blob: dc42b44c2aa14b409fc9307afb6e7458773c14aa [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the AF_INET socket handler.
7 *
8 * Version: @(#)sock.h 1.0.4 05/13/93
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
14 *
15 * Fixes:
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
19 * than the reverse.
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
33 *
34 *
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
39 */
40#ifndef _SOCK_H
41#define _SOCK_H
42
Ilpo Järvinen172589c2007-08-28 15:50:33 -070043#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/list.h>
45#include <linux/timer.h>
46#include <linux/cache.h>
47#include <linux/module.h>
Ingo Molnara5b5bb92006-07-03 00:25:35 -070048#include <linux/lockdep.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#include <linux/netdevice.h>
50#include <linux/skbuff.h> /* struct sk_buff */
Al Virod7fe0f22006-12-03 23:15:30 -050051#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070052#include <linux/security.h>
53
54#include <linux/filter.h>
55
56#include <asm/atomic.h>
57#include <net/dst.h>
58#include <net/checksum.h>
59
60/*
61 * This structure really needs to be cleaned up.
62 * Most of it is for TCP, and not used by any of
63 * the other protocols.
64 */
65
66/* Define this to get the SOCK_DBG debugging facility. */
67#define SOCK_DEBUGGING
68#ifdef SOCK_DEBUGGING
69#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
70 printk(KERN_DEBUG msg); } while (0)
71#else
Stephen Hemminger4cd90292008-03-21 15:54:53 -070072/* Validate arguments and do nothing */
73static void inline int __attribute__ ((format (printf, 2, 3)))
74SOCK_DEBUG(struct sock *sk, const char *msg, ...)
75{
76}
Linus Torvalds1da177e2005-04-16 15:20:36 -070077#endif
78
79/* This is the per-socket lock. The spinlock provides a synchronization
80 * between user contexts and software interrupt processing, whereas the
81 * mini-semaphore synchronizes multiple users amongst themselves.
82 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070083typedef struct {
84 spinlock_t slock;
John Heffnerd2e91172007-09-12 10:44:19 +020085 int owned;
Linus Torvalds1da177e2005-04-16 15:20:36 -070086 wait_queue_head_t wq;
Ingo Molnara5b5bb92006-07-03 00:25:35 -070087 /*
88 * We express the mutex-alike socket_lock semantics
89 * to the lock validator by explicitly managing
90 * the slock as a lock variant (in addition to
91 * the slock itself):
92 */
93#ifdef CONFIG_DEBUG_LOCK_ALLOC
94 struct lockdep_map dep_map;
95#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070096} socket_lock_t;
97
Linus Torvalds1da177e2005-04-16 15:20:36 -070098struct sock;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -070099struct proto;
Denis V. Lunev0eeb8ff2007-12-04 01:15:45 -0800100struct net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101
102/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700103 * struct sock_common - minimal network layer representation of sockets
104 * @skc_family: network address family
105 * @skc_state: Connection state
106 * @skc_reuse: %SO_REUSEADDR setting
107 * @skc_bound_dev_if: bound device index if != 0
108 * @skc_node: main hash linkage for various protocol lookup tables
109 * @skc_bind_node: bind hash linkage for various protocol lookup tables
110 * @skc_refcnt: reference count
Eric Dumazet81c3d542005-10-03 14:13:38 -0700111 * @skc_hash: hash value used with various protocol lookup tables
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700112 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200113 * @skc_net: reference to the network namespace of this socket
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700114 *
115 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700116 * for struct sock and struct inet_timewait_sock.
117 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118struct sock_common {
119 unsigned short skc_family;
120 volatile unsigned char skc_state;
121 unsigned char skc_reuse;
122 int skc_bound_dev_if;
123 struct hlist_node skc_node;
124 struct hlist_node skc_bind_node;
125 atomic_t skc_refcnt;
Eric Dumazet81c3d542005-10-03 14:13:38 -0700126 unsigned int skc_hash;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700127 struct proto *skc_prot;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900128#ifdef CONFIG_NET_NS
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200129 struct net *skc_net;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900130#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131};
132
133/**
134 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700135 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700136 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
137 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
138 * @sk_lock: synchronizer
139 * @sk_rcvbuf: size of receive buffer in bytes
140 * @sk_sleep: sock wait queue
141 * @sk_dst_cache: destination cache
142 * @sk_dst_lock: destination cache lock
143 * @sk_policy: flow policy
144 * @sk_rmem_alloc: receive queue bytes committed
145 * @sk_receive_queue: incoming packets
146 * @sk_wmem_alloc: transmit queue bytes committed
147 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700148 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700149 * @sk_omem_alloc: "o" is "option" or "other"
150 * @sk_wmem_queued: persistent queue size
151 * @sk_forward_alloc: space allocated forward
152 * @sk_allocation: allocation mode
153 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800154 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
155 * %SO_OOBINLINE settings
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700156 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
157 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Herbert Xubcd76112006-06-30 13:36:35 -0700158 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700159 * @sk_gso_max_size: Maximum GSO segment size to build
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700160 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700161 * @sk_backlog: always used with the per-socket spinlock held
162 * @sk_callback_lock: used with the callbacks in the end of this struct
163 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800164 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
165 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700166 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800167 * @sk_err_soft: errors that don't cause failure but are the cause of a
168 * persistent failure not just 'timed out'
169 * @sk_drops: raw drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700170 * @sk_ack_backlog: current listen backlog
171 * @sk_max_ack_backlog: listen backlog set in listen()
172 * @sk_priority: %SO_PRIORITY setting
173 * @sk_type: socket type (%SOCK_STREAM, etc)
174 * @sk_protocol: which protocol this socket belongs in this network family
175 * @sk_peercred: %SO_PEERCRED setting
176 * @sk_rcvlowat: %SO_RCVLOWAT setting
177 * @sk_rcvtimeo: %SO_RCVTIMEO setting
178 * @sk_sndtimeo: %SO_SNDTIMEO setting
179 * @sk_filter: socket filtering instructions
180 * @sk_protinfo: private area, net family specific, when not using slab
181 * @sk_timer: sock cleanup timer
182 * @sk_stamp: time stamp of last packet received
183 * @sk_socket: Identd and reporting IO signals
184 * @sk_user_data: RPC layer private data
185 * @sk_sndmsg_page: cached page for sendmsg
186 * @sk_sndmsg_off: cached offset for sendmsg
187 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700188 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800189 * @sk_mark: generic packet mark
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700190 * @sk_write_pending: a write to stream socket waits to start
191 * @sk_state_change: callback to indicate change in the state of the sock
192 * @sk_data_ready: callback to indicate there is data to be processed
193 * @sk_write_space: callback to indicate there is bf sending space available
194 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
195 * @sk_backlog_rcv: callback to process the backlog
196 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 */
198struct sock {
199 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700200 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 * don't add nothing before this first member (__sk_common) --acme
202 */
203 struct sock_common __sk_common;
204#define sk_family __sk_common.skc_family
205#define sk_state __sk_common.skc_state
206#define sk_reuse __sk_common.skc_reuse
207#define sk_bound_dev_if __sk_common.skc_bound_dev_if
208#define sk_node __sk_common.skc_node
209#define sk_bind_node __sk_common.skc_bind_node
210#define sk_refcnt __sk_common.skc_refcnt
Eric Dumazet81c3d542005-10-03 14:13:38 -0700211#define sk_hash __sk_common.skc_hash
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700212#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200213#define sk_net __sk_common.skc_net
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 unsigned char sk_shutdown : 2,
215 sk_no_check : 2,
216 sk_userlocks : 4;
217 unsigned char sk_protocol;
218 unsigned short sk_type;
219 int sk_rcvbuf;
220 socket_lock_t sk_lock;
Eric Dumazetfa438cc2007-03-04 16:05:44 -0800221 /*
222 * The backlog queue is special, it is always used with
223 * the per-socket spinlock held and requires low latency
224 * access. Therefore we special case it's implementation.
225 */
226 struct {
227 struct sk_buff *head;
228 struct sk_buff *tail;
229 } sk_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 wait_queue_head_t *sk_sleep;
231 struct dst_entry *sk_dst_cache;
232 struct xfrm_policy *sk_policy[2];
233 rwlock_t sk_dst_lock;
234 atomic_t sk_rmem_alloc;
235 atomic_t sk_wmem_alloc;
236 atomic_t sk_omem_alloc;
Arnaldo Carvalho de Melo4e07a912007-05-29 13:17:47 -0700237 int sk_sndbuf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 struct sk_buff_head sk_receive_queue;
239 struct sk_buff_head sk_write_queue;
Chris Leech97fc2f02006-05-23 17:55:33 -0700240 struct sk_buff_head sk_async_wait_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 int sk_wmem_queued;
242 int sk_forward_alloc;
Al Viro7d877f32005-10-21 03:20:43 -0400243 gfp_t sk_allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 int sk_route_caps;
Herbert Xubcd76112006-06-30 13:36:35 -0700245 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700246 unsigned int sk_gso_max_size;
David S. Miller9932cf92006-03-24 15:12:37 -0800247 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 unsigned long sk_flags;
249 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700251 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 rwlock_t sk_callback_lock;
253 int sk_err,
254 sk_err_soft;
Wang Chen33c732c2007-11-13 20:30:01 -0800255 atomic_t sk_drops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 unsigned short sk_ack_backlog;
257 unsigned short sk_max_ack_backlog;
258 __u32 sk_priority;
259 struct ucred sk_peercred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 long sk_rcvtimeo;
261 long sk_sndtimeo;
262 struct sk_filter *sk_filter;
263 void *sk_protinfo;
264 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700265 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 struct socket *sk_socket;
267 void *sk_user_data;
268 struct page *sk_sndmsg_page;
269 struct sk_buff *sk_send_head;
270 __u32 sk_sndmsg_off;
271 int sk_write_pending;
272 void *sk_security;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800273 __u32 sk_mark;
274 /* XXX 4 bytes hole on 64 bit */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 void (*sk_state_change)(struct sock *sk);
276 void (*sk_data_ready)(struct sock *sk, int bytes);
277 void (*sk_write_space)(struct sock *sk);
278 void (*sk_error_report)(struct sock *sk);
279 int (*sk_backlog_rcv)(struct sock *sk,
280 struct sk_buff *skb);
281 void (*sk_destruct)(struct sock *sk);
282};
283
284/*
285 * Hashed lists helper routines
286 */
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700287static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288{
289 return hlist_entry(head->first, struct sock, sk_node);
290}
291
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700292static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293{
294 return hlist_empty(head) ? NULL : __sk_head(head);
295}
296
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700297static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298{
299 return sk->sk_node.next ?
300 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
301}
302
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700303static inline int sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304{
305 return hlist_unhashed(&sk->sk_node);
306}
307
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700308static inline int sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309{
Akinobu Mitada753be2006-04-28 15:21:23 -0700310 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311}
312
313static __inline__ void sk_node_init(struct hlist_node *node)
314{
315 node->pprev = NULL;
316}
317
318static __inline__ void __sk_del_node(struct sock *sk)
319{
320 __hlist_del(&sk->sk_node);
321}
322
323static __inline__ int __sk_del_node_init(struct sock *sk)
324{
325 if (sk_hashed(sk)) {
326 __sk_del_node(sk);
327 sk_node_init(&sk->sk_node);
328 return 1;
329 }
330 return 0;
331}
332
333/* Grab socket reference count. This operation is valid only
334 when sk is ALREADY grabbed f.e. it is found in hash table
335 or a list and the lookup is made under lock preventing hash table
336 modifications.
337 */
338
339static inline void sock_hold(struct sock *sk)
340{
341 atomic_inc(&sk->sk_refcnt);
342}
343
344/* Ungrab socket in the context, which assumes that socket refcnt
345 cannot hit zero, f.e. it is true in context of any socketcall.
346 */
347static inline void __sock_put(struct sock *sk)
348{
349 atomic_dec(&sk->sk_refcnt);
350}
351
352static __inline__ int sk_del_node_init(struct sock *sk)
353{
354 int rc = __sk_del_node_init(sk);
355
356 if (rc) {
357 /* paranoid for a while -acme */
358 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
359 __sock_put(sk);
360 }
361 return rc;
362}
363
364static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
365{
366 hlist_add_head(&sk->sk_node, list);
367}
368
369static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
370{
371 sock_hold(sk);
372 __sk_add_node(sk, list);
373}
374
375static __inline__ void __sk_del_bind_node(struct sock *sk)
376{
377 __hlist_del(&sk->sk_bind_node);
378}
379
380static __inline__ void sk_add_bind_node(struct sock *sk,
381 struct hlist_head *list)
382{
383 hlist_add_head(&sk->sk_bind_node, list);
384}
385
386#define sk_for_each(__sk, node, list) \
387 hlist_for_each_entry(__sk, node, list, sk_node)
388#define sk_for_each_from(__sk, node) \
389 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
390 hlist_for_each_entry_from(__sk, node, sk_node)
391#define sk_for_each_continue(__sk, node) \
392 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
393 hlist_for_each_entry_continue(__sk, node, sk_node)
394#define sk_for_each_safe(__sk, node, tmp, list) \
395 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
396#define sk_for_each_bound(__sk, node, list) \
397 hlist_for_each_entry(__sk, node, list, sk_bind_node)
398
399/* Sock flags */
400enum sock_flags {
401 SOCK_DEAD,
402 SOCK_DONE,
403 SOCK_URGINLINE,
404 SOCK_KEEPOPEN,
405 SOCK_LINGER,
406 SOCK_DESTROY,
407 SOCK_BROADCAST,
408 SOCK_TIMESTAMP,
409 SOCK_ZAPPED,
410 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
411 SOCK_DBG, /* %SO_DEBUG setting */
412 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700413 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
415 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
416};
417
Ralf Baechle53b924b2005-08-23 10:11:30 -0700418static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
419{
420 nsk->sk_flags = osk->sk_flags;
421}
422
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
424{
425 __set_bit(flag, &sk->sk_flags);
426}
427
428static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
429{
430 __clear_bit(flag, &sk->sk_flags);
431}
432
433static inline int sock_flag(struct sock *sk, enum sock_flags flag)
434{
435 return test_bit(flag, &sk->sk_flags);
436}
437
438static inline void sk_acceptq_removed(struct sock *sk)
439{
440 sk->sk_ack_backlog--;
441}
442
443static inline void sk_acceptq_added(struct sock *sk)
444{
445 sk->sk_ack_backlog++;
446}
447
448static inline int sk_acceptq_is_full(struct sock *sk)
449{
David S. Miller64a14652007-03-06 11:21:05 -0800450 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451}
452
453/*
454 * Compute minimal free write space needed to queue new packets.
455 */
456static inline int sk_stream_min_wspace(struct sock *sk)
457{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800458 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459}
460
461static inline int sk_stream_wspace(struct sock *sk)
462{
463 return sk->sk_sndbuf - sk->sk_wmem_queued;
464}
465
466extern void sk_stream_write_space(struct sock *sk);
467
468static inline int sk_stream_memory_free(struct sock *sk)
469{
470 return sk->sk_wmem_queued < sk->sk_sndbuf;
471}
472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473/* The per-socket spinlock must be held here. */
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800474static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
475{
476 if (!sk->sk_backlog.tail) {
477 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
478 } else {
479 sk->sk_backlog.tail->next = skb;
480 sk->sk_backlog.tail = skb;
481 }
482 skb->next = NULL;
483}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700485#define sk_wait_event(__sk, __timeo, __condition) \
486 ({ int __rc; \
487 release_sock(__sk); \
488 __rc = __condition; \
489 if (!__rc) { \
490 *(__timeo) = schedule_timeout(*(__timeo)); \
491 } \
492 lock_sock(__sk); \
493 __rc = __condition; \
494 __rc; \
495 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496
497extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
498extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
499extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
500extern int sk_stream_error(struct sock *sk, int flags, int err);
501extern void sk_stream_kill_queues(struct sock *sk);
502
503extern int sk_wait_data(struct sock *sk, long *timeo);
504
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700505struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800506struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800507struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700508struct raw_hashinfo;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700509
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510/* Networking protocol blocks we attach to sockets.
511 * socket layer -> transport layer interface
512 * transport -> network interface is defined by struct inet_proto
513 */
514struct proto {
515 void (*close)(struct sock *sk,
516 long timeout);
517 int (*connect)(struct sock *sk,
518 struct sockaddr *uaddr,
519 int addr_len);
520 int (*disconnect)(struct sock *sk, int flags);
521
522 struct sock * (*accept) (struct sock *sk, int flags, int *err);
523
524 int (*ioctl)(struct sock *sk, int cmd,
525 unsigned long arg);
526 int (*init)(struct sock *sk);
527 int (*destroy)(struct sock *sk);
528 void (*shutdown)(struct sock *sk, int how);
529 int (*setsockopt)(struct sock *sk, int level,
530 int optname, char __user *optval,
531 int optlen);
532 int (*getsockopt)(struct sock *sk, int level,
533 int optname, char __user *optval,
534 int __user *option);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800535 int (*compat_setsockopt)(struct sock *sk,
536 int level,
537 int optname, char __user *optval,
538 int optlen);
539 int (*compat_getsockopt)(struct sock *sk,
540 int level,
541 int optname, char __user *optval,
542 int __user *option);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
544 struct msghdr *msg, size_t len);
545 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
546 struct msghdr *msg,
547 size_t len, int noblock, int flags,
548 int *addr_len);
549 int (*sendpage)(struct sock *sk, struct page *page,
550 int offset, size_t size, int flags);
551 int (*bind)(struct sock *sk,
552 struct sockaddr *uaddr, int addr_len);
553
554 int (*backlog_rcv) (struct sock *sk,
555 struct sk_buff *skb);
556
557 /* Keeping track of sk's, looking them up, and port selection methods. */
558 void (*hash)(struct sock *sk);
559 void (*unhash)(struct sock *sk);
560 int (*get_port)(struct sock *sk, unsigned short snum);
561
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800562 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800563#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700564 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800565#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800566
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 /* Memory pressure */
568 void (*enter_memory_pressure)(void);
569 atomic_t *memory_allocated; /* Current allocated memory. */
570 atomic_t *sockets_allocated; /* Current number of sockets. */
571 /*
572 * Pressure flag: try to collapse.
573 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800574 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 * is strict, actions are advisory and have some latency.
576 */
577 int *memory_pressure;
578 int *sysctl_mem;
579 int *sysctl_wmem;
580 int *sysctl_rmem;
581 int max_header;
582
Christoph Lametere18b8902006-12-06 20:33:20 -0800583 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 unsigned int obj_size;
585
Arnaldo Carvalho de Melo0a5578c2005-08-09 20:11:41 -0700586 atomic_t *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700587
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700588 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800589 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700590
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700591 union {
592 struct inet_hashinfo *hashinfo;
593 struct hlist_head *udp_hash;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700594 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700595 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800596
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 struct module *owner;
598
599 char name[32];
600
601 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700602#ifdef SOCK_REFCNT_DEBUG
603 atomic_t socks;
604#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605};
606
607extern int proto_register(struct proto *prot, int alloc_slab);
608extern void proto_unregister(struct proto *prot);
609
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700610#ifdef SOCK_REFCNT_DEBUG
611static inline void sk_refcnt_debug_inc(struct sock *sk)
612{
613 atomic_inc(&sk->sk_prot->socks);
614}
615
616static inline void sk_refcnt_debug_dec(struct sock *sk)
617{
618 atomic_dec(&sk->sk_prot->socks);
619 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
620 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
621}
622
623static inline void sk_refcnt_debug_release(const struct sock *sk)
624{
625 if (atomic_read(&sk->sk_refcnt) != 1)
626 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
627 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
628}
629#else /* SOCK_REFCNT_DEBUG */
630#define sk_refcnt_debug_inc(sk) do { } while (0)
631#define sk_refcnt_debug_dec(sk) do { } while (0)
632#define sk_refcnt_debug_release(sk) do { } while (0)
633#endif /* SOCK_REFCNT_DEBUG */
634
Eric Dumazet65f76512008-01-03 20:46:48 -0800635
636#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637/* Called with local bh disabled */
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700638extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
639extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -0800640#else
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700641static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
642 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -0800643{
644}
Eric Dumazet65f76512008-01-03 20:46:48 -0800645#endif
646
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -0700648/* With per-bucket locks this operation is not-atomic, so that
649 * this version is not worse.
650 */
651static inline void __sk_prot_rehash(struct sock *sk)
652{
653 sk->sk_prot->unhash(sk);
654 sk->sk_prot->hash(sk);
655}
656
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657/* About 10 seconds */
658#define SOCK_DESTROY_TIME (10*HZ)
659
660/* Sockets 0-1023 can't be bound to unless you are superuser */
661#define PROT_SOCK 1024
662
663#define SHUTDOWN_MASK 3
664#define RCV_SHUTDOWN 1
665#define SEND_SHUTDOWN 2
666
667#define SOCK_SNDBUF_LOCK 1
668#define SOCK_RCVBUF_LOCK 2
669#define SOCK_BINDADDR_LOCK 4
670#define SOCK_BINDPORT_LOCK 8
671
672/* sock_iocb: used to kick off async processing of socket ios */
673struct sock_iocb {
674 struct list_head list;
675
676 int flags;
677 int size;
678 struct socket *sock;
679 struct sock *sk;
680 struct scm_cookie *scm;
681 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 struct kiocb *kiocb;
683};
684
685static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
686{
687 return (struct sock_iocb *)iocb->private;
688}
689
690static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
691{
692 return si->kiocb;
693}
694
695struct socket_alloc {
696 struct socket socket;
697 struct inode vfs_inode;
698};
699
700static inline struct socket *SOCKET_I(struct inode *inode)
701{
702 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
703}
704
705static inline struct inode *SOCK_INODE(struct socket *socket)
706{
707 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
708}
709
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800710/*
711 * Functions for memory accounting
712 */
713extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
714extern void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800716#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
717#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
718#define SK_MEM_SEND 0
719#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800721static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800723 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724}
725
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800726static inline int sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800728 /* return true if protocol supports memory accounting */
729 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730}
731
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800732static inline int sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800734 if (!sk_has_account(sk))
735 return 1;
Herbert Xud80d99d2005-09-01 17:48:23 -0700736 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800737 __sk_mem_schedule(sk, size, SK_MEM_SEND);
738}
739
740static inline int sk_rmem_schedule(struct sock *sk, int size)
741{
742 if (!sk_has_account(sk))
743 return 1;
744 return size <= sk->sk_forward_alloc ||
745 __sk_mem_schedule(sk, size, SK_MEM_RECV);
746}
747
748static inline void sk_mem_reclaim(struct sock *sk)
749{
750 if (!sk_has_account(sk))
751 return;
752 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
753 __sk_mem_reclaim(sk);
754}
755
David S. Miller9993e7d2008-01-10 21:56:38 -0800756static inline void sk_mem_reclaim_partial(struct sock *sk)
757{
758 if (!sk_has_account(sk))
759 return;
760 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
761 __sk_mem_reclaim(sk);
762}
763
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800764static inline void sk_mem_charge(struct sock *sk, int size)
765{
766 if (!sk_has_account(sk))
767 return;
768 sk->sk_forward_alloc -= size;
769}
770
771static inline void sk_mem_uncharge(struct sock *sk, int size)
772{
773 if (!sk_has_account(sk))
774 return;
775 sk->sk_forward_alloc += size;
776}
777
778static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
779{
780 skb_truesize_check(skb);
781 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
782 sk->sk_wmem_queued -= skb->truesize;
783 sk_mem_uncharge(sk, skb->truesize);
784 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -0700785}
786
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787/* Used by processes to "lock" a socket state, so that
788 * interrupts and bottom half handlers won't change it
789 * from under us. It essentially blocks any incoming
790 * packets, so that we won't get any new data or any
791 * packets that change the state of the socket.
792 *
793 * While locked, BH processing will add new packets to
794 * the backlog queue. This queue is processed by the
795 * owner of the socket lock right before it is released.
796 *
797 * Since ~2.3.5 it is also exclusive sleep lock serializing
798 * accesses from user process context.
799 */
John Heffnerd2e91172007-09-12 10:44:19 +0200800#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801
Peter Zijlstraed075362006-12-06 20:35:24 -0800802/*
803 * Macro so as to not evaluate some arguments when
804 * lockdep is not enabled.
805 *
806 * Mark both the sk_lock and the sk_lock.slock as a
807 * per-address-family lock class.
808 */
809#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
810do { \
John Heffnerd2e91172007-09-12 10:44:19 +0200811 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -0800812 init_waitqueue_head(&sk->sk_lock.wq); \
813 spin_lock_init(&(sk)->sk_lock.slock); \
814 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
815 sizeof((sk)->sk_lock)); \
816 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
817 (skey), (sname)); \
818 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
819} while (0)
820
Harvey Harrison41380932007-12-12 10:46:51 -0800821extern void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -0800822
823static inline void lock_sock(struct sock *sk)
824{
825 lock_sock_nested(sk, 0);
826}
827
Harvey Harrison41380932007-12-12 10:46:51 -0800828extern void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829
830/* BH context may only use the following locking interface. */
831#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -0700832#define bh_lock_sock_nested(__sk) \
833 spin_lock_nested(&((__sk)->sk_lock.slock), \
834 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
836
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700837extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +0100838 gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700839 struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840extern void sk_free(struct sock *sk);
Denis V. Lunevedf02082008-02-29 11:18:32 -0800841extern void sk_release_kernel(struct sock *sk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -0700842extern struct sock *sk_clone(const struct sock *sk,
Al Virodd0fc662005-10-07 07:46:04 +0100843 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844
845extern struct sk_buff *sock_wmalloc(struct sock *sk,
846 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +0100847 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848extern struct sk_buff *sock_rmalloc(struct sock *sk,
849 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +0100850 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851extern void sock_wfree(struct sk_buff *skb);
852extern void sock_rfree(struct sk_buff *skb);
853
854extern int sock_setsockopt(struct socket *sock, int level,
855 int op, char __user *optval,
856 int optlen);
857
858extern int sock_getsockopt(struct socket *sock, int level,
859 int op, char __user *optval,
860 int __user *optlen);
861extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
862 unsigned long size,
863 int noblock,
864 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700865extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +0100866 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867extern void sock_kfree_s(struct sock *sk, void *mem, int size);
868extern void sk_send_sigurg(struct sock *sk);
869
870/*
871 * Functions to fill in entries in struct proto_ops when a protocol
872 * does not implement a particular function.
873 */
874extern int sock_no_bind(struct socket *,
875 struct sockaddr *, int);
876extern int sock_no_connect(struct socket *,
877 struct sockaddr *, int, int);
878extern int sock_no_socketpair(struct socket *,
879 struct socket *);
880extern int sock_no_accept(struct socket *,
881 struct socket *, int);
882extern int sock_no_getname(struct socket *,
883 struct sockaddr *, int *, int);
884extern unsigned int sock_no_poll(struct file *, struct socket *,
885 struct poll_table_struct *);
886extern int sock_no_ioctl(struct socket *, unsigned int,
887 unsigned long);
888extern int sock_no_listen(struct socket *, int);
889extern int sock_no_shutdown(struct socket *, int);
890extern int sock_no_getsockopt(struct socket *, int , int,
891 char __user *, int __user *);
892extern int sock_no_setsockopt(struct socket *, int, int,
893 char __user *, int);
894extern int sock_no_sendmsg(struct kiocb *, struct socket *,
895 struct msghdr *, size_t);
896extern int sock_no_recvmsg(struct kiocb *, struct socket *,
897 struct msghdr *, size_t, int);
898extern int sock_no_mmap(struct file *file,
899 struct socket *sock,
900 struct vm_area_struct *vma);
901extern ssize_t sock_no_sendpage(struct socket *sock,
902 struct page *page,
903 int offset, size_t size,
904 int flags);
905
906/*
907 * Functions to fill in entries in struct proto_ops when a protocol
908 * uses the inet style.
909 */
910extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
911 char __user *optval, int __user *optlen);
912extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
913 struct msghdr *msg, size_t size, int flags);
914extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
915 char __user *optval, int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800916extern int compat_sock_common_getsockopt(struct socket *sock, int level,
917 int optname, char __user *optval, int __user *optlen);
918extern int compat_sock_common_setsockopt(struct socket *sock, int level,
919 int optname, char __user *optval, int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920
921extern void sk_common_release(struct sock *sk);
922
923/*
924 * Default socket callbacks and setup code
925 */
926
927/* Initialise core socket variables */
928extern void sock_init_data(struct socket *sock, struct sock *sk);
929
930/**
Paul Bonserdc9b3342006-11-23 17:56:13 -0800931 * sk_filter_release: Release a socket filter
932 * @sk: socket
933 * @fp: filter to remove
934 *
935 * Remove a filter from a socket and release its resources.
936 */
937
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -0700938static inline void sk_filter_release(struct sk_filter *fp)
939{
940 if (atomic_dec_and_test(&fp->refcnt))
Pavel Emelyanov47e958e2007-10-17 21:22:42 -0700941 kfree(fp);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -0700942}
943
944static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945{
946 unsigned int size = sk_filter_len(fp);
947
948 atomic_sub(size, &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -0700949 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950}
951
952static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
953{
954 atomic_inc(&fp->refcnt);
955 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
956}
957
958/*
959 * Socket reference counting postulates.
960 *
961 * * Each user of socket SHOULD hold a reference count.
962 * * Each access point to socket (an hash table bucket, reference from a list,
963 * running timer, skb in flight MUST hold a reference count.
964 * * When reference count hits 0, it means it will never increase back.
965 * * When reference count hits 0, it means that no references from
966 * outside exist to this socket and current process on current CPU
967 * is last user and may/should destroy this socket.
968 * * sk_free is called from any context: process, BH, IRQ. When
969 * it is called, socket has no references from outside -> sk_free
970 * may release descendant resources allocated by the socket, but
971 * to the time when it is called, socket is NOT referenced by any
972 * hash tables, lists etc.
973 * * Packets, delivered from outside (from network or from another process)
974 * and enqueued on receive/error queues SHOULD NOT grab reference count,
975 * when they sit in queue. Otherwise, packets will leak to hole, when
976 * socket is looked up by one cpu and unhasing is made by another CPU.
977 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
978 * (leak to backlog). Packet socket does all the processing inside
979 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
980 * use separate SMP lock, so that they are prone too.
981 */
982
983/* Ungrab socket and destroy it, if it was the last reference. */
984static inline void sock_put(struct sock *sk)
985{
986 if (atomic_dec_and_test(&sk->sk_refcnt))
987 sk_free(sk);
988}
989
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -0200990extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
991 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -0200992
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993/* Detach socket from process context.
994 * Announce socket dead, detach it from wait queue and inode.
995 * Note that parent inode held reference count on this struct sock,
996 * we do not release it in this function, because protocol
997 * probably wants some additional cleanups or even continuing
998 * to work with this socket (TCP).
999 */
1000static inline void sock_orphan(struct sock *sk)
1001{
1002 write_lock_bh(&sk->sk_callback_lock);
1003 sock_set_flag(sk, SOCK_DEAD);
1004 sk->sk_socket = NULL;
1005 sk->sk_sleep = NULL;
1006 write_unlock_bh(&sk->sk_callback_lock);
1007}
1008
1009static inline void sock_graft(struct sock *sk, struct socket *parent)
1010{
1011 write_lock_bh(&sk->sk_callback_lock);
1012 sk->sk_sleep = &parent->wait;
1013 parent->sk = sk;
1014 sk->sk_socket = parent;
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001015 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 write_unlock_bh(&sk->sk_callback_lock);
1017}
1018
1019extern int sock_i_uid(struct sock *sk);
1020extern unsigned long sock_i_ino(struct sock *sk);
1021
1022static inline struct dst_entry *
1023__sk_dst_get(struct sock *sk)
1024{
1025 return sk->sk_dst_cache;
1026}
1027
1028static inline struct dst_entry *
1029sk_dst_get(struct sock *sk)
1030{
1031 struct dst_entry *dst;
1032
1033 read_lock(&sk->sk_dst_lock);
1034 dst = sk->sk_dst_cache;
1035 if (dst)
1036 dst_hold(dst);
1037 read_unlock(&sk->sk_dst_lock);
1038 return dst;
1039}
1040
1041static inline void
1042__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1043{
1044 struct dst_entry *old_dst;
1045
1046 old_dst = sk->sk_dst_cache;
1047 sk->sk_dst_cache = dst;
1048 dst_release(old_dst);
1049}
1050
1051static inline void
1052sk_dst_set(struct sock *sk, struct dst_entry *dst)
1053{
1054 write_lock(&sk->sk_dst_lock);
1055 __sk_dst_set(sk, dst);
1056 write_unlock(&sk->sk_dst_lock);
1057}
1058
1059static inline void
1060__sk_dst_reset(struct sock *sk)
1061{
1062 struct dst_entry *old_dst;
1063
1064 old_dst = sk->sk_dst_cache;
1065 sk->sk_dst_cache = NULL;
1066 dst_release(old_dst);
1067}
1068
1069static inline void
1070sk_dst_reset(struct sock *sk)
1071{
1072 write_lock(&sk->sk_dst_lock);
1073 __sk_dst_reset(sk);
1074 write_unlock(&sk->sk_dst_lock);
1075}
1076
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001077extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001079extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080
Herbert Xubcd76112006-06-30 13:36:35 -07001081static inline int sk_can_gso(const struct sock *sk)
1082{
1083 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1084}
1085
Andi Kleen99580892007-04-20 17:12:43 -07001086extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001087
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1089 struct sk_buff *skb, struct page *page,
1090 int off, int copy)
1091{
1092 if (skb->ip_summed == CHECKSUM_NONE) {
1093 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001094 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 page_address(page) + off,
1096 copy, 0, &err);
1097 if (err)
1098 return err;
1099 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1100 } else if (copy_from_user(page_address(page) + off, from, copy))
1101 return -EFAULT;
1102
1103 skb->len += copy;
1104 skb->data_len += copy;
1105 skb->truesize += copy;
1106 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001107 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 return 0;
1109}
1110
1111/*
1112 * Queue a received datagram if it will fit. Stream and sequenced
1113 * protocols can't normally use this as they need to fit buffers in
1114 * and play with them.
1115 *
1116 * Inlined as it's very short and called for pretty much every
1117 * packet ever received.
1118 */
1119
1120static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1121{
1122 sock_hold(sk);
1123 skb->sk = sk;
1124 skb->destructor = sock_wfree;
1125 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1126}
1127
1128static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1129{
1130 skb->sk = sk;
1131 skb->destructor = sock_rfree;
1132 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001133 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134}
1135
1136extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1137 unsigned long expires);
1138
1139extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1140
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001141extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142
1143static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1144{
1145 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1146 number of warnings when compiling with -W --ANK
1147 */
1148 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1149 (unsigned)sk->sk_rcvbuf)
1150 return -ENOMEM;
1151 skb_set_owner_r(skb, sk);
1152 skb_queue_tail(&sk->sk_error_queue, skb);
1153 if (!sock_flag(sk, SOCK_DEAD))
1154 sk->sk_data_ready(sk, skb->len);
1155 return 0;
1156}
1157
1158/*
1159 * Recover an error report and clear atomically
1160 */
1161
1162static inline int sock_error(struct sock *sk)
1163{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001164 int err;
1165 if (likely(!sk->sk_err))
1166 return 0;
1167 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 return -err;
1169}
1170
1171static inline unsigned long sock_wspace(struct sock *sk)
1172{
1173 int amt = 0;
1174
1175 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1176 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1177 if (amt < 0)
1178 amt = 0;
1179 }
1180 return amt;
1181}
1182
1183static inline void sk_wake_async(struct sock *sk, int how, int band)
1184{
1185 if (sk->sk_socket && sk->sk_socket->fasync_list)
1186 sock_wake_async(sk->sk_socket, how, band);
1187}
1188
1189#define SOCK_MIN_SNDBUF 2048
1190#define SOCK_MIN_RCVBUF 256
1191
1192static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1193{
1194 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001195 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1197 }
1198}
1199
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11001200struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201
1202static inline struct page *sk_stream_alloc_page(struct sock *sk)
1203{
1204 struct page *page = NULL;
1205
Herbert Xuef015782005-09-01 17:48:59 -07001206 page = alloc_pages(sk->sk_allocation, 0);
1207 if (!page) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208 sk->sk_prot->enter_memory_pressure();
1209 sk_stream_moderate_sndbuf(sk);
1210 }
1211 return page;
1212}
1213
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214/*
1215 * Default write policy as shown to user space via poll/select/SIGIO
1216 */
1217static inline int sock_writeable(const struct sock *sk)
1218{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001219 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220}
1221
Al Virodd0fc662005-10-07 07:46:04 +01001222static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223{
Patrick McHardy44981212007-02-27 09:56:42 -08001224 return in_atomic() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225}
1226
1227static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1228{
1229 return noblock ? 0 : sk->sk_rcvtimeo;
1230}
1231
1232static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1233{
1234 return noblock ? 0 : sk->sk_sndtimeo;
1235}
1236
1237static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1238{
1239 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1240}
1241
1242/* Alas, with timeout socket operations are not restartable.
1243 * Compare this to poll().
1244 */
1245static inline int sock_intr_errno(long timeo)
1246{
1247 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1248}
1249
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001250extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1251 struct sk_buff *skb);
1252
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253static __inline__ void
1254sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1255{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001256 ktime_t kt = skb->tstamp;
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001257
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001258 if (sock_flag(sk, SOCK_RCVTSTAMP))
1259 __sock_recv_timestamp(msg, sk, skb);
1260 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001261 sk->sk_stamp = kt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262}
1263
1264/**
1265 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001266 * @sk: socket to eat this skb from
1267 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07001268 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 *
1270 * This routine must be called with interrupts disabled or with the socket
1271 * locked so that the sk_buff queue operation is ok.
1272*/
Chris Leech624d1162006-05-23 18:01:28 -07001273#ifdef CONFIG_NET_DMA
1274static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1275{
1276 __skb_unlink(skb, &sk->sk_receive_queue);
1277 if (!copied_early)
1278 __kfree_skb(skb);
1279 else
1280 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1281}
1282#else
1283static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284{
1285 __skb_unlink(skb, &sk->sk_receive_queue);
1286 __kfree_skb(skb);
1287}
Chris Leech624d1162006-05-23 18:01:28 -07001288#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001290static inline
1291struct net *sock_net(const struct sock *sk)
1292{
1293#ifdef CONFIG_NET_NS
1294 return sk->sk_net;
1295#else
1296 return &init_net;
1297#endif
1298}
1299
1300static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07001301void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001302{
1303#ifdef CONFIG_NET_NS
1304 sk->sk_net = net;
1305#endif
1306}
1307
Denis V. Lunevedf02082008-02-29 11:18:32 -08001308/*
1309 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1310 * They should not hold a referrence to a namespace in order to allow
1311 * to stop it.
1312 * Sockets after sk_change_net should be released using sk_release_kernel
1313 */
1314static inline void sk_change_net(struct sock *sk, struct net *net)
1315{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001316 put_net(sock_net(sk));
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07001317 sock_net_set(sk, hold_net(net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08001318}
1319
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320extern void sock_enable_timestamp(struct sock *sk);
1321extern int sock_get_timestamp(struct sock *, struct timeval __user *);
Eric Dumazetae40eb12007-03-18 17:33:16 -07001322extern int sock_get_timestampns(struct sock *, struct timespec __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323
1324/*
1325 * Enable debug/info messages
1326 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08001327extern int net_msg_warn;
1328#define NETDEBUG(fmt, args...) \
1329 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08001331#define LIMIT_NETDEBUG(fmt, args...) \
1332 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333
1334/*
1335 * Macros for sleeping on a socket. Use them like this:
1336 *
1337 * SOCK_SLEEP_PRE(sk)
1338 * if (condition)
1339 * schedule();
1340 * SOCK_SLEEP_POST(sk)
1341 *
1342 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1343 * and when the last use of them in DECnet has gone, I'm intending to
1344 * remove them.
1345 */
1346
1347#define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1348 DECLARE_WAITQUEUE(wait, tsk); \
1349 tsk->state = TASK_INTERRUPTIBLE; \
1350 add_wait_queue((sk)->sk_sleep, &wait); \
1351 release_sock(sk);
1352
1353#define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1354 remove_wait_queue((sk)->sk_sleep, &wait); \
1355 lock_sock(sk); \
1356 }
1357
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358extern __u32 sysctl_wmem_max;
1359extern __u32 sysctl_rmem_max;
1360
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001361extern void sk_init(void);
1362
David S. Miller6baf1f42005-09-05 18:14:11 -07001363extern int sysctl_optmem_max;
1364
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001365extern __u32 sysctl_wmem_default;
1366extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001367
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368#endif /* _SOCK_H */