blob: 453c79d0915ba00ca8b33ff160ba64cb7c58a521 [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>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070059#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
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 Melo8feaf0c2005-08-09 20:09:30 -070096struct proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
98/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -070099 * struct sock_common - minimal network layer representation of sockets
100 * @skc_family: network address family
101 * @skc_state: Connection state
102 * @skc_reuse: %SO_REUSEADDR setting
103 * @skc_bound_dev_if: bound device index if != 0
104 * @skc_node: main hash linkage for various protocol lookup tables
105 * @skc_bind_node: bind hash linkage for various protocol lookup tables
106 * @skc_refcnt: reference count
Eric Dumazet81c3d542005-10-03 14:13:38 -0700107 * @skc_hash: hash value used with various protocol lookup tables
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700108 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200109 * @skc_net: reference to the network namespace of this socket
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700110 *
111 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700112 * for struct sock and struct inet_timewait_sock.
113 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114struct sock_common {
115 unsigned short skc_family;
116 volatile unsigned char skc_state;
117 unsigned char skc_reuse;
118 int skc_bound_dev_if;
119 struct hlist_node skc_node;
120 struct hlist_node skc_bind_node;
121 atomic_t skc_refcnt;
Eric Dumazet81c3d542005-10-03 14:13:38 -0700122 unsigned int skc_hash;
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700123 struct proto *skc_prot;
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200124 struct net *skc_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125};
126
127/**
128 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700129 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700130 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
131 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
132 * @sk_lock: synchronizer
133 * @sk_rcvbuf: size of receive buffer in bytes
134 * @sk_sleep: sock wait queue
135 * @sk_dst_cache: destination cache
136 * @sk_dst_lock: destination cache lock
137 * @sk_policy: flow policy
138 * @sk_rmem_alloc: receive queue bytes committed
139 * @sk_receive_queue: incoming packets
140 * @sk_wmem_alloc: transmit queue bytes committed
141 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700142 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700143 * @sk_omem_alloc: "o" is "option" or "other"
144 * @sk_wmem_queued: persistent queue size
145 * @sk_forward_alloc: space allocated forward
146 * @sk_allocation: allocation mode
147 * @sk_sndbuf: size of send buffer in bytes
148 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
149 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
150 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Herbert Xubcd76112006-06-30 13:36:35 -0700151 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700152 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700153 * @sk_backlog: always used with the per-socket spinlock held
154 * @sk_callback_lock: used with the callbacks in the end of this struct
155 * @sk_error_queue: rarely used
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700156 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700157 * @sk_err: last error
158 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
159 * @sk_ack_backlog: current listen backlog
160 * @sk_max_ack_backlog: listen backlog set in listen()
161 * @sk_priority: %SO_PRIORITY setting
162 * @sk_type: socket type (%SOCK_STREAM, etc)
163 * @sk_protocol: which protocol this socket belongs in this network family
164 * @sk_peercred: %SO_PEERCRED setting
165 * @sk_rcvlowat: %SO_RCVLOWAT setting
166 * @sk_rcvtimeo: %SO_RCVTIMEO setting
167 * @sk_sndtimeo: %SO_SNDTIMEO setting
168 * @sk_filter: socket filtering instructions
169 * @sk_protinfo: private area, net family specific, when not using slab
170 * @sk_timer: sock cleanup timer
171 * @sk_stamp: time stamp of last packet received
172 * @sk_socket: Identd and reporting IO signals
173 * @sk_user_data: RPC layer private data
174 * @sk_sndmsg_page: cached page for sendmsg
175 * @sk_sndmsg_off: cached offset for sendmsg
176 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700177 * @sk_security: used by security modules
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700178 * @sk_write_pending: a write to stream socket waits to start
179 * @sk_state_change: callback to indicate change in the state of the sock
180 * @sk_data_ready: callback to indicate there is data to be processed
181 * @sk_write_space: callback to indicate there is bf sending space available
182 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
183 * @sk_backlog_rcv: callback to process the backlog
184 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 */
186struct sock {
187 /*
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700188 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 * don't add nothing before this first member (__sk_common) --acme
190 */
191 struct sock_common __sk_common;
192#define sk_family __sk_common.skc_family
193#define sk_state __sk_common.skc_state
194#define sk_reuse __sk_common.skc_reuse
195#define sk_bound_dev_if __sk_common.skc_bound_dev_if
196#define sk_node __sk_common.skc_node
197#define sk_bind_node __sk_common.skc_bind_node
198#define sk_refcnt __sk_common.skc_refcnt
Eric Dumazet81c3d542005-10-03 14:13:38 -0700199#define sk_hash __sk_common.skc_hash
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700200#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200201#define sk_net __sk_common.skc_net
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 unsigned char sk_shutdown : 2,
203 sk_no_check : 2,
204 sk_userlocks : 4;
205 unsigned char sk_protocol;
206 unsigned short sk_type;
207 int sk_rcvbuf;
208 socket_lock_t sk_lock;
Eric Dumazetfa438cc2007-03-04 16:05:44 -0800209 /*
210 * The backlog queue is special, it is always used with
211 * the per-socket spinlock held and requires low latency
212 * access. Therefore we special case it's implementation.
213 */
214 struct {
215 struct sk_buff *head;
216 struct sk_buff *tail;
217 } sk_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 wait_queue_head_t *sk_sleep;
219 struct dst_entry *sk_dst_cache;
220 struct xfrm_policy *sk_policy[2];
221 rwlock_t sk_dst_lock;
222 atomic_t sk_rmem_alloc;
223 atomic_t sk_wmem_alloc;
224 atomic_t sk_omem_alloc;
Arnaldo Carvalho de Melo4e07a912007-05-29 13:17:47 -0700225 int sk_sndbuf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 struct sk_buff_head sk_receive_queue;
227 struct sk_buff_head sk_write_queue;
Chris Leech97fc2f02006-05-23 17:55:33 -0700228 struct sk_buff_head sk_async_wait_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 int sk_wmem_queued;
230 int sk_forward_alloc;
Al Viro7d877f32005-10-21 03:20:43 -0400231 gfp_t sk_allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 int sk_route_caps;
Herbert Xubcd76112006-06-30 13:36:35 -0700233 int sk_gso_type;
David S. Miller9932cf92006-03-24 15:12:37 -0800234 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 unsigned long sk_flags;
236 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700238 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 rwlock_t sk_callback_lock;
240 int sk_err,
241 sk_err_soft;
242 unsigned short sk_ack_backlog;
243 unsigned short sk_max_ack_backlog;
244 __u32 sk_priority;
245 struct ucred sk_peercred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 long sk_rcvtimeo;
247 long sk_sndtimeo;
248 struct sk_filter *sk_filter;
249 void *sk_protinfo;
250 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700251 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 struct socket *sk_socket;
253 void *sk_user_data;
254 struct page *sk_sndmsg_page;
255 struct sk_buff *sk_send_head;
256 __u32 sk_sndmsg_off;
257 int sk_write_pending;
258 void *sk_security;
259 void (*sk_state_change)(struct sock *sk);
260 void (*sk_data_ready)(struct sock *sk, int bytes);
261 void (*sk_write_space)(struct sock *sk);
262 void (*sk_error_report)(struct sock *sk);
263 int (*sk_backlog_rcv)(struct sock *sk,
264 struct sk_buff *skb);
265 void (*sk_destruct)(struct sock *sk);
266};
267
268/*
269 * Hashed lists helper routines
270 */
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700271static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272{
273 return hlist_entry(head->first, struct sock, sk_node);
274}
275
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700276static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277{
278 return hlist_empty(head) ? NULL : __sk_head(head);
279}
280
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700281static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282{
283 return sk->sk_node.next ?
284 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
285}
286
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700287static inline int sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288{
289 return hlist_unhashed(&sk->sk_node);
290}
291
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700292static inline int sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293{
Akinobu Mitada753be2006-04-28 15:21:23 -0700294 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295}
296
297static __inline__ void sk_node_init(struct hlist_node *node)
298{
299 node->pprev = NULL;
300}
301
302static __inline__ void __sk_del_node(struct sock *sk)
303{
304 __hlist_del(&sk->sk_node);
305}
306
307static __inline__ int __sk_del_node_init(struct sock *sk)
308{
309 if (sk_hashed(sk)) {
310 __sk_del_node(sk);
311 sk_node_init(&sk->sk_node);
312 return 1;
313 }
314 return 0;
315}
316
317/* Grab socket reference count. This operation is valid only
318 when sk is ALREADY grabbed f.e. it is found in hash table
319 or a list and the lookup is made under lock preventing hash table
320 modifications.
321 */
322
323static inline void sock_hold(struct sock *sk)
324{
325 atomic_inc(&sk->sk_refcnt);
326}
327
328/* Ungrab socket in the context, which assumes that socket refcnt
329 cannot hit zero, f.e. it is true in context of any socketcall.
330 */
331static inline void __sock_put(struct sock *sk)
332{
333 atomic_dec(&sk->sk_refcnt);
334}
335
336static __inline__ int sk_del_node_init(struct sock *sk)
337{
338 int rc = __sk_del_node_init(sk);
339
340 if (rc) {
341 /* paranoid for a while -acme */
342 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
343 __sock_put(sk);
344 }
345 return rc;
346}
347
348static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
349{
350 hlist_add_head(&sk->sk_node, list);
351}
352
353static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
354{
355 sock_hold(sk);
356 __sk_add_node(sk, list);
357}
358
359static __inline__ void __sk_del_bind_node(struct sock *sk)
360{
361 __hlist_del(&sk->sk_bind_node);
362}
363
364static __inline__ void sk_add_bind_node(struct sock *sk,
365 struct hlist_head *list)
366{
367 hlist_add_head(&sk->sk_bind_node, list);
368}
369
370#define sk_for_each(__sk, node, list) \
371 hlist_for_each_entry(__sk, node, list, sk_node)
372#define sk_for_each_from(__sk, node) \
373 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
374 hlist_for_each_entry_from(__sk, node, sk_node)
375#define sk_for_each_continue(__sk, node) \
376 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
377 hlist_for_each_entry_continue(__sk, node, sk_node)
378#define sk_for_each_safe(__sk, node, tmp, list) \
379 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
380#define sk_for_each_bound(__sk, node, list) \
381 hlist_for_each_entry(__sk, node, list, sk_bind_node)
382
383/* Sock flags */
384enum sock_flags {
385 SOCK_DEAD,
386 SOCK_DONE,
387 SOCK_URGINLINE,
388 SOCK_KEEPOPEN,
389 SOCK_LINGER,
390 SOCK_DESTROY,
391 SOCK_BROADCAST,
392 SOCK_TIMESTAMP,
393 SOCK_ZAPPED,
394 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
395 SOCK_DBG, /* %SO_DEBUG setting */
396 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700397 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
399 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
400};
401
Ralf Baechle53b924b2005-08-23 10:11:30 -0700402static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
403{
404 nsk->sk_flags = osk->sk_flags;
405}
406
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
408{
409 __set_bit(flag, &sk->sk_flags);
410}
411
412static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
413{
414 __clear_bit(flag, &sk->sk_flags);
415}
416
417static inline int sock_flag(struct sock *sk, enum sock_flags flag)
418{
419 return test_bit(flag, &sk->sk_flags);
420}
421
422static inline void sk_acceptq_removed(struct sock *sk)
423{
424 sk->sk_ack_backlog--;
425}
426
427static inline void sk_acceptq_added(struct sock *sk)
428{
429 sk->sk_ack_backlog++;
430}
431
432static inline int sk_acceptq_is_full(struct sock *sk)
433{
David S. Miller64a14652007-03-06 11:21:05 -0800434 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435}
436
437/*
438 * Compute minimal free write space needed to queue new packets.
439 */
440static inline int sk_stream_min_wspace(struct sock *sk)
441{
442 return sk->sk_wmem_queued / 2;
443}
444
445static inline int sk_stream_wspace(struct sock *sk)
446{
447 return sk->sk_sndbuf - sk->sk_wmem_queued;
448}
449
450extern void sk_stream_write_space(struct sock *sk);
451
452static inline int sk_stream_memory_free(struct sock *sk)
453{
454 return sk->sk_wmem_queued < sk->sk_sndbuf;
455}
456
457extern void sk_stream_rfree(struct sk_buff *skb);
458
459static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
460{
461 skb->sk = sk;
462 skb->destructor = sk_stream_rfree;
463 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
464 sk->sk_forward_alloc -= skb->truesize;
465}
466
467static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
468{
David S. Millerdc6de332006-04-20 00:10:50 -0700469 skb_truesize_check(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
471 sk->sk_wmem_queued -= skb->truesize;
472 sk->sk_forward_alloc += skb->truesize;
473 __kfree_skb(skb);
474}
475
476/* The per-socket spinlock must be held here. */
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800477static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
478{
479 if (!sk->sk_backlog.tail) {
480 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
481 } else {
482 sk->sk_backlog.tail->next = skb;
483 sk->sk_backlog.tail = skb;
484 }
485 skb->next = NULL;
486}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700488#define sk_wait_event(__sk, __timeo, __condition) \
489 ({ int __rc; \
490 release_sock(__sk); \
491 __rc = __condition; \
492 if (!__rc) { \
493 *(__timeo) = schedule_timeout(*(__timeo)); \
494 } \
495 lock_sock(__sk); \
496 __rc = __condition; \
497 __rc; \
498 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
500extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
501extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
502extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
503extern int sk_stream_error(struct sock *sk, int flags, int err);
504extern void sk_stream_kill_queues(struct sock *sk);
505
506extern int sk_wait_data(struct sock *sk, long *timeo);
507
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700508struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800509struct timewait_sock_ops;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700510
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511/* Networking protocol blocks we attach to sockets.
512 * socket layer -> transport layer interface
513 * transport -> network interface is defined by struct inet_proto
514 */
515struct proto {
516 void (*close)(struct sock *sk,
517 long timeout);
518 int (*connect)(struct sock *sk,
519 struct sockaddr *uaddr,
520 int addr_len);
521 int (*disconnect)(struct sock *sk, int flags);
522
523 struct sock * (*accept) (struct sock *sk, int flags, int *err);
524
525 int (*ioctl)(struct sock *sk, int cmd,
526 unsigned long arg);
527 int (*init)(struct sock *sk);
528 int (*destroy)(struct sock *sk);
529 void (*shutdown)(struct sock *sk, int how);
530 int (*setsockopt)(struct sock *sk, int level,
531 int optname, char __user *optval,
532 int optlen);
533 int (*getsockopt)(struct sock *sk, int level,
534 int optname, char __user *optval,
535 int __user *option);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800536 int (*compat_setsockopt)(struct sock *sk,
537 int level,
538 int optname, char __user *optval,
539 int optlen);
540 int (*compat_getsockopt)(struct sock *sk,
541 int level,
542 int optname, char __user *optval,
543 int __user *option);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
545 struct msghdr *msg, size_t len);
546 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
547 struct msghdr *msg,
548 size_t len, int noblock, int flags,
549 int *addr_len);
550 int (*sendpage)(struct sock *sk, struct page *page,
551 int offset, size_t size, int flags);
552 int (*bind)(struct sock *sk,
553 struct sockaddr *uaddr, int addr_len);
554
555 int (*backlog_rcv) (struct sock *sk,
556 struct sk_buff *skb);
557
558 /* Keeping track of sk's, looking them up, and port selection methods. */
559 void (*hash)(struct sock *sk);
560 void (*unhash)(struct sock *sk);
561 int (*get_port)(struct sock *sk, unsigned short snum);
562
563 /* Memory pressure */
564 void (*enter_memory_pressure)(void);
565 atomic_t *memory_allocated; /* Current allocated memory. */
566 atomic_t *sockets_allocated; /* Current number of sockets. */
567 /*
568 * Pressure flag: try to collapse.
569 * Technical note: it is used by multiple contexts non atomically.
570 * All the sk_stream_mem_schedule() is of this nature: accounting
571 * is strict, actions are advisory and have some latency.
572 */
573 int *memory_pressure;
574 int *sysctl_mem;
575 int *sysctl_wmem;
576 int *sysctl_rmem;
577 int max_header;
578
Christoph Lametere18b8902006-12-06 20:33:20 -0800579 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 unsigned int obj_size;
581
Arnaldo Carvalho de Melo0a5578c2005-08-09 20:11:41 -0700582 atomic_t *orphan_count;
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -0700583
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700584 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800585 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700586
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 struct module *owner;
588
589 char name[32];
590
591 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700592#ifdef SOCK_REFCNT_DEBUG
593 atomic_t socks;
594#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 struct {
596 int inuse;
597 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
598 } stats[NR_CPUS];
599};
600
601extern int proto_register(struct proto *prot, int alloc_slab);
602extern void proto_unregister(struct proto *prot);
603
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700604#ifdef SOCK_REFCNT_DEBUG
605static inline void sk_refcnt_debug_inc(struct sock *sk)
606{
607 atomic_inc(&sk->sk_prot->socks);
608}
609
610static inline void sk_refcnt_debug_dec(struct sock *sk)
611{
612 atomic_dec(&sk->sk_prot->socks);
613 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
614 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
615}
616
617static inline void sk_refcnt_debug_release(const struct sock *sk)
618{
619 if (atomic_read(&sk->sk_refcnt) != 1)
620 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
621 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
622}
623#else /* SOCK_REFCNT_DEBUG */
624#define sk_refcnt_debug_inc(sk) do { } while (0)
625#define sk_refcnt_debug_dec(sk) do { } while (0)
626#define sk_refcnt_debug_release(sk) do { } while (0)
627#endif /* SOCK_REFCNT_DEBUG */
628
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629/* Called with local bh disabled */
630static __inline__ void sock_prot_inc_use(struct proto *prot)
631{
632 prot->stats[smp_processor_id()].inuse++;
633}
634
635static __inline__ void sock_prot_dec_use(struct proto *prot)
636{
637 prot->stats[smp_processor_id()].inuse--;
638}
639
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -0700640/* With per-bucket locks this operation is not-atomic, so that
641 * this version is not worse.
642 */
643static inline void __sk_prot_rehash(struct sock *sk)
644{
645 sk->sk_prot->unhash(sk);
646 sk->sk_prot->hash(sk);
647}
648
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649/* About 10 seconds */
650#define SOCK_DESTROY_TIME (10*HZ)
651
652/* Sockets 0-1023 can't be bound to unless you are superuser */
653#define PROT_SOCK 1024
654
655#define SHUTDOWN_MASK 3
656#define RCV_SHUTDOWN 1
657#define SEND_SHUTDOWN 2
658
659#define SOCK_SNDBUF_LOCK 1
660#define SOCK_RCVBUF_LOCK 2
661#define SOCK_BINDADDR_LOCK 4
662#define SOCK_BINDPORT_LOCK 8
663
664/* sock_iocb: used to kick off async processing of socket ios */
665struct sock_iocb {
666 struct list_head list;
667
668 int flags;
669 int size;
670 struct socket *sock;
671 struct sock *sk;
672 struct scm_cookie *scm;
673 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 struct kiocb *kiocb;
675};
676
677static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
678{
679 return (struct sock_iocb *)iocb->private;
680}
681
682static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
683{
684 return si->kiocb;
685}
686
687struct socket_alloc {
688 struct socket socket;
689 struct inode vfs_inode;
690};
691
692static inline struct socket *SOCKET_I(struct inode *inode)
693{
694 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
695}
696
697static inline struct inode *SOCK_INODE(struct socket *socket)
698{
699 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
700}
701
702extern void __sk_stream_mem_reclaim(struct sock *sk);
703extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
704
705#define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
706
707static inline int sk_stream_pages(int amt)
708{
Ilpo Järvinen172589c2007-08-28 15:50:33 -0700709 return DIV_ROUND_UP(amt, SK_STREAM_MEM_QUANTUM);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710}
711
712static inline void sk_stream_mem_reclaim(struct sock *sk)
713{
714 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
715 __sk_stream_mem_reclaim(sk);
716}
717
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
719{
720 return (int)skb->truesize <= sk->sk_forward_alloc ||
721 sk_stream_mem_schedule(sk, skb->truesize, 1);
722}
723
Herbert Xud80d99d2005-09-01 17:48:23 -0700724static inline int sk_stream_wmem_schedule(struct sock *sk, int size)
725{
726 return size <= sk->sk_forward_alloc ||
727 sk_stream_mem_schedule(sk, size, 0);
728}
729
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730/* Used by processes to "lock" a socket state, so that
731 * interrupts and bottom half handlers won't change it
732 * from under us. It essentially blocks any incoming
733 * packets, so that we won't get any new data or any
734 * packets that change the state of the socket.
735 *
736 * While locked, BH processing will add new packets to
737 * the backlog queue. This queue is processed by the
738 * owner of the socket lock right before it is released.
739 *
740 * Since ~2.3.5 it is also exclusive sleep lock serializing
741 * accesses from user process context.
742 */
John Heffnerd2e91172007-09-12 10:44:19 +0200743#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744
Peter Zijlstraed075362006-12-06 20:35:24 -0800745/*
746 * Macro so as to not evaluate some arguments when
747 * lockdep is not enabled.
748 *
749 * Mark both the sk_lock and the sk_lock.slock as a
750 * per-address-family lock class.
751 */
752#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
753do { \
John Heffnerd2e91172007-09-12 10:44:19 +0200754 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -0800755 init_waitqueue_head(&sk->sk_lock.wq); \
756 spin_lock_init(&(sk)->sk_lock.slock); \
757 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
758 sizeof((sk)->sk_lock)); \
759 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
760 (skey), (sname)); \
761 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
762} while (0)
763
Peter Zijlstrafcc70d52006-11-08 22:44:35 -0800764extern void FASTCALL(lock_sock_nested(struct sock *sk, int subclass));
765
766static inline void lock_sock(struct sock *sk)
767{
768 lock_sock_nested(sk, 0);
769}
770
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771extern void FASTCALL(release_sock(struct sock *sk));
772
773/* BH context may only use the following locking interface. */
774#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -0700775#define bh_lock_sock_nested(__sk) \
776 spin_lock_nested(&((__sk)->sk_lock.slock), \
777 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
779
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700780extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +0100781 gfp_t priority,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 struct proto *prot, int zero_it);
783extern void sk_free(struct sock *sk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -0700784extern struct sock *sk_clone(const struct sock *sk,
Al Virodd0fc662005-10-07 07:46:04 +0100785 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786
787extern struct sk_buff *sock_wmalloc(struct sock *sk,
788 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +0100789 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790extern struct sk_buff *sock_rmalloc(struct sock *sk,
791 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +0100792 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793extern void sock_wfree(struct sk_buff *skb);
794extern void sock_rfree(struct sk_buff *skb);
795
796extern int sock_setsockopt(struct socket *sock, int level,
797 int op, char __user *optval,
798 int optlen);
799
800extern int sock_getsockopt(struct socket *sock, int level,
801 int op, char __user *optval,
802 int __user *optlen);
803extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
804 unsigned long size,
805 int noblock,
806 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700807extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +0100808 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809extern void sock_kfree_s(struct sock *sk, void *mem, int size);
810extern void sk_send_sigurg(struct sock *sk);
811
812/*
813 * Functions to fill in entries in struct proto_ops when a protocol
814 * does not implement a particular function.
815 */
816extern int sock_no_bind(struct socket *,
817 struct sockaddr *, int);
818extern int sock_no_connect(struct socket *,
819 struct sockaddr *, int, int);
820extern int sock_no_socketpair(struct socket *,
821 struct socket *);
822extern int sock_no_accept(struct socket *,
823 struct socket *, int);
824extern int sock_no_getname(struct socket *,
825 struct sockaddr *, int *, int);
826extern unsigned int sock_no_poll(struct file *, struct socket *,
827 struct poll_table_struct *);
828extern int sock_no_ioctl(struct socket *, unsigned int,
829 unsigned long);
830extern int sock_no_listen(struct socket *, int);
831extern int sock_no_shutdown(struct socket *, int);
832extern int sock_no_getsockopt(struct socket *, int , int,
833 char __user *, int __user *);
834extern int sock_no_setsockopt(struct socket *, int, int,
835 char __user *, int);
836extern int sock_no_sendmsg(struct kiocb *, struct socket *,
837 struct msghdr *, size_t);
838extern int sock_no_recvmsg(struct kiocb *, struct socket *,
839 struct msghdr *, size_t, int);
840extern int sock_no_mmap(struct file *file,
841 struct socket *sock,
842 struct vm_area_struct *vma);
843extern ssize_t sock_no_sendpage(struct socket *sock,
844 struct page *page,
845 int offset, size_t size,
846 int flags);
847
848/*
849 * Functions to fill in entries in struct proto_ops when a protocol
850 * uses the inet style.
851 */
852extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
853 char __user *optval, int __user *optlen);
854extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
855 struct msghdr *msg, size_t size, int flags);
856extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
857 char __user *optval, int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800858extern int compat_sock_common_getsockopt(struct socket *sock, int level,
859 int optname, char __user *optval, int __user *optlen);
860extern int compat_sock_common_setsockopt(struct socket *sock, int level,
861 int optname, char __user *optval, int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862
863extern void sk_common_release(struct sock *sk);
864
865/*
866 * Default socket callbacks and setup code
867 */
868
869/* Initialise core socket variables */
870extern void sock_init_data(struct socket *sock, struct sock *sk);
871
872/**
873 * sk_filter - run a packet through a socket filter
874 * @sk: sock associated with &sk_buff
875 * @skb: buffer to filter
876 * @needlock: set to 1 if the sock is not locked by caller.
877 *
878 * Run the filter code and then cut skb->data to correct size returned by
879 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
880 * than pkt_len we keep whole skb->data. This is the socket level
881 * wrapper to sk_run_filter. It returns 0 if the packet should
882 * be accepted or -EPERM if the packet should be tossed.
883 *
884 */
885
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700886static inline int sk_filter(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887{
888 int err;
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700889 struct sk_filter *filter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
891 err = security_sock_rcv_skb(sk, skb);
892 if (err)
893 return err;
894
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700895 rcu_read_lock_bh();
896 filter = sk->sk_filter;
897 if (filter) {
898 unsigned int pkt_len = sk_run_filter(skb, filter->insns,
899 filter->len);
900 err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 }
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700902 rcu_read_unlock_bh();
903
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 return err;
905}
906
907/**
Paul Bonserdc9b3342006-11-23 17:56:13 -0800908 * sk_filter_rcu_free: Free a socket filter
909 * @rcu: rcu_head that contains the sk_filter to free
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 */
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700911static inline void sk_filter_rcu_free(struct rcu_head *rcu)
912{
913 struct sk_filter *fp = container_of(rcu, struct sk_filter, rcu);
914 kfree(fp);
915}
916
Paul Bonserdc9b3342006-11-23 17:56:13 -0800917/**
918 * sk_filter_release: Release a socket filter
919 * @sk: socket
920 * @fp: filter to remove
921 *
922 * Remove a filter from a socket and release its resources.
923 */
924
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
926{
927 unsigned int size = sk_filter_len(fp);
928
929 atomic_sub(size, &sk->sk_omem_alloc);
930
931 if (atomic_dec_and_test(&fp->refcnt))
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700932 call_rcu_bh(&fp->rcu, sk_filter_rcu_free);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933}
934
935static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
936{
937 atomic_inc(&fp->refcnt);
938 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
939}
940
941/*
942 * Socket reference counting postulates.
943 *
944 * * Each user of socket SHOULD hold a reference count.
945 * * Each access point to socket (an hash table bucket, reference from a list,
946 * running timer, skb in flight MUST hold a reference count.
947 * * When reference count hits 0, it means it will never increase back.
948 * * When reference count hits 0, it means that no references from
949 * outside exist to this socket and current process on current CPU
950 * is last user and may/should destroy this socket.
951 * * sk_free is called from any context: process, BH, IRQ. When
952 * it is called, socket has no references from outside -> sk_free
953 * may release descendant resources allocated by the socket, but
954 * to the time when it is called, socket is NOT referenced by any
955 * hash tables, lists etc.
956 * * Packets, delivered from outside (from network or from another process)
957 * and enqueued on receive/error queues SHOULD NOT grab reference count,
958 * when they sit in queue. Otherwise, packets will leak to hole, when
959 * socket is looked up by one cpu and unhasing is made by another CPU.
960 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
961 * (leak to backlog). Packet socket does all the processing inside
962 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
963 * use separate SMP lock, so that they are prone too.
964 */
965
966/* Ungrab socket and destroy it, if it was the last reference. */
967static inline void sock_put(struct sock *sk)
968{
969 if (atomic_dec_and_test(&sk->sk_refcnt))
970 sk_free(sk);
971}
972
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -0200973extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
974 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -0200975
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976/* Detach socket from process context.
977 * Announce socket dead, detach it from wait queue and inode.
978 * Note that parent inode held reference count on this struct sock,
979 * we do not release it in this function, because protocol
980 * probably wants some additional cleanups or even continuing
981 * to work with this socket (TCP).
982 */
983static inline void sock_orphan(struct sock *sk)
984{
985 write_lock_bh(&sk->sk_callback_lock);
986 sock_set_flag(sk, SOCK_DEAD);
987 sk->sk_socket = NULL;
988 sk->sk_sleep = NULL;
989 write_unlock_bh(&sk->sk_callback_lock);
990}
991
992static inline void sock_graft(struct sock *sk, struct socket *parent)
993{
994 write_lock_bh(&sk->sk_callback_lock);
995 sk->sk_sleep = &parent->wait;
996 parent->sk = sk;
997 sk->sk_socket = parent;
Venkat Yekkirala4237c752006-07-24 23:32:50 -0700998 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 write_unlock_bh(&sk->sk_callback_lock);
1000}
1001
Venkat Yekkirala892c1412006-08-04 23:08:56 -07001002static inline void sock_copy(struct sock *nsk, const struct sock *osk)
1003{
1004#ifdef CONFIG_SECURITY_NETWORK
1005 void *sptr = nsk->sk_security;
1006#endif
1007
1008 memcpy(nsk, osk, osk->sk_prot->obj_size);
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001009 get_net(nsk->sk_net);
Venkat Yekkirala892c1412006-08-04 23:08:56 -07001010#ifdef CONFIG_SECURITY_NETWORK
1011 nsk->sk_security = sptr;
1012 security_sk_clone(osk, nsk);
1013#endif
1014}
1015
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016extern int sock_i_uid(struct sock *sk);
1017extern unsigned long sock_i_ino(struct sock *sk);
1018
1019static inline struct dst_entry *
1020__sk_dst_get(struct sock *sk)
1021{
1022 return sk->sk_dst_cache;
1023}
1024
1025static inline struct dst_entry *
1026sk_dst_get(struct sock *sk)
1027{
1028 struct dst_entry *dst;
1029
1030 read_lock(&sk->sk_dst_lock);
1031 dst = sk->sk_dst_cache;
1032 if (dst)
1033 dst_hold(dst);
1034 read_unlock(&sk->sk_dst_lock);
1035 return dst;
1036}
1037
1038static inline void
1039__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1040{
1041 struct dst_entry *old_dst;
1042
1043 old_dst = sk->sk_dst_cache;
1044 sk->sk_dst_cache = dst;
1045 dst_release(old_dst);
1046}
1047
1048static inline void
1049sk_dst_set(struct sock *sk, struct dst_entry *dst)
1050{
1051 write_lock(&sk->sk_dst_lock);
1052 __sk_dst_set(sk, dst);
1053 write_unlock(&sk->sk_dst_lock);
1054}
1055
1056static inline void
1057__sk_dst_reset(struct sock *sk)
1058{
1059 struct dst_entry *old_dst;
1060
1061 old_dst = sk->sk_dst_cache;
1062 sk->sk_dst_cache = NULL;
1063 dst_release(old_dst);
1064}
1065
1066static inline void
1067sk_dst_reset(struct sock *sk)
1068{
1069 write_lock(&sk->sk_dst_lock);
1070 __sk_dst_reset(sk);
1071 write_unlock(&sk->sk_dst_lock);
1072}
1073
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001074extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001076extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077
Herbert Xubcd76112006-06-30 13:36:35 -07001078static inline int sk_can_gso(const struct sock *sk)
1079{
1080 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1081}
1082
Andi Kleen99580892007-04-20 17:12:43 -07001083extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001084
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1086{
1087 sk->sk_wmem_queued += skb->truesize;
1088 sk->sk_forward_alloc -= skb->truesize;
1089}
1090
1091static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1092 struct sk_buff *skb, struct page *page,
1093 int off, int copy)
1094{
1095 if (skb->ip_summed == CHECKSUM_NONE) {
1096 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001097 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 page_address(page) + off,
1099 copy, 0, &err);
1100 if (err)
1101 return err;
1102 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1103 } else if (copy_from_user(page_address(page) + off, from, copy))
1104 return -EFAULT;
1105
1106 skb->len += copy;
1107 skb->data_len += copy;
1108 skb->truesize += copy;
1109 sk->sk_wmem_queued += copy;
1110 sk->sk_forward_alloc -= copy;
1111 return 0;
1112}
1113
1114/*
1115 * Queue a received datagram if it will fit. Stream and sequenced
1116 * protocols can't normally use this as they need to fit buffers in
1117 * and play with them.
1118 *
1119 * Inlined as it's very short and called for pretty much every
1120 * packet ever received.
1121 */
1122
1123static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1124{
1125 sock_hold(sk);
1126 skb->sk = sk;
1127 skb->destructor = sock_wfree;
1128 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1129}
1130
1131static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1132{
1133 skb->sk = sk;
1134 skb->destructor = sock_rfree;
1135 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1136}
1137
1138extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1139 unsigned long expires);
1140
1141extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1142
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001143extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144
1145static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1146{
1147 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1148 number of warnings when compiling with -W --ANK
1149 */
1150 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1151 (unsigned)sk->sk_rcvbuf)
1152 return -ENOMEM;
1153 skb_set_owner_r(skb, sk);
1154 skb_queue_tail(&sk->sk_error_queue, skb);
1155 if (!sock_flag(sk, SOCK_DEAD))
1156 sk->sk_data_ready(sk, skb->len);
1157 return 0;
1158}
1159
1160/*
1161 * Recover an error report and clear atomically
1162 */
1163
1164static inline int sock_error(struct sock *sk)
1165{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001166 int err;
1167 if (likely(!sk->sk_err))
1168 return 0;
1169 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 return -err;
1171}
1172
1173static inline unsigned long sock_wspace(struct sock *sk)
1174{
1175 int amt = 0;
1176
1177 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1178 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1179 if (amt < 0)
1180 amt = 0;
1181 }
1182 return amt;
1183}
1184
1185static inline void sk_wake_async(struct sock *sk, int how, int band)
1186{
1187 if (sk->sk_socket && sk->sk_socket->fasync_list)
1188 sock_wake_async(sk->sk_socket, how, band);
1189}
1190
1191#define SOCK_MIN_SNDBUF 2048
1192#define SOCK_MIN_RCVBUF 256
1193
1194static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1195{
1196 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1197 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1198 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1199 }
1200}
1201
1202static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
Victor Fusco86a76ca2005-07-08 14:57:47 -07001203 int size, int mem,
Al Virodd0fc662005-10-07 07:46:04 +01001204 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205{
David S. Millerc65f7f02005-07-05 15:17:25 -07001206 struct sk_buff *skb;
1207 int hdr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208
David S. Millerc65f7f02005-07-05 15:17:25 -07001209 hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header);
David S. Millerd179cd12005-08-17 14:57:30 -07001210 skb = alloc_skb_fclone(size + hdr_len, gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 if (skb) {
1212 skb->truesize += mem;
Herbert Xud80d99d2005-09-01 17:48:23 -07001213 if (sk_stream_wmem_schedule(sk, skb->truesize)) {
David S. Millerc65f7f02005-07-05 15:17:25 -07001214 skb_reserve(skb, hdr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 return skb;
1216 }
1217 __kfree_skb(skb);
1218 } else {
1219 sk->sk_prot->enter_memory_pressure();
1220 sk_stream_moderate_sndbuf(sk);
1221 }
1222 return NULL;
1223}
1224
1225static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
Victor Fusco86a76ca2005-07-08 14:57:47 -07001226 int size,
Al Virodd0fc662005-10-07 07:46:04 +01001227 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228{
1229 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1230}
1231
1232static inline struct page *sk_stream_alloc_page(struct sock *sk)
1233{
1234 struct page *page = NULL;
1235
Herbert Xuef015782005-09-01 17:48:59 -07001236 page = alloc_pages(sk->sk_allocation, 0);
1237 if (!page) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238 sk->sk_prot->enter_memory_pressure();
1239 sk_stream_moderate_sndbuf(sk);
1240 }
1241 return page;
1242}
1243
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244/*
1245 * Default write policy as shown to user space via poll/select/SIGIO
1246 */
1247static inline int sock_writeable(const struct sock *sk)
1248{
1249 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1250}
1251
Al Virodd0fc662005-10-07 07:46:04 +01001252static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253{
Patrick McHardy44981212007-02-27 09:56:42 -08001254 return in_atomic() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255}
1256
1257static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1258{
1259 return noblock ? 0 : sk->sk_rcvtimeo;
1260}
1261
1262static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1263{
1264 return noblock ? 0 : sk->sk_sndtimeo;
1265}
1266
1267static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1268{
1269 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1270}
1271
1272/* Alas, with timeout socket operations are not restartable.
1273 * Compare this to poll().
1274 */
1275static inline int sock_intr_errno(long timeo)
1276{
1277 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1278}
1279
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001280extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1281 struct sk_buff *skb);
1282
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283static __inline__ void
1284sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1285{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001286 ktime_t kt = skb->tstamp;
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001287
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001288 if (sock_flag(sk, SOCK_RCVTSTAMP))
1289 __sock_recv_timestamp(msg, sk, skb);
1290 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001291 sk->sk_stamp = kt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292}
1293
1294/**
1295 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001296 * @sk: socket to eat this skb from
1297 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07001298 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299 *
1300 * This routine must be called with interrupts disabled or with the socket
1301 * locked so that the sk_buff queue operation is ok.
1302*/
Chris Leech624d1162006-05-23 18:01:28 -07001303#ifdef CONFIG_NET_DMA
1304static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1305{
1306 __skb_unlink(skb, &sk->sk_receive_queue);
1307 if (!copied_early)
1308 __kfree_skb(skb);
1309 else
1310 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1311}
1312#else
1313static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314{
1315 __skb_unlink(skb, &sk->sk_receive_queue);
1316 __kfree_skb(skb);
1317}
Chris Leech624d1162006-05-23 18:01:28 -07001318#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319
1320extern 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
1358static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1359{
1360 if (valbool)
1361 sock_set_flag(sk, bit);
1362 else
1363 sock_reset_flag(sk, bit);
1364}
1365
1366extern __u32 sysctl_wmem_max;
1367extern __u32 sysctl_rmem_max;
1368
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001369extern void sk_init(void);
1370
1371#ifdef CONFIG_SYSCTL
1372extern struct ctl_table core_table[];
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001373#endif
1374
David S. Miller6baf1f42005-09-05 18:14:11 -07001375extern int sysctl_optmem_max;
1376
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001377extern __u32 sysctl_wmem_default;
1378extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001379
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380#endif /* _SOCK_H */