has_account(sk))
1283 return;
1284 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1285 __sk_mem_reclaim(sk);
1286}
1287
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001288static inline void sk_mem_charge(struct sock *sk, int size)
1289{
1290 if (!sk_has_account(sk))
1291 return;
1292 sk->sk_forward_alloc -= size;
1293}
1294
1295static inline void sk_mem_uncharge(struct sock *sk, int size)
1296{
1297 if (!sk_has_account(sk))
1298 return;
1299 sk->sk_forward_alloc += size;
1300}
1301
1302static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1303{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001304 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1305 sk->sk_wmem_queued -= skb->truesize;
1306 sk_mem_uncharge(sk, skb->truesize);
1307 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001308}
1309
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310/* Used by processes to "lock" a socket state, so that
1311 * interrupts and bottom half handlers won't change it
1312 * from under us. It essentially blocks any incoming
1313 * packets, so that we won't get any new data or any
1314 * packets that change the state of the socket.
1315 *
1316 * While locked, BH processing will add new packets to
1317 * the backlog queue. This queue is processed by the
1318 * owner of the socket lock right before it is released.
1319 *
1320 * Since ~2.3.5 it is also exclusive sleep lock serializing
1321 * accesses from user process context.
1322 */
John Heffnerd2e91172007-09-12 10:44:19 +02001323#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324
Peter Zijlstraed075362006-12-06 20:35:24 -08001325/*
1326 * Macro so as to not evaluate some arguments when
1327 * lockdep is not enabled.
1328 *
1329 * Mark both the sk_lock and the sk_lock.slock as a
1330 * per-address-family lock class.
1331 */
1332#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1333do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001334 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001335 init_waitqueue_head(&sk->sk_lock.wq); \
1336 spin_lock_init(&(sk)->sk_lock.slock); \
1337 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1338 sizeof((sk)->sk_lock)); \
1339 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1340 (skey), (sname)); \
1341 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1342} while (0)
1343
Harvey Harrison41380932007-12-12 10:46:51 -08001344extern void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001345
1346static inline void lock_sock(struct sock *sk)
1347{
1348 lock_sock_nested(sk, 0);
1349}
1350
Harvey Harrison41380932007-12-12 10:46:51 -08001351extern void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
1353/* BH context may only use the following locking interface. */
1354#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001355#define bh_lock_sock_nested(__sk) \
1356 spin_lock_nested(&((__sk)->sk_lock.slock), \
1357 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1359
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001360extern bool lock_sock_fast(struct sock *sk);
1361/**
1362 * unlock_sock_fast - complement of lock_sock_fast
1363 * @sk: socket
1364 * @slow: slow mode
1365 *
1366 * fast unlock socket for user context.
1367 * If slow mode is on, we call regular release_sock()
1368 */
1369static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001370{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001371 if (slow)
1372 release_sock(sk);
1373 else
1374 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001375}
1376
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001377
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001378extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +01001379 gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001380 struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381extern void sk_free(struct sock *sk);
Denis V. Lunevedf02082008-02-29 11:18:32 -08001382extern void sk_release_kernel(struct sock *sk);
Eric Dumazete56c57d2011-11-08 17:07:07 -05001383extern struct sock *sk_clone_lock(const struct sock *sk,
1384 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385
1386extern struct sk_buff *sock_wmalloc(struct sock *sk,
1387 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001388 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389extern struct sk_buff *sock_rmalloc(struct sock *sk,
1390 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001391 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392extern void sock_wfree(struct sk_buff *skb);
1393extern void sock_rfree(struct sk_buff *skb);
1394
1395extern int sock_setsockopt(struct socket *sock, int level,
1396 int op, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07001397 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398
1399extern int sock_getsockopt(struct socket *sock, int level,
1400 int op, char __user *optval,
1401 int __user *optlen);
1402extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1403 unsigned long size,
1404 int noblock,
1405 int *errcode);
Herbert Xu4cc7f682009-02-04 16:55:54 -08001406extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1407 unsigned long header_len,
1408 unsigned long data_len,
1409 int noblock,
1410 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -07001411extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +01001412 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1414extern void sk_send_sigurg(struct sock *sk);
1415
Herbert Xuf8451722010-05-24 00:12:34 -07001416#ifdef CONFIG_CGROUPS
1417extern void sock_update_classid(struct sock *sk);
1418#else
1419static inline void sock_update_classid(struct sock *sk)
1420{
1421}
1422#endif
1423
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424/*
1425 * Functions to fill in entries in struct proto_ops when a protocol
1426 * does not implement a particular function.
1427 */
1428extern int sock_no_bind(struct socket *,
1429 struct sockaddr *, int);
1430extern int sock_no_connect(struct socket *,
1431 struct sockaddr *, int, int);
1432extern int sock_no_socketpair(struct socket *,
1433 struct socket *);
1434extern int sock_no_accept(struct socket *,
1435 struct socket *, int);
1436extern int sock_no_getname(struct socket *,
1437 struct sockaddr *, int *, int);
1438extern unsigned int sock_no_poll(struct file *, struct socket *,
1439 struct poll_table_struct *);
1440extern int sock_no_ioctl(struct socket *, unsigned int,
1441 unsigned long);
1442extern int sock_no_listen(struct socket *, int);
1443extern int sock_no_shutdown(struct socket *, int);
1444extern int sock_no_getsockopt(struct socket *, int , int,
1445 char __user *, int __user *);
1446extern int sock_no_setsockopt(struct socket *, int, int,
David S. Millerb7058842009-09-30 16:12:20 -07001447 char __user *, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1449 struct msghdr *, size_t);
1450extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1451 struct msghdr *, size_t, int);
1452extern int sock_no_mmap(struct file *file,
1453 struct socket *sock,
1454 struct vm_area_struct *vma);
1455extern ssize_t sock_no_sendpage(struct socket *sock,
1456 struct page *page,
1457 int offset, size_t size,
1458 int flags);
1459
1460/*
1461 * Functions to fill in entries in struct proto_ops when a protocol
1462 * uses the inet style.
1463 */
1464extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1465 char __user *optval, int __user *optlen);
1466extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1467 struct msghdr *msg, size_t size, int flags);
1468extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001469 char __user *optval, unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001470extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1471 int optname, char __user *optval, int __user *optlen);
1472extern int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001473 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474
1475extern void sk_common_release(struct sock *sk);
1476
1477/*
1478 * Default socket callbacks and setup code
1479 */
1480
1481/* Initialise core socket variables */
1482extern void sock_init_data(struct socket *sock, struct sock *sk);
1483
Eric Dumazet46bcf142010-12-06 09:29:43 -08001484extern void sk_filter_release_rcu(struct rcu_head *rcu);
1485
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001487 * sk_filter_release - release a socket filter
Paul Bonserdc9b3342006-11-23 17:56:13 -08001488 * @fp: filter to remove
1489 *
1490 * Remove a filter from a socket and release its resources.
1491 */
1492
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001493static inline void sk_filter_release(struct sk_filter *fp)
1494{
1495 if (atomic_dec_and_test(&fp->refcnt))
Eric Dumazet80f8f102011-01-18 07:46:52 +00001496 call_rcu(&fp->rcu, sk_filter_release_rcu);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001497}
1498
1499static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500{
1501 unsigned int size = sk_filter_len(fp);
1502
1503 atomic_sub(size, &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001504 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505}
1506
1507static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1508{
1509 atomic_inc(&fp->refcnt);
1510 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1511}
1512
1513/*
1514 * Socket reference counting postulates.
1515 *
1516 * * Each user of socket SHOULD hold a reference count.
1517 * * Each access point to socket (an hash table bucket, reference from a list,
1518 * running timer, skb in flight MUST hold a reference count.
1519 * * When reference count hits 0, it means it will never increase back.
1520 * * When reference count hits 0, it means that no references from
1521 * outside exist to this socket and current process on current CPU
1522 * is last user and may/should destroy this socket.
1523 * * sk_free is called from any context: process, BH, IRQ. When
1524 * it is called, socket has no references from outside -> sk_free
1525 * may release descendant resources allocated by the socket, but
1526 * to the time when it is called, socket is NOT referenced by any
1527 * hash tables, lists etc.
1528 * * Packets, delivered from outside (from network or from another process)
1529 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1530 * when they sit in queue. Otherwise, packets will leak to hole, when
1531 * socket is looked up by one cpu and unhasing is made by another CPU.
1532 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1533 * (leak to backlog). Packet socket does all the processing inside
1534 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1535 * use separate SMP lock, so that they are prone too.
1536 */
1537
1538/* Ungrab socket and destroy it, if it was the last reference. */
1539static inline void sock_put(struct sock *sk)
1540{
1541 if (atomic_dec_and_test(&sk->sk_refcnt))
1542 sk_free(sk);
1543}
1544
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -02001545extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1546 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001547
Krishna Kumare022f0b2009-10-19 23:46:20 +00001548static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1549{
1550 sk->sk_tx_queue_mapping = tx_queue;
1551}
1552
1553static inline void sk_tx_queue_clear(struct sock *sk)
1554{
1555 sk->sk_tx_queue_mapping = -1;
1556}
1557
1558static inline int sk_tx_queue_get(const struct sock *sk)
1559{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001560 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001561}
1562
David S. Miller972692e2008-06-17 22:41:38 -07001563static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1564{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001565 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001566 sk->sk_socket = sock;
1567}
1568
Eric Dumazetaa395142010-04-20 13:03:51 +00001569static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1570{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001571 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1572 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001573}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574/* Detach socket from process context.
1575 * Announce socket dead, detach it from wait queue and inode.
1576 * Note that parent inode held reference count on this struct sock,
1577 * we do not release it in this function, because protocol
1578 * probably wants some additional cleanups or even continuing
1579 * to work with this socket (TCP).
1580 */
1581static inline void sock_orphan(struct sock *sk)
1582{
1583 write_lock_bh(&sk->sk_callback_lock);
1584 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001585 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001586 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 write_unlock_bh(&sk->sk_callback_lock);
1588}
1589
1590static inline void sock_graft(struct sock *sk, struct socket *parent)
1591{
1592 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001593 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001595 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001596 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 write_unlock_bh(&sk->sk_callback_lock);
1598}
1599
1600extern int sock_i_uid(struct sock *sk);
1601extern unsigned long sock_i_ino(struct sock *sk);
1602
1603static inline struct dst_entry *
1604__sk_dst_get(struct sock *sk)
1605{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001606 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001607 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608}
1609
1610static inline struct dst_entry *
1611sk_dst_get(struct sock *sk)
1612{
1613 struct dst_entry *dst;
1614
Eric Dumazetb6c67122010-04-08 23:03:29 +00001615 rcu_read_lock();
1616 dst = rcu_dereference(sk->sk_dst_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617 if (dst)
1618 dst_hold(dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001619 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 return dst;
1621}
1622
Eric Dumazetb6c67122010-04-08 23:03:29 +00001623extern void sk_reset_txq(struct sock *sk);
1624
1625static inline void dst_negative_advice(struct sock *sk)
1626{
1627 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1628
1629 if (dst && dst->ops->negative_advice) {
1630 ndst = dst->ops->negative_advice(dst);
1631
1632 if (ndst != dst) {
1633 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1634 sk_reset_txq(sk);
1635 }
1636 }
1637}
1638
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639static inline void
1640__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1641{
1642 struct dst_entry *old_dst;
1643
Krishna Kumare022f0b2009-10-19 23:46:20 +00001644 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001645 /*
1646 * This can be called while sk is owned by the caller only,
1647 * with no state that can be checked in a rcu_dereference_check() cond
1648 */
1649 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001650 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 dst_release(old_dst);
1652}
1653
1654static inline void
1655sk_dst_set(struct sock *sk, struct dst_entry *dst)
1656{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001657 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 __sk_dst_set(sk, dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001659 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660}
1661
1662static inline void
1663__sk_dst_reset(struct sock *sk)
1664{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001665 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666}
1667
1668static inline void
1669sk_dst_reset(struct sock *sk)
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_reset(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001673 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674}
1675
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001676extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001678extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679
Herbert Xubcd76112006-06-30 13:36:35 -07001680static inline int sk_can_gso(const struct sock *sk)
1681{
1682 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1683}
1684
Andi Kleen99580892007-04-20 17:12:43 -07001685extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001686
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001687static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001688{
1689 sk->sk_route_nocaps |= flags;
1690 sk->sk_route_caps &= ~flags;
1691}
1692
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001693static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1694 char __user *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001695 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001696{
1697 if (skb->ip_summed == CHECKSUM_NONE) {
1698 int err = 0;
1699 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1700 if (err)
1701 return err;
Wei Yongjun912d3982011-04-06 18:40:12 +00001702 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001703 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1704 if (!access_ok(VERIFY_READ, from, copy) ||
1705 __copy_from_user_nocache(to, from, copy))
1706 return -EFAULT;
1707 } else if (copy_from_user(to, from, copy))
1708 return -EFAULT;
1709
1710 return 0;
1711}
1712
1713static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1714 char __user *from, int copy)
1715{
Wei Yongjun912d3982011-04-06 18:40:12 +00001716 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001717
Wei Yongjun912d3982011-04-06 18:40:12 +00001718 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1719 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001720 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001721 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001722
1723 return err;
1724}
1725
1726static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1727 struct sk_buff *skb,
1728 struct page *page,
1729 int off, int copy)
1730{
1731 int err;
1732
Wei Yongjun912d3982011-04-06 18:40:12 +00001733 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1734 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001735 if (err)
1736 return err;
1737
1738 skb->len += copy;
1739 skb->data_len += copy;
1740 skb->truesize += copy;
1741 sk->sk_wmem_queued += copy;
1742 sk_mem_charge(sk, copy);
1743 return 0;
1744}
1745
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1747 struct sk_buff *skb, struct page *page,
1748 int off, int copy)
1749{
1750 if (skb->ip_summed == CHECKSUM_NONE) {
1751 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001752 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 page_address(page) + off,
1754 copy, 0, &err);
1755 if (err)
1756 return err;
1757 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1758 } else if (copy_from_user(page_address(page) + off, from, copy))
1759 return -EFAULT;
1760
1761 skb->len += copy;
1762 skb->data_len += copy;
1763 skb->truesize += copy;
1764 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001765 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 return 0;
1767}
1768
Eric Dumazetc5640392009-06-16 10:12:03 +00001769/**
1770 * sk_wmem_alloc_get - returns write allocations
1771 * @sk: socket
1772 *
1773 * Returns sk_wmem_alloc minus initial offset of one
1774 */
1775static inline int sk_wmem_alloc_get(const struct sock *sk)
1776{
1777 return atomic_read(&sk->sk_wmem_alloc) - 1;
1778}
1779
1780/**
1781 * sk_rmem_alloc_get - returns read allocations
1782 * @sk: socket
1783 *
1784 * Returns sk_rmem_alloc
1785 */
1786static inline int sk_rmem_alloc_get(const struct sock *sk)
1787{
1788 return atomic_read(&sk->sk_rmem_alloc);
1789}
1790
1791/**
1792 * sk_has_allocations - check if allocations are outstanding
1793 * @sk: socket
1794 *
1795 * Returns true if socket has write or read allocations
1796 */
1797static inline int sk_has_allocations(const struct sock *sk)
1798{
1799 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1800}
1801
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001802/**
Eric Dumazet43815482010-04-29 11:01:49 +00001803 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001804 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001805 *
Eric Dumazet43815482010-04-29 11:01:49 +00001806 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001807 *
Eric Dumazet43815482010-04-29 11:01:49 +00001808 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001809 * barrier call. They were added due to the race found within the tcp code.
1810 *
1811 * Consider following tcp code paths:
1812 *
1813 * CPU1 CPU2
1814 *
1815 * sys_select receive packet
1816 * ... ...
1817 * __add_wait_queue update tp->rcv_nxt
1818 * ... ...
1819 * tp->rcv_nxt check sock_def_readable
1820 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001821 * schedule rcu_read_lock();
1822 * wq = rcu_dereference(sk->sk_wq);
1823 * if (wq && waitqueue_active(&wq->wait))
1824 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001825 * ...
1826 * }
1827 *
1828 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1829 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1830 * could then endup calling schedule and sleep forever if there are no more
1831 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001832 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001833 */
Eric Dumazet43815482010-04-29 11:01:49 +00001834static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001835{
Eric Dumazet43815482010-04-29 11:01:49 +00001836
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001837 /*
1838 * We need to be sure we are in sync with the
1839 * add_wait_queue modifications to the wait queue.
1840 *
1841 * This memory barrier is paired in the sock_poll_wait.
1842 */
Eric Dumazet43815482010-04-29 11:01:49 +00001843 smp_mb();
1844 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001845}
1846
1847/**
1848 * sock_poll_wait - place memory barrier behind the poll_wait call.
1849 * @filp: file
1850 * @wait_address: socket wait queue
1851 * @p: poll_table
1852 *
Eric Dumazet43815482010-04-29 11:01:49 +00001853 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001854 */
1855static inline void sock_poll_wait(struct file *filp,
1856 wait_queue_head_t *wait_address, poll_table *p)
1857{
Hans Verkuil626cf232012-03-23 15:02:27 -07001858 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001859 poll_wait(filp, wait_address, p);
1860 /*
1861 * We need to be sure we are in sync with the
1862 * socket flags modification.
1863 *
Eric Dumazet43815482010-04-29 11:01:49 +00001864 * This memory barrier is paired in the wq_has_sleeper.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001865 */
1866 smp_mb();
1867 }
1868}
1869
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870/*
1871 * Queue a received datagram if it will fit. Stream and sequenced
1872 * protocols can't normally use this as they need to fit buffers in
1873 * and play with them.
1874 *
1875 * Inlined as it's very short and called for pretty much every
1876 * packet ever received.
1877 */
1878
1879static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1880{
Herbert Xud55d87f2009-06-22 02:25:25 +00001881 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 skb->sk = sk;
1883 skb->destructor = sock_wfree;
Eric Dumazet2b85a342009-06-11 02:55:43 -07001884 /*
1885 * We used to take a refcount on sk, but following operation
1886 * is enough to guarantee sk_free() wont free this sock until
1887 * all in-flight packets are completed
1888 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1890}
1891
1892static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1893{
Herbert Xud55d87f2009-06-22 02:25:25 +00001894 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 skb->sk = sk;
1896 skb->destructor = sock_rfree;
1897 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001898 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899}
1900
1901extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1902 unsigned long expires);
1903
1904extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1905
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001906extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907
Eric Dumazetb1faf562010-05-31 23:44:05 -07001908extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909
1910/*
1911 * Recover an error report and clear atomically
1912 */
1913
1914static inline int sock_error(struct sock *sk)
1915{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001916 int err;
1917 if (likely(!sk->sk_err))
1918 return 0;
1919 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 return -err;
1921}
1922
1923static inline unsigned long sock_wspace(struct sock *sk)
1924{
1925 int amt = 0;
1926
1927 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1928 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1929 if (amt < 0)
1930 amt = 0;
1931 }
1932 return amt;
1933}
1934
1935static inline void sk_wake_async(struct sock *sk, int how, int band)
1936{
Eric Dumazetbcdce712009-10-06 17:28:29 -07001937 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938 sock_wake_async(sk->sk_socket, how, band);
1939}
1940
1941#define SOCK_MIN_SNDBUF 2048
Eric Dumazet7a91b4342010-09-26 18:53:07 -07001942/*
1943 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1944 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1945 */
1946#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947
1948static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1949{
1950 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001951 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1953 }
1954}
1955
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11001956struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
1958static inline struct page *sk_stream_alloc_page(struct sock *sk)
1959{
1960 struct page *page = NULL;
1961
Herbert Xuef015782005-09-01 17:48:59 -07001962 page = alloc_pages(sk->sk_allocation, 0);
1963 if (!page) {
Glauber Costa180d8cd2011-12-11 21:47:02 +00001964 sk_enter_memory_pressure(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 sk_stream_moderate_sndbuf(sk);
1966 }
1967 return page;
1968}
1969
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970/*
1971 * Default write policy as shown to user space via poll/select/SIGIO
1972 */
1973static inline int sock_writeable(const struct sock *sk)
1974{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001975 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976}
1977
Al Virodd0fc662005-10-07 07:46:04 +01001978static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979{
Andrew Morton99709372009-02-12 16:43:17 -08001980 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981}
1982
1983static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1984{
1985 return noblock ? 0 : sk->sk_rcvtimeo;
1986}
1987
1988static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1989{
1990 return noblock ? 0 : sk->sk_sndtimeo;
1991}
1992
1993static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1994{
1995 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1996}
1997
1998/* Alas, with timeout socket operations are not restartable.
1999 * Compare this to poll().
2000 */
2001static inline int sock_intr_errno(long timeo)
2002{
2003 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2004}
2005
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002006extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2007 struct sk_buff *skb);
Johannes Berg6e3e9392011-11-09 10:15:42 +01002008extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2009 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002010
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011static __inline__ void
2012sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2013{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002014 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002015 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002016
Patrick Ohly20d49472009-02-12 05:03:38 +00002017 /*
2018 * generate control messages if
2019 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2020 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2021 * - software time stamp available and wanted
2022 * (SOCK_TIMESTAMPING_SOFTWARE)
2023 * - hardware time stamps available and wanted
2024 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2025 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2026 */
2027 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2028 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2029 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2030 (hwtstamps->hwtstamp.tv64 &&
2031 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2032 (hwtstamps->syststamp.tv64 &&
2033 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002034 __sock_recv_timestamp(msg, sk, skb);
2035 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002036 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002037
2038 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2039 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040}
2041
Eric Dumazet767dd032010-04-28 19:14:43 +00002042extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2043 struct sk_buff *skb);
2044
2045static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2046 struct sk_buff *skb)
2047{
2048#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2049 (1UL << SOCK_RCVTSTAMP) | \
2050 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2051 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2052 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2053 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2054
2055 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2056 __sock_recv_ts_and_drops(msg, sk, skb);
2057 else
2058 sk->sk_stamp = skb->tstamp;
2059}
Neil Horman3b885782009-10-12 13:26:31 -07002060
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002062 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002063 * @sk: socket sending this packet
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002064 * @tx_flags: filled with instructions for time stamping
Patrick Ohly20d49472009-02-12 05:03:38 +00002065 *
2066 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2067 * parameters are invalid.
2068 */
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002069extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
Patrick Ohly20d49472009-02-12 05:03:38 +00002070
2071/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002073 * @sk: socket to eat this skb from
2074 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07002075 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076 *
2077 * This routine must be called with interrupts disabled or with the socket
2078 * locked so that the sk_buff queue operation is ok.
2079*/
Chris Leech624d1162006-05-23 18:01:28 -07002080#ifdef CONFIG_NET_DMA
2081static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2082{
2083 __skb_unlink(skb, &sk->sk_receive_queue);
2084 if (!copied_early)
2085 __kfree_skb(skb);
2086 else
2087 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2088}
2089#else
2090static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091{
2092 __skb_unlink(skb, &sk->sk_receive_queue);
2093 __kfree_skb(skb);
2094}
Chris Leech624d1162006-05-23 18:01:28 -07002095#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002097static inline
2098struct net *sock_net(const struct sock *sk)
2099{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002100 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002101}
2102
2103static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002104void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002105{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002106 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002107}
2108
Denis V. Lunevedf02082008-02-29 11:18:32 -08002109/*
2110 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002111 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08002112 * to stop it.
2113 * Sockets after sk_change_net should be released using sk_release_kernel
2114 */
2115static inline void sk_change_net(struct sock *sk, struct net *net)
2116{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002117 put_net(sock_net(sk));
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07002118 sock_net_set(sk, hold_net(net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08002119}
2120
KOVACS Krisztian23542612008-10-07 12:41:01 -07002121static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2122{
2123 if (unlikely(skb->sk)) {
2124 struct sock *sk = skb->sk;
2125
2126 skb->destructor = NULL;
2127 skb->sk = NULL;
2128 return sk;
2129 }
2130 return NULL;
2131}
2132
Patrick Ohly20d49472009-02-12 05:03:38 +00002133extern void sock_enable_timestamp(struct sock *sk, int flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134extern int sock_get_timestamp(struct sock *, struct timeval __user *);
Eric Dumazetae40eb12007-03-18 17:33:16 -07002135extern int sock_get_timestampns(struct sock *, struct timespec __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136
2137/*
2138 * Enable debug/info messages
2139 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002140extern int net_msg_warn;
2141#define NETDEBUG(fmt, args...) \
2142 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002144#define LIMIT_NETDEBUG(fmt, args...) \
2145 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147extern __u32 sysctl_wmem_max;
2148extern __u32 sysctl_rmem_max;
2149
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002150extern void sk_init(void);
2151
David S. Miller6baf1f42005-09-05 18:14:11 -07002152extern int sysctl_optmem_max;
2153
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002154extern __u32 sysctl_wmem_default;
2155extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002156
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157#endif /* _SOCK_H */