blob: 03684e702d13464cbc0a48b95354dc540341f504 [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include <linux/list.h>
44#include <linux/timer.h>
45#include <linux/cache.h>
46#include <linux/module.h>
Ingo Molnara5b5bb92006-07-03 00:25:35 -070047#include <linux/lockdep.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#include <linux/netdevice.h>
49#include <linux/skbuff.h> /* struct sk_buff */
Al Virod7fe0f22006-12-03 23:15:30 -050050#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070051#include <linux/security.h>
52
53#include <linux/filter.h>
54
55#include <asm/atomic.h>
56#include <net/dst.h>
57#include <net/checksum.h>
58
59/*
60 * This structure really needs to be cleaned up.
61 * Most of it is for TCP, and not used by any of
62 * the other protocols.
63 */
64
65/* Define this to get the SOCK_DBG debugging facility. */
66#define SOCK_DEBUGGING
67#ifdef SOCK_DEBUGGING
68#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
69 printk(KERN_DEBUG msg); } while (0)
70#else
71#define SOCK_DEBUG(sk, msg...) do { } while (0)
72#endif
73
74/* This is the per-socket lock. The spinlock provides a synchronization
75 * between user contexts and software interrupt processing, whereas the
76 * mini-semaphore synchronizes multiple users amongst themselves.
77 */
78struct sock_iocb;
79typedef struct {
80 spinlock_t slock;
81 struct sock_iocb *owner;
82 wait_queue_head_t wq;
Ingo Molnara5b5bb92006-07-03 00:25:35 -070083 /*
84 * We express the mutex-alike socket_lock semantics
85 * to the lock validator by explicitly managing
86 * the slock as a lock variant (in addition to
87 * the slock itself):
88 */
89#ifdef CONFIG_DEBUG_LOCK_ALLOC
90 struct lockdep_map dep_map;
91#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070092} socket_lock_t;
93
Linus Torvalds1da177e2005-04-16 15:20:36 -070094struct sock;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -070095struct proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
97/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -070098 * struct sock_common - minimal network layer representation of sockets
99 * @skc_family: network address family
100 * @skc_state: Connection state
101 * @skc_reuse: %SO_REUSEADDR setting
102 * @skc_bound_dev_if: bound device index if != 0
103 * @skc_node: main hash linkage for various protocol lookup tables
104 * @skc_bind_node: bind hash linkage for various protocol lookup tables
105 * @skc_refcnt: reference count
Eric Dumazet81c3d542005-10-03 14:13:38 -0700106 * @skc_hash: hash value used with various protocol lookup tables
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700107 * @skc_prot: protocol handlers inside a network family
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700108 *
109 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700110 * for struct sock and struct inet_timewait_sock.
111 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112struct sock_common {
113 unsigned short skc_family;
114 volatile unsigned char skc_state;
115 unsigned char skc_reuse;
116 int skc_bound_dev_if;
117 struct hlist_node skc_node;
118 struct hlist_node skc_bind_node;
119 atomic_t skc_refcnt;
Eric Dumazet81c3d542005-10-03 14:13:38 -0700120 unsigned int skc_hash;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700121 struct proto *skc_prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122};
123
124/**
125 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700126 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700127 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
128 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
129 * @sk_lock: synchronizer
130 * @sk_rcvbuf: size of receive buffer in bytes
131 * @sk_sleep: sock wait queue
132 * @sk_dst_cache: destination cache
133 * @sk_dst_lock: destination cache lock
134 * @sk_policy: flow policy
135 * @sk_rmem_alloc: receive queue bytes committed
136 * @sk_receive_queue: incoming packets
137 * @sk_wmem_alloc: transmit queue bytes committed
138 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700139 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700140 * @sk_omem_alloc: "o" is "option" or "other"
141 * @sk_wmem_queued: persistent queue size
142 * @sk_forward_alloc: space allocated forward
143 * @sk_allocation: allocation mode
144 * @sk_sndbuf: size of send buffer in bytes
145 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
146 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
147 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Herbert Xubcd76112006-06-30 13:36:35 -0700148 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700149 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700150 * @sk_backlog: always used with the per-socket spinlock held
151 * @sk_callback_lock: used with the callbacks in the end of this struct
152 * @sk_error_queue: rarely used
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700153 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700154 * @sk_err: last error
155 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
156 * @sk_ack_backlog: current listen backlog
157 * @sk_max_ack_backlog: listen backlog set in listen()
158 * @sk_priority: %SO_PRIORITY setting
159 * @sk_type: socket type (%SOCK_STREAM, etc)
160 * @sk_protocol: which protocol this socket belongs in this network family
161 * @sk_peercred: %SO_PEERCRED setting
162 * @sk_rcvlowat: %SO_RCVLOWAT setting
163 * @sk_rcvtimeo: %SO_RCVTIMEO setting
164 * @sk_sndtimeo: %SO_SNDTIMEO setting
165 * @sk_filter: socket filtering instructions
166 * @sk_protinfo: private area, net family specific, when not using slab
167 * @sk_timer: sock cleanup timer
168 * @sk_stamp: time stamp of last packet received
169 * @sk_socket: Identd and reporting IO signals
170 * @sk_user_data: RPC layer private data
171 * @sk_sndmsg_page: cached page for sendmsg
172 * @sk_sndmsg_off: cached offset for sendmsg
173 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700174 * @sk_security: used by security modules
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700175 * @sk_write_pending: a write to stream socket waits to start
176 * @sk_state_change: callback to indicate change in the state of the sock
177 * @sk_data_ready: callback to indicate there is data to be processed
178 * @sk_write_space: callback to indicate there is bf sending space available
179 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
180 * @sk_backlog_rcv: callback to process the backlog
181 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 */
183struct sock {
184 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700185 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 * don't add nothing before this first member (__sk_common) --acme
187 */
188 struct sock_common __sk_common;
189#define sk_family __sk_common.skc_family
190#define sk_state __sk_common.skc_state
191#define sk_reuse __sk_common.skc_reuse
192#define sk_bound_dev_if __sk_common.skc_bound_dev_if
193#define sk_node __sk_common.skc_node
194#define sk_bind_node __sk_common.skc_bind_node
195#define sk_refcnt __sk_common.skc_refcnt
Eric Dumazet81c3d542005-10-03 14:13:38 -0700196#define sk_hash __sk_common.skc_hash
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700197#define sk_prot __sk_common.skc_prot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 unsigned char sk_shutdown : 2,
199 sk_no_check : 2,
200 sk_userlocks : 4;
201 unsigned char sk_protocol;
202 unsigned short sk_type;
203 int sk_rcvbuf;
204 socket_lock_t sk_lock;
205 wait_queue_head_t *sk_sleep;
206 struct dst_entry *sk_dst_cache;
207 struct xfrm_policy *sk_policy[2];
208 rwlock_t sk_dst_lock;
209 atomic_t sk_rmem_alloc;
210 atomic_t sk_wmem_alloc;
211 atomic_t sk_omem_alloc;
212 struct sk_buff_head sk_receive_queue;
213 struct sk_buff_head sk_write_queue;
Chris Leech97fc2f02006-05-23 17:55:33 -0700214 struct sk_buff_head sk_async_wait_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 int sk_wmem_queued;
216 int sk_forward_alloc;
Al Viro7d877f32005-10-21 03:20:43 -0400217 gfp_t sk_allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 int sk_sndbuf;
219 int sk_route_caps;
Herbert Xubcd76112006-06-30 13:36:35 -0700220 int sk_gso_type;
David S. Miller9932cf92006-03-24 15:12:37 -0800221 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 unsigned long sk_flags;
223 unsigned long sk_lingertime;
224 /*
225 * The backlog queue is special, it is always used with
226 * the per-socket spinlock held and requires low latency
227 * access. Therefore we special case it's implementation.
228 */
229 struct {
230 struct sk_buff *head;
231 struct sk_buff *tail;
232 } sk_backlog;
233 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700234 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 rwlock_t sk_callback_lock;
236 int sk_err,
237 sk_err_soft;
238 unsigned short sk_ack_backlog;
239 unsigned short sk_max_ack_backlog;
240 __u32 sk_priority;
241 struct ucred sk_peercred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 long sk_rcvtimeo;
243 long sk_sndtimeo;
244 struct sk_filter *sk_filter;
245 void *sk_protinfo;
246 struct timer_list sk_timer;
247 struct timeval sk_stamp;
248 struct socket *sk_socket;
249 void *sk_user_data;
250 struct page *sk_sndmsg_page;
251 struct sk_buff *sk_send_head;
252 __u32 sk_sndmsg_off;
253 int sk_write_pending;
254 void *sk_security;
255 void (*sk_state_change)(struct sock *sk);
256 void (*sk_data_ready)(struct sock *sk, int bytes);
257 void (*sk_write_space)(struct sock *sk);
258 void (*sk_error_report)(struct sock *sk);
259 int (*sk_backlog_rcv)(struct sock *sk,
260 struct sk_buff *skb);
261 void (*sk_destruct)(struct sock *sk);
262};
263
264/*
265 * Hashed lists helper routines
266 */
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700267static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268{
269 return hlist_entry(head->first, struct sock, sk_node);
270}
271
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700272static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273{
274 return hlist_empty(head) ? NULL : __sk_head(head);
275}
276
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700277static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278{
279 return sk->sk_node.next ?
280 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
281}
282
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700283static inline int sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284{
285 return hlist_unhashed(&sk->sk_node);
286}
287
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700288static inline int sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289{
Akinobu Mitada753be2006-04-28 15:21:23 -0700290 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291}
292
293static __inline__ void sk_node_init(struct hlist_node *node)
294{
295 node->pprev = NULL;
296}
297
298static __inline__ void __sk_del_node(struct sock *sk)
299{
300 __hlist_del(&sk->sk_node);
301}
302
303static __inline__ int __sk_del_node_init(struct sock *sk)
304{
305 if (sk_hashed(sk)) {
306 __sk_del_node(sk);
307 sk_node_init(&sk->sk_node);
308 return 1;
309 }
310 return 0;
311}
312
313/* Grab socket reference count. This operation is valid only
314 when sk is ALREADY grabbed f.e. it is found in hash table
315 or a list and the lookup is made under lock preventing hash table
316 modifications.
317 */
318
319static inline void sock_hold(struct sock *sk)
320{
321 atomic_inc(&sk->sk_refcnt);
322}
323
324/* Ungrab socket in the context, which assumes that socket refcnt
325 cannot hit zero, f.e. it is true in context of any socketcall.
326 */
327static inline void __sock_put(struct sock *sk)
328{
329 atomic_dec(&sk->sk_refcnt);
330}
331
332static __inline__ int sk_del_node_init(struct sock *sk)
333{
334 int rc = __sk_del_node_init(sk);
335
336 if (rc) {
337 /* paranoid for a while -acme */
338 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
339 __sock_put(sk);
340 }
341 return rc;
342}
343
344static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
345{
346 hlist_add_head(&sk->sk_node, list);
347}
348
349static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
350{
351 sock_hold(sk);
352 __sk_add_node(sk, list);
353}
354
355static __inline__ void __sk_del_bind_node(struct sock *sk)
356{
357 __hlist_del(&sk->sk_bind_node);
358}
359
360static __inline__ void sk_add_bind_node(struct sock *sk,
361 struct hlist_head *list)
362{
363 hlist_add_head(&sk->sk_bind_node, list);
364}
365
366#define sk_for_each(__sk, node, list) \
367 hlist_for_each_entry(__sk, node, list, sk_node)
368#define sk_for_each_from(__sk, node) \
369 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
370 hlist_for_each_entry_from(__sk, node, sk_node)
371#define sk_for_each_continue(__sk, node) \
372 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
373 hlist_for_each_entry_continue(__sk, node, sk_node)
374#define sk_for_each_safe(__sk, node, tmp, list) \
375 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
376#define sk_for_each_bound(__sk, node, list) \
377 hlist_for_each_entry(__sk, node, list, sk_bind_node)
378
379/* Sock flags */
380enum sock_flags {
381 SOCK_DEAD,
382 SOCK_DONE,
383 SOCK_URGINLINE,
384 SOCK_KEEPOPEN,
385 SOCK_LINGER,
386 SOCK_DESTROY,
387 SOCK_BROADCAST,
388 SOCK_TIMESTAMP,
389 SOCK_ZAPPED,
390 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
391 SOCK_DBG, /* %SO_DEBUG setting */
392 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
394 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
395};
396
Ralf Baechle53b924b2005-08-23 10:11:30 -0700397static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
398{
399 nsk->sk_flags = osk->sk_flags;
400}
401
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
403{
404 __set_bit(flag, &sk->sk_flags);
405}
406
407static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
408{
409 __clear_bit(flag, &sk->sk_flags);
410}
411
412static inline int sock_flag(struct sock *sk, enum sock_flags flag)
413{
414 return test_bit(flag, &sk->sk_flags);
415}
416
417static inline void sk_acceptq_removed(struct sock *sk)
418{
419 sk->sk_ack_backlog--;
420}
421
422static inline void sk_acceptq_added(struct sock *sk)
423{
424 sk->sk_ack_backlog++;
425}
426
427static inline int sk_acceptq_is_full(struct sock *sk)
428{
429 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
430}
431
432/*
433 * Compute minimal free write space needed to queue new packets.
434 */
435static inline int sk_stream_min_wspace(struct sock *sk)
436{
437 return sk->sk_wmem_queued / 2;
438}
439
440static inline int sk_stream_wspace(struct sock *sk)
441{
442 return sk->sk_sndbuf - sk->sk_wmem_queued;
443}
444
445extern void sk_stream_write_space(struct sock *sk);
446
447static inline int sk_stream_memory_free(struct sock *sk)
448{
449 return sk->sk_wmem_queued < sk->sk_sndbuf;
450}
451
452extern void sk_stream_rfree(struct sk_buff *skb);
453
454static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
455{
456 skb->sk = sk;
457 skb->destructor = sk_stream_rfree;
458 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
459 sk->sk_forward_alloc -= skb->truesize;
460}
461
462static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
463{
David S. Millerdc6de332006-04-20 00:10:50 -0700464 skb_truesize_check(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
466 sk->sk_wmem_queued -= skb->truesize;
467 sk->sk_forward_alloc += skb->truesize;
468 __kfree_skb(skb);
469}
470
471/* The per-socket spinlock must be held here. */
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800472static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
473{
474 if (!sk->sk_backlog.tail) {
475 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
476 } else {
477 sk->sk_backlog.tail->next = skb;
478 sk->sk_backlog.tail = skb;
479 }
480 skb->next = NULL;
481}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482
483#define sk_wait_event(__sk, __timeo, __condition) \
484({ int rc; \
485 release_sock(__sk); \
486 rc = __condition; \
487 if (!rc) { \
488 *(__timeo) = schedule_timeout(*(__timeo)); \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 } \
490 lock_sock(__sk); \
Alexey Kuznetsov265a9282006-03-17 16:05:43 -0800491 rc = __condition; \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 rc; \
493})
494
495extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
496extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
497extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
498extern int sk_stream_error(struct sock *sk, int flags, int err);
499extern void sk_stream_kill_queues(struct sock *sk);
500
501extern int sk_wait_data(struct sock *sk, long *timeo);
502
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700503struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800504struct timewait_sock_ops;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700505
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506/* Networking protocol blocks we attach to sockets.
507 * socket layer -> transport layer interface
508 * transport -> network interface is defined by struct inet_proto
509 */
510struct proto {
511 void (*close)(struct sock *sk,
512 long timeout);
513 int (*connect)(struct sock *sk,
514 struct sockaddr *uaddr,
515 int addr_len);
516 int (*disconnect)(struct sock *sk, int flags);
517
518 struct sock * (*accept) (struct sock *sk, int flags, int *err);
519
520 int (*ioctl)(struct sock *sk, int cmd,
521 unsigned long arg);
522 int (*init)(struct sock *sk);
523 int (*destroy)(struct sock *sk);
524 void (*shutdown)(struct sock *sk, int how);
525 int (*setsockopt)(struct sock *sk, int level,
526 int optname, char __user *optval,
527 int optlen);
528 int (*getsockopt)(struct sock *sk, int level,
529 int optname, char __user *optval,
530 int __user *option);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800531 int (*compat_setsockopt)(struct sock *sk,
532 int level,
533 int optname, char __user *optval,
534 int optlen);
535 int (*compat_getsockopt)(struct sock *sk,
536 int level,
537 int optname, char __user *optval,
538 int __user *option);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
540 struct msghdr *msg, size_t len);
541 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
542 struct msghdr *msg,
543 size_t len, int noblock, int flags,
544 int *addr_len);
545 int (*sendpage)(struct sock *sk, struct page *page,
546 int offset, size_t size, int flags);
547 int (*bind)(struct sock *sk,
548 struct sockaddr *uaddr, int addr_len);
549
550 int (*backlog_rcv) (struct sock *sk,
551 struct sk_buff *skb);
552
553 /* Keeping track of sk's, looking them up, and port selection methods. */
554 void (*hash)(struct sock *sk);
555 void (*unhash)(struct sock *sk);
556 int (*get_port)(struct sock *sk, unsigned short snum);
557
558 /* Memory pressure */
559 void (*enter_memory_pressure)(void);
560 atomic_t *memory_allocated; /* Current allocated memory. */
561 atomic_t *sockets_allocated; /* Current number of sockets. */
562 /*
563 * Pressure flag: try to collapse.
564 * Technical note: it is used by multiple contexts non atomically.
565 * All the sk_stream_mem_schedule() is of this nature: accounting
566 * is strict, actions are advisory and have some latency.
567 */
568 int *memory_pressure;
569 int *sysctl_mem;
570 int *sysctl_wmem;
571 int *sysctl_rmem;
572 int max_header;
573
Christoph Lametere18b8902006-12-06 20:33:20 -0800574 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 unsigned int obj_size;
576
Arnaldo Carvalho de Melo0a5578c2005-08-09 20:11:41 -0700577 atomic_t *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700578
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700579 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800580 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700581
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 struct module *owner;
583
584 char name[32];
585
586 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700587#ifdef SOCK_REFCNT_DEBUG
588 atomic_t socks;
589#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 struct {
591 int inuse;
592 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
593 } stats[NR_CPUS];
594};
595
596extern int proto_register(struct proto *prot, int alloc_slab);
597extern void proto_unregister(struct proto *prot);
598
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700599#ifdef SOCK_REFCNT_DEBUG
600static inline void sk_refcnt_debug_inc(struct sock *sk)
601{
602 atomic_inc(&sk->sk_prot->socks);
603}
604
605static inline void sk_refcnt_debug_dec(struct sock *sk)
606{
607 atomic_dec(&sk->sk_prot->socks);
608 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
609 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
610}
611
612static inline void sk_refcnt_debug_release(const struct sock *sk)
613{
614 if (atomic_read(&sk->sk_refcnt) != 1)
615 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
616 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
617}
618#else /* SOCK_REFCNT_DEBUG */
619#define sk_refcnt_debug_inc(sk) do { } while (0)
620#define sk_refcnt_debug_dec(sk) do { } while (0)
621#define sk_refcnt_debug_release(sk) do { } while (0)
622#endif /* SOCK_REFCNT_DEBUG */
623
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624/* Called with local bh disabled */
625static __inline__ void sock_prot_inc_use(struct proto *prot)
626{
627 prot->stats[smp_processor_id()].inuse++;
628}
629
630static __inline__ void sock_prot_dec_use(struct proto *prot)
631{
632 prot->stats[smp_processor_id()].inuse--;
633}
634
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -0700635/* With per-bucket locks this operation is not-atomic, so that
636 * this version is not worse.
637 */
638static inline void __sk_prot_rehash(struct sock *sk)
639{
640 sk->sk_prot->unhash(sk);
641 sk->sk_prot->hash(sk);
642}
643
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644/* About 10 seconds */
645#define SOCK_DESTROY_TIME (10*HZ)
646
647/* Sockets 0-1023 can't be bound to unless you are superuser */
648#define PROT_SOCK 1024
649
650#define SHUTDOWN_MASK 3
651#define RCV_SHUTDOWN 1
652#define SEND_SHUTDOWN 2
653
654#define SOCK_SNDBUF_LOCK 1
655#define SOCK_RCVBUF_LOCK 2
656#define SOCK_BINDADDR_LOCK 4
657#define SOCK_BINDPORT_LOCK 8
658
659/* sock_iocb: used to kick off async processing of socket ios */
660struct sock_iocb {
661 struct list_head list;
662
663 int flags;
664 int size;
665 struct socket *sock;
666 struct sock *sk;
667 struct scm_cookie *scm;
668 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 struct kiocb *kiocb;
670};
671
672static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
673{
674 return (struct sock_iocb *)iocb->private;
675}
676
677static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
678{
679 return si->kiocb;
680}
681
682struct socket_alloc {
683 struct socket socket;
684 struct inode vfs_inode;
685};
686
687static inline struct socket *SOCKET_I(struct inode *inode)
688{
689 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
690}
691
692static inline struct inode *SOCK_INODE(struct socket *socket)
693{
694 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
695}
696
697extern void __sk_stream_mem_reclaim(struct sock *sk);
698extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
699
700#define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
701
702static inline int sk_stream_pages(int amt)
703{
704 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
705}
706
707static inline void sk_stream_mem_reclaim(struct sock *sk)
708{
709 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
710 __sk_stream_mem_reclaim(sk);
711}
712
713static inline void sk_stream_writequeue_purge(struct sock *sk)
714{
715 struct sk_buff *skb;
716
717 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
718 sk_stream_free_skb(sk, skb);
719 sk_stream_mem_reclaim(sk);
720}
721
722static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
723{
724 return (int)skb->truesize <= sk->sk_forward_alloc ||
725 sk_stream_mem_schedule(sk, skb->truesize, 1);
726}
727
Herbert Xud80d99d2005-09-01 17:48:23 -0700728static inline int sk_stream_wmem_schedule(struct sock *sk, int size)
729{
730 return size <= sk->sk_forward_alloc ||
731 sk_stream_mem_schedule(sk, size, 0);
732}
733
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734/* Used by processes to "lock" a socket state, so that
735 * interrupts and bottom half handlers won't change it
736 * from under us. It essentially blocks any incoming
737 * packets, so that we won't get any new data or any
738 * packets that change the state of the socket.
739 *
740 * While locked, BH processing will add new packets to
741 * the backlog queue. This queue is processed by the
742 * owner of the socket lock right before it is released.
743 *
744 * Since ~2.3.5 it is also exclusive sleep lock serializing
745 * accesses from user process context.
746 */
747#define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
748
Peter Zijlstraed075362006-12-06 20:35:24 -0800749/*
750 * Macro so as to not evaluate some arguments when
751 * lockdep is not enabled.
752 *
753 * Mark both the sk_lock and the sk_lock.slock as a
754 * per-address-family lock class.
755 */
756#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
757do { \
758 sk->sk_lock.owner = NULL; \
759 init_waitqueue_head(&sk->sk_lock.wq); \
760 spin_lock_init(&(sk)->sk_lock.slock); \
761 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
762 sizeof((sk)->sk_lock)); \
763 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
764 (skey), (sname)); \
765 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
766} while (0)
767
Peter Zijlstrafcc70d52006-11-08 22:44:35 -0800768extern void FASTCALL(lock_sock_nested(struct sock *sk, int subclass));
769
770static inline void lock_sock(struct sock *sk)
771{
772 lock_sock_nested(sk, 0);
773}
774
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775extern void FASTCALL(release_sock(struct sock *sk));
776
777/* BH context may only use the following locking interface. */
778#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -0700779#define bh_lock_sock_nested(__sk) \
780 spin_lock_nested(&((__sk)->sk_lock.slock), \
781 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
783
Victor Fusco86a76ca2005-07-08 14:57:47 -0700784extern struct sock *sk_alloc(int family,
Al Virodd0fc662005-10-07 07:46:04 +0100785 gfp_t priority,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 struct proto *prot, int zero_it);
787extern void sk_free(struct sock *sk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -0700788extern struct sock *sk_clone(const struct sock *sk,
Al Virodd0fc662005-10-07 07:46:04 +0100789 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790
791extern struct sk_buff *sock_wmalloc(struct sock *sk,
792 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +0100793 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794extern struct sk_buff *sock_rmalloc(struct sock *sk,
795 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +0100796 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797extern void sock_wfree(struct sk_buff *skb);
798extern void sock_rfree(struct sk_buff *skb);
799
800extern int sock_setsockopt(struct socket *sock, int level,
801 int op, char __user *optval,
802 int optlen);
803
804extern int sock_getsockopt(struct socket *sock, int level,
805 int op, char __user *optval,
806 int __user *optlen);
807extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
808 unsigned long size,
809 int noblock,
810 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700811extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +0100812 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813extern void sock_kfree_s(struct sock *sk, void *mem, int size);
814extern void sk_send_sigurg(struct sock *sk);
815
816/*
817 * Functions to fill in entries in struct proto_ops when a protocol
818 * does not implement a particular function.
819 */
820extern int sock_no_bind(struct socket *,
821 struct sockaddr *, int);
822extern int sock_no_connect(struct socket *,
823 struct sockaddr *, int, int);
824extern int sock_no_socketpair(struct socket *,
825 struct socket *);
826extern int sock_no_accept(struct socket *,
827 struct socket *, int);
828extern int sock_no_getname(struct socket *,
829 struct sockaddr *, int *, int);
830extern unsigned int sock_no_poll(struct file *, struct socket *,
831 struct poll_table_struct *);
832extern int sock_no_ioctl(struct socket *, unsigned int,
833 unsigned long);
834extern int sock_no_listen(struct socket *, int);
835extern int sock_no_shutdown(struct socket *, int);
836extern int sock_no_getsockopt(struct socket *, int , int,
837 char __user *, int __user *);
838extern int sock_no_setsockopt(struct socket *, int, int,
839 char __user *, int);
840extern int sock_no_sendmsg(struct kiocb *, struct socket *,
841 struct msghdr *, size_t);
842extern int sock_no_recvmsg(struct kiocb *, struct socket *,
843 struct msghdr *, size_t, int);
844extern int sock_no_mmap(struct file *file,
845 struct socket *sock,
846 struct vm_area_struct *vma);
847extern ssize_t sock_no_sendpage(struct socket *sock,
848 struct page *page,
849 int offset, size_t size,
850 int flags);
851
852/*
853 * Functions to fill in entries in struct proto_ops when a protocol
854 * uses the inet style.
855 */
856extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
857 char __user *optval, int __user *optlen);
858extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
859 struct msghdr *msg, size_t size, int flags);
860extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
861 char __user *optval, int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800862extern int compat_sock_common_getsockopt(struct socket *sock, int level,
863 int optname, char __user *optval, int __user *optlen);
864extern int compat_sock_common_setsockopt(struct socket *sock, int level,
865 int optname, char __user *optval, int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866
867extern void sk_common_release(struct sock *sk);
868
869/*
870 * Default socket callbacks and setup code
871 */
872
873/* Initialise core socket variables */
874extern void sock_init_data(struct socket *sock, struct sock *sk);
875
876/**
877 * sk_filter - run a packet through a socket filter
878 * @sk: sock associated with &sk_buff
879 * @skb: buffer to filter
880 * @needlock: set to 1 if the sock is not locked by caller.
881 *
882 * Run the filter code and then cut skb->data to correct size returned by
883 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
884 * than pkt_len we keep whole skb->data. This is the socket level
885 * wrapper to sk_run_filter. It returns 0 if the packet should
886 * be accepted or -EPERM if the packet should be tossed.
887 *
888 */
889
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700890static inline int sk_filter(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891{
892 int err;
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700893 struct sk_filter *filter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894
895 err = security_sock_rcv_skb(sk, skb);
896 if (err)
897 return err;
898
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700899 rcu_read_lock_bh();
900 filter = sk->sk_filter;
901 if (filter) {
902 unsigned int pkt_len = sk_run_filter(skb, filter->insns,
903 filter->len);
904 err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 }
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700906 rcu_read_unlock_bh();
907
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 return err;
909}
910
911/**
Paul Bonserdc9b3342006-11-23 17:56:13 -0800912 * sk_filter_rcu_free: Free a socket filter
913 * @rcu: rcu_head that contains the sk_filter to free
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 */
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700915static inline void sk_filter_rcu_free(struct rcu_head *rcu)
916{
917 struct sk_filter *fp = container_of(rcu, struct sk_filter, rcu);
918 kfree(fp);
919}
920
Paul Bonserdc9b3342006-11-23 17:56:13 -0800921/**
922 * sk_filter_release: Release a socket filter
923 * @sk: socket
924 * @fp: filter to remove
925 *
926 * Remove a filter from a socket and release its resources.
927 */
928
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
930{
931 unsigned int size = sk_filter_len(fp);
932
933 atomic_sub(size, &sk->sk_omem_alloc);
934
935 if (atomic_dec_and_test(&fp->refcnt))
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700936 call_rcu_bh(&fp->rcu, sk_filter_rcu_free);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937}
938
939static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
940{
941 atomic_inc(&fp->refcnt);
942 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
943}
944
945/*
946 * Socket reference counting postulates.
947 *
948 * * Each user of socket SHOULD hold a reference count.
949 * * Each access point to socket (an hash table bucket, reference from a list,
950 * running timer, skb in flight MUST hold a reference count.
951 * * When reference count hits 0, it means it will never increase back.
952 * * When reference count hits 0, it means that no references from
953 * outside exist to this socket and current process on current CPU
954 * is last user and may/should destroy this socket.
955 * * sk_free is called from any context: process, BH, IRQ. When
956 * it is called, socket has no references from outside -> sk_free
957 * may release descendant resources allocated by the socket, but
958 * to the time when it is called, socket is NOT referenced by any
959 * hash tables, lists etc.
960 * * Packets, delivered from outside (from network or from another process)
961 * and enqueued on receive/error queues SHOULD NOT grab reference count,
962 * when they sit in queue. Otherwise, packets will leak to hole, when
963 * socket is looked up by one cpu and unhasing is made by another CPU.
964 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
965 * (leak to backlog). Packet socket does all the processing inside
966 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
967 * use separate SMP lock, so that they are prone too.
968 */
969
970/* Ungrab socket and destroy it, if it was the last reference. */
971static inline void sock_put(struct sock *sk)
972{
973 if (atomic_dec_and_test(&sk->sk_refcnt))
974 sk_free(sk);
975}
976
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -0200977extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
978 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -0200979
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980/* Detach socket from process context.
981 * Announce socket dead, detach it from wait queue and inode.
982 * Note that parent inode held reference count on this struct sock,
983 * we do not release it in this function, because protocol
984 * probably wants some additional cleanups or even continuing
985 * to work with this socket (TCP).
986 */
987static inline void sock_orphan(struct sock *sk)
988{
989 write_lock_bh(&sk->sk_callback_lock);
990 sock_set_flag(sk, SOCK_DEAD);
991 sk->sk_socket = NULL;
992 sk->sk_sleep = NULL;
993 write_unlock_bh(&sk->sk_callback_lock);
994}
995
996static inline void sock_graft(struct sock *sk, struct socket *parent)
997{
998 write_lock_bh(&sk->sk_callback_lock);
999 sk->sk_sleep = &parent->wait;
1000 parent->sk = sk;
1001 sk->sk_socket = parent;
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001002 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 write_unlock_bh(&sk->sk_callback_lock);
1004}
1005
Venkat Yekkirala892c1412006-08-04 23:08:56 -07001006static inline void sock_copy(struct sock *nsk, const struct sock *osk)
1007{
1008#ifdef CONFIG_SECURITY_NETWORK
1009 void *sptr = nsk->sk_security;
1010#endif
1011
1012 memcpy(nsk, osk, osk->sk_prot->obj_size);
1013#ifdef CONFIG_SECURITY_NETWORK
1014 nsk->sk_security = sptr;
1015 security_sk_clone(osk, nsk);
1016#endif
1017}
1018
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019extern int sock_i_uid(struct sock *sk);
1020extern unsigned long sock_i_ino(struct sock *sk);
1021
1022static inline struct dst_entry *
1023__sk_dst_get(struct sock *sk)
1024{
1025 return sk->sk_dst_cache;
1026}
1027
1028static inline struct dst_entry *
1029sk_dst_get(struct sock *sk)
1030{
1031 struct dst_entry *dst;
1032
1033 read_lock(&sk->sk_dst_lock);
1034 dst = sk->sk_dst_cache;
1035 if (dst)
1036 dst_hold(dst);
1037 read_unlock(&sk->sk_dst_lock);
1038 return dst;
1039}
1040
1041static inline void
1042__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1043{
1044 struct dst_entry *old_dst;
1045
1046 old_dst = sk->sk_dst_cache;
1047 sk->sk_dst_cache = dst;
1048 dst_release(old_dst);
1049}
1050
1051static inline void
1052sk_dst_set(struct sock *sk, struct dst_entry *dst)
1053{
1054 write_lock(&sk->sk_dst_lock);
1055 __sk_dst_set(sk, dst);
1056 write_unlock(&sk->sk_dst_lock);
1057}
1058
1059static inline void
1060__sk_dst_reset(struct sock *sk)
1061{
1062 struct dst_entry *old_dst;
1063
1064 old_dst = sk->sk_dst_cache;
1065 sk->sk_dst_cache = NULL;
1066 dst_release(old_dst);
1067}
1068
1069static inline void
1070sk_dst_reset(struct sock *sk)
1071{
1072 write_lock(&sk->sk_dst_lock);
1073 __sk_dst_reset(sk);
1074 write_unlock(&sk->sk_dst_lock);
1075}
1076
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001077extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001079extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080
Herbert Xubcd76112006-06-30 13:36:35 -07001081static inline int sk_can_gso(const struct sock *sk)
1082{
1083 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1084}
1085
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001086static inline void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1087{
1088 __sk_dst_set(sk, dst);
1089 sk->sk_route_caps = dst->dev->features;
Herbert Xu37c31852006-06-22 03:07:29 -07001090 if (sk->sk_route_caps & NETIF_F_GSO)
Herbert Xubcd76112006-06-30 13:36:35 -07001091 sk->sk_route_caps |= NETIF_F_GSO_MASK;
1092 if (sk_can_gso(sk)) {
Michael Chanb0da85372006-06-29 12:30:00 -07001093 if (dst->header_len)
Herbert Xubcd76112006-06-30 13:36:35 -07001094 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
Herbert Xu37c31852006-06-22 03:07:29 -07001095 else
1096 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001097 }
1098}
1099
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1101{
1102 sk->sk_wmem_queued += skb->truesize;
1103 sk->sk_forward_alloc -= skb->truesize;
1104}
1105
1106static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1107 struct sk_buff *skb, struct page *page,
1108 int off, int copy)
1109{
1110 if (skb->ip_summed == CHECKSUM_NONE) {
1111 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001112 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 page_address(page) + off,
1114 copy, 0, &err);
1115 if (err)
1116 return err;
1117 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1118 } else if (copy_from_user(page_address(page) + off, from, copy))
1119 return -EFAULT;
1120
1121 skb->len += copy;
1122 skb->data_len += copy;
1123 skb->truesize += copy;
1124 sk->sk_wmem_queued += copy;
1125 sk->sk_forward_alloc -= copy;
1126 return 0;
1127}
1128
1129/*
1130 * Queue a received datagram if it will fit. Stream and sequenced
1131 * protocols can't normally use this as they need to fit buffers in
1132 * and play with them.
1133 *
1134 * Inlined as it's very short and called for pretty much every
1135 * packet ever received.
1136 */
1137
1138static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1139{
1140 sock_hold(sk);
1141 skb->sk = sk;
1142 skb->destructor = sock_wfree;
1143 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1144}
1145
1146static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1147{
1148 skb->sk = sk;
1149 skb->destructor = sock_rfree;
1150 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1151}
1152
1153extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1154 unsigned long expires);
1155
1156extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1157
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001158extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159
1160static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1161{
1162 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1163 number of warnings when compiling with -W --ANK
1164 */
1165 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1166 (unsigned)sk->sk_rcvbuf)
1167 return -ENOMEM;
1168 skb_set_owner_r(skb, sk);
1169 skb_queue_tail(&sk->sk_error_queue, skb);
1170 if (!sock_flag(sk, SOCK_DEAD))
1171 sk->sk_data_ready(sk, skb->len);
1172 return 0;
1173}
1174
1175/*
1176 * Recover an error report and clear atomically
1177 */
1178
1179static inline int sock_error(struct sock *sk)
1180{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001181 int err;
1182 if (likely(!sk->sk_err))
1183 return 0;
1184 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 return -err;
1186}
1187
1188static inline unsigned long sock_wspace(struct sock *sk)
1189{
1190 int amt = 0;
1191
1192 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1193 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1194 if (amt < 0)
1195 amt = 0;
1196 }
1197 return amt;
1198}
1199
1200static inline void sk_wake_async(struct sock *sk, int how, int band)
1201{
1202 if (sk->sk_socket && sk->sk_socket->fasync_list)
1203 sock_wake_async(sk->sk_socket, how, band);
1204}
1205
1206#define SOCK_MIN_SNDBUF 2048
1207#define SOCK_MIN_RCVBUF 256
1208
1209static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1210{
1211 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1212 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1213 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1214 }
1215}
1216
1217static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
Victor Fusco86a76ca2005-07-08 14:57:47 -07001218 int size, int mem,
Al Virodd0fc662005-10-07 07:46:04 +01001219 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220{
David S. Millerc65f7f02005-07-05 15:17:25 -07001221 struct sk_buff *skb;
1222 int hdr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223
David S. Millerc65f7f02005-07-05 15:17:25 -07001224 hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header);
David S. Millerd179cd12005-08-17 14:57:30 -07001225 skb = alloc_skb_fclone(size + hdr_len, gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 if (skb) {
1227 skb->truesize += mem;
Herbert Xud80d99d2005-09-01 17:48:23 -07001228 if (sk_stream_wmem_schedule(sk, skb->truesize)) {
David S. Millerc65f7f02005-07-05 15:17:25 -07001229 skb_reserve(skb, hdr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 return skb;
1231 }
1232 __kfree_skb(skb);
1233 } else {
1234 sk->sk_prot->enter_memory_pressure();
1235 sk_stream_moderate_sndbuf(sk);
1236 }
1237 return NULL;
1238}
1239
1240static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
Victor Fusco86a76ca2005-07-08 14:57:47 -07001241 int size,
Al Virodd0fc662005-10-07 07:46:04 +01001242 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243{
1244 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1245}
1246
1247static inline struct page *sk_stream_alloc_page(struct sock *sk)
1248{
1249 struct page *page = NULL;
1250
Herbert Xuef015782005-09-01 17:48:59 -07001251 page = alloc_pages(sk->sk_allocation, 0);
1252 if (!page) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 sk->sk_prot->enter_memory_pressure();
1254 sk_stream_moderate_sndbuf(sk);
1255 }
1256 return page;
1257}
1258
1259#define sk_stream_for_retrans_queue(skb, sk) \
1260 for (skb = (sk)->sk_write_queue.next; \
1261 (skb != (sk)->sk_send_head) && \
1262 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1263 skb = skb->next)
1264
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001265/*from STCP for fast SACK Process*/
1266#define sk_stream_for_retrans_queue_from(skb, sk) \
1267 for (; (skb != (sk)->sk_send_head) && \
1268 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1269 skb = skb->next)
1270
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271/*
1272 * Default write policy as shown to user space via poll/select/SIGIO
1273 */
1274static inline int sock_writeable(const struct sock *sk)
1275{
1276 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1277}
1278
Al Virodd0fc662005-10-07 07:46:04 +01001279static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280{
1281 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1282}
1283
1284static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1285{
1286 return noblock ? 0 : sk->sk_rcvtimeo;
1287}
1288
1289static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1290{
1291 return noblock ? 0 : sk->sk_sndtimeo;
1292}
1293
1294static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1295{
1296 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1297}
1298
1299/* Alas, with timeout socket operations are not restartable.
1300 * Compare this to poll().
1301 */
1302static inline int sock_intr_errno(long timeo)
1303{
1304 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1305}
1306
1307static __inline__ void
1308sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1309{
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001310 struct timeval stamp;
1311
1312 skb_get_timestamp(skb, &stamp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 if (sock_flag(sk, SOCK_RCVTSTAMP)) {
1314 /* Race occurred between timestamp enabling and packet
1315 receiving. Fill in the current time for now. */
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001316 if (stamp.tv_sec == 0)
1317 do_gettimeofday(&stamp);
1318 skb_set_timestamp(skb, &stamp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001320 &stamp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 } else
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001322 sk->sk_stamp = stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323}
1324
1325/**
1326 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001327 * @sk: socket to eat this skb from
1328 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07001329 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 *
1331 * This routine must be called with interrupts disabled or with the socket
1332 * locked so that the sk_buff queue operation is ok.
1333*/
Chris Leech624d1162006-05-23 18:01:28 -07001334#ifdef CONFIG_NET_DMA
1335static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1336{
1337 __skb_unlink(skb, &sk->sk_receive_queue);
1338 if (!copied_early)
1339 __kfree_skb(skb);
1340 else
1341 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1342}
1343#else
1344static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345{
1346 __skb_unlink(skb, &sk->sk_receive_queue);
1347 __kfree_skb(skb);
1348}
Chris Leech624d1162006-05-23 18:01:28 -07001349#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350
1351extern void sock_enable_timestamp(struct sock *sk);
1352extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1353
1354/*
1355 * Enable debug/info messages
1356 */
1357
Stephen Hemminger0dec4562006-02-02 20:40:09 -08001358#ifdef CONFIG_NETDEBUG
Patrick McHardy64ce2072005-08-09 20:50:53 -07001359#define NETDEBUG(fmt, args...) printk(fmt,##args)
1360#define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0)
Stephen Hemminger0dec4562006-02-02 20:40:09 -08001361#else
1362#define NETDEBUG(fmt, args...) do { } while (0)
1363#define LIMIT_NETDEBUG(fmt, args...) do { } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364#endif
1365
1366/*
1367 * Macros for sleeping on a socket. Use them like this:
1368 *
1369 * SOCK_SLEEP_PRE(sk)
1370 * if (condition)
1371 * schedule();
1372 * SOCK_SLEEP_POST(sk)
1373 *
1374 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1375 * and when the last use of them in DECnet has gone, I'm intending to
1376 * remove them.
1377 */
1378
1379#define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1380 DECLARE_WAITQUEUE(wait, tsk); \
1381 tsk->state = TASK_INTERRUPTIBLE; \
1382 add_wait_queue((sk)->sk_sleep, &wait); \
1383 release_sock(sk);
1384
1385#define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1386 remove_wait_queue((sk)->sk_sleep, &wait); \
1387 lock_sock(sk); \
1388 }
1389
1390static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1391{
1392 if (valbool)
1393 sock_set_flag(sk, bit);
1394 else
1395 sock_reset_flag(sk, bit);
1396}
1397
1398extern __u32 sysctl_wmem_max;
1399extern __u32 sysctl_rmem_max;
1400
1401#ifdef CONFIG_NET
1402int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1403#else
1404static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
1405{
1406 return -ENODEV;
1407}
1408#endif
1409
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001410extern void sk_init(void);
1411
1412#ifdef CONFIG_SYSCTL
1413extern struct ctl_table core_table[];
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001414#endif
1415
David S. Miller6baf1f42005-09-05 18:14:11 -07001416extern int sysctl_optmem_max;
1417
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001418extern __u32 sysctl_wmem_default;
1419extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001420
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421#endif /* _SOCK_H */