blob: abb6e0f0c3c3e59b79f23fc72d2e9d8d389e3774 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070056
57#include <linux/filter.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080058#include <linux/rculist_nulls.h>
Jiri Olsaa57de0b2009-07-08 12:09:13 +000059#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
Eric Dumazetc31504d2010-11-15 19:58:26 +000061#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <net/dst.h>
63#include <net/checksum.h>
64
65/*
66 * This structure really needs to be cleaned up.
67 * Most of it is for TCP, and not used by any of
68 * the other protocols.
69 */
70
71/* Define this to get the SOCK_DBG debugging facility. */
72#define SOCK_DEBUGGING
73#ifdef SOCK_DEBUGGING
74#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
75 printk(KERN_DEBUG msg); } while (0)
76#else
Stephen Hemminger4cd90292008-03-21 15:54:53 -070077/* Validate arguments and do nothing */
Joe Perchesb9075fa2011-10-31 17:11:33 -070078static inline __printf(2, 3)
79void SOCK_DEBUG(struct sock *sk, const char *msg, ...)
Stephen Hemminger4cd90292008-03-21 15:54:53 -070080{
81}
Linus Torvalds1da177e2005-04-16 15:20:36 -070082#endif
83
84/* This is the per-socket lock. The spinlock provides a synchronization
85 * between user contexts and software interrupt processing, whereas the
86 * mini-semaphore synchronizes multiple users amongst themselves.
87 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070088typedef struct {
89 spinlock_t slock;
John Heffnerd2e91172007-09-12 10:44:19 +020090 int owned;
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 wait_queue_head_t wq;
Ingo Molnara5b5bb92006-07-03 00:25:35 -070092 /*
93 * We express the mutex-alike socket_lock semantics
94 * to the lock validator by explicitly managing
95 * the slock as a lock variant (in addition to
96 * the slock itself):
97 */
98#ifdef CONFIG_DEBUG_LOCK_ALLOC
99 struct lockdep_map dep_map;
100#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101} socket_lock_t;
102
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103struct sock;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700104struct proto;
Denis V. Lunev0eeb8ff2007-12-04 01:15:45 -0800105struct net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106
107/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700108 * struct sock_common - minimal network layer representation of sockets
Eric Dumazet68835ab2010-11-30 19:04:07 +0000109 * @skc_daddr: Foreign IPv4 addr
110 * @skc_rcv_saddr: Bound local IPv4 addr
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000111 * @skc_hash: hash value used with various protocol lookup tables
Eric Dumazetd4cada42009-11-08 10:17:30 +0000112 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700113 * @skc_family: network address family
114 * @skc_state: Connection state
115 * @skc_reuse: %SO_REUSEADDR setting
116 * @skc_bound_dev_if: bound device index if != 0
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700117 * @skc_bind_node: bind hash linkage for various protocol lookup tables
Eric Dumazet512615b2009-11-08 10:17:58 +0000118 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700119 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200120 * @skc_net: reference to the network namespace of this socket
Eric Dumazet68835ab2010-11-30 19:04:07 +0000121 * @skc_node: main hash linkage for various protocol lookup tables
122 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
123 * @skc_tx_queue_mapping: tx queue number for this connection
124 * @skc_refcnt: reference count
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700125 *
126 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700127 * for struct sock and struct inet_timewait_sock.
128 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129struct sock_common {
Eric Dumazet68835ab2010-11-30 19:04:07 +0000130 /* skc_daddr and skc_rcv_saddr must be grouped :
131 * cf INET_MATCH() and INET_TW_MATCH()
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000132 */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000133 __be32 skc_daddr;
134 __be32 skc_rcv_saddr;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000135
Eric Dumazetd4cada42009-11-08 10:17:30 +0000136 union {
137 unsigned int skc_hash;
138 __u16 skc_u16hashes[2];
139 };
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000140 unsigned short skc_family;
141 volatile unsigned char skc_state;
142 unsigned char skc_reuse;
143 int skc_bound_dev_if;
Eric Dumazet512615b2009-11-08 10:17:58 +0000144 union {
145 struct hlist_node skc_bind_node;
146 struct hlist_nulls_node skc_portaddr_node;
147 };
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700148 struct proto *skc_prot;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900149#ifdef CONFIG_NET_NS
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200150 struct net *skc_net;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900151#endif
Eric Dumazet68835ab2010-11-30 19:04:07 +0000152 /*
153 * fields between dontcopy_begin/dontcopy_end
154 * are not copied in sock_copy()
155 */
Randy Dunlap928c41e2011-01-08 17:39:21 +0000156 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000157 int skc_dontcopy_begin[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000158 /* public: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000159 union {
160 struct hlist_node skc_node;
161 struct hlist_nulls_node skc_nulls_node;
162 };
163 int skc_tx_queue_mapping;
164 atomic_t skc_refcnt;
Randy Dunlap928c41e2011-01-08 17:39:21 +0000165 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000166 int skc_dontcopy_end[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000167 /* public: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168};
169
170/**
171 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700172 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700173 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
174 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
175 * @sk_lock: synchronizer
176 * @sk_rcvbuf: size of receive buffer in bytes
Eric Dumazet43815482010-04-29 11:01:49 +0000177 * @sk_wq: sock wait queue and async head
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700178 * @sk_dst_cache: destination cache
179 * @sk_dst_lock: destination cache lock
180 * @sk_policy: flow policy
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700181 * @sk_receive_queue: incoming packets
182 * @sk_wmem_alloc: transmit queue bytes committed
183 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700184 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700185 * @sk_omem_alloc: "o" is "option" or "other"
186 * @sk_wmem_queued: persistent queue size
187 * @sk_forward_alloc: space allocated forward
188 * @sk_allocation: allocation mode
189 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800190 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
Patrick Ohly20d49472009-02-12 05:03:38 +0000191 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700192 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
193 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700194 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700195 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700196 * @sk_gso_max_size: Maximum GSO segment size to build
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700197 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700198 * @sk_backlog: always used with the per-socket spinlock held
199 * @sk_callback_lock: used with the callbacks in the end of this struct
200 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800201 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
202 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700203 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800204 * @sk_err_soft: errors that don't cause failure but are the cause of a
205 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700206 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700207 * @sk_ack_backlog: current listen backlog
208 * @sk_max_ack_backlog: listen backlog set in listen()
209 * @sk_priority: %SO_PRIORITY setting
210 * @sk_type: socket type (%SOCK_STREAM, etc)
211 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000212 * @sk_peer_pid: &struct pid for this socket's peer
213 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700214 * @sk_rcvlowat: %SO_RCVLOWAT setting
215 * @sk_rcvtimeo: %SO_RCVTIMEO setting
216 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700217 * @sk_rxhash: flow hash received from netif layer
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700218 * @sk_filter: socket filtering instructions
219 * @sk_protinfo: private area, net family specific, when not using slab
220 * @sk_timer: sock cleanup timer
221 * @sk_stamp: time stamp of last packet received
222 * @sk_socket: Identd and reporting IO signals
223 * @sk_user_data: RPC layer private data
224 * @sk_sndmsg_page: cached page for sendmsg
225 * @sk_sndmsg_off: cached offset for sendmsg
226 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700227 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800228 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000229 * @sk_classid: this socket's cgroup classid
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700230 * @sk_write_pending: a write to stream socket waits to start
231 * @sk_state_change: callback to indicate change in the state of the sock
232 * @sk_data_ready: callback to indicate there is data to be processed
233 * @sk_write_space: callback to indicate there is bf sending space available
234 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
235 * @sk_backlog_rcv: callback to process the backlog
236 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 */
238struct sock {
239 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700240 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 * don't add nothing before this first member (__sk_common) --acme
242 */
243 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000244#define sk_node __sk_common.skc_node
245#define sk_nulls_node __sk_common.skc_nulls_node
246#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000247#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000248
Eric Dumazet68835ab2010-11-30 19:04:07 +0000249#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
250#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000251#define sk_hash __sk_common.skc_hash
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252#define sk_family __sk_common.skc_family
253#define sk_state __sk_common.skc_state
254#define sk_reuse __sk_common.skc_reuse
255#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700257#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200258#define sk_net __sk_common.skc_net
Eric Dumazetb178bb32010-11-16 05:56:04 +0000259 socket_lock_t sk_lock;
260 struct sk_buff_head sk_receive_queue;
261 /*
262 * The backlog queue is special, it is always used with
263 * the per-socket spinlock held and requires low latency
264 * access. Therefore we special case it's implementation.
265 * Note : rmem_alloc is in this structure to fill a hole
266 * on 64bit arches, not because its logically part of
267 * backlog.
268 */
269 struct {
270 atomic_t rmem_alloc;
271 int len;
272 struct sk_buff *head;
273 struct sk_buff *tail;
274 } sk_backlog;
275#define sk_rmem_alloc sk_backlog.rmem_alloc
276 int sk_forward_alloc;
277#ifdef CONFIG_RPS
278 __u32 sk_rxhash;
279#endif
280 atomic_t sk_drops;
281 int sk_rcvbuf;
282
283 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000284 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000285
286#ifdef CONFIG_NET_DMA
287 struct sk_buff_head sk_async_wait_queue;
288#endif
289
290#ifdef CONFIG_XFRM
291 struct xfrm_policy *sk_policy[2];
292#endif
293 unsigned long sk_flags;
294 struct dst_entry *sk_dst_cache;
295 spinlock_t sk_dst_lock;
296 atomic_t sk_wmem_alloc;
297 atomic_t sk_omem_alloc;
298 int sk_sndbuf;
299 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100300 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000301 unsigned int sk_shutdown : 2,
302 sk_no_check : 2,
303 sk_userlocks : 4,
304 sk_protocol : 8,
305 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100306 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400308 gfp_t sk_allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 int sk_route_caps;
Eric Dumazeta4654192010-05-16 00:36:33 -0700310 int sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700311 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700312 unsigned int sk_gso_max_size;
David S. Miller9932cf92006-03-24 15:12:37 -0800313 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700316 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 rwlock_t sk_callback_lock;
318 int sk_err,
319 sk_err_soft;
320 unsigned short sk_ack_backlog;
321 unsigned short sk_max_ack_backlog;
322 __u32 sk_priority;
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000323 struct pid *sk_peer_pid;
324 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 long sk_rcvtimeo;
326 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 void *sk_protinfo;
328 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700329 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 struct socket *sk_socket;
331 void *sk_user_data;
332 struct page *sk_sndmsg_page;
333 struct sk_buff *sk_send_head;
334 __u32 sk_sndmsg_off;
335 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800336#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800338#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800339 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700340 u32 sk_classid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 void (*sk_state_change)(struct sock *sk);
342 void (*sk_data_ready)(struct sock *sk, int bytes);
343 void (*sk_write_space)(struct sock *sk);
344 void (*sk_error_report)(struct sock *sk);
345 int (*sk_backlog_rcv)(struct sock *sk,
346 struct sk_buff *skb);
347 void (*sk_destruct)(struct sock *sk);
348};
349
350/*
351 * Hashed lists helper routines
352 */
Li Zefanc4146642010-02-08 23:18:45 +0000353static inline struct sock *sk_entry(const struct hlist_node *node)
354{
355 return hlist_entry(node, struct sock, sk_node);
356}
357
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700358static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359{
360 return hlist_entry(head->first, struct sock, sk_node);
361}
362
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700363static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364{
365 return hlist_empty(head) ? NULL : __sk_head(head);
366}
367
Eric Dumazet88ab1932008-11-16 19:39:21 -0800368static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
369{
370 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
371}
372
373static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
374{
375 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
376}
377
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700378static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379{
380 return sk->sk_node.next ?
381 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
382}
383
Eric Dumazet88ab1932008-11-16 19:39:21 -0800384static inline struct sock *sk_nulls_next(const struct sock *sk)
385{
386 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
387 hlist_nulls_entry(sk->sk_nulls_node.next,
388 struct sock, sk_nulls_node) :
389 NULL;
390}
391
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700392static inline int sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393{
394 return hlist_unhashed(&sk->sk_node);
395}
396
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700397static inline int sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398{
Akinobu Mitada753be2006-04-28 15:21:23 -0700399 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400}
401
402static __inline__ void sk_node_init(struct hlist_node *node)
403{
404 node->pprev = NULL;
405}
406
Eric Dumazet88ab1932008-11-16 19:39:21 -0800407static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
408{
409 node->pprev = NULL;
410}
411
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412static __inline__ void __sk_del_node(struct sock *sk)
413{
414 __hlist_del(&sk->sk_node);
415}
416
stephen hemminger808f5112010-02-22 07:57:18 +0000417/* NB: equivalent to hlist_del_init_rcu */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418static __inline__ int __sk_del_node_init(struct sock *sk)
419{
420 if (sk_hashed(sk)) {
421 __sk_del_node(sk);
422 sk_node_init(&sk->sk_node);
423 return 1;
424 }
425 return 0;
426}
427
428/* Grab socket reference count. This operation is valid only
429 when sk is ALREADY grabbed f.e. it is found in hash table
430 or a list and the lookup is made under lock preventing hash table
431 modifications.
432 */
433
434static inline void sock_hold(struct sock *sk)
435{
436 atomic_inc(&sk->sk_refcnt);
437}
438
439/* Ungrab socket in the context, which assumes that socket refcnt
440 cannot hit zero, f.e. it is true in context of any socketcall.
441 */
442static inline void __sock_put(struct sock *sk)
443{
444 atomic_dec(&sk->sk_refcnt);
445}
446
447static __inline__ int sk_del_node_init(struct sock *sk)
448{
449 int rc = __sk_del_node_init(sk);
450
451 if (rc) {
452 /* paranoid for a while -acme */
453 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
454 __sock_put(sk);
455 }
456 return rc;
457}
stephen hemminger808f5112010-02-22 07:57:18 +0000458#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459
Eric Dumazet88ab1932008-11-16 19:39:21 -0800460static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700461{
462 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800463 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700464 return 1;
465 }
466 return 0;
467}
468
Eric Dumazet88ab1932008-11-16 19:39:21 -0800469static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700470{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800471 int rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700472
473 if (rc) {
474 /* paranoid for a while -acme */
475 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
476 __sock_put(sk);
477 }
478 return rc;
479}
480
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
482{
483 hlist_add_head(&sk->sk_node, list);
484}
485
486static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
487{
488 sock_hold(sk);
489 __sk_add_node(sk, list);
490}
491
stephen hemminger808f5112010-02-22 07:57:18 +0000492static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
493{
494 sock_hold(sk);
495 hlist_add_head_rcu(&sk->sk_node, list);
496}
497
Eric Dumazet88ab1932008-11-16 19:39:21 -0800498static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700499{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800500 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700501}
502
Eric Dumazet88ab1932008-11-16 19:39:21 -0800503static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700504{
505 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800506 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700507}
508
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509static __inline__ void __sk_del_bind_node(struct sock *sk)
510{
511 __hlist_del(&sk->sk_bind_node);
512}
513
514static __inline__ void sk_add_bind_node(struct sock *sk,
515 struct hlist_head *list)
516{
517 hlist_add_head(&sk->sk_bind_node, list);
518}
519
520#define sk_for_each(__sk, node, list) \
521 hlist_for_each_entry(__sk, node, list, sk_node)
stephen hemminger808f5112010-02-22 07:57:18 +0000522#define sk_for_each_rcu(__sk, node, list) \
523 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800524#define sk_nulls_for_each(__sk, node, list) \
525 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
526#define sk_nulls_for_each_rcu(__sk, node, list) \
527 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528#define sk_for_each_from(__sk, node) \
529 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
530 hlist_for_each_entry_from(__sk, node, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800531#define sk_nulls_for_each_from(__sk, node) \
532 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
533 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534#define sk_for_each_safe(__sk, node, tmp, list) \
535 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
536#define sk_for_each_bound(__sk, node, list) \
537 hlist_for_each_entry(__sk, node, list, sk_bind_node)
538
539/* Sock flags */
540enum sock_flags {
541 SOCK_DEAD,
542 SOCK_DONE,
543 SOCK_URGINLINE,
544 SOCK_KEEPOPEN,
545 SOCK_LINGER,
546 SOCK_DESTROY,
547 SOCK_BROADCAST,
548 SOCK_TIMESTAMP,
549 SOCK_ZAPPED,
550 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
551 SOCK_DBG, /* %SO_DEBUG setting */
552 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700553 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
555 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Patrick Ohly20d49472009-02-12 05:03:38 +0000556 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
557 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
558 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
559 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
560 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
561 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
562 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700563 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700564 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000565 SOCK_ZEROCOPY, /* buffers from userspace */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566};
567
Ralf Baechle53b924b2005-08-23 10:11:30 -0700568static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
569{
570 nsk->sk_flags = osk->sk_flags;
571}
572
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
574{
575 __set_bit(flag, &sk->sk_flags);
576}
577
578static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
579{
580 __clear_bit(flag, &sk->sk_flags);
581}
582
583static inline int sock_flag(struct sock *sk, enum sock_flags flag)
584{
585 return test_bit(flag, &sk->sk_flags);
586}
587
588static inline void sk_acceptq_removed(struct sock *sk)
589{
590 sk->sk_ack_backlog--;
591}
592
593static inline void sk_acceptq_added(struct sock *sk)
594{
595 sk->sk_ack_backlog++;
596}
597
598static inline int sk_acceptq_is_full(struct sock *sk)
599{
David S. Miller64a14652007-03-06 11:21:05 -0800600 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601}
602
603/*
604 * Compute minimal free write space needed to queue new packets.
605 */
606static inline int sk_stream_min_wspace(struct sock *sk)
607{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800608 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609}
610
611static inline int sk_stream_wspace(struct sock *sk)
612{
613 return sk->sk_sndbuf - sk->sk_wmem_queued;
614}
615
616extern void sk_stream_write_space(struct sock *sk);
617
618static inline int sk_stream_memory_free(struct sock *sk)
619{
620 return sk->sk_wmem_queued < sk->sk_sndbuf;
621}
622
Zhu Yi8eae9392010-03-04 18:01:40 +0000623/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000624static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800625{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000626 /* dont let skb dst not refcounted, we are going to leave rcu lock */
627 skb_dst_force(skb);
628
629 if (!sk->sk_backlog.tail)
630 sk->sk_backlog.head = skb;
631 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800632 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000633
634 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800635 skb->next = NULL;
636}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637
Eric Dumazetc3774112010-04-27 15:13:20 -0700638/*
639 * Take into account size of receive queue and backlog queue
640 */
641static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
642{
643 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
644
645 return qsize + skb->truesize > sk->sk_rcvbuf;
646}
647
Zhu Yi8eae9392010-03-04 18:01:40 +0000648/* The per-socket spinlock must be held here. */
Zhu Yi40456352010-03-07 16:21:39 +0000649static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Zhu Yi8eae9392010-03-04 18:01:40 +0000650{
Eric Dumazetc3774112010-04-27 15:13:20 -0700651 if (sk_rcvqueues_full(sk, skb))
Zhu Yi8eae9392010-03-04 18:01:40 +0000652 return -ENOBUFS;
653
Zhu Yia3a858f2010-03-04 18:01:47 +0000654 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000655 sk->sk_backlog.len += skb->truesize;
656 return 0;
657}
658
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700659static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
660{
661 return sk->sk_backlog_rcv(sk, skb);
662}
663
David S. Millerc58dc012010-04-27 15:05:31 -0700664static inline void sock_rps_record_flow(const struct sock *sk)
665{
666#ifdef CONFIG_RPS
667 struct rps_sock_flow_table *sock_flow_table;
668
669 rcu_read_lock();
670 sock_flow_table = rcu_dereference(rps_sock_flow_table);
671 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
672 rcu_read_unlock();
673#endif
674}
675
676static inline void sock_rps_reset_flow(const struct sock *sk)
677{
678#ifdef CONFIG_RPS
679 struct rps_sock_flow_table *sock_flow_table;
680
681 rcu_read_lock();
682 sock_flow_table = rcu_dereference(rps_sock_flow_table);
683 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
684 rcu_read_unlock();
685#endif
686}
687
Tom Herbertbdeab992011-08-14 19:45:55 +0000688static inline void sock_rps_save_rxhash(struct sock *sk,
689 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700690{
691#ifdef CONFIG_RPS
Tom Herbertbdeab992011-08-14 19:45:55 +0000692 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
David S. Millerc58dc012010-04-27 15:05:31 -0700693 sock_rps_reset_flow(sk);
Tom Herbertbdeab992011-08-14 19:45:55 +0000694 sk->sk_rxhash = skb->rxhash;
David S. Millerc58dc012010-04-27 15:05:31 -0700695 }
696#endif
697}
698
Tom Herbertbdeab992011-08-14 19:45:55 +0000699static inline void sock_rps_reset_rxhash(struct sock *sk)
700{
701#ifdef CONFIG_RPS
702 sock_rps_reset_flow(sk);
703 sk->sk_rxhash = 0;
704#endif
705}
706
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700707#define sk_wait_event(__sk, __timeo, __condition) \
708 ({ int __rc; \
709 release_sock(__sk); \
710 __rc = __condition; \
711 if (!__rc) { \
712 *(__timeo) = schedule_timeout(*(__timeo)); \
713 } \
714 lock_sock(__sk); \
715 __rc = __condition; \
716 __rc; \
717 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718
719extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
720extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
721extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
722extern int sk_stream_error(struct sock *sk, int flags, int err);
723extern void sk_stream_kill_queues(struct sock *sk);
724
725extern int sk_wait_data(struct sock *sk, long *timeo);
726
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700727struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800728struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800729struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700730struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400731struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700732
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733/* Networking protocol blocks we attach to sockets.
734 * socket layer -> transport layer interface
735 * transport -> network interface is defined by struct inet_proto
736 */
737struct proto {
738 void (*close)(struct sock *sk,
739 long timeout);
740 int (*connect)(struct sock *sk,
741 struct sockaddr *uaddr,
742 int addr_len);
743 int (*disconnect)(struct sock *sk, int flags);
744
745 struct sock * (*accept) (struct sock *sk, int flags, int *err);
746
747 int (*ioctl)(struct sock *sk, int cmd,
748 unsigned long arg);
749 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700750 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 void (*shutdown)(struct sock *sk, int how);
752 int (*setsockopt)(struct sock *sk, int level,
753 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700754 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 int (*getsockopt)(struct sock *sk, int level,
756 int optname, char __user *optval,
757 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700758#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800759 int (*compat_setsockopt)(struct sock *sk,
760 int level,
761 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700762 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800763 int (*compat_getsockopt)(struct sock *sk,
764 int level,
765 int optname, char __user *optval,
766 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000767 int (*compat_ioctl)(struct sock *sk,
768 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700769#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
771 struct msghdr *msg, size_t len);
772 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
773 struct msghdr *msg,
774 size_t len, int noblock, int flags,
775 int *addr_len);
776 int (*sendpage)(struct sock *sk, struct page *page,
777 int offset, size_t size, int flags);
778 int (*bind)(struct sock *sk,
779 struct sockaddr *uaddr, int addr_len);
780
781 int (*backlog_rcv) (struct sock *sk,
782 struct sk_buff *skb);
783
784 /* Keeping track of sk's, looking them up, and port selection methods. */
785 void (*hash)(struct sock *sk);
786 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000787 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800789 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800791 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800792#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700793 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800794#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800795
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700797 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +0000798 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -0800799 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 /*
801 * Pressure flag: try to collapse.
802 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800803 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 * is strict, actions are advisory and have some latency.
805 */
806 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000807 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 int *sysctl_wmem;
809 int *sysctl_rmem;
810 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +0000811 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
Eric Dumazet271b72c2008-10-29 02:11:14 -0700813 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700815 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816
Eric Dumazetdd24c002008-11-25 21:17:14 -0800817 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700818
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700819 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800820 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700821
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700822 union {
823 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -0700824 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700825 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700826 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800827
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 struct module *owner;
829
830 char name[32];
831
832 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700833#ifdef SOCK_REFCNT_DEBUG
834 atomic_t socks;
835#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836};
837
838extern int proto_register(struct proto *prot, int alloc_slab);
839extern void proto_unregister(struct proto *prot);
840
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700841#ifdef SOCK_REFCNT_DEBUG
842static inline void sk_refcnt_debug_inc(struct sock *sk)
843{
844 atomic_inc(&sk->sk_prot->socks);
845}
846
847static inline void sk_refcnt_debug_dec(struct sock *sk)
848{
849 atomic_dec(&sk->sk_prot->socks);
850 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
851 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
852}
853
854static inline void sk_refcnt_debug_release(const struct sock *sk)
855{
856 if (atomic_read(&sk->sk_refcnt) != 1)
857 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
858 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
859}
860#else /* SOCK_REFCNT_DEBUG */
861#define sk_refcnt_debug_inc(sk) do { } while (0)
862#define sk_refcnt_debug_dec(sk) do { } while (0)
863#define sk_refcnt_debug_release(sk) do { } while (0)
864#endif /* SOCK_REFCNT_DEBUG */
865
Eric Dumazet65f76512008-01-03 20:46:48 -0800866
867#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868/* Called with local bh disabled */
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700869extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
870extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -0800871#else
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700872static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
873 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -0800874{
875}
Eric Dumazet65f76512008-01-03 20:46:48 -0800876#endif
877
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -0700879/* With per-bucket locks this operation is not-atomic, so that
880 * this version is not worse.
881 */
882static inline void __sk_prot_rehash(struct sock *sk)
883{
884 sk->sk_prot->unhash(sk);
885 sk->sk_prot->hash(sk);
886}
887
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800888void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
889
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890/* About 10 seconds */
891#define SOCK_DESTROY_TIME (10*HZ)
892
893/* Sockets 0-1023 can't be bound to unless you are superuser */
894#define PROT_SOCK 1024
895
896#define SHUTDOWN_MASK 3
897#define RCV_SHUTDOWN 1
898#define SEND_SHUTDOWN 2
899
900#define SOCK_SNDBUF_LOCK 1
901#define SOCK_RCVBUF_LOCK 2
902#define SOCK_BINDADDR_LOCK 4
903#define SOCK_BINDPORT_LOCK 8
904
905/* sock_iocb: used to kick off async processing of socket ios */
906struct sock_iocb {
907 struct list_head list;
908
909 int flags;
910 int size;
911 struct socket *sock;
912 struct sock *sk;
913 struct scm_cookie *scm;
914 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 struct kiocb *kiocb;
916};
917
918static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
919{
920 return (struct sock_iocb *)iocb->private;
921}
922
923static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
924{
925 return si->kiocb;
926}
927
928struct socket_alloc {
929 struct socket socket;
930 struct inode vfs_inode;
931};
932
933static inline struct socket *SOCKET_I(struct inode *inode)
934{
935 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
936}
937
938static inline struct inode *SOCK_INODE(struct socket *socket)
939{
940 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
941}
942
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800943/*
944 * Functions for memory accounting
945 */
946extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
947extern void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800949#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
950#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
951#define SK_MEM_SEND 0
952#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800954static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800956 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957}
958
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800959static inline int sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800961 /* return true if protocol supports memory accounting */
962 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963}
964
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800965static inline int sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800967 if (!sk_has_account(sk))
968 return 1;
Herbert Xud80d99d2005-09-01 17:48:23 -0700969 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800970 __sk_mem_schedule(sk, size, SK_MEM_SEND);
971}
972
973static inline int sk_rmem_schedule(struct sock *sk, int size)
974{
975 if (!sk_has_account(sk))
976 return 1;
977 return size <= sk->sk_forward_alloc ||
978 __sk_mem_schedule(sk, size, SK_MEM_RECV);
979}
980
981static inline void sk_mem_reclaim(struct sock *sk)
982{
983 if (!sk_has_account(sk))
984 return;
985 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
986 __sk_mem_reclaim(sk);
987}
988
David S. Miller9993e7d2008-01-10 21:56:38 -0800989static inline void sk_mem_reclaim_partial(struct sock *sk)
990{
991 if (!sk_has_account(sk))
992 return;
993 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
994 __sk_mem_reclaim(sk);
995}
996
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800997static inline void sk_mem_charge(struct sock *sk, int size)
998{
999 if (!sk_has_account(sk))
1000 return;
1001 sk->sk_forward_alloc -= size;
1002}
1003
1004static inline void sk_mem_uncharge(struct sock *sk, int size)
1005{
1006 if (!sk_has_account(sk))
1007 return;
1008 sk->sk_forward_alloc += size;
1009}
1010
1011static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1012{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001013 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1014 sk->sk_wmem_queued -= skb->truesize;
1015 sk_mem_uncharge(sk, skb->truesize);
1016 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001017}
1018
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019/* Used by processes to "lock" a socket state, so that
1020 * interrupts and bottom half handlers won't change it
1021 * from under us. It essentially blocks any incoming
1022 * packets, so that we won't get any new data or any
1023 * packets that change the state of the socket.
1024 *
1025 * While locked, BH processing will add new packets to
1026 * the backlog queue. This queue is processed by the
1027 * owner of the socket lock right before it is released.
1028 *
1029 * Since ~2.3.5 it is also exclusive sleep lock serializing
1030 * accesses from user process context.
1031 */
John Heffnerd2e91172007-09-12 10:44:19 +02001032#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033
Peter Zijlstraed075362006-12-06 20:35:24 -08001034/*
1035 * Macro so as to not evaluate some arguments when
1036 * lockdep is not enabled.
1037 *
1038 * Mark both the sk_lock and the sk_lock.slock as a
1039 * per-address-family lock class.
1040 */
1041#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1042do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001043 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001044 init_waitqueue_head(&sk->sk_lock.wq); \
1045 spin_lock_init(&(sk)->sk_lock.slock); \
1046 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1047 sizeof((sk)->sk_lock)); \
1048 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1049 (skey), (sname)); \
1050 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1051} while (0)
1052
Harvey Harrison41380932007-12-12 10:46:51 -08001053extern void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001054
1055static inline void lock_sock(struct sock *sk)
1056{
1057 lock_sock_nested(sk, 0);
1058}
1059
Harvey Harrison41380932007-12-12 10:46:51 -08001060extern void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061
1062/* BH context may only use the following locking interface. */
1063#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001064#define bh_lock_sock_nested(__sk) \
1065 spin_lock_nested(&((__sk)->sk_lock.slock), \
1066 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1068
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001069extern bool lock_sock_fast(struct sock *sk);
1070/**
1071 * unlock_sock_fast - complement of lock_sock_fast
1072 * @sk: socket
1073 * @slow: slow mode
1074 *
1075 * fast unlock socket for user context.
1076 * If slow mode is on, we call regular release_sock()
1077 */
1078static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001079{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001080 if (slow)
1081 release_sock(sk);
1082 else
1083 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001084}
1085
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001086
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001087extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +01001088 gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001089 struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090extern void sk_free(struct sock *sk);
Denis V. Lunevedf02082008-02-29 11:18:32 -08001091extern void sk_release_kernel(struct sock *sk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001092extern struct sock *sk_clone(const struct sock *sk,
Al Virodd0fc662005-10-07 07:46:04 +01001093 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094
1095extern struct sk_buff *sock_wmalloc(struct sock *sk,
1096 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001097 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098extern struct sk_buff *sock_rmalloc(struct sock *sk,
1099 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001100 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101extern void sock_wfree(struct sk_buff *skb);
1102extern void sock_rfree(struct sk_buff *skb);
1103
1104extern int sock_setsockopt(struct socket *sock, int level,
1105 int op, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07001106 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
1108extern int sock_getsockopt(struct socket *sock, int level,
1109 int op, char __user *optval,
1110 int __user *optlen);
1111extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1112 unsigned long size,
1113 int noblock,
1114 int *errcode);
Herbert Xu4cc7f682009-02-04 16:55:54 -08001115extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1116 unsigned long header_len,
1117 unsigned long data_len,
1118 int noblock,
1119 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -07001120extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +01001121 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1123extern void sk_send_sigurg(struct sock *sk);
1124
Herbert Xuf8451722010-05-24 00:12:34 -07001125#ifdef CONFIG_CGROUPS
1126extern void sock_update_classid(struct sock *sk);
1127#else
1128static inline void sock_update_classid(struct sock *sk)
1129{
1130}
1131#endif
1132
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133/*
1134 * Functions to fill in entries in struct proto_ops when a protocol
1135 * does not implement a particular function.
1136 */
1137extern int sock_no_bind(struct socket *,
1138 struct sockaddr *, int);
1139extern int sock_no_connect(struct socket *,
1140 struct sockaddr *, int, int);
1141extern int sock_no_socketpair(struct socket *,
1142 struct socket *);
1143extern int sock_no_accept(struct socket *,
1144 struct socket *, int);
1145extern int sock_no_getname(struct socket *,
1146 struct sockaddr *, int *, int);
1147extern unsigned int sock_no_poll(struct file *, struct socket *,
1148 struct poll_table_struct *);
1149extern int sock_no_ioctl(struct socket *, unsigned int,
1150 unsigned long);
1151extern int sock_no_listen(struct socket *, int);
1152extern int sock_no_shutdown(struct socket *, int);
1153extern int sock_no_getsockopt(struct socket *, int , int,
1154 char __user *, int __user *);
1155extern int sock_no_setsockopt(struct socket *, int, int,
David S. Millerb7058842009-09-30 16:12:20 -07001156 char __user *, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1158 struct msghdr *, size_t);
1159extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1160 struct msghdr *, size_t, int);
1161extern int sock_no_mmap(struct file *file,
1162 struct socket *sock,
1163 struct vm_area_struct *vma);
1164extern ssize_t sock_no_sendpage(struct socket *sock,
1165 struct page *page,
1166 int offset, size_t size,
1167 int flags);
1168
1169/*
1170 * Functions to fill in entries in struct proto_ops when a protocol
1171 * uses the inet style.
1172 */
1173extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1174 char __user *optval, int __user *optlen);
1175extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1176 struct msghdr *msg, size_t size, int flags);
1177extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001178 char __user *optval, unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001179extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1180 int optname, char __user *optval, int __user *optlen);
1181extern int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001182 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
1184extern void sk_common_release(struct sock *sk);
1185
1186/*
1187 * Default socket callbacks and setup code
1188 */
1189
1190/* Initialise core socket variables */
1191extern void sock_init_data(struct socket *sock, struct sock *sk);
1192
Eric Dumazet46bcf142010-12-06 09:29:43 -08001193extern void sk_filter_release_rcu(struct rcu_head *rcu);
1194
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001196 * sk_filter_release - release a socket filter
Paul Bonserdc9b3342006-11-23 17:56:13 -08001197 * @fp: filter to remove
1198 *
1199 * Remove a filter from a socket and release its resources.
1200 */
1201
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001202static inline void sk_filter_release(struct sk_filter *fp)
1203{
1204 if (atomic_dec_and_test(&fp->refcnt))
Eric Dumazet80f8f102011-01-18 07:46:52 +00001205 call_rcu(&fp->rcu, sk_filter_release_rcu);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001206}
1207
1208static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209{
1210 unsigned int size = sk_filter_len(fp);
1211
1212 atomic_sub(size, &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001213 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214}
1215
1216static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1217{
1218 atomic_inc(&fp->refcnt);
1219 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1220}
1221
1222/*
1223 * Socket reference counting postulates.
1224 *
1225 * * Each user of socket SHOULD hold a reference count.
1226 * * Each access point to socket (an hash table bucket, reference from a list,
1227 * running timer, skb in flight MUST hold a reference count.
1228 * * When reference count hits 0, it means it will never increase back.
1229 * * When reference count hits 0, it means that no references from
1230 * outside exist to this socket and current process on current CPU
1231 * is last user and may/should destroy this socket.
1232 * * sk_free is called from any context: process, BH, IRQ. When
1233 * it is called, socket has no references from outside -> sk_free
1234 * may release descendant resources allocated by the socket, but
1235 * to the time when it is called, socket is NOT referenced by any
1236 * hash tables, lists etc.
1237 * * Packets, delivered from outside (from network or from another process)
1238 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1239 * when they sit in queue. Otherwise, packets will leak to hole, when
1240 * socket is looked up by one cpu and unhasing is made by another CPU.
1241 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1242 * (leak to backlog). Packet socket does all the processing inside
1243 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1244 * use separate SMP lock, so that they are prone too.
1245 */
1246
1247/* Ungrab socket and destroy it, if it was the last reference. */
1248static inline void sock_put(struct sock *sk)
1249{
1250 if (atomic_dec_and_test(&sk->sk_refcnt))
1251 sk_free(sk);
1252}
1253
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -02001254extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1255 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001256
Krishna Kumare022f0b2009-10-19 23:46:20 +00001257static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1258{
1259 sk->sk_tx_queue_mapping = tx_queue;
1260}
1261
1262static inline void sk_tx_queue_clear(struct sock *sk)
1263{
1264 sk->sk_tx_queue_mapping = -1;
1265}
1266
1267static inline int sk_tx_queue_get(const struct sock *sk)
1268{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001269 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001270}
1271
David S. Miller972692e2008-06-17 22:41:38 -07001272static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1273{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001274 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001275 sk->sk_socket = sock;
1276}
1277
Eric Dumazetaa395142010-04-20 13:03:51 +00001278static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1279{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001280 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1281 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001282}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283/* Detach socket from process context.
1284 * Announce socket dead, detach it from wait queue and inode.
1285 * Note that parent inode held reference count on this struct sock,
1286 * we do not release it in this function, because protocol
1287 * probably wants some additional cleanups or even continuing
1288 * to work with this socket (TCP).
1289 */
1290static inline void sock_orphan(struct sock *sk)
1291{
1292 write_lock_bh(&sk->sk_callback_lock);
1293 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001294 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001295 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 write_unlock_bh(&sk->sk_callback_lock);
1297}
1298
1299static inline void sock_graft(struct sock *sk, struct socket *parent)
1300{
1301 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001302 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001304 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001305 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 write_unlock_bh(&sk->sk_callback_lock);
1307}
1308
1309extern int sock_i_uid(struct sock *sk);
1310extern unsigned long sock_i_ino(struct sock *sk);
1311
1312static inline struct dst_entry *
1313__sk_dst_get(struct sock *sk)
1314{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001315 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001316 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317}
1318
1319static inline struct dst_entry *
1320sk_dst_get(struct sock *sk)
1321{
1322 struct dst_entry *dst;
1323
Eric Dumazetb6c67122010-04-08 23:03:29 +00001324 rcu_read_lock();
1325 dst = rcu_dereference(sk->sk_dst_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 if (dst)
1327 dst_hold(dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001328 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 return dst;
1330}
1331
Eric Dumazetb6c67122010-04-08 23:03:29 +00001332extern void sk_reset_txq(struct sock *sk);
1333
1334static inline void dst_negative_advice(struct sock *sk)
1335{
1336 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1337
1338 if (dst && dst->ops->negative_advice) {
1339 ndst = dst->ops->negative_advice(dst);
1340
1341 if (ndst != dst) {
1342 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1343 sk_reset_txq(sk);
1344 }
1345 }
1346}
1347
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348static inline void
1349__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1350{
1351 struct dst_entry *old_dst;
1352
Krishna Kumare022f0b2009-10-19 23:46:20 +00001353 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001354 /*
1355 * This can be called while sk is owned by the caller only,
1356 * with no state that can be checked in a rcu_dereference_check() cond
1357 */
1358 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001359 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 dst_release(old_dst);
1361}
1362
1363static inline void
1364sk_dst_set(struct sock *sk, struct dst_entry *dst)
1365{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001366 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 __sk_dst_set(sk, dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001368 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369}
1370
1371static inline void
1372__sk_dst_reset(struct sock *sk)
1373{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001374 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375}
1376
1377static inline void
1378sk_dst_reset(struct sock *sk)
1379{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001380 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 __sk_dst_reset(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001382 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383}
1384
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001385extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001387extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388
Herbert Xubcd76112006-06-30 13:36:35 -07001389static inline int sk_can_gso(const struct sock *sk)
1390{
1391 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1392}
1393
Andi Kleen99580892007-04-20 17:12:43 -07001394extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001395
Eric Dumazeta4654192010-05-16 00:36:33 -07001396static inline void sk_nocaps_add(struct sock *sk, int flags)
1397{
1398 sk->sk_route_nocaps |= flags;
1399 sk->sk_route_caps &= ~flags;
1400}
1401
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001402static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1403 char __user *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001404 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001405{
1406 if (skb->ip_summed == CHECKSUM_NONE) {
1407 int err = 0;
1408 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1409 if (err)
1410 return err;
Wei Yongjun912d3982011-04-06 18:40:12 +00001411 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001412 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1413 if (!access_ok(VERIFY_READ, from, copy) ||
1414 __copy_from_user_nocache(to, from, copy))
1415 return -EFAULT;
1416 } else if (copy_from_user(to, from, copy))
1417 return -EFAULT;
1418
1419 return 0;
1420}
1421
1422static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1423 char __user *from, int copy)
1424{
Wei Yongjun912d3982011-04-06 18:40:12 +00001425 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001426
Wei Yongjun912d3982011-04-06 18:40:12 +00001427 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1428 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001429 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001430 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001431
1432 return err;
1433}
1434
1435static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1436 struct sk_buff *skb,
1437 struct page *page,
1438 int off, int copy)
1439{
1440 int err;
1441
Wei Yongjun912d3982011-04-06 18:40:12 +00001442 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1443 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001444 if (err)
1445 return err;
1446
1447 skb->len += copy;
1448 skb->data_len += copy;
1449 skb->truesize += copy;
1450 sk->sk_wmem_queued += copy;
1451 sk_mem_charge(sk, copy);
1452 return 0;
1453}
1454
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1456 struct sk_buff *skb, struct page *page,
1457 int off, int copy)
1458{
1459 if (skb->ip_summed == CHECKSUM_NONE) {
1460 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001461 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 page_address(page) + off,
1463 copy, 0, &err);
1464 if (err)
1465 return err;
1466 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1467 } else if (copy_from_user(page_address(page) + off, from, copy))
1468 return -EFAULT;
1469
1470 skb->len += copy;
1471 skb->data_len += copy;
1472 skb->truesize += copy;
1473 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001474 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 return 0;
1476}
1477
Eric Dumazetc5640392009-06-16 10:12:03 +00001478/**
1479 * sk_wmem_alloc_get - returns write allocations
1480 * @sk: socket
1481 *
1482 * Returns sk_wmem_alloc minus initial offset of one
1483 */
1484static inline int sk_wmem_alloc_get(const struct sock *sk)
1485{
1486 return atomic_read(&sk->sk_wmem_alloc) - 1;
1487}
1488
1489/**
1490 * sk_rmem_alloc_get - returns read allocations
1491 * @sk: socket
1492 *
1493 * Returns sk_rmem_alloc
1494 */
1495static inline int sk_rmem_alloc_get(const struct sock *sk)
1496{
1497 return atomic_read(&sk->sk_rmem_alloc);
1498}
1499
1500/**
1501 * sk_has_allocations - check if allocations are outstanding
1502 * @sk: socket
1503 *
1504 * Returns true if socket has write or read allocations
1505 */
1506static inline int sk_has_allocations(const struct sock *sk)
1507{
1508 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1509}
1510
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001511/**
Eric Dumazet43815482010-04-29 11:01:49 +00001512 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001513 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001514 *
Eric Dumazet43815482010-04-29 11:01:49 +00001515 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001516 *
Eric Dumazet43815482010-04-29 11:01:49 +00001517 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001518 * barrier call. They were added due to the race found within the tcp code.
1519 *
1520 * Consider following tcp code paths:
1521 *
1522 * CPU1 CPU2
1523 *
1524 * sys_select receive packet
1525 * ... ...
1526 * __add_wait_queue update tp->rcv_nxt
1527 * ... ...
1528 * tp->rcv_nxt check sock_def_readable
1529 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001530 * schedule rcu_read_lock();
1531 * wq = rcu_dereference(sk->sk_wq);
1532 * if (wq && waitqueue_active(&wq->wait))
1533 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001534 * ...
1535 * }
1536 *
1537 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1538 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1539 * could then endup calling schedule and sleep forever if there are no more
1540 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001541 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001542 */
Eric Dumazet43815482010-04-29 11:01:49 +00001543static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001544{
Eric Dumazet43815482010-04-29 11:01:49 +00001545
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001546 /*
1547 * We need to be sure we are in sync with the
1548 * add_wait_queue modifications to the wait queue.
1549 *
1550 * This memory barrier is paired in the sock_poll_wait.
1551 */
Eric Dumazet43815482010-04-29 11:01:49 +00001552 smp_mb();
1553 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001554}
1555
1556/**
1557 * sock_poll_wait - place memory barrier behind the poll_wait call.
1558 * @filp: file
1559 * @wait_address: socket wait queue
1560 * @p: poll_table
1561 *
Eric Dumazet43815482010-04-29 11:01:49 +00001562 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001563 */
1564static inline void sock_poll_wait(struct file *filp,
1565 wait_queue_head_t *wait_address, poll_table *p)
1566{
1567 if (p && wait_address) {
1568 poll_wait(filp, wait_address, p);
1569 /*
1570 * We need to be sure we are in sync with the
1571 * socket flags modification.
1572 *
Eric Dumazet43815482010-04-29 11:01:49 +00001573 * This memory barrier is paired in the wq_has_sleeper.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001574 */
1575 smp_mb();
1576 }
1577}
1578
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579/*
1580 * Queue a received datagram if it will fit. Stream and sequenced
1581 * protocols can't normally use this as they need to fit buffers in
1582 * and play with them.
1583 *
1584 * Inlined as it's very short and called for pretty much every
1585 * packet ever received.
1586 */
1587
1588static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1589{
Herbert Xud55d87f2009-06-22 02:25:25 +00001590 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 skb->sk = sk;
1592 skb->destructor = sock_wfree;
Eric Dumazet2b85a342009-06-11 02:55:43 -07001593 /*
1594 * We used to take a refcount on sk, but following operation
1595 * is enough to guarantee sk_free() wont free this sock until
1596 * all in-flight packets are completed
1597 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1599}
1600
1601static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1602{
Herbert Xud55d87f2009-06-22 02:25:25 +00001603 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 skb->sk = sk;
1605 skb->destructor = sock_rfree;
1606 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001607 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608}
1609
1610extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1611 unsigned long expires);
1612
1613extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1614
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001615extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616
Eric Dumazetb1faf562010-05-31 23:44:05 -07001617extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618
1619/*
1620 * Recover an error report and clear atomically
1621 */
1622
1623static inline int sock_error(struct sock *sk)
1624{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001625 int err;
1626 if (likely(!sk->sk_err))
1627 return 0;
1628 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 return -err;
1630}
1631
1632static inline unsigned long sock_wspace(struct sock *sk)
1633{
1634 int amt = 0;
1635
1636 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1637 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1638 if (amt < 0)
1639 amt = 0;
1640 }
1641 return amt;
1642}
1643
1644static inline void sk_wake_async(struct sock *sk, int how, int band)
1645{
Eric Dumazetbcdce712009-10-06 17:28:29 -07001646 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 sock_wake_async(sk->sk_socket, how, band);
1648}
1649
1650#define SOCK_MIN_SNDBUF 2048
Eric Dumazet7a91b4342010-09-26 18:53:07 -07001651/*
1652 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1653 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1654 */
1655#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656
1657static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1658{
1659 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001660 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1662 }
1663}
1664
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11001665struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666
1667static inline struct page *sk_stream_alloc_page(struct sock *sk)
1668{
1669 struct page *page = NULL;
1670
Herbert Xuef015782005-09-01 17:48:59 -07001671 page = alloc_pages(sk->sk_allocation, 0);
1672 if (!page) {
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001673 sk->sk_prot->enter_memory_pressure(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 sk_stream_moderate_sndbuf(sk);
1675 }
1676 return page;
1677}
1678
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679/*
1680 * Default write policy as shown to user space via poll/select/SIGIO
1681 */
1682static inline int sock_writeable(const struct sock *sk)
1683{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001684 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685}
1686
Al Virodd0fc662005-10-07 07:46:04 +01001687static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688{
Andrew Morton99709372009-02-12 16:43:17 -08001689 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690}
1691
1692static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1693{
1694 return noblock ? 0 : sk->sk_rcvtimeo;
1695}
1696
1697static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1698{
1699 return noblock ? 0 : sk->sk_sndtimeo;
1700}
1701
1702static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1703{
1704 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1705}
1706
1707/* Alas, with timeout socket operations are not restartable.
1708 * Compare this to poll().
1709 */
1710static inline int sock_intr_errno(long timeo)
1711{
1712 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1713}
1714
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001715extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1716 struct sk_buff *skb);
1717
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718static __inline__ void
1719sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1720{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001721 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00001722 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001723
Patrick Ohly20d49472009-02-12 05:03:38 +00001724 /*
1725 * generate control messages if
1726 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1727 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1728 * - software time stamp available and wanted
1729 * (SOCK_TIMESTAMPING_SOFTWARE)
1730 * - hardware time stamps available and wanted
1731 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1732 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1733 */
1734 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1735 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1736 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1737 (hwtstamps->hwtstamp.tv64 &&
1738 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1739 (hwtstamps->syststamp.tv64 &&
1740 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001741 __sock_recv_timestamp(msg, sk, skb);
1742 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001743 sk->sk_stamp = kt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744}
1745
Eric Dumazet767dd032010-04-28 19:14:43 +00001746extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1747 struct sk_buff *skb);
1748
1749static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1750 struct sk_buff *skb)
1751{
1752#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
1753 (1UL << SOCK_RCVTSTAMP) | \
1754 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
1755 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
1756 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
1757 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
1758
1759 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
1760 __sock_recv_ts_and_drops(msg, sk, skb);
1761 else
1762 sk->sk_stamp = skb->tstamp;
1763}
Neil Horman3b885782009-10-12 13:26:31 -07001764
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765/**
Patrick Ohly20d49472009-02-12 05:03:38 +00001766 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00001767 * @sk: socket sending this packet
Oliver Hartkopp2244d072010-08-17 08:59:14 +00001768 * @tx_flags: filled with instructions for time stamping
Patrick Ohly20d49472009-02-12 05:03:38 +00001769 *
1770 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1771 * parameters are invalid.
1772 */
Oliver Hartkopp2244d072010-08-17 08:59:14 +00001773extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
Patrick Ohly20d49472009-02-12 05:03:38 +00001774
1775/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001777 * @sk: socket to eat this skb from
1778 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07001779 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 *
1781 * This routine must be called with interrupts disabled or with the socket
1782 * locked so that the sk_buff queue operation is ok.
1783*/
Chris Leech624d1162006-05-23 18:01:28 -07001784#ifdef CONFIG_NET_DMA
1785static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1786{
1787 __skb_unlink(skb, &sk->sk_receive_queue);
1788 if (!copied_early)
1789 __kfree_skb(skb);
1790 else
1791 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1792}
1793#else
1794static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795{
1796 __skb_unlink(skb, &sk->sk_receive_queue);
1797 __kfree_skb(skb);
1798}
Chris Leech624d1162006-05-23 18:01:28 -07001799#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001801static inline
1802struct net *sock_net(const struct sock *sk)
1803{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001804 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001805}
1806
1807static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07001808void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001809{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001810 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001811}
1812
Denis V. Lunevedf02082008-02-29 11:18:32 -08001813/*
1814 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03001815 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08001816 * to stop it.
1817 * Sockets after sk_change_net should be released using sk_release_kernel
1818 */
1819static inline void sk_change_net(struct sock *sk, struct net *net)
1820{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001821 put_net(sock_net(sk));
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07001822 sock_net_set(sk, hold_net(net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08001823}
1824
KOVACS Krisztian23542612008-10-07 12:41:01 -07001825static inline struct sock *skb_steal_sock(struct sk_buff *skb)
1826{
1827 if (unlikely(skb->sk)) {
1828 struct sock *sk = skb->sk;
1829
1830 skb->destructor = NULL;
1831 skb->sk = NULL;
1832 return sk;
1833 }
1834 return NULL;
1835}
1836
Patrick Ohly20d49472009-02-12 05:03:38 +00001837extern void sock_enable_timestamp(struct sock *sk, int flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838extern int sock_get_timestamp(struct sock *, struct timeval __user *);
Eric Dumazetae40eb12007-03-18 17:33:16 -07001839extern int sock_get_timestampns(struct sock *, struct timespec __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840
1841/*
1842 * Enable debug/info messages
1843 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08001844extern int net_msg_warn;
1845#define NETDEBUG(fmt, args...) \
1846 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08001848#define LIMIT_NETDEBUG(fmt, args...) \
1849 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851extern __u32 sysctl_wmem_max;
1852extern __u32 sysctl_rmem_max;
1853
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001854extern void sk_init(void);
1855
David S. Miller6baf1f42005-09-05 18:14:11 -07001856extern int sysctl_optmem_max;
1857
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001858extern __u32 sysctl_wmem_default;
1859extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001860
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861#endif /* _SOCK_H */