xref: /linux/include/net/sock.h (revision 363c55cae53742f3f685a1814912c6d4fda245b4)
1 /*
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  *
10  * Authors:	Ross Biro
11  *		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 
43 #include <linux/kernel.h>
44 #include <linux/list.h>
45 #include <linux/list_nulls.h>
46 #include <linux/timer.h>
47 #include <linux/cache.h>
48 #include <linux/module.h>
49 #include <linux/lockdep.h>
50 #include <linux/netdevice.h>
51 #include <linux/skbuff.h>	/* struct sk_buff */
52 #include <linux/mm.h>
53 #include <linux/security.h>
54 
55 #include <linux/filter.h>
56 #include <linux/rculist_nulls.h>
57 
58 #include <asm/atomic.h>
59 #include <net/dst.h>
60 #include <net/checksum.h>
61 
62 /*
63  * This structure really needs to be cleaned up.
64  * Most of it is for TCP, and not used by any of
65  * the other protocols.
66  */
67 
68 /* Define this to get the SOCK_DBG debugging facility. */
69 #define SOCK_DEBUGGING
70 #ifdef SOCK_DEBUGGING
71 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
72 					printk(KERN_DEBUG msg); } while (0)
73 #else
74 /* Validate arguments and do nothing */
75 static void inline int __attribute__ ((format (printf, 2, 3)))
76 SOCK_DEBUG(struct sock *sk, const char *msg, ...)
77 {
78 }
79 #endif
80 
81 /* This is the per-socket lock.  The spinlock provides a synchronization
82  * between user contexts and software interrupt processing, whereas the
83  * mini-semaphore synchronizes multiple users amongst themselves.
84  */
85 typedef struct {
86 	spinlock_t		slock;
87 	int			owned;
88 	wait_queue_head_t	wq;
89 	/*
90 	 * We express the mutex-alike socket_lock semantics
91 	 * to the lock validator by explicitly managing
92 	 * the slock as a lock variant (in addition to
93 	 * the slock itself):
94 	 */
95 #ifdef CONFIG_DEBUG_LOCK_ALLOC
96 	struct lockdep_map dep_map;
97 #endif
98 } socket_lock_t;
99 
100 struct sock;
101 struct proto;
102 struct net;
103 
104 /**
105  *	struct sock_common - minimal network layer representation of sockets
106  *	@skc_family: network address family
107  *	@skc_state: Connection state
108  *	@skc_reuse: %SO_REUSEADDR setting
109  *	@skc_bound_dev_if: bound device index if != 0
110  *	@skc_node: main hash linkage for various protocol lookup tables
111  *	@skc_nulls_node: main hash linkage for UDP/UDP-Lite protocol
112  *	@skc_bind_node: bind hash linkage for various protocol lookup tables
113  *	@skc_refcnt: reference count
114  *	@skc_hash: hash value used with various protocol lookup tables
115  *	@skc_prot: protocol handlers inside a network family
116  *	@skc_net: reference to the network namespace of this socket
117  *
118  *	This is the minimal network layer representation of sockets, the header
119  *	for struct sock and struct inet_timewait_sock.
120  */
121 struct sock_common {
122 	unsigned short		skc_family;
123 	volatile unsigned char	skc_state;
124 	unsigned char		skc_reuse;
125 	int			skc_bound_dev_if;
126 	union {
127 		struct hlist_node	skc_node;
128 		struct hlist_nulls_node skc_nulls_node;
129 	};
130 	struct hlist_node	skc_bind_node;
131 	atomic_t		skc_refcnt;
132 	unsigned int		skc_hash;
133 	struct proto		*skc_prot;
134 #ifdef CONFIG_NET_NS
135 	struct net	 	*skc_net;
136 #endif
137 };
138 
139 /**
140   *	struct sock - network layer representation of sockets
141   *	@__sk_common: shared layout with inet_timewait_sock
142   *	@sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
143   *	@sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
144   *	@sk_lock:	synchronizer
145   *	@sk_rcvbuf: size of receive buffer in bytes
146   *	@sk_sleep: sock wait queue
147   *	@sk_dst_cache: destination cache
148   *	@sk_dst_lock: destination cache lock
149   *	@sk_policy: flow policy
150   *	@sk_rmem_alloc: receive queue bytes committed
151   *	@sk_receive_queue: incoming packets
152   *	@sk_wmem_alloc: transmit queue bytes committed
153   *	@sk_write_queue: Packet sending queue
154   *	@sk_async_wait_queue: DMA copied packets
155   *	@sk_omem_alloc: "o" is "option" or "other"
156   *	@sk_wmem_queued: persistent queue size
157   *	@sk_forward_alloc: space allocated forward
158   *	@sk_allocation: allocation mode
159   *	@sk_sndbuf: size of send buffer in bytes
160   *	@sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
161   *		   %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
162   *	@sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
163   *	@sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
164   *	@sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
165   *	@sk_gso_max_size: Maximum GSO segment size to build
166   *	@sk_lingertime: %SO_LINGER l_linger setting
167   *	@sk_backlog: always used with the per-socket spinlock held
168   *	@sk_callback_lock: used with the callbacks in the end of this struct
169   *	@sk_error_queue: rarely used
170   *	@sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
171   *			  IPV6_ADDRFORM for instance)
172   *	@sk_err: last error
173   *	@sk_err_soft: errors that don't cause failure but are the cause of a
174   *		      persistent failure not just 'timed out'
175   *	@sk_drops: raw/udp drops counter
176   *	@sk_ack_backlog: current listen backlog
177   *	@sk_max_ack_backlog: listen backlog set in listen()
178   *	@sk_priority: %SO_PRIORITY setting
179   *	@sk_type: socket type (%SOCK_STREAM, etc)
180   *	@sk_protocol: which protocol this socket belongs in this network family
181   *	@sk_peercred: %SO_PEERCRED setting
182   *	@sk_rcvlowat: %SO_RCVLOWAT setting
183   *	@sk_rcvtimeo: %SO_RCVTIMEO setting
184   *	@sk_sndtimeo: %SO_SNDTIMEO setting
185   *	@sk_filter: socket filtering instructions
186   *	@sk_protinfo: private area, net family specific, when not using slab
187   *	@sk_timer: sock cleanup timer
188   *	@sk_stamp: time stamp of last packet received
189   *	@sk_socket: Identd and reporting IO signals
190   *	@sk_user_data: RPC layer private data
191   *	@sk_sndmsg_page: cached page for sendmsg
192   *	@sk_sndmsg_off: cached offset for sendmsg
193   *	@sk_send_head: front of stuff to transmit
194   *	@sk_security: used by security modules
195   *	@sk_mark: generic packet mark
196   *	@sk_write_pending: a write to stream socket waits to start
197   *	@sk_state_change: callback to indicate change in the state of the sock
198   *	@sk_data_ready: callback to indicate there is data to be processed
199   *	@sk_write_space: callback to indicate there is bf sending space available
200   *	@sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
201   *	@sk_backlog_rcv: callback to process the backlog
202   *	@sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
203  */
204 struct sock {
205 	/*
206 	 * Now struct inet_timewait_sock also uses sock_common, so please just
207 	 * don't add nothing before this first member (__sk_common) --acme
208 	 */
209 	struct sock_common	__sk_common;
210 #define sk_family		__sk_common.skc_family
211 #define sk_state		__sk_common.skc_state
212 #define sk_reuse		__sk_common.skc_reuse
213 #define sk_bound_dev_if		__sk_common.skc_bound_dev_if
214 #define sk_node			__sk_common.skc_node
215 #define sk_nulls_node		__sk_common.skc_nulls_node
216 #define sk_bind_node		__sk_common.skc_bind_node
217 #define sk_refcnt		__sk_common.skc_refcnt
218 #define sk_hash			__sk_common.skc_hash
219 #define sk_prot			__sk_common.skc_prot
220 #define sk_net			__sk_common.skc_net
221 	kmemcheck_bitfield_begin(flags);
222 	unsigned char		sk_shutdown : 2,
223 				sk_no_check : 2,
224 				sk_userlocks : 4;
225 	kmemcheck_bitfield_end(flags);
226 	unsigned char		sk_protocol;
227 	unsigned short		sk_type;
228 	int			sk_rcvbuf;
229 	socket_lock_t		sk_lock;
230 	/*
231 	 * The backlog queue is special, it is always used with
232 	 * the per-socket spinlock held and requires low latency
233 	 * access. Therefore we special case it's implementation.
234 	 */
235 	struct {
236 		struct sk_buff *head;
237 		struct sk_buff *tail;
238 	} sk_backlog;
239 	wait_queue_head_t	*sk_sleep;
240 	struct dst_entry	*sk_dst_cache;
241 #ifdef CONFIG_XFRM
242 	struct xfrm_policy	*sk_policy[2];
243 #endif
244 	rwlock_t		sk_dst_lock;
245 	atomic_t		sk_rmem_alloc;
246 	atomic_t		sk_wmem_alloc;
247 	atomic_t		sk_omem_alloc;
248 	int			sk_sndbuf;
249 	struct sk_buff_head	sk_receive_queue;
250 	struct sk_buff_head	sk_write_queue;
251 #ifdef CONFIG_NET_DMA
252 	struct sk_buff_head	sk_async_wait_queue;
253 #endif
254 	int			sk_wmem_queued;
255 	int			sk_forward_alloc;
256 	gfp_t			sk_allocation;
257 	int			sk_route_caps;
258 	int			sk_gso_type;
259 	unsigned int		sk_gso_max_size;
260 	int			sk_rcvlowat;
261 	unsigned long 		sk_flags;
262 	unsigned long	        sk_lingertime;
263 	struct sk_buff_head	sk_error_queue;
264 	struct proto		*sk_prot_creator;
265 	rwlock_t		sk_callback_lock;
266 	int			sk_err,
267 				sk_err_soft;
268 	atomic_t		sk_drops;
269 	unsigned short		sk_ack_backlog;
270 	unsigned short		sk_max_ack_backlog;
271 	__u32			sk_priority;
272 	struct ucred		sk_peercred;
273 	long			sk_rcvtimeo;
274 	long			sk_sndtimeo;
275 	struct sk_filter      	*sk_filter;
276 	void			*sk_protinfo;
277 	struct timer_list	sk_timer;
278 	ktime_t			sk_stamp;
279 	struct socket		*sk_socket;
280 	void			*sk_user_data;
281 	struct page		*sk_sndmsg_page;
282 	struct sk_buff		*sk_send_head;
283 	__u32			sk_sndmsg_off;
284 	int			sk_write_pending;
285 #ifdef CONFIG_SECURITY
286 	void			*sk_security;
287 #endif
288 	__u32			sk_mark;
289 	/* XXX 4 bytes hole on 64 bit */
290 	void			(*sk_state_change)(struct sock *sk);
291 	void			(*sk_data_ready)(struct sock *sk, int bytes);
292 	void			(*sk_write_space)(struct sock *sk);
293 	void			(*sk_error_report)(struct sock *sk);
294   	int			(*sk_backlog_rcv)(struct sock *sk,
295 						  struct sk_buff *skb);
296 	void                    (*sk_destruct)(struct sock *sk);
297 };
298 
299 /*
300  * Hashed lists helper routines
301  */
302 static inline struct sock *__sk_head(const struct hlist_head *head)
303 {
304 	return hlist_entry(head->first, struct sock, sk_node);
305 }
306 
307 static inline struct sock *sk_head(const struct hlist_head *head)
308 {
309 	return hlist_empty(head) ? NULL : __sk_head(head);
310 }
311 
312 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
313 {
314 	return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
315 }
316 
317 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
318 {
319 	return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
320 }
321 
322 static inline struct sock *sk_next(const struct sock *sk)
323 {
324 	return sk->sk_node.next ?
325 		hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
326 }
327 
328 static inline struct sock *sk_nulls_next(const struct sock *sk)
329 {
330 	return (!is_a_nulls(sk->sk_nulls_node.next)) ?
331 		hlist_nulls_entry(sk->sk_nulls_node.next,
332 				  struct sock, sk_nulls_node) :
333 		NULL;
334 }
335 
336 static inline int sk_unhashed(const struct sock *sk)
337 {
338 	return hlist_unhashed(&sk->sk_node);
339 }
340 
341 static inline int sk_hashed(const struct sock *sk)
342 {
343 	return !sk_unhashed(sk);
344 }
345 
346 static __inline__ void sk_node_init(struct hlist_node *node)
347 {
348 	node->pprev = NULL;
349 }
350 
351 static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
352 {
353 	node->pprev = NULL;
354 }
355 
356 static __inline__ void __sk_del_node(struct sock *sk)
357 {
358 	__hlist_del(&sk->sk_node);
359 }
360 
361 static __inline__ int __sk_del_node_init(struct sock *sk)
362 {
363 	if (sk_hashed(sk)) {
364 		__sk_del_node(sk);
365 		sk_node_init(&sk->sk_node);
366 		return 1;
367 	}
368 	return 0;
369 }
370 
371 /* Grab socket reference count. This operation is valid only
372    when sk is ALREADY grabbed f.e. it is found in hash table
373    or a list and the lookup is made under lock preventing hash table
374    modifications.
375  */
376 
377 static inline void sock_hold(struct sock *sk)
378 {
379 	atomic_inc(&sk->sk_refcnt);
380 }
381 
382 /* Ungrab socket in the context, which assumes that socket refcnt
383    cannot hit zero, f.e. it is true in context of any socketcall.
384  */
385 static inline void __sock_put(struct sock *sk)
386 {
387 	atomic_dec(&sk->sk_refcnt);
388 }
389 
390 static __inline__ int sk_del_node_init(struct sock *sk)
391 {
392 	int rc = __sk_del_node_init(sk);
393 
394 	if (rc) {
395 		/* paranoid for a while -acme */
396 		WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
397 		__sock_put(sk);
398 	}
399 	return rc;
400 }
401 
402 static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
403 {
404 	if (sk_hashed(sk)) {
405 		hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
406 		return 1;
407 	}
408 	return 0;
409 }
410 
411 static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
412 {
413 	int rc = __sk_nulls_del_node_init_rcu(sk);
414 
415 	if (rc) {
416 		/* paranoid for a while -acme */
417 		WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
418 		__sock_put(sk);
419 	}
420 	return rc;
421 }
422 
423 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
424 {
425 	hlist_add_head(&sk->sk_node, list);
426 }
427 
428 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
429 {
430 	sock_hold(sk);
431 	__sk_add_node(sk, list);
432 }
433 
434 static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
435 {
436 	hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
437 }
438 
439 static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
440 {
441 	sock_hold(sk);
442 	__sk_nulls_add_node_rcu(sk, list);
443 }
444 
445 static __inline__ void __sk_del_bind_node(struct sock *sk)
446 {
447 	__hlist_del(&sk->sk_bind_node);
448 }
449 
450 static __inline__ void sk_add_bind_node(struct sock *sk,
451 					struct hlist_head *list)
452 {
453 	hlist_add_head(&sk->sk_bind_node, list);
454 }
455 
456 #define sk_for_each(__sk, node, list) \
457 	hlist_for_each_entry(__sk, node, list, sk_node)
458 #define sk_nulls_for_each(__sk, node, list) \
459 	hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
460 #define sk_nulls_for_each_rcu(__sk, node, list) \
461 	hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
462 #define sk_for_each_from(__sk, node) \
463 	if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
464 		hlist_for_each_entry_from(__sk, node, sk_node)
465 #define sk_nulls_for_each_from(__sk, node) \
466 	if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
467 		hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
468 #define sk_for_each_continue(__sk, node) \
469 	if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
470 		hlist_for_each_entry_continue(__sk, node, sk_node)
471 #define sk_for_each_safe(__sk, node, tmp, list) \
472 	hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
473 #define sk_for_each_bound(__sk, node, list) \
474 	hlist_for_each_entry(__sk, node, list, sk_bind_node)
475 
476 /* Sock flags */
477 enum sock_flags {
478 	SOCK_DEAD,
479 	SOCK_DONE,
480 	SOCK_URGINLINE,
481 	SOCK_KEEPOPEN,
482 	SOCK_LINGER,
483 	SOCK_DESTROY,
484 	SOCK_BROADCAST,
485 	SOCK_TIMESTAMP,
486 	SOCK_ZAPPED,
487 	SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
488 	SOCK_DBG, /* %SO_DEBUG setting */
489 	SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
490 	SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
491 	SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
492 	SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
493 	SOCK_TIMESTAMPING_TX_HARDWARE,  /* %SOF_TIMESTAMPING_TX_HARDWARE */
494 	SOCK_TIMESTAMPING_TX_SOFTWARE,  /* %SOF_TIMESTAMPING_TX_SOFTWARE */
495 	SOCK_TIMESTAMPING_RX_HARDWARE,  /* %SOF_TIMESTAMPING_RX_HARDWARE */
496 	SOCK_TIMESTAMPING_RX_SOFTWARE,  /* %SOF_TIMESTAMPING_RX_SOFTWARE */
497 	SOCK_TIMESTAMPING_SOFTWARE,     /* %SOF_TIMESTAMPING_SOFTWARE */
498 	SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
499 	SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
500 };
501 
502 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
503 {
504 	nsk->sk_flags = osk->sk_flags;
505 }
506 
507 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
508 {
509 	__set_bit(flag, &sk->sk_flags);
510 }
511 
512 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
513 {
514 	__clear_bit(flag, &sk->sk_flags);
515 }
516 
517 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
518 {
519 	return test_bit(flag, &sk->sk_flags);
520 }
521 
522 static inline void sk_acceptq_removed(struct sock *sk)
523 {
524 	sk->sk_ack_backlog--;
525 }
526 
527 static inline void sk_acceptq_added(struct sock *sk)
528 {
529 	sk->sk_ack_backlog++;
530 }
531 
532 static inline int sk_acceptq_is_full(struct sock *sk)
533 {
534 	return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
535 }
536 
537 /*
538  * Compute minimal free write space needed to queue new packets.
539  */
540 static inline int sk_stream_min_wspace(struct sock *sk)
541 {
542 	return sk->sk_wmem_queued >> 1;
543 }
544 
545 static inline int sk_stream_wspace(struct sock *sk)
546 {
547 	return sk->sk_sndbuf - sk->sk_wmem_queued;
548 }
549 
550 extern void sk_stream_write_space(struct sock *sk);
551 
552 static inline int sk_stream_memory_free(struct sock *sk)
553 {
554 	return sk->sk_wmem_queued < sk->sk_sndbuf;
555 }
556 
557 /* The per-socket spinlock must be held here. */
558 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
559 {
560 	if (!sk->sk_backlog.tail) {
561 		sk->sk_backlog.head = sk->sk_backlog.tail = skb;
562 	} else {
563 		sk->sk_backlog.tail->next = skb;
564 		sk->sk_backlog.tail = skb;
565 	}
566 	skb->next = NULL;
567 }
568 
569 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
570 {
571 	return sk->sk_backlog_rcv(sk, skb);
572 }
573 
574 #define sk_wait_event(__sk, __timeo, __condition)			\
575 	({	int __rc;						\
576 		release_sock(__sk);					\
577 		__rc = __condition;					\
578 		if (!__rc) {						\
579 			*(__timeo) = schedule_timeout(*(__timeo));	\
580 		}							\
581 		lock_sock(__sk);					\
582 		__rc = __condition;					\
583 		__rc;							\
584 	})
585 
586 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
587 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
588 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
589 extern int sk_stream_error(struct sock *sk, int flags, int err);
590 extern void sk_stream_kill_queues(struct sock *sk);
591 
592 extern int sk_wait_data(struct sock *sk, long *timeo);
593 
594 struct request_sock_ops;
595 struct timewait_sock_ops;
596 struct inet_hashinfo;
597 struct raw_hashinfo;
598 
599 /* Networking protocol blocks we attach to sockets.
600  * socket layer -> transport layer interface
601  * transport -> network interface is defined by struct inet_proto
602  */
603 struct proto {
604 	void			(*close)(struct sock *sk,
605 					long timeout);
606 	int			(*connect)(struct sock *sk,
607 				        struct sockaddr *uaddr,
608 					int addr_len);
609 	int			(*disconnect)(struct sock *sk, int flags);
610 
611 	struct sock *		(*accept) (struct sock *sk, int flags, int *err);
612 
613 	int			(*ioctl)(struct sock *sk, int cmd,
614 					 unsigned long arg);
615 	int			(*init)(struct sock *sk);
616 	void			(*destroy)(struct sock *sk);
617 	void			(*shutdown)(struct sock *sk, int how);
618 	int			(*setsockopt)(struct sock *sk, int level,
619 					int optname, char __user *optval,
620 					int optlen);
621 	int			(*getsockopt)(struct sock *sk, int level,
622 					int optname, char __user *optval,
623 					int __user *option);
624 #ifdef CONFIG_COMPAT
625 	int			(*compat_setsockopt)(struct sock *sk,
626 					int level,
627 					int optname, char __user *optval,
628 					int optlen);
629 	int			(*compat_getsockopt)(struct sock *sk,
630 					int level,
631 					int optname, char __user *optval,
632 					int __user *option);
633 #endif
634 	int			(*sendmsg)(struct kiocb *iocb, struct sock *sk,
635 					   struct msghdr *msg, size_t len);
636 	int			(*recvmsg)(struct kiocb *iocb, struct sock *sk,
637 					   struct msghdr *msg,
638 					size_t len, int noblock, int flags,
639 					int *addr_len);
640 	int			(*sendpage)(struct sock *sk, struct page *page,
641 					int offset, size_t size, int flags);
642 	int			(*bind)(struct sock *sk,
643 					struct sockaddr *uaddr, int addr_len);
644 
645 	int			(*backlog_rcv) (struct sock *sk,
646 						struct sk_buff *skb);
647 
648 	/* Keeping track of sk's, looking them up, and port selection methods. */
649 	void			(*hash)(struct sock *sk);
650 	void			(*unhash)(struct sock *sk);
651 	int			(*get_port)(struct sock *sk, unsigned short snum);
652 
653 	/* Keeping track of sockets in use */
654 #ifdef CONFIG_PROC_FS
655 	unsigned int		inuse_idx;
656 #endif
657 
658 	/* Memory pressure */
659 	void			(*enter_memory_pressure)(struct sock *sk);
660 	atomic_t		*memory_allocated;	/* Current allocated memory. */
661 	struct percpu_counter	*sockets_allocated;	/* Current number of sockets. */
662 	/*
663 	 * Pressure flag: try to collapse.
664 	 * Technical note: it is used by multiple contexts non atomically.
665 	 * All the __sk_mem_schedule() is of this nature: accounting
666 	 * is strict, actions are advisory and have some latency.
667 	 */
668 	int			*memory_pressure;
669 	int			*sysctl_mem;
670 	int			*sysctl_wmem;
671 	int			*sysctl_rmem;
672 	int			max_header;
673 
674 	struct kmem_cache	*slab;
675 	unsigned int		obj_size;
676 	int			slab_flags;
677 
678 	struct percpu_counter	*orphan_count;
679 
680 	struct request_sock_ops	*rsk_prot;
681 	struct timewait_sock_ops *twsk_prot;
682 
683 	union {
684 		struct inet_hashinfo	*hashinfo;
685 		struct udp_table	*udp_table;
686 		struct raw_hashinfo	*raw_hash;
687 	} h;
688 
689 	struct module		*owner;
690 
691 	char			name[32];
692 
693 	struct list_head	node;
694 #ifdef SOCK_REFCNT_DEBUG
695 	atomic_t		socks;
696 #endif
697 };
698 
699 extern int proto_register(struct proto *prot, int alloc_slab);
700 extern void proto_unregister(struct proto *prot);
701 
702 #ifdef SOCK_REFCNT_DEBUG
703 static inline void sk_refcnt_debug_inc(struct sock *sk)
704 {
705 	atomic_inc(&sk->sk_prot->socks);
706 }
707 
708 static inline void sk_refcnt_debug_dec(struct sock *sk)
709 {
710 	atomic_dec(&sk->sk_prot->socks);
711 	printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
712 	       sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
713 }
714 
715 static inline void sk_refcnt_debug_release(const struct sock *sk)
716 {
717 	if (atomic_read(&sk->sk_refcnt) != 1)
718 		printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
719 		       sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
720 }
721 #else /* SOCK_REFCNT_DEBUG */
722 #define sk_refcnt_debug_inc(sk) do { } while (0)
723 #define sk_refcnt_debug_dec(sk) do { } while (0)
724 #define sk_refcnt_debug_release(sk) do { } while (0)
725 #endif /* SOCK_REFCNT_DEBUG */
726 
727 
728 #ifdef CONFIG_PROC_FS
729 /* Called with local bh disabled */
730 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
731 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
732 #else
733 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
734 		int inc)
735 {
736 }
737 #endif
738 
739 
740 /* With per-bucket locks this operation is not-atomic, so that
741  * this version is not worse.
742  */
743 static inline void __sk_prot_rehash(struct sock *sk)
744 {
745 	sk->sk_prot->unhash(sk);
746 	sk->sk_prot->hash(sk);
747 }
748 
749 /* About 10 seconds */
750 #define SOCK_DESTROY_TIME (10*HZ)
751 
752 /* Sockets 0-1023 can't be bound to unless you are superuser */
753 #define PROT_SOCK	1024
754 
755 #define SHUTDOWN_MASK	3
756 #define RCV_SHUTDOWN	1
757 #define SEND_SHUTDOWN	2
758 
759 #define SOCK_SNDBUF_LOCK	1
760 #define SOCK_RCVBUF_LOCK	2
761 #define SOCK_BINDADDR_LOCK	4
762 #define SOCK_BINDPORT_LOCK	8
763 
764 /* sock_iocb: used to kick off async processing of socket ios */
765 struct sock_iocb {
766 	struct list_head	list;
767 
768 	int			flags;
769 	int			size;
770 	struct socket		*sock;
771 	struct sock		*sk;
772 	struct scm_cookie	*scm;
773 	struct msghdr		*msg, async_msg;
774 	struct kiocb		*kiocb;
775 };
776 
777 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
778 {
779 	return (struct sock_iocb *)iocb->private;
780 }
781 
782 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
783 {
784 	return si->kiocb;
785 }
786 
787 struct socket_alloc {
788 	struct socket socket;
789 	struct inode vfs_inode;
790 };
791 
792 static inline struct socket *SOCKET_I(struct inode *inode)
793 {
794 	return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
795 }
796 
797 static inline struct inode *SOCK_INODE(struct socket *socket)
798 {
799 	return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
800 }
801 
802 /*
803  * Functions for memory accounting
804  */
805 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
806 extern void __sk_mem_reclaim(struct sock *sk);
807 
808 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
809 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
810 #define SK_MEM_SEND	0
811 #define SK_MEM_RECV	1
812 
813 static inline int sk_mem_pages(int amt)
814 {
815 	return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
816 }
817 
818 static inline int sk_has_account(struct sock *sk)
819 {
820 	/* return true if protocol supports memory accounting */
821 	return !!sk->sk_prot->memory_allocated;
822 }
823 
824 static inline int sk_wmem_schedule(struct sock *sk, int size)
825 {
826 	if (!sk_has_account(sk))
827 		return 1;
828 	return size <= sk->sk_forward_alloc ||
829 		__sk_mem_schedule(sk, size, SK_MEM_SEND);
830 }
831 
832 static inline int sk_rmem_schedule(struct sock *sk, int size)
833 {
834 	if (!sk_has_account(sk))
835 		return 1;
836 	return size <= sk->sk_forward_alloc ||
837 		__sk_mem_schedule(sk, size, SK_MEM_RECV);
838 }
839 
840 static inline void sk_mem_reclaim(struct sock *sk)
841 {
842 	if (!sk_has_account(sk))
843 		return;
844 	if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
845 		__sk_mem_reclaim(sk);
846 }
847 
848 static inline void sk_mem_reclaim_partial(struct sock *sk)
849 {
850 	if (!sk_has_account(sk))
851 		return;
852 	if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
853 		__sk_mem_reclaim(sk);
854 }
855 
856 static inline void sk_mem_charge(struct sock *sk, int size)
857 {
858 	if (!sk_has_account(sk))
859 		return;
860 	sk->sk_forward_alloc -= size;
861 }
862 
863 static inline void sk_mem_uncharge(struct sock *sk, int size)
864 {
865 	if (!sk_has_account(sk))
866 		return;
867 	sk->sk_forward_alloc += size;
868 }
869 
870 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
871 {
872 	sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
873 	sk->sk_wmem_queued -= skb->truesize;
874 	sk_mem_uncharge(sk, skb->truesize);
875 	__kfree_skb(skb);
876 }
877 
878 /* Used by processes to "lock" a socket state, so that
879  * interrupts and bottom half handlers won't change it
880  * from under us. It essentially blocks any incoming
881  * packets, so that we won't get any new data or any
882  * packets that change the state of the socket.
883  *
884  * While locked, BH processing will add new packets to
885  * the backlog queue.  This queue is processed by the
886  * owner of the socket lock right before it is released.
887  *
888  * Since ~2.3.5 it is also exclusive sleep lock serializing
889  * accesses from user process context.
890  */
891 #define sock_owned_by_user(sk)	((sk)->sk_lock.owned)
892 
893 /*
894  * Macro so as to not evaluate some arguments when
895  * lockdep is not enabled.
896  *
897  * Mark both the sk_lock and the sk_lock.slock as a
898  * per-address-family lock class.
899  */
900 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) 	\
901 do {									\
902 	sk->sk_lock.owned = 0;						\
903 	init_waitqueue_head(&sk->sk_lock.wq);				\
904 	spin_lock_init(&(sk)->sk_lock.slock);				\
905 	debug_check_no_locks_freed((void *)&(sk)->sk_lock,		\
906 			sizeof((sk)->sk_lock));				\
907 	lockdep_set_class_and_name(&(sk)->sk_lock.slock,		\
908 		       	(skey), (sname));				\
909 	lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0);	\
910 } while (0)
911 
912 extern void lock_sock_nested(struct sock *sk, int subclass);
913 
914 static inline void lock_sock(struct sock *sk)
915 {
916 	lock_sock_nested(sk, 0);
917 }
918 
919 extern void release_sock(struct sock *sk);
920 
921 /* BH context may only use the following locking interface. */
922 #define bh_lock_sock(__sk)	spin_lock(&((__sk)->sk_lock.slock))
923 #define bh_lock_sock_nested(__sk) \
924 				spin_lock_nested(&((__sk)->sk_lock.slock), \
925 				SINGLE_DEPTH_NESTING)
926 #define bh_unlock_sock(__sk)	spin_unlock(&((__sk)->sk_lock.slock))
927 
928 extern struct sock		*sk_alloc(struct net *net, int family,
929 					  gfp_t priority,
930 					  struct proto *prot);
931 extern void			sk_free(struct sock *sk);
932 extern void			sk_release_kernel(struct sock *sk);
933 extern struct sock		*sk_clone(const struct sock *sk,
934 					  const gfp_t priority);
935 
936 extern struct sk_buff		*sock_wmalloc(struct sock *sk,
937 					      unsigned long size, int force,
938 					      gfp_t priority);
939 extern struct sk_buff		*sock_rmalloc(struct sock *sk,
940 					      unsigned long size, int force,
941 					      gfp_t priority);
942 extern void			sock_wfree(struct sk_buff *skb);
943 extern void			sock_rfree(struct sk_buff *skb);
944 
945 extern int			sock_setsockopt(struct socket *sock, int level,
946 						int op, char __user *optval,
947 						int optlen);
948 
949 extern int			sock_getsockopt(struct socket *sock, int level,
950 						int op, char __user *optval,
951 						int __user *optlen);
952 extern struct sk_buff 		*sock_alloc_send_skb(struct sock *sk,
953 						     unsigned long size,
954 						     int noblock,
955 						     int *errcode);
956 extern struct sk_buff 		*sock_alloc_send_pskb(struct sock *sk,
957 						      unsigned long header_len,
958 						      unsigned long data_len,
959 						      int noblock,
960 						      int *errcode);
961 extern void *sock_kmalloc(struct sock *sk, int size,
962 			  gfp_t priority);
963 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
964 extern void sk_send_sigurg(struct sock *sk);
965 
966 /*
967  * Functions to fill in entries in struct proto_ops when a protocol
968  * does not implement a particular function.
969  */
970 extern int                      sock_no_bind(struct socket *,
971 					     struct sockaddr *, int);
972 extern int                      sock_no_connect(struct socket *,
973 						struct sockaddr *, int, int);
974 extern int                      sock_no_socketpair(struct socket *,
975 						   struct socket *);
976 extern int                      sock_no_accept(struct socket *,
977 					       struct socket *, int);
978 extern int                      sock_no_getname(struct socket *,
979 						struct sockaddr *, int *, int);
980 extern unsigned int             sock_no_poll(struct file *, struct socket *,
981 					     struct poll_table_struct *);
982 extern int                      sock_no_ioctl(struct socket *, unsigned int,
983 					      unsigned long);
984 extern int			sock_no_listen(struct socket *, int);
985 extern int                      sock_no_shutdown(struct socket *, int);
986 extern int			sock_no_getsockopt(struct socket *, int , int,
987 						   char __user *, int __user *);
988 extern int			sock_no_setsockopt(struct socket *, int, int,
989 						   char __user *, int);
990 extern int                      sock_no_sendmsg(struct kiocb *, struct socket *,
991 						struct msghdr *, size_t);
992 extern int                      sock_no_recvmsg(struct kiocb *, struct socket *,
993 						struct msghdr *, size_t, int);
994 extern int			sock_no_mmap(struct file *file,
995 					     struct socket *sock,
996 					     struct vm_area_struct *vma);
997 extern ssize_t			sock_no_sendpage(struct socket *sock,
998 						struct page *page,
999 						int offset, size_t size,
1000 						int flags);
1001 
1002 /*
1003  * Functions to fill in entries in struct proto_ops when a protocol
1004  * uses the inet style.
1005  */
1006 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1007 				  char __user *optval, int __user *optlen);
1008 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1009 			       struct msghdr *msg, size_t size, int flags);
1010 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1011 				  char __user *optval, int optlen);
1012 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1013 		int optname, char __user *optval, int __user *optlen);
1014 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1015 		int optname, char __user *optval, int optlen);
1016 
1017 extern void sk_common_release(struct sock *sk);
1018 
1019 /*
1020  *	Default socket callbacks and setup code
1021  */
1022 
1023 /* Initialise core socket variables */
1024 extern void sock_init_data(struct socket *sock, struct sock *sk);
1025 
1026 /**
1027  *	sk_filter_release: Release a socket filter
1028  *	@fp: filter to remove
1029  *
1030  *	Remove a filter from a socket and release its resources.
1031  */
1032 
1033 static inline void sk_filter_release(struct sk_filter *fp)
1034 {
1035 	if (atomic_dec_and_test(&fp->refcnt))
1036 		kfree(fp);
1037 }
1038 
1039 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1040 {
1041 	unsigned int size = sk_filter_len(fp);
1042 
1043 	atomic_sub(size, &sk->sk_omem_alloc);
1044 	sk_filter_release(fp);
1045 }
1046 
1047 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1048 {
1049 	atomic_inc(&fp->refcnt);
1050 	atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1051 }
1052 
1053 /*
1054  * Socket reference counting postulates.
1055  *
1056  * * Each user of socket SHOULD hold a reference count.
1057  * * Each access point to socket (an hash table bucket, reference from a list,
1058  *   running timer, skb in flight MUST hold a reference count.
1059  * * When reference count hits 0, it means it will never increase back.
1060  * * When reference count hits 0, it means that no references from
1061  *   outside exist to this socket and current process on current CPU
1062  *   is last user and may/should destroy this socket.
1063  * * sk_free is called from any context: process, BH, IRQ. When
1064  *   it is called, socket has no references from outside -> sk_free
1065  *   may release descendant resources allocated by the socket, but
1066  *   to the time when it is called, socket is NOT referenced by any
1067  *   hash tables, lists etc.
1068  * * Packets, delivered from outside (from network or from another process)
1069  *   and enqueued on receive/error queues SHOULD NOT grab reference count,
1070  *   when they sit in queue. Otherwise, packets will leak to hole, when
1071  *   socket is looked up by one cpu and unhasing is made by another CPU.
1072  *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
1073  *   (leak to backlog). Packet socket does all the processing inside
1074  *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1075  *   use separate SMP lock, so that they are prone too.
1076  */
1077 
1078 /* Ungrab socket and destroy it, if it was the last reference. */
1079 static inline void sock_put(struct sock *sk)
1080 {
1081 	if (atomic_dec_and_test(&sk->sk_refcnt))
1082 		sk_free(sk);
1083 }
1084 
1085 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1086 			  const int nested);
1087 
1088 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1089 {
1090 	sk->sk_socket = sock;
1091 }
1092 
1093 /* Detach socket from process context.
1094  * Announce socket dead, detach it from wait queue and inode.
1095  * Note that parent inode held reference count on this struct sock,
1096  * we do not release it in this function, because protocol
1097  * probably wants some additional cleanups or even continuing
1098  * to work with this socket (TCP).
1099  */
1100 static inline void sock_orphan(struct sock *sk)
1101 {
1102 	write_lock_bh(&sk->sk_callback_lock);
1103 	sock_set_flag(sk, SOCK_DEAD);
1104 	sk_set_socket(sk, NULL);
1105 	sk->sk_sleep  = NULL;
1106 	write_unlock_bh(&sk->sk_callback_lock);
1107 }
1108 
1109 static inline void sock_graft(struct sock *sk, struct socket *parent)
1110 {
1111 	write_lock_bh(&sk->sk_callback_lock);
1112 	sk->sk_sleep = &parent->wait;
1113 	parent->sk = sk;
1114 	sk_set_socket(sk, parent);
1115 	security_sock_graft(sk, parent);
1116 	write_unlock_bh(&sk->sk_callback_lock);
1117 }
1118 
1119 extern int sock_i_uid(struct sock *sk);
1120 extern unsigned long sock_i_ino(struct sock *sk);
1121 
1122 static inline struct dst_entry *
1123 __sk_dst_get(struct sock *sk)
1124 {
1125 	return sk->sk_dst_cache;
1126 }
1127 
1128 static inline struct dst_entry *
1129 sk_dst_get(struct sock *sk)
1130 {
1131 	struct dst_entry *dst;
1132 
1133 	read_lock(&sk->sk_dst_lock);
1134 	dst = sk->sk_dst_cache;
1135 	if (dst)
1136 		dst_hold(dst);
1137 	read_unlock(&sk->sk_dst_lock);
1138 	return dst;
1139 }
1140 
1141 static inline void
1142 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1143 {
1144 	struct dst_entry *old_dst;
1145 
1146 	old_dst = sk->sk_dst_cache;
1147 	sk->sk_dst_cache = dst;
1148 	dst_release(old_dst);
1149 }
1150 
1151 static inline void
1152 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1153 {
1154 	write_lock(&sk->sk_dst_lock);
1155 	__sk_dst_set(sk, dst);
1156 	write_unlock(&sk->sk_dst_lock);
1157 }
1158 
1159 static inline void
1160 __sk_dst_reset(struct sock *sk)
1161 {
1162 	struct dst_entry *old_dst;
1163 
1164 	old_dst = sk->sk_dst_cache;
1165 	sk->sk_dst_cache = NULL;
1166 	dst_release(old_dst);
1167 }
1168 
1169 static inline void
1170 sk_dst_reset(struct sock *sk)
1171 {
1172 	write_lock(&sk->sk_dst_lock);
1173 	__sk_dst_reset(sk);
1174 	write_unlock(&sk->sk_dst_lock);
1175 }
1176 
1177 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1178 
1179 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1180 
1181 static inline int sk_can_gso(const struct sock *sk)
1182 {
1183 	return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1184 }
1185 
1186 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1187 
1188 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1189 				   struct sk_buff *skb, struct page *page,
1190 				   int off, int copy)
1191 {
1192 	if (skb->ip_summed == CHECKSUM_NONE) {
1193 		int err = 0;
1194 		__wsum csum = csum_and_copy_from_user(from,
1195 						     page_address(page) + off,
1196 							    copy, 0, &err);
1197 		if (err)
1198 			return err;
1199 		skb->csum = csum_block_add(skb->csum, csum, skb->len);
1200 	} else if (copy_from_user(page_address(page) + off, from, copy))
1201 		return -EFAULT;
1202 
1203 	skb->len	     += copy;
1204 	skb->data_len	     += copy;
1205 	skb->truesize	     += copy;
1206 	sk->sk_wmem_queued   += copy;
1207 	sk_mem_charge(sk, copy);
1208 	return 0;
1209 }
1210 
1211 /*
1212  * 	Queue a received datagram if it will fit. Stream and sequenced
1213  *	protocols can't normally use this as they need to fit buffers in
1214  *	and play with them.
1215  *
1216  * 	Inlined as it's very short and called for pretty much every
1217  *	packet ever received.
1218  */
1219 
1220 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1221 {
1222 	skb->sk = sk;
1223 	skb->destructor = sock_wfree;
1224 	/*
1225 	 * We used to take a refcount on sk, but following operation
1226 	 * is enough to guarantee sk_free() wont free this sock until
1227 	 * all in-flight packets are completed
1228 	 */
1229 	atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1230 }
1231 
1232 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1233 {
1234 	skb->sk = sk;
1235 	skb->destructor = sock_rfree;
1236 	atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1237 	sk_mem_charge(sk, skb->truesize);
1238 }
1239 
1240 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1241 			   unsigned long expires);
1242 
1243 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1244 
1245 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1246 
1247 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1248 {
1249 	/* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1250 	   number of warnings when compiling with -W --ANK
1251 	 */
1252 	if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1253 	    (unsigned)sk->sk_rcvbuf)
1254 		return -ENOMEM;
1255 	skb_set_owner_r(skb, sk);
1256 	skb_queue_tail(&sk->sk_error_queue, skb);
1257 	if (!sock_flag(sk, SOCK_DEAD))
1258 		sk->sk_data_ready(sk, skb->len);
1259 	return 0;
1260 }
1261 
1262 /*
1263  *	Recover an error report and clear atomically
1264  */
1265 
1266 static inline int sock_error(struct sock *sk)
1267 {
1268 	int err;
1269 	if (likely(!sk->sk_err))
1270 		return 0;
1271 	err = xchg(&sk->sk_err, 0);
1272 	return -err;
1273 }
1274 
1275 static inline unsigned long sock_wspace(struct sock *sk)
1276 {
1277 	int amt = 0;
1278 
1279 	if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1280 		amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1281 		if (amt < 0)
1282 			amt = 0;
1283 	}
1284 	return amt;
1285 }
1286 
1287 static inline void sk_wake_async(struct sock *sk, int how, int band)
1288 {
1289 	if (sk->sk_socket && sk->sk_socket->fasync_list)
1290 		sock_wake_async(sk->sk_socket, how, band);
1291 }
1292 
1293 #define SOCK_MIN_SNDBUF 2048
1294 #define SOCK_MIN_RCVBUF 256
1295 
1296 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1297 {
1298 	if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1299 		sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1300 		sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1301 	}
1302 }
1303 
1304 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1305 
1306 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1307 {
1308 	struct page *page = NULL;
1309 
1310 	page = alloc_pages(sk->sk_allocation, 0);
1311 	if (!page) {
1312 		sk->sk_prot->enter_memory_pressure(sk);
1313 		sk_stream_moderate_sndbuf(sk);
1314 	}
1315 	return page;
1316 }
1317 
1318 /*
1319  *	Default write policy as shown to user space via poll/select/SIGIO
1320  */
1321 static inline int sock_writeable(const struct sock *sk)
1322 {
1323 	return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1324 }
1325 
1326 static inline gfp_t gfp_any(void)
1327 {
1328 	return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1329 }
1330 
1331 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1332 {
1333 	return noblock ? 0 : sk->sk_rcvtimeo;
1334 }
1335 
1336 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1337 {
1338 	return noblock ? 0 : sk->sk_sndtimeo;
1339 }
1340 
1341 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1342 {
1343 	return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1344 }
1345 
1346 /* Alas, with timeout socket operations are not restartable.
1347  * Compare this to poll().
1348  */
1349 static inline int sock_intr_errno(long timeo)
1350 {
1351 	return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1352 }
1353 
1354 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1355 	struct sk_buff *skb);
1356 
1357 static __inline__ void
1358 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1359 {
1360 	ktime_t kt = skb->tstamp;
1361 	struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
1362 
1363 	/*
1364 	 * generate control messages if
1365 	 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1366 	 *   or SOCK_TIMESTAMPING_RX_SOFTWARE)
1367 	 * - software time stamp available and wanted
1368 	 *   (SOCK_TIMESTAMPING_SOFTWARE)
1369 	 * - hardware time stamps available and wanted
1370 	 *   (SOCK_TIMESTAMPING_SYS_HARDWARE or
1371 	 *   SOCK_TIMESTAMPING_RAW_HARDWARE)
1372 	 */
1373 	if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1374 	    sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1375 	    (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1376 	    (hwtstamps->hwtstamp.tv64 &&
1377 	     sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1378 	    (hwtstamps->syststamp.tv64 &&
1379 	     sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
1380 		__sock_recv_timestamp(msg, sk, skb);
1381 	else
1382 		sk->sk_stamp = kt;
1383 }
1384 
1385 /**
1386  * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1387  * @msg:	outgoing packet
1388  * @sk:		socket sending this packet
1389  * @shtx:	filled with instructions for time stamping
1390  *
1391  * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1392  * parameters are invalid.
1393  */
1394 extern int sock_tx_timestamp(struct msghdr *msg,
1395 			     struct sock *sk,
1396 			     union skb_shared_tx *shtx);
1397 
1398 
1399 /**
1400  * sk_eat_skb - Release a skb if it is no longer needed
1401  * @sk: socket to eat this skb from
1402  * @skb: socket buffer to eat
1403  * @copied_early: flag indicating whether DMA operations copied this data early
1404  *
1405  * This routine must be called with interrupts disabled or with the socket
1406  * locked so that the sk_buff queue operation is ok.
1407 */
1408 #ifdef CONFIG_NET_DMA
1409 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1410 {
1411 	__skb_unlink(skb, &sk->sk_receive_queue);
1412 	if (!copied_early)
1413 		__kfree_skb(skb);
1414 	else
1415 		__skb_queue_tail(&sk->sk_async_wait_queue, skb);
1416 }
1417 #else
1418 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1419 {
1420 	__skb_unlink(skb, &sk->sk_receive_queue);
1421 	__kfree_skb(skb);
1422 }
1423 #endif
1424 
1425 static inline
1426 struct net *sock_net(const struct sock *sk)
1427 {
1428 #ifdef CONFIG_NET_NS
1429 	return sk->sk_net;
1430 #else
1431 	return &init_net;
1432 #endif
1433 }
1434 
1435 static inline
1436 void sock_net_set(struct sock *sk, struct net *net)
1437 {
1438 #ifdef CONFIG_NET_NS
1439 	sk->sk_net = net;
1440 #endif
1441 }
1442 
1443 /*
1444  * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1445  * They should not hold a referrence to a namespace in order to allow
1446  * to stop it.
1447  * Sockets after sk_change_net should be released using sk_release_kernel
1448  */
1449 static inline void sk_change_net(struct sock *sk, struct net *net)
1450 {
1451 	put_net(sock_net(sk));
1452 	sock_net_set(sk, hold_net(net));
1453 }
1454 
1455 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
1456 {
1457 	if (unlikely(skb->sk)) {
1458 		struct sock *sk = skb->sk;
1459 
1460 		skb->destructor = NULL;
1461 		skb->sk = NULL;
1462 		return sk;
1463 	}
1464 	return NULL;
1465 }
1466 
1467 extern void sock_enable_timestamp(struct sock *sk, int flag);
1468 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1469 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1470 
1471 /*
1472  *	Enable debug/info messages
1473  */
1474 extern int net_msg_warn;
1475 #define NETDEBUG(fmt, args...) \
1476 	do { if (net_msg_warn) printk(fmt,##args); } while (0)
1477 
1478 #define LIMIT_NETDEBUG(fmt, args...) \
1479 	do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1480 
1481 extern __u32 sysctl_wmem_max;
1482 extern __u32 sysctl_rmem_max;
1483 
1484 extern void sk_init(void);
1485 
1486 extern int sysctl_optmem_max;
1487 
1488 extern __u32 sysctl_wmem_default;
1489 extern __u32 sysctl_rmem_default;
1490 
1491 #endif	/* _SOCK_H */
1492