blob: fd98760876513ffb4020a37a1ec43f1704085c80 [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>
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +080050#include <linux/pcounter.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070051#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>
54
55#include <linux/filter.h>
56
57#include <asm/atomic.h>
58#include <net/dst.h>
59#include <net/checksum.h>
60
61/*
62 * This structure really needs to be cleaned up.
63 * Most of it is for TCP, and not used by any of
64 * the other protocols.
65 */
66
67/* Define this to get the SOCK_DBG debugging facility. */
68#define SOCK_DEBUGGING
69#ifdef SOCK_DEBUGGING
70#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
71 printk(KERN_DEBUG msg); } while (0)
72#else
73#define SOCK_DEBUG(sk, msg...) do { } while (0)
74#endif
75
76/* This is the per-socket lock. The spinlock provides a synchronization
77 * between user contexts and software interrupt processing, whereas the
78 * mini-semaphore synchronizes multiple users amongst themselves.
79 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070080typedef struct {
81 spinlock_t slock;
John Heffnerd2e91172007-09-12 10:44:19 +020082 int owned;
Linus Torvalds1da177e2005-04-16 15:20:36 -070083 wait_queue_head_t wq;
Ingo Molnara5b5bb92006-07-03 00:25:35 -070084 /*
85 * We express the mutex-alike socket_lock semantics
86 * to the lock validator by explicitly managing
87 * the slock as a lock variant (in addition to
88 * the slock itself):
89 */
90#ifdef CONFIG_DEBUG_LOCK_ALLOC
91 struct lockdep_map dep_map;
92#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070093} socket_lock_t;
94
Linus Torvalds1da177e2005-04-16 15:20:36 -070095struct sock;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -070096struct proto;
Denis V. Lunev0eeb8ff2007-12-04 01:15:45 -080097struct net;
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
99/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700100 * struct sock_common - minimal network layer representation of sockets
101 * @skc_family: network address family
102 * @skc_state: Connection state
103 * @skc_reuse: %SO_REUSEADDR setting
104 * @skc_bound_dev_if: bound device index if != 0
105 * @skc_node: main hash linkage for various protocol lookup tables
106 * @skc_bind_node: bind hash linkage for various protocol lookup tables
107 * @skc_refcnt: reference count
Eric Dumazet81c3d542005-10-03 14:13:38 -0700108 * @skc_hash: hash value used with various protocol lookup tables
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700109 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200110 * @skc_net: reference to the network namespace of this socket
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700111 *
112 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700113 * for struct sock and struct inet_timewait_sock.
114 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115struct sock_common {
116 unsigned short skc_family;
117 volatile unsigned char skc_state;
118 unsigned char skc_reuse;
119 int skc_bound_dev_if;
120 struct hlist_node skc_node;
121 struct hlist_node skc_bind_node;
122 atomic_t skc_refcnt;
Eric Dumazet81c3d542005-10-03 14:13:38 -0700123 unsigned int skc_hash;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700124 struct proto *skc_prot;
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200125 struct net *skc_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126};
127
128/**
129 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700130 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700131 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
132 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
133 * @sk_lock: synchronizer
134 * @sk_rcvbuf: size of receive buffer in bytes
135 * @sk_sleep: sock wait queue
136 * @sk_dst_cache: destination cache
137 * @sk_dst_lock: destination cache lock
138 * @sk_policy: flow policy
139 * @sk_rmem_alloc: receive queue bytes committed
140 * @sk_receive_queue: incoming packets
141 * @sk_wmem_alloc: transmit queue bytes committed
142 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700143 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700144 * @sk_omem_alloc: "o" is "option" or "other"
145 * @sk_wmem_queued: persistent queue size
146 * @sk_forward_alloc: space allocated forward
147 * @sk_allocation: allocation mode
148 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800149 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
150 * %SO_OOBINLINE settings
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700151 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
152 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Herbert Xubcd76112006-06-30 13:36:35 -0700153 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700154 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700155 * @sk_backlog: always used with the per-socket spinlock held
156 * @sk_callback_lock: used with the callbacks in the end of this struct
157 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800158 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
159 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700160 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800161 * @sk_err_soft: errors that don't cause failure but are the cause of a
162 * persistent failure not just 'timed out'
163 * @sk_drops: raw drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700164 * @sk_ack_backlog: current listen backlog
165 * @sk_max_ack_backlog: listen backlog set in listen()
166 * @sk_priority: %SO_PRIORITY setting
167 * @sk_type: socket type (%SOCK_STREAM, etc)
168 * @sk_protocol: which protocol this socket belongs in this network family
169 * @sk_peercred: %SO_PEERCRED setting
170 * @sk_rcvlowat: %SO_RCVLOWAT setting
171 * @sk_rcvtimeo: %SO_RCVTIMEO setting
172 * @sk_sndtimeo: %SO_SNDTIMEO setting
173 * @sk_filter: socket filtering instructions
174 * @sk_protinfo: private area, net family specific, when not using slab
175 * @sk_timer: sock cleanup timer
176 * @sk_stamp: time stamp of last packet received
177 * @sk_socket: Identd and reporting IO signals
178 * @sk_user_data: RPC layer private data
179 * @sk_sndmsg_page: cached page for sendmsg
180 * @sk_sndmsg_off: cached offset for sendmsg
181 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700182 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800183 * @sk_mark: generic packet mark
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700184 * @sk_write_pending: a write to stream socket waits to start
185 * @sk_state_change: callback to indicate change in the state of the sock
186 * @sk_data_ready: callback to indicate there is data to be processed
187 * @sk_write_space: callback to indicate there is bf sending space available
188 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
189 * @sk_backlog_rcv: callback to process the backlog
190 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 */
192struct sock {
193 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700194 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 * don't add nothing before this first member (__sk_common) --acme
196 */
197 struct sock_common __sk_common;
198#define sk_family __sk_common.skc_family
199#define sk_state __sk_common.skc_state
200#define sk_reuse __sk_common.skc_reuse
201#define sk_bound_dev_if __sk_common.skc_bound_dev_if
202#define sk_node __sk_common.skc_node
203#define sk_bind_node __sk_common.skc_bind_node
204#define sk_refcnt __sk_common.skc_refcnt
Eric Dumazet81c3d542005-10-03 14:13:38 -0700205#define sk_hash __sk_common.skc_hash
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700206#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200207#define sk_net __sk_common.skc_net
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 unsigned char sk_shutdown : 2,
209 sk_no_check : 2,
210 sk_userlocks : 4;
211 unsigned char sk_protocol;
212 unsigned short sk_type;
213 int sk_rcvbuf;
214 socket_lock_t sk_lock;
Eric Dumazetfa438cc2007-03-04 16:05:44 -0800215 /*
216 * The backlog queue is special, it is always used with
217 * the per-socket spinlock held and requires low latency
218 * access. Therefore we special case it's implementation.
219 */
220 struct {
221 struct sk_buff *head;
222 struct sk_buff *tail;
223 } sk_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 wait_queue_head_t *sk_sleep;
225 struct dst_entry *sk_dst_cache;
226 struct xfrm_policy *sk_policy[2];
227 rwlock_t sk_dst_lock;
228 atomic_t sk_rmem_alloc;
229 atomic_t sk_wmem_alloc;
230 atomic_t sk_omem_alloc;
Arnaldo Carvalho de Melo4e07a912007-05-29 13:17:47 -0700231 int sk_sndbuf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 struct sk_buff_head sk_receive_queue;
233 struct sk_buff_head sk_write_queue;
Chris Leech97fc2f02006-05-23 17:55:33 -0700234 struct sk_buff_head sk_async_wait_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 int sk_wmem_queued;
236 int sk_forward_alloc;
Al Viro7d877f32005-10-21 03:20:43 -0400237 gfp_t sk_allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 int sk_route_caps;
Herbert Xubcd76112006-06-30 13:36:35 -0700239 int sk_gso_type;
David S. Miller9932cf92006-03-24 15:12:37 -0800240 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 unsigned long sk_flags;
242 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700244 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 rwlock_t sk_callback_lock;
246 int sk_err,
247 sk_err_soft;
Wang Chen33c732c2007-11-13 20:30:01 -0800248 atomic_t sk_drops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 unsigned short sk_ack_backlog;
250 unsigned short sk_max_ack_backlog;
251 __u32 sk_priority;
252 struct ucred sk_peercred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 long sk_rcvtimeo;
254 long sk_sndtimeo;
255 struct sk_filter *sk_filter;
256 void *sk_protinfo;
257 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700258 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 struct socket *sk_socket;
260 void *sk_user_data;
261 struct page *sk_sndmsg_page;
262 struct sk_buff *sk_send_head;
263 __u32 sk_sndmsg_off;
264 int sk_write_pending;
265 void *sk_security;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800266 __u32 sk_mark;
267 /* XXX 4 bytes hole on 64 bit */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 void (*sk_state_change)(struct sock *sk);
269 void (*sk_data_ready)(struct sock *sk, int bytes);
270 void (*sk_write_space)(struct sock *sk);
271 void (*sk_error_report)(struct sock *sk);
272 int (*sk_backlog_rcv)(struct sock *sk,
273 struct sk_buff *skb);
274 void (*sk_destruct)(struct sock *sk);
275};
276
277/*
278 * Hashed lists helper routines
279 */
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700280static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281{
282 return hlist_entry(head->first, struct sock, sk_node);
283}
284
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700285static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286{
287 return hlist_empty(head) ? NULL : __sk_head(head);
288}
289
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700290static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291{
292 return sk->sk_node.next ?
293 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
294}
295
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700296static inline int sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297{
298 return hlist_unhashed(&sk->sk_node);
299}
300
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700301static inline int sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302{
Akinobu Mitada753be2006-04-28 15:21:23 -0700303 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304}
305
306static __inline__ void sk_node_init(struct hlist_node *node)
307{
308 node->pprev = NULL;
309}
310
311static __inline__ void __sk_del_node(struct sock *sk)
312{
313 __hlist_del(&sk->sk_node);
314}
315
316static __inline__ int __sk_del_node_init(struct sock *sk)
317{
318 if (sk_hashed(sk)) {
319 __sk_del_node(sk);
320 sk_node_init(&sk->sk_node);
321 return 1;
322 }
323 return 0;
324}
325
326/* Grab socket reference count. This operation is valid only
327 when sk is ALREADY grabbed f.e. it is found in hash table
328 or a list and the lookup is made under lock preventing hash table
329 modifications.
330 */
331
332static inline void sock_hold(struct sock *sk)
333{
334 atomic_inc(&sk->sk_refcnt);
335}
336
337/* Ungrab socket in the context, which assumes that socket refcnt
338 cannot hit zero, f.e. it is true in context of any socketcall.
339 */
340static inline void __sock_put(struct sock *sk)
341{
342 atomic_dec(&sk->sk_refcnt);
343}
344
345static __inline__ int sk_del_node_init(struct sock *sk)
346{
347 int rc = __sk_del_node_init(sk);
348
349 if (rc) {
350 /* paranoid for a while -acme */
351 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
352 __sock_put(sk);
353 }
354 return rc;
355}
356
357static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
358{
359 hlist_add_head(&sk->sk_node, list);
360}
361
362static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
363{
364 sock_hold(sk);
365 __sk_add_node(sk, list);
366}
367
368static __inline__ void __sk_del_bind_node(struct sock *sk)
369{
370 __hlist_del(&sk->sk_bind_node);
371}
372
373static __inline__ void sk_add_bind_node(struct sock *sk,
374 struct hlist_head *list)
375{
376 hlist_add_head(&sk->sk_bind_node, list);
377}
378
379#define sk_for_each(__sk, node, list) \
380 hlist_for_each_entry(__sk, node, list, sk_node)
381#define sk_for_each_from(__sk, node) \
382 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
383 hlist_for_each_entry_from(__sk, node, sk_node)
384#define sk_for_each_continue(__sk, node) \
385 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
386 hlist_for_each_entry_continue(__sk, node, sk_node)
387#define sk_for_each_safe(__sk, node, tmp, list) \
388 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
389#define sk_for_each_bound(__sk, node, list) \
390 hlist_for_each_entry(__sk, node, list, sk_bind_node)
391
392/* Sock flags */
393enum sock_flags {
394 SOCK_DEAD,
395 SOCK_DONE,
396 SOCK_URGINLINE,
397 SOCK_KEEPOPEN,
398 SOCK_LINGER,
399 SOCK_DESTROY,
400 SOCK_BROADCAST,
401 SOCK_TIMESTAMP,
402 SOCK_ZAPPED,
403 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
404 SOCK_DBG, /* %SO_DEBUG setting */
405 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700406 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
408 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
409};
410
Ralf Baechle53b924b2005-08-23 10:11:30 -0700411static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
412{
413 nsk->sk_flags = osk->sk_flags;
414}
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
417{
418 __set_bit(flag, &sk->sk_flags);
419}
420
421static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
422{
423 __clear_bit(flag, &sk->sk_flags);
424}
425
426static inline int sock_flag(struct sock *sk, enum sock_flags flag)
427{
428 return test_bit(flag, &sk->sk_flags);
429}
430
431static inline void sk_acceptq_removed(struct sock *sk)
432{
433 sk->sk_ack_backlog--;
434}
435
436static inline void sk_acceptq_added(struct sock *sk)
437{
438 sk->sk_ack_backlog++;
439}
440
441static inline int sk_acceptq_is_full(struct sock *sk)
442{
David S. Miller64a14652007-03-06 11:21:05 -0800443 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444}
445
446/*
447 * Compute minimal free write space needed to queue new packets.
448 */
449static inline int sk_stream_min_wspace(struct sock *sk)
450{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800451 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452}
453
454static inline int sk_stream_wspace(struct sock *sk)
455{
456 return sk->sk_sndbuf - sk->sk_wmem_queued;
457}
458
459extern void sk_stream_write_space(struct sock *sk);
460
461static inline int sk_stream_memory_free(struct sock *sk)
462{
463 return sk->sk_wmem_queued < sk->sk_sndbuf;
464}
465
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466/* The per-socket spinlock must be held here. */
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800467static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
468{
469 if (!sk->sk_backlog.tail) {
470 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
471 } else {
472 sk->sk_backlog.tail->next = skb;
473 sk->sk_backlog.tail = skb;
474 }
475 skb->next = NULL;
476}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700478#define sk_wait_event(__sk, __timeo, __condition) \
479 ({ int __rc; \
480 release_sock(__sk); \
481 __rc = __condition; \
482 if (!__rc) { \
483 *(__timeo) = schedule_timeout(*(__timeo)); \
484 } \
485 lock_sock(__sk); \
486 __rc = __condition; \
487 __rc; \
488 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489
490extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
491extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
492extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
493extern int sk_stream_error(struct sock *sk, int flags, int err);
494extern void sk_stream_kill_queues(struct sock *sk);
495
496extern int sk_wait_data(struct sock *sk, long *timeo);
497
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700498struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800499struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800500struct inet_hashinfo;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700501
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502/* Networking protocol blocks we attach to sockets.
503 * socket layer -> transport layer interface
504 * transport -> network interface is defined by struct inet_proto
505 */
506struct proto {
507 void (*close)(struct sock *sk,
508 long timeout);
509 int (*connect)(struct sock *sk,
510 struct sockaddr *uaddr,
511 int addr_len);
512 int (*disconnect)(struct sock *sk, int flags);
513
514 struct sock * (*accept) (struct sock *sk, int flags, int *err);
515
516 int (*ioctl)(struct sock *sk, int cmd,
517 unsigned long arg);
518 int (*init)(struct sock *sk);
519 int (*destroy)(struct sock *sk);
520 void (*shutdown)(struct sock *sk, int how);
521 int (*setsockopt)(struct sock *sk, int level,
522 int optname, char __user *optval,
523 int optlen);
524 int (*getsockopt)(struct sock *sk, int level,
525 int optname, char __user *optval,
526 int __user *option);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800527 int (*compat_setsockopt)(struct sock *sk,
528 int level,
529 int optname, char __user *optval,
530 int optlen);
531 int (*compat_getsockopt)(struct sock *sk,
532 int level,
533 int optname, char __user *optval,
534 int __user *option);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
536 struct msghdr *msg, size_t len);
537 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
538 struct msghdr *msg,
539 size_t len, int noblock, int flags,
540 int *addr_len);
541 int (*sendpage)(struct sock *sk, struct page *page,
542 int offset, size_t size, int flags);
543 int (*bind)(struct sock *sk,
544 struct sockaddr *uaddr, int addr_len);
545
546 int (*backlog_rcv) (struct sock *sk,
547 struct sk_buff *skb);
548
549 /* Keeping track of sk's, looking them up, and port selection methods. */
550 void (*hash)(struct sock *sk);
551 void (*unhash)(struct sock *sk);
552 int (*get_port)(struct sock *sk, unsigned short snum);
553
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800554 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800555#ifdef CONFIG_PROC_FS
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800556 struct pcounter inuse;
Eric Dumazet65f76512008-01-03 20:46:48 -0800557#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800558
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 /* Memory pressure */
560 void (*enter_memory_pressure)(void);
561 atomic_t *memory_allocated; /* Current allocated memory. */
562 atomic_t *sockets_allocated; /* Current number of sockets. */
563 /*
564 * Pressure flag: try to collapse.
565 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800566 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 * is strict, actions are advisory and have some latency.
568 */
569 int *memory_pressure;
570 int *sysctl_mem;
571 int *sysctl_wmem;
572 int *sysctl_rmem;
573 int max_header;
574
Christoph Lametere18b8902006-12-06 20:33:20 -0800575 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 unsigned int obj_size;
577
Arnaldo Carvalho de Melo0a5578c2005-08-09 20:11:41 -0700578 atomic_t *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700579
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700580 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800581 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700582
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800583 struct inet_hashinfo *hashinfo;
584
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 struct module *owner;
586
587 char name[32];
588
589 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700590#ifdef SOCK_REFCNT_DEBUG
591 atomic_t socks;
592#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593};
594
595extern int proto_register(struct proto *prot, int alloc_slab);
596extern void proto_unregister(struct proto *prot);
597
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700598#ifdef SOCK_REFCNT_DEBUG
599static inline void sk_refcnt_debug_inc(struct sock *sk)
600{
601 atomic_inc(&sk->sk_prot->socks);
602}
603
604static inline void sk_refcnt_debug_dec(struct sock *sk)
605{
606 atomic_dec(&sk->sk_prot->socks);
607 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
608 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
609}
610
611static inline void sk_refcnt_debug_release(const struct sock *sk)
612{
613 if (atomic_read(&sk->sk_refcnt) != 1)
614 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
615 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
616}
617#else /* SOCK_REFCNT_DEBUG */
618#define sk_refcnt_debug_inc(sk) do { } while (0)
619#define sk_refcnt_debug_dec(sk) do { } while (0)
620#define sk_refcnt_debug_release(sk) do { } while (0)
621#endif /* SOCK_REFCNT_DEBUG */
622
Eric Dumazet65f76512008-01-03 20:46:48 -0800623
624#ifdef CONFIG_PROC_FS
625# define DEFINE_PROTO_INUSE(NAME) DEFINE_PCOUNTER(NAME)
626# define REF_PROTO_INUSE(NAME) PCOUNTER_MEMBER_INITIALIZER(NAME, .inuse)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627/* Called with local bh disabled */
Eric Dumazet65f76512008-01-03 20:46:48 -0800628static inline void sock_prot_inuse_add(struct proto *prot, int inc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629{
Eric Dumazet65f76512008-01-03 20:46:48 -0800630 pcounter_add(&prot->inuse, inc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631}
Eric Dumazet65f76512008-01-03 20:46:48 -0800632static inline int sock_prot_inuse_init(struct proto *proto)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633{
Eric Dumazet65f76512008-01-03 20:46:48 -0800634 return pcounter_alloc(&proto->inuse);
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800635}
Eric Dumazet65f76512008-01-03 20:46:48 -0800636static inline int sock_prot_inuse_get(struct proto *proto)
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800637{
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800638 return pcounter_getval(&proto->inuse);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639}
Eric Dumazet65f76512008-01-03 20:46:48 -0800640static inline void sock_prot_inuse_free(struct proto *proto)
641{
642 pcounter_free(&proto->inuse);
643}
644#else
645# define DEFINE_PROTO_INUSE(NAME)
646# define REF_PROTO_INUSE(NAME)
647static void inline sock_prot_inuse_add(struct proto *prot, int inc)
648{
649}
650static int inline sock_prot_inuse_init(struct proto *proto)
651{
652 return 0;
653}
654static void inline sock_prot_inuse_free(struct proto *proto)
655{
656}
657#endif
658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -0700660/* With per-bucket locks this operation is not-atomic, so that
661 * this version is not worse.
662 */
663static inline void __sk_prot_rehash(struct sock *sk)
664{
665 sk->sk_prot->unhash(sk);
666 sk->sk_prot->hash(sk);
667}
668
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669/* About 10 seconds */
670#define SOCK_DESTROY_TIME (10*HZ)
671
672/* Sockets 0-1023 can't be bound to unless you are superuser */
673#define PROT_SOCK 1024
674
675#define SHUTDOWN_MASK 3
676#define RCV_SHUTDOWN 1
677#define SEND_SHUTDOWN 2
678
679#define SOCK_SNDBUF_LOCK 1
680#define SOCK_RCVBUF_LOCK 2
681#define SOCK_BINDADDR_LOCK 4
682#define SOCK_BINDPORT_LOCK 8
683
684/* sock_iocb: used to kick off async processing of socket ios */
685struct sock_iocb {
686 struct list_head list;
687
688 int flags;
689 int size;
690 struct socket *sock;
691 struct sock *sk;
692 struct scm_cookie *scm;
693 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 struct kiocb *kiocb;
695};
696
697static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
698{
699 return (struct sock_iocb *)iocb->private;
700}
701
702static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
703{
704 return si->kiocb;
705}
706
707struct socket_alloc {
708 struct socket socket;
709 struct inode vfs_inode;
710};
711
712static inline struct socket *SOCKET_I(struct inode *inode)
713{
714 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
715}
716
717static inline struct inode *SOCK_INODE(struct socket *socket)
718{
719 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
720}
721
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800722/*
723 * Functions for memory accounting
724 */
725extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
726extern void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800728#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
729#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
730#define SK_MEM_SEND 0
731#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800733static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800735 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736}
737
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800738static inline int sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800740 /* return true if protocol supports memory accounting */
741 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742}
743
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800744static inline int sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745{
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800746 if (!sk_has_account(sk))
747 return 1;
Herbert Xud80d99d2005-09-01 17:48:23 -0700748 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800749 __sk_mem_schedule(sk, size, SK_MEM_SEND);
750}
751
752static inline int sk_rmem_schedule(struct sock *sk, int size)
753{
754 if (!sk_has_account(sk))
755 return 1;
756 return size <= sk->sk_forward_alloc ||
757 __sk_mem_schedule(sk, size, SK_MEM_RECV);
758}
759
760static inline void sk_mem_reclaim(struct sock *sk)
761{
762 if (!sk_has_account(sk))
763 return;
764 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
765 __sk_mem_reclaim(sk);
766}
767
David S. Miller9993e7d2008-01-10 21:56:38 -0800768static inline void sk_mem_reclaim_partial(struct sock *sk)
769{
770 if (!sk_has_account(sk))
771 return;
772 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
773 __sk_mem_reclaim(sk);
774}
775
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800776static inline void sk_mem_charge(struct sock *sk, int size)
777{
778 if (!sk_has_account(sk))
779 return;
780 sk->sk_forward_alloc -= size;
781}
782
783static inline void sk_mem_uncharge(struct sock *sk, int size)
784{
785 if (!sk_has_account(sk))
786 return;
787 sk->sk_forward_alloc += size;
788}
789
790static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
791{
792 skb_truesize_check(skb);
793 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
794 sk->sk_wmem_queued -= skb->truesize;
795 sk_mem_uncharge(sk, skb->truesize);
796 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -0700797}
798
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799/* Used by processes to "lock" a socket state, so that
800 * interrupts and bottom half handlers won't change it
801 * from under us. It essentially blocks any incoming
802 * packets, so that we won't get any new data or any
803 * packets that change the state of the socket.
804 *
805 * While locked, BH processing will add new packets to
806 * the backlog queue. This queue is processed by the
807 * owner of the socket lock right before it is released.
808 *
809 * Since ~2.3.5 it is also exclusive sleep lock serializing
810 * accesses from user process context.
811 */
John Heffnerd2e91172007-09-12 10:44:19 +0200812#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813
Peter Zijlstraed075362006-12-06 20:35:24 -0800814/*
815 * Macro so as to not evaluate some arguments when
816 * lockdep is not enabled.
817 *
818 * Mark both the sk_lock and the sk_lock.slock as a
819 * per-address-family lock class.
820 */
821#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
822do { \
John Heffnerd2e91172007-09-12 10:44:19 +0200823 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -0800824 init_waitqueue_head(&sk->sk_lock.wq); \
825 spin_lock_init(&(sk)->sk_lock.slock); \
826 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
827 sizeof((sk)->sk_lock)); \
828 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
829 (skey), (sname)); \
830 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
831} while (0)
832
Harvey Harrison41380932007-12-12 10:46:51 -0800833extern void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -0800834
835static inline void lock_sock(struct sock *sk)
836{
837 lock_sock_nested(sk, 0);
838}
839
Harvey Harrison41380932007-12-12 10:46:51 -0800840extern void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
842/* BH context may only use the following locking interface. */
843#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -0700844#define bh_lock_sock_nested(__sk) \
845 spin_lock_nested(&((__sk)->sk_lock.slock), \
846 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
848
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700849extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +0100850 gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700851 struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852extern void sk_free(struct sock *sk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -0700853extern struct sock *sk_clone(const struct sock *sk,
Al Virodd0fc662005-10-07 07:46:04 +0100854 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855
856extern struct sk_buff *sock_wmalloc(struct sock *sk,
857 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +0100858 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859extern struct sk_buff *sock_rmalloc(struct sock *sk,
860 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +0100861 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862extern void sock_wfree(struct sk_buff *skb);
863extern void sock_rfree(struct sk_buff *skb);
864
865extern int sock_setsockopt(struct socket *sock, int level,
866 int op, char __user *optval,
867 int optlen);
868
869extern int sock_getsockopt(struct socket *sock, int level,
870 int op, char __user *optval,
871 int __user *optlen);
872extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
873 unsigned long size,
874 int noblock,
875 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700876extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +0100877 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878extern void sock_kfree_s(struct sock *sk, void *mem, int size);
879extern void sk_send_sigurg(struct sock *sk);
880
881/*
882 * Functions to fill in entries in struct proto_ops when a protocol
883 * does not implement a particular function.
884 */
885extern int sock_no_bind(struct socket *,
886 struct sockaddr *, int);
887extern int sock_no_connect(struct socket *,
888 struct sockaddr *, int, int);
889extern int sock_no_socketpair(struct socket *,
890 struct socket *);
891extern int sock_no_accept(struct socket *,
892 struct socket *, int);
893extern int sock_no_getname(struct socket *,
894 struct sockaddr *, int *, int);
895extern unsigned int sock_no_poll(struct file *, struct socket *,
896 struct poll_table_struct *);
897extern int sock_no_ioctl(struct socket *, unsigned int,
898 unsigned long);
899extern int sock_no_listen(struct socket *, int);
900extern int sock_no_shutdown(struct socket *, int);
901extern int sock_no_getsockopt(struct socket *, int , int,
902 char __user *, int __user *);
903extern int sock_no_setsockopt(struct socket *, int, int,
904 char __user *, int);
905extern int sock_no_sendmsg(struct kiocb *, struct socket *,
906 struct msghdr *, size_t);
907extern int sock_no_recvmsg(struct kiocb *, struct socket *,
908 struct msghdr *, size_t, int);
909extern int sock_no_mmap(struct file *file,
910 struct socket *sock,
911 struct vm_area_struct *vma);
912extern ssize_t sock_no_sendpage(struct socket *sock,
913 struct page *page,
914 int offset, size_t size,
915 int flags);
916
917/*
918 * Functions to fill in entries in struct proto_ops when a protocol
919 * uses the inet style.
920 */
921extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
922 char __user *optval, int __user *optlen);
923extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
924 struct msghdr *msg, size_t size, int flags);
925extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
926 char __user *optval, int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800927extern int compat_sock_common_getsockopt(struct socket *sock, int level,
928 int optname, char __user *optval, int __user *optlen);
929extern int compat_sock_common_setsockopt(struct socket *sock, int level,
930 int optname, char __user *optval, int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931
932extern void sk_common_release(struct sock *sk);
933
934/*
935 * Default socket callbacks and setup code
936 */
937
938/* Initialise core socket variables */
939extern void sock_init_data(struct socket *sock, struct sock *sk);
940
941/**
942 * sk_filter - run a packet through a socket filter
943 * @sk: sock associated with &sk_buff
944 * @skb: buffer to filter
945 * @needlock: set to 1 if the sock is not locked by caller.
946 *
947 * Run the filter code and then cut skb->data to correct size returned by
948 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
949 * than pkt_len we keep whole skb->data. This is the socket level
950 * wrapper to sk_run_filter. It returns 0 if the packet should
951 * be accepted or -EPERM if the packet should be tossed.
952 *
953 */
954
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700955static inline int sk_filter(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956{
957 int err;
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700958 struct sk_filter *filter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959
960 err = security_sock_rcv_skb(sk, skb);
961 if (err)
962 return err;
963
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700964 rcu_read_lock_bh();
Eric Dumazet9d3e4442008-01-08 23:41:28 -0800965 filter = rcu_dereference(sk->sk_filter);
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700966 if (filter) {
967 unsigned int pkt_len = sk_run_filter(skb, filter->insns,
968 filter->len);
969 err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 }
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700971 rcu_read_unlock_bh();
972
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 return err;
974}
975
976/**
Paul Bonserdc9b3342006-11-23 17:56:13 -0800977 * sk_filter_release: Release a socket filter
978 * @sk: socket
979 * @fp: filter to remove
980 *
981 * Remove a filter from a socket and release its resources.
982 */
983
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -0700984static inline void sk_filter_release(struct sk_filter *fp)
985{
986 if (atomic_dec_and_test(&fp->refcnt))
Pavel Emelyanov47e958e2007-10-17 21:22:42 -0700987 kfree(fp);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -0700988}
989
990static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991{
992 unsigned int size = sk_filter_len(fp);
993
994 atomic_sub(size, &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -0700995 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996}
997
998static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
999{
1000 atomic_inc(&fp->refcnt);
1001 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1002}
1003
1004/*
1005 * Socket reference counting postulates.
1006 *
1007 * * Each user of socket SHOULD hold a reference count.
1008 * * Each access point to socket (an hash table bucket, reference from a list,
1009 * running timer, skb in flight MUST hold a reference count.
1010 * * When reference count hits 0, it means it will never increase back.
1011 * * When reference count hits 0, it means that no references from
1012 * outside exist to this socket and current process on current CPU
1013 * is last user and may/should destroy this socket.
1014 * * sk_free is called from any context: process, BH, IRQ. When
1015 * it is called, socket has no references from outside -> sk_free
1016 * may release descendant resources allocated by the socket, but
1017 * to the time when it is called, socket is NOT referenced by any
1018 * hash tables, lists etc.
1019 * * Packets, delivered from outside (from network or from another process)
1020 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1021 * when they sit in queue. Otherwise, packets will leak to hole, when
1022 * socket is looked up by one cpu and unhasing is made by another CPU.
1023 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1024 * (leak to backlog). Packet socket does all the processing inside
1025 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1026 * use separate SMP lock, so that they are prone too.
1027 */
1028
1029/* Ungrab socket and destroy it, if it was the last reference. */
1030static inline void sock_put(struct sock *sk)
1031{
1032 if (atomic_dec_and_test(&sk->sk_refcnt))
1033 sk_free(sk);
1034}
1035
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -02001036extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1037 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001038
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039/* Detach socket from process context.
1040 * Announce socket dead, detach it from wait queue and inode.
1041 * Note that parent inode held reference count on this struct sock,
1042 * we do not release it in this function, because protocol
1043 * probably wants some additional cleanups or even continuing
1044 * to work with this socket (TCP).
1045 */
1046static inline void sock_orphan(struct sock *sk)
1047{
1048 write_lock_bh(&sk->sk_callback_lock);
1049 sock_set_flag(sk, SOCK_DEAD);
1050 sk->sk_socket = NULL;
1051 sk->sk_sleep = NULL;
1052 write_unlock_bh(&sk->sk_callback_lock);
1053}
1054
1055static inline void sock_graft(struct sock *sk, struct socket *parent)
1056{
1057 write_lock_bh(&sk->sk_callback_lock);
1058 sk->sk_sleep = &parent->wait;
1059 parent->sk = sk;
1060 sk->sk_socket = parent;
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001061 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 write_unlock_bh(&sk->sk_callback_lock);
1063}
1064
1065extern int sock_i_uid(struct sock *sk);
1066extern unsigned long sock_i_ino(struct sock *sk);
1067
1068static inline struct dst_entry *
1069__sk_dst_get(struct sock *sk)
1070{
1071 return sk->sk_dst_cache;
1072}
1073
1074static inline struct dst_entry *
1075sk_dst_get(struct sock *sk)
1076{
1077 struct dst_entry *dst;
1078
1079 read_lock(&sk->sk_dst_lock);
1080 dst = sk->sk_dst_cache;
1081 if (dst)
1082 dst_hold(dst);
1083 read_unlock(&sk->sk_dst_lock);
1084 return dst;
1085}
1086
1087static inline void
1088__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1089{
1090 struct dst_entry *old_dst;
1091
1092 old_dst = sk->sk_dst_cache;
1093 sk->sk_dst_cache = dst;
1094 dst_release(old_dst);
1095}
1096
1097static inline void
1098sk_dst_set(struct sock *sk, struct dst_entry *dst)
1099{
1100 write_lock(&sk->sk_dst_lock);
1101 __sk_dst_set(sk, dst);
1102 write_unlock(&sk->sk_dst_lock);
1103}
1104
1105static inline void
1106__sk_dst_reset(struct sock *sk)
1107{
1108 struct dst_entry *old_dst;
1109
1110 old_dst = sk->sk_dst_cache;
1111 sk->sk_dst_cache = NULL;
1112 dst_release(old_dst);
1113}
1114
1115static inline void
1116sk_dst_reset(struct sock *sk)
1117{
1118 write_lock(&sk->sk_dst_lock);
1119 __sk_dst_reset(sk);
1120 write_unlock(&sk->sk_dst_lock);
1121}
1122
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001123extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001125extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126
Herbert Xubcd76112006-06-30 13:36:35 -07001127static inline int sk_can_gso(const struct sock *sk)
1128{
1129 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1130}
1131
Andi Kleen99580892007-04-20 17:12:43 -07001132extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001133
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1135 struct sk_buff *skb, struct page *page,
1136 int off, int copy)
1137{
1138 if (skb->ip_summed == CHECKSUM_NONE) {
1139 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001140 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 page_address(page) + off,
1142 copy, 0, &err);
1143 if (err)
1144 return err;
1145 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1146 } else if (copy_from_user(page_address(page) + off, from, copy))
1147 return -EFAULT;
1148
1149 skb->len += copy;
1150 skb->data_len += copy;
1151 skb->truesize += copy;
1152 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001153 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 return 0;
1155}
1156
1157/*
1158 * Queue a received datagram if it will fit. Stream and sequenced
1159 * protocols can't normally use this as they need to fit buffers in
1160 * and play with them.
1161 *
1162 * Inlined as it's very short and called for pretty much every
1163 * packet ever received.
1164 */
1165
1166static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1167{
1168 sock_hold(sk);
1169 skb->sk = sk;
1170 skb->destructor = sock_wfree;
1171 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1172}
1173
1174static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1175{
1176 skb->sk = sk;
1177 skb->destructor = sock_rfree;
1178 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001179 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180}
1181
1182extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1183 unsigned long expires);
1184
1185extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1186
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001187extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188
1189static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1190{
1191 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1192 number of warnings when compiling with -W --ANK
1193 */
1194 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1195 (unsigned)sk->sk_rcvbuf)
1196 return -ENOMEM;
1197 skb_set_owner_r(skb, sk);
1198 skb_queue_tail(&sk->sk_error_queue, skb);
1199 if (!sock_flag(sk, SOCK_DEAD))
1200 sk->sk_data_ready(sk, skb->len);
1201 return 0;
1202}
1203
1204/*
1205 * Recover an error report and clear atomically
1206 */
1207
1208static inline int sock_error(struct sock *sk)
1209{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001210 int err;
1211 if (likely(!sk->sk_err))
1212 return 0;
1213 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 return -err;
1215}
1216
1217static inline unsigned long sock_wspace(struct sock *sk)
1218{
1219 int amt = 0;
1220
1221 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1222 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1223 if (amt < 0)
1224 amt = 0;
1225 }
1226 return amt;
1227}
1228
1229static inline void sk_wake_async(struct sock *sk, int how, int band)
1230{
1231 if (sk->sk_socket && sk->sk_socket->fasync_list)
1232 sock_wake_async(sk->sk_socket, how, band);
1233}
1234
1235#define SOCK_MIN_SNDBUF 2048
1236#define SOCK_MIN_RCVBUF 256
1237
1238static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1239{
1240 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001241 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1243 }
1244}
1245
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11001246struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247
1248static inline struct page *sk_stream_alloc_page(struct sock *sk)
1249{
1250 struct page *page = NULL;
1251
Herbert Xuef015782005-09-01 17:48:59 -07001252 page = alloc_pages(sk->sk_allocation, 0);
1253 if (!page) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 sk->sk_prot->enter_memory_pressure();
1255 sk_stream_moderate_sndbuf(sk);
1256 }
1257 return page;
1258}
1259
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260/*
1261 * Default write policy as shown to user space via poll/select/SIGIO
1262 */
1263static inline int sock_writeable(const struct sock *sk)
1264{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001265 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266}
1267
Al Virodd0fc662005-10-07 07:46:04 +01001268static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269{
Patrick McHardy44981212007-02-27 09:56:42 -08001270 return in_atomic() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271}
1272
1273static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1274{
1275 return noblock ? 0 : sk->sk_rcvtimeo;
1276}
1277
1278static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1279{
1280 return noblock ? 0 : sk->sk_sndtimeo;
1281}
1282
1283static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1284{
1285 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1286}
1287
1288/* Alas, with timeout socket operations are not restartable.
1289 * Compare this to poll().
1290 */
1291static inline int sock_intr_errno(long timeo)
1292{
1293 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1294}
1295
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001296extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1297 struct sk_buff *skb);
1298
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299static __inline__ void
1300sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1301{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001302 ktime_t kt = skb->tstamp;
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001303
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001304 if (sock_flag(sk, SOCK_RCVTSTAMP))
1305 __sock_recv_timestamp(msg, sk, skb);
1306 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001307 sk->sk_stamp = kt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308}
1309
1310/**
1311 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001312 * @sk: socket to eat this skb from
1313 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07001314 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 *
1316 * This routine must be called with interrupts disabled or with the socket
1317 * locked so that the sk_buff queue operation is ok.
1318*/
Chris Leech624d1162006-05-23 18:01:28 -07001319#ifdef CONFIG_NET_DMA
1320static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1321{
1322 __skb_unlink(skb, &sk->sk_receive_queue);
1323 if (!copied_early)
1324 __kfree_skb(skb);
1325 else
1326 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1327}
1328#else
1329static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330{
1331 __skb_unlink(skb, &sk->sk_receive_queue);
1332 __kfree_skb(skb);
1333}
Chris Leech624d1162006-05-23 18:01:28 -07001334#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335
1336extern void sock_enable_timestamp(struct sock *sk);
1337extern int sock_get_timestamp(struct sock *, struct timeval __user *);
Eric Dumazetae40eb12007-03-18 17:33:16 -07001338extern int sock_get_timestampns(struct sock *, struct timespec __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339
1340/*
1341 * Enable debug/info messages
1342 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08001343extern int net_msg_warn;
1344#define NETDEBUG(fmt, args...) \
1345 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08001347#define LIMIT_NETDEBUG(fmt, args...) \
1348 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349
1350/*
1351 * Macros for sleeping on a socket. Use them like this:
1352 *
1353 * SOCK_SLEEP_PRE(sk)
1354 * if (condition)
1355 * schedule();
1356 * SOCK_SLEEP_POST(sk)
1357 *
1358 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1359 * and when the last use of them in DECnet has gone, I'm intending to
1360 * remove them.
1361 */
1362
1363#define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1364 DECLARE_WAITQUEUE(wait, tsk); \
1365 tsk->state = TASK_INTERRUPTIBLE; \
1366 add_wait_queue((sk)->sk_sleep, &wait); \
1367 release_sock(sk);
1368
1369#define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1370 remove_wait_queue((sk)->sk_sleep, &wait); \
1371 lock_sock(sk); \
1372 }
1373
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374extern __u32 sysctl_wmem_max;
1375extern __u32 sysctl_rmem_max;
1376
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001377extern void sk_init(void);
1378
David S. Miller6baf1f42005-09-05 18:14:11 -07001379extern int sysctl_optmem_max;
1380
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001381extern __u32 sysctl_wmem_default;
1382extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001383
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384#endif /* _SOCK_H */