xref: /linux/include/net/sock.h (revision 2624f124b3b5d550ab2fbef7ee3bc0e1fed09722)
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/config.h>
44 #include <linux/list.h>
45 #include <linux/timer.h>
46 #include <linux/cache.h>
47 #include <linux/module.h>
48 #include <linux/netdevice.h>
49 #include <linux/skbuff.h>	/* struct sk_buff */
50 #include <linux/security.h>
51 
52 #include <linux/filter.h>
53 
54 #include <asm/atomic.h>
55 #include <net/dst.h>
56 #include <net/checksum.h>
57 
58 /*
59  * This structure really needs to be cleaned up.
60  * Most of it is for TCP, and not used by any of
61  * the other protocols.
62  */
63 
64 /* Define this to get the SOCK_DBG debugging facility. */
65 #define SOCK_DEBUGGING
66 #ifdef SOCK_DEBUGGING
67 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
68 					printk(KERN_DEBUG msg); } while (0)
69 #else
70 #define SOCK_DEBUG(sk, msg...) do { } while (0)
71 #endif
72 
73 /* This is the per-socket lock.  The spinlock provides a synchronization
74  * between user contexts and software interrupt processing, whereas the
75  * mini-semaphore synchronizes multiple users amongst themselves.
76  */
77 struct sock_iocb;
78 typedef struct {
79 	spinlock_t		slock;
80 	struct sock_iocb	*owner;
81 	wait_queue_head_t	wq;
82 } socket_lock_t;
83 
84 #define sock_lock_init(__sk) \
85 do {	spin_lock_init(&((__sk)->sk_lock.slock)); \
86 	(__sk)->sk_lock.owner = NULL; \
87 	init_waitqueue_head(&((__sk)->sk_lock.wq)); \
88 } while(0)
89 
90 struct sock;
91 struct proto;
92 
93 /**
94  *	struct sock_common - minimal network layer representation of sockets
95  *	@skc_family: network address family
96  *	@skc_state: Connection state
97  *	@skc_reuse: %SO_REUSEADDR setting
98  *	@skc_bound_dev_if: bound device index if != 0
99  *	@skc_node: main hash linkage for various protocol lookup tables
100  *	@skc_bind_node: bind hash linkage for various protocol lookup tables
101  *	@skc_refcnt: reference count
102  *	@skc_prot: protocol handlers inside a network family
103  *
104  *	This is the minimal network layer representation of sockets, the header
105  *	for struct sock and struct inet_timewait_sock.
106  */
107 struct sock_common {
108 	unsigned short		skc_family;
109 	volatile unsigned char	skc_state;
110 	unsigned char		skc_reuse;
111 	int			skc_bound_dev_if;
112 	struct hlist_node	skc_node;
113 	struct hlist_node	skc_bind_node;
114 	atomic_t		skc_refcnt;
115 	struct proto		*skc_prot;
116 };
117 
118 /**
119   *	struct sock - network layer representation of sockets
120   *	@__sk_common: shared layout with inet_timewait_sock
121   *	@sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
122   *	@sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
123   *	@sk_lock:	synchronizer
124   *	@sk_rcvbuf: size of receive buffer in bytes
125   *	@sk_sleep: sock wait queue
126   *	@sk_dst_cache: destination cache
127   *	@sk_dst_lock: destination cache lock
128   *	@sk_policy: flow policy
129   *	@sk_rmem_alloc: receive queue bytes committed
130   *	@sk_receive_queue: incoming packets
131   *	@sk_wmem_alloc: transmit queue bytes committed
132   *	@sk_write_queue: Packet sending queue
133   *	@sk_omem_alloc: "o" is "option" or "other"
134   *	@sk_wmem_queued: persistent queue size
135   *	@sk_forward_alloc: space allocated forward
136   *	@sk_allocation: allocation mode
137   *	@sk_sndbuf: size of send buffer in bytes
138   *	@sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
139   *	@sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
140   *	@sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
141   *	@sk_lingertime: %SO_LINGER l_linger setting
142   *	@sk_hashent: hash entry in several tables (e.g. inet_hashinfo.ehash)
143   *	@sk_backlog: always used with the per-socket spinlock held
144   *	@sk_callback_lock: used with the callbacks in the end of this struct
145   *	@sk_error_queue: rarely used
146   *	@sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
147   *	@sk_err: last error
148   *	@sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
149   *	@sk_ack_backlog: current listen backlog
150   *	@sk_max_ack_backlog: listen backlog set in listen()
151   *	@sk_priority: %SO_PRIORITY setting
152   *	@sk_type: socket type (%SOCK_STREAM, etc)
153   *	@sk_protocol: which protocol this socket belongs in this network family
154   *	@sk_peercred: %SO_PEERCRED setting
155   *	@sk_rcvlowat: %SO_RCVLOWAT setting
156   *	@sk_rcvtimeo: %SO_RCVTIMEO setting
157   *	@sk_sndtimeo: %SO_SNDTIMEO setting
158   *	@sk_filter: socket filtering instructions
159   *	@sk_protinfo: private area, net family specific, when not using slab
160   *	@sk_timer: sock cleanup timer
161   *	@sk_stamp: time stamp of last packet received
162   *	@sk_socket: Identd and reporting IO signals
163   *	@sk_user_data: RPC layer private data
164   *	@sk_sndmsg_page: cached page for sendmsg
165   *	@sk_sndmsg_off: cached offset for sendmsg
166   *	@sk_send_head: front of stuff to transmit
167   *	@sk_security: used by security modules
168   *	@sk_write_pending: a write to stream socket waits to start
169   *	@sk_state_change: callback to indicate change in the state of the sock
170   *	@sk_data_ready: callback to indicate there is data to be processed
171   *	@sk_write_space: callback to indicate there is bf sending space available
172   *	@sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
173   *	@sk_backlog_rcv: callback to process the backlog
174   *	@sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
175  */
176 struct sock {
177 	/*
178 	 * Now struct inet_timewait_sock also uses sock_common, so please just
179 	 * don't add nothing before this first member (__sk_common) --acme
180 	 */
181 	struct sock_common	__sk_common;
182 #define sk_family		__sk_common.skc_family
183 #define sk_state		__sk_common.skc_state
184 #define sk_reuse		__sk_common.skc_reuse
185 #define sk_bound_dev_if		__sk_common.skc_bound_dev_if
186 #define sk_node			__sk_common.skc_node
187 #define sk_bind_node		__sk_common.skc_bind_node
188 #define sk_refcnt		__sk_common.skc_refcnt
189 #define sk_prot			__sk_common.skc_prot
190 	unsigned char		sk_shutdown : 2,
191 				sk_no_check : 2,
192 				sk_userlocks : 4;
193 	unsigned char		sk_protocol;
194 	unsigned short		sk_type;
195 	int			sk_rcvbuf;
196 	socket_lock_t		sk_lock;
197 	wait_queue_head_t	*sk_sleep;
198 	struct dst_entry	*sk_dst_cache;
199 	struct xfrm_policy	*sk_policy[2];
200 	rwlock_t		sk_dst_lock;
201 	atomic_t		sk_rmem_alloc;
202 	atomic_t		sk_wmem_alloc;
203 	atomic_t		sk_omem_alloc;
204 	struct sk_buff_head	sk_receive_queue;
205 	struct sk_buff_head	sk_write_queue;
206 	int			sk_wmem_queued;
207 	int			sk_forward_alloc;
208 	unsigned int		sk_allocation;
209 	int			sk_sndbuf;
210 	int			sk_route_caps;
211 	int			sk_hashent;
212 	unsigned long 		sk_flags;
213 	unsigned long	        sk_lingertime;
214 	/*
215 	 * The backlog queue is special, it is always used with
216 	 * the per-socket spinlock held and requires low latency
217 	 * access. Therefore we special case it's implementation.
218 	 */
219 	struct {
220 		struct sk_buff *head;
221 		struct sk_buff *tail;
222 	} sk_backlog;
223 	struct sk_buff_head	sk_error_queue;
224 	struct proto		*sk_prot_creator;
225 	rwlock_t		sk_callback_lock;
226 	int			sk_err,
227 				sk_err_soft;
228 	unsigned short		sk_ack_backlog;
229 	unsigned short		sk_max_ack_backlog;
230 	__u32			sk_priority;
231 	struct ucred		sk_peercred;
232 	int			sk_rcvlowat;
233 	long			sk_rcvtimeo;
234 	long			sk_sndtimeo;
235 	struct sk_filter      	*sk_filter;
236 	void			*sk_protinfo;
237 	struct timer_list	sk_timer;
238 	struct timeval		sk_stamp;
239 	struct socket		*sk_socket;
240 	void			*sk_user_data;
241 	struct page		*sk_sndmsg_page;
242 	struct sk_buff		*sk_send_head;
243 	__u32			sk_sndmsg_off;
244 	int			sk_write_pending;
245 	void			*sk_security;
246 	void			(*sk_state_change)(struct sock *sk);
247 	void			(*sk_data_ready)(struct sock *sk, int bytes);
248 	void			(*sk_write_space)(struct sock *sk);
249 	void			(*sk_error_report)(struct sock *sk);
250   	int			(*sk_backlog_rcv)(struct sock *sk,
251 						  struct sk_buff *skb);
252 	void                    (*sk_destruct)(struct sock *sk);
253 };
254 
255 /*
256  * Hashed lists helper routines
257  */
258 static inline struct sock *__sk_head(const struct hlist_head *head)
259 {
260 	return hlist_entry(head->first, struct sock, sk_node);
261 }
262 
263 static inline struct sock *sk_head(const struct hlist_head *head)
264 {
265 	return hlist_empty(head) ? NULL : __sk_head(head);
266 }
267 
268 static inline struct sock *sk_next(const struct sock *sk)
269 {
270 	return sk->sk_node.next ?
271 		hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
272 }
273 
274 static inline int sk_unhashed(const struct sock *sk)
275 {
276 	return hlist_unhashed(&sk->sk_node);
277 }
278 
279 static inline int sk_hashed(const struct sock *sk)
280 {
281 	return sk->sk_node.pprev != NULL;
282 }
283 
284 static __inline__ void sk_node_init(struct hlist_node *node)
285 {
286 	node->pprev = NULL;
287 }
288 
289 static __inline__ void __sk_del_node(struct sock *sk)
290 {
291 	__hlist_del(&sk->sk_node);
292 }
293 
294 static __inline__ int __sk_del_node_init(struct sock *sk)
295 {
296 	if (sk_hashed(sk)) {
297 		__sk_del_node(sk);
298 		sk_node_init(&sk->sk_node);
299 		return 1;
300 	}
301 	return 0;
302 }
303 
304 /* Grab socket reference count. This operation is valid only
305    when sk is ALREADY grabbed f.e. it is found in hash table
306    or a list and the lookup is made under lock preventing hash table
307    modifications.
308  */
309 
310 static inline void sock_hold(struct sock *sk)
311 {
312 	atomic_inc(&sk->sk_refcnt);
313 }
314 
315 /* Ungrab socket in the context, which assumes that socket refcnt
316    cannot hit zero, f.e. it is true in context of any socketcall.
317  */
318 static inline void __sock_put(struct sock *sk)
319 {
320 	atomic_dec(&sk->sk_refcnt);
321 }
322 
323 static __inline__ int sk_del_node_init(struct sock *sk)
324 {
325 	int rc = __sk_del_node_init(sk);
326 
327 	if (rc) {
328 		/* paranoid for a while -acme */
329 		WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
330 		__sock_put(sk);
331 	}
332 	return rc;
333 }
334 
335 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
336 {
337 	hlist_add_head(&sk->sk_node, list);
338 }
339 
340 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
341 {
342 	sock_hold(sk);
343 	__sk_add_node(sk, list);
344 }
345 
346 static __inline__ void __sk_del_bind_node(struct sock *sk)
347 {
348 	__hlist_del(&sk->sk_bind_node);
349 }
350 
351 static __inline__ void sk_add_bind_node(struct sock *sk,
352 					struct hlist_head *list)
353 {
354 	hlist_add_head(&sk->sk_bind_node, list);
355 }
356 
357 #define sk_for_each(__sk, node, list) \
358 	hlist_for_each_entry(__sk, node, list, sk_node)
359 #define sk_for_each_from(__sk, node) \
360 	if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
361 		hlist_for_each_entry_from(__sk, node, sk_node)
362 #define sk_for_each_continue(__sk, node) \
363 	if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
364 		hlist_for_each_entry_continue(__sk, node, sk_node)
365 #define sk_for_each_safe(__sk, node, tmp, list) \
366 	hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
367 #define sk_for_each_bound(__sk, node, list) \
368 	hlist_for_each_entry(__sk, node, list, sk_bind_node)
369 
370 /* Sock flags */
371 enum sock_flags {
372 	SOCK_DEAD,
373 	SOCK_DONE,
374 	SOCK_URGINLINE,
375 	SOCK_KEEPOPEN,
376 	SOCK_LINGER,
377 	SOCK_DESTROY,
378 	SOCK_BROADCAST,
379 	SOCK_TIMESTAMP,
380 	SOCK_ZAPPED,
381 	SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
382 	SOCK_DBG, /* %SO_DEBUG setting */
383 	SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
384 	SOCK_NO_LARGESEND, /* whether to sent large segments or not */
385 	SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
386 	SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
387 };
388 
389 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
390 {
391 	nsk->sk_flags = osk->sk_flags;
392 }
393 
394 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
395 {
396 	__set_bit(flag, &sk->sk_flags);
397 }
398 
399 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
400 {
401 	__clear_bit(flag, &sk->sk_flags);
402 }
403 
404 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
405 {
406 	return test_bit(flag, &sk->sk_flags);
407 }
408 
409 static inline void sk_acceptq_removed(struct sock *sk)
410 {
411 	sk->sk_ack_backlog--;
412 }
413 
414 static inline void sk_acceptq_added(struct sock *sk)
415 {
416 	sk->sk_ack_backlog++;
417 }
418 
419 static inline int sk_acceptq_is_full(struct sock *sk)
420 {
421 	return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
422 }
423 
424 /*
425  * Compute minimal free write space needed to queue new packets.
426  */
427 static inline int sk_stream_min_wspace(struct sock *sk)
428 {
429 	return sk->sk_wmem_queued / 2;
430 }
431 
432 static inline int sk_stream_wspace(struct sock *sk)
433 {
434 	return sk->sk_sndbuf - sk->sk_wmem_queued;
435 }
436 
437 extern void sk_stream_write_space(struct sock *sk);
438 
439 static inline int sk_stream_memory_free(struct sock *sk)
440 {
441 	return sk->sk_wmem_queued < sk->sk_sndbuf;
442 }
443 
444 extern void sk_stream_rfree(struct sk_buff *skb);
445 
446 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
447 {
448 	skb->sk = sk;
449 	skb->destructor = sk_stream_rfree;
450 	atomic_add(skb->truesize, &sk->sk_rmem_alloc);
451 	sk->sk_forward_alloc -= skb->truesize;
452 }
453 
454 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
455 {
456 	sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
457 	sk->sk_wmem_queued   -= skb->truesize;
458 	sk->sk_forward_alloc += skb->truesize;
459 	__kfree_skb(skb);
460 }
461 
462 /* The per-socket spinlock must be held here. */
463 #define sk_add_backlog(__sk, __skb)				\
464 do {	if (!(__sk)->sk_backlog.tail) {				\
465 		(__sk)->sk_backlog.head =			\
466 		     (__sk)->sk_backlog.tail = (__skb);		\
467 	} else {						\
468 		((__sk)->sk_backlog.tail)->next = (__skb);	\
469 		(__sk)->sk_backlog.tail = (__skb);		\
470 	}							\
471 	(__skb)->next = NULL;					\
472 } while(0)
473 
474 #define sk_wait_event(__sk, __timeo, __condition)		\
475 ({	int rc;							\
476 	release_sock(__sk);					\
477 	rc = __condition;					\
478 	if (!rc) {						\
479 		*(__timeo) = schedule_timeout(*(__timeo));	\
480 		rc = __condition;				\
481 	}							\
482 	lock_sock(__sk);					\
483 	rc;							\
484 })
485 
486 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
487 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
488 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
489 extern int sk_stream_error(struct sock *sk, int flags, int err);
490 extern void sk_stream_kill_queues(struct sock *sk);
491 
492 extern int sk_wait_data(struct sock *sk, long *timeo);
493 
494 struct request_sock_ops;
495 
496 /* Networking protocol blocks we attach to sockets.
497  * socket layer -> transport layer interface
498  * transport -> network interface is defined by struct inet_proto
499  */
500 struct proto {
501 	void			(*close)(struct sock *sk,
502 					long timeout);
503 	int			(*connect)(struct sock *sk,
504 				        struct sockaddr *uaddr,
505 					int addr_len);
506 	int			(*disconnect)(struct sock *sk, int flags);
507 
508 	struct sock *		(*accept) (struct sock *sk, int flags, int *err);
509 
510 	int			(*ioctl)(struct sock *sk, int cmd,
511 					 unsigned long arg);
512 	int			(*init)(struct sock *sk);
513 	int			(*destroy)(struct sock *sk);
514 	void			(*shutdown)(struct sock *sk, int how);
515 	int			(*setsockopt)(struct sock *sk, int level,
516 					int optname, char __user *optval,
517 					int optlen);
518 	int			(*getsockopt)(struct sock *sk, int level,
519 					int optname, char __user *optval,
520 					int __user *option);
521 	int			(*sendmsg)(struct kiocb *iocb, struct sock *sk,
522 					   struct msghdr *msg, size_t len);
523 	int			(*recvmsg)(struct kiocb *iocb, struct sock *sk,
524 					   struct msghdr *msg,
525 					size_t len, int noblock, int flags,
526 					int *addr_len);
527 	int			(*sendpage)(struct sock *sk, struct page *page,
528 					int offset, size_t size, int flags);
529 	int			(*bind)(struct sock *sk,
530 					struct sockaddr *uaddr, int addr_len);
531 
532 	int			(*backlog_rcv) (struct sock *sk,
533 						struct sk_buff *skb);
534 
535 	/* Keeping track of sk's, looking them up, and port selection methods. */
536 	void			(*hash)(struct sock *sk);
537 	void			(*unhash)(struct sock *sk);
538 	int			(*get_port)(struct sock *sk, unsigned short snum);
539 
540 	/* Memory pressure */
541 	void			(*enter_memory_pressure)(void);
542 	atomic_t		*memory_allocated;	/* Current allocated memory. */
543 	atomic_t		*sockets_allocated;	/* Current number of sockets. */
544 	/*
545 	 * Pressure flag: try to collapse.
546 	 * Technical note: it is used by multiple contexts non atomically.
547 	 * All the sk_stream_mem_schedule() is of this nature: accounting
548 	 * is strict, actions are advisory and have some latency.
549 	 */
550 	int			*memory_pressure;
551 	int			*sysctl_mem;
552 	int			*sysctl_wmem;
553 	int			*sysctl_rmem;
554 	int			max_header;
555 
556 	kmem_cache_t		*slab;
557 	unsigned int		obj_size;
558 
559 	kmem_cache_t		*twsk_slab;
560 	unsigned int		twsk_obj_size;
561 	atomic_t		*orphan_count;
562 
563 	struct request_sock_ops	*rsk_prot;
564 
565 	struct module		*owner;
566 
567 	char			name[32];
568 
569 	struct list_head	node;
570 #ifdef SOCK_REFCNT_DEBUG
571 	atomic_t		socks;
572 #endif
573 	struct {
574 		int inuse;
575 		u8  __pad[SMP_CACHE_BYTES - sizeof(int)];
576 	} stats[NR_CPUS];
577 };
578 
579 extern int proto_register(struct proto *prot, int alloc_slab);
580 extern void proto_unregister(struct proto *prot);
581 
582 #ifdef SOCK_REFCNT_DEBUG
583 static inline void sk_refcnt_debug_inc(struct sock *sk)
584 {
585 	atomic_inc(&sk->sk_prot->socks);
586 }
587 
588 static inline void sk_refcnt_debug_dec(struct sock *sk)
589 {
590 	atomic_dec(&sk->sk_prot->socks);
591 	printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
592 	       sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
593 }
594 
595 static inline void sk_refcnt_debug_release(const struct sock *sk)
596 {
597 	if (atomic_read(&sk->sk_refcnt) != 1)
598 		printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
599 		       sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
600 }
601 #else /* SOCK_REFCNT_DEBUG */
602 #define sk_refcnt_debug_inc(sk) do { } while (0)
603 #define sk_refcnt_debug_dec(sk) do { } while (0)
604 #define sk_refcnt_debug_release(sk) do { } while (0)
605 #endif /* SOCK_REFCNT_DEBUG */
606 
607 /* Called with local bh disabled */
608 static __inline__ void sock_prot_inc_use(struct proto *prot)
609 {
610 	prot->stats[smp_processor_id()].inuse++;
611 }
612 
613 static __inline__ void sock_prot_dec_use(struct proto *prot)
614 {
615 	prot->stats[smp_processor_id()].inuse--;
616 }
617 
618 /* With per-bucket locks this operation is not-atomic, so that
619  * this version is not worse.
620  */
621 static inline void __sk_prot_rehash(struct sock *sk)
622 {
623 	sk->sk_prot->unhash(sk);
624 	sk->sk_prot->hash(sk);
625 }
626 
627 /* About 10 seconds */
628 #define SOCK_DESTROY_TIME (10*HZ)
629 
630 /* Sockets 0-1023 can't be bound to unless you are superuser */
631 #define PROT_SOCK	1024
632 
633 #define SHUTDOWN_MASK	3
634 #define RCV_SHUTDOWN	1
635 #define SEND_SHUTDOWN	2
636 
637 #define SOCK_SNDBUF_LOCK	1
638 #define SOCK_RCVBUF_LOCK	2
639 #define SOCK_BINDADDR_LOCK	4
640 #define SOCK_BINDPORT_LOCK	8
641 
642 /* sock_iocb: used to kick off async processing of socket ios */
643 struct sock_iocb {
644 	struct list_head	list;
645 
646 	int			flags;
647 	int			size;
648 	struct socket		*sock;
649 	struct sock		*sk;
650 	struct scm_cookie	*scm;
651 	struct msghdr		*msg, async_msg;
652 	struct iovec		async_iov;
653 	struct kiocb		*kiocb;
654 };
655 
656 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
657 {
658 	return (struct sock_iocb *)iocb->private;
659 }
660 
661 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
662 {
663 	return si->kiocb;
664 }
665 
666 struct socket_alloc {
667 	struct socket socket;
668 	struct inode vfs_inode;
669 };
670 
671 static inline struct socket *SOCKET_I(struct inode *inode)
672 {
673 	return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
674 }
675 
676 static inline struct inode *SOCK_INODE(struct socket *socket)
677 {
678 	return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
679 }
680 
681 extern void __sk_stream_mem_reclaim(struct sock *sk);
682 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
683 
684 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
685 
686 static inline int sk_stream_pages(int amt)
687 {
688 	return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
689 }
690 
691 static inline void sk_stream_mem_reclaim(struct sock *sk)
692 {
693 	if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
694 		__sk_stream_mem_reclaim(sk);
695 }
696 
697 static inline void sk_stream_writequeue_purge(struct sock *sk)
698 {
699 	struct sk_buff *skb;
700 
701 	while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
702 		sk_stream_free_skb(sk, skb);
703 	sk_stream_mem_reclaim(sk);
704 }
705 
706 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
707 {
708 	return (int)skb->truesize <= sk->sk_forward_alloc ||
709 		sk_stream_mem_schedule(sk, skb->truesize, 1);
710 }
711 
712 static inline int sk_stream_wmem_schedule(struct sock *sk, int size)
713 {
714 	return size <= sk->sk_forward_alloc ||
715 	       sk_stream_mem_schedule(sk, size, 0);
716 }
717 
718 /* Used by processes to "lock" a socket state, so that
719  * interrupts and bottom half handlers won't change it
720  * from under us. It essentially blocks any incoming
721  * packets, so that we won't get any new data or any
722  * packets that change the state of the socket.
723  *
724  * While locked, BH processing will add new packets to
725  * the backlog queue.  This queue is processed by the
726  * owner of the socket lock right before it is released.
727  *
728  * Since ~2.3.5 it is also exclusive sleep lock serializing
729  * accesses from user process context.
730  */
731 #define sock_owned_by_user(sk)	((sk)->sk_lock.owner)
732 
733 extern void FASTCALL(lock_sock(struct sock *sk));
734 extern void FASTCALL(release_sock(struct sock *sk));
735 
736 /* BH context may only use the following locking interface. */
737 #define bh_lock_sock(__sk)	spin_lock(&((__sk)->sk_lock.slock))
738 #define bh_unlock_sock(__sk)	spin_unlock(&((__sk)->sk_lock.slock))
739 
740 extern struct sock		*sk_alloc(int family,
741 					  unsigned int __nocast priority,
742 					  struct proto *prot, int zero_it);
743 extern void			sk_free(struct sock *sk);
744 extern struct sock		*sk_clone(const struct sock *sk,
745 					  const unsigned int __nocast priority);
746 
747 extern struct sk_buff		*sock_wmalloc(struct sock *sk,
748 					      unsigned long size, int force,
749 					      unsigned int __nocast priority);
750 extern struct sk_buff		*sock_rmalloc(struct sock *sk,
751 					      unsigned long size, int force,
752 					      unsigned int __nocast priority);
753 extern void			sock_wfree(struct sk_buff *skb);
754 extern void			sock_rfree(struct sk_buff *skb);
755 
756 extern int			sock_setsockopt(struct socket *sock, int level,
757 						int op, char __user *optval,
758 						int optlen);
759 
760 extern int			sock_getsockopt(struct socket *sock, int level,
761 						int op, char __user *optval,
762 						int __user *optlen);
763 extern struct sk_buff 		*sock_alloc_send_skb(struct sock *sk,
764 						     unsigned long size,
765 						     int noblock,
766 						     int *errcode);
767 extern void *sock_kmalloc(struct sock *sk, int size,
768 			  unsigned int __nocast priority);
769 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
770 extern void sk_send_sigurg(struct sock *sk);
771 
772 /*
773  * Functions to fill in entries in struct proto_ops when a protocol
774  * does not implement a particular function.
775  */
776 extern int                      sock_no_bind(struct socket *,
777 					     struct sockaddr *, int);
778 extern int                      sock_no_connect(struct socket *,
779 						struct sockaddr *, int, int);
780 extern int                      sock_no_socketpair(struct socket *,
781 						   struct socket *);
782 extern int                      sock_no_accept(struct socket *,
783 					       struct socket *, int);
784 extern int                      sock_no_getname(struct socket *,
785 						struct sockaddr *, int *, int);
786 extern unsigned int             sock_no_poll(struct file *, struct socket *,
787 					     struct poll_table_struct *);
788 extern int                      sock_no_ioctl(struct socket *, unsigned int,
789 					      unsigned long);
790 extern int			sock_no_listen(struct socket *, int);
791 extern int                      sock_no_shutdown(struct socket *, int);
792 extern int			sock_no_getsockopt(struct socket *, int , int,
793 						   char __user *, int __user *);
794 extern int			sock_no_setsockopt(struct socket *, int, int,
795 						   char __user *, int);
796 extern int                      sock_no_sendmsg(struct kiocb *, struct socket *,
797 						struct msghdr *, size_t);
798 extern int                      sock_no_recvmsg(struct kiocb *, struct socket *,
799 						struct msghdr *, size_t, int);
800 extern int			sock_no_mmap(struct file *file,
801 					     struct socket *sock,
802 					     struct vm_area_struct *vma);
803 extern ssize_t			sock_no_sendpage(struct socket *sock,
804 						struct page *page,
805 						int offset, size_t size,
806 						int flags);
807 
808 /*
809  * Functions to fill in entries in struct proto_ops when a protocol
810  * uses the inet style.
811  */
812 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
813 				  char __user *optval, int __user *optlen);
814 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
815 			       struct msghdr *msg, size_t size, int flags);
816 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
817 				  char __user *optval, int optlen);
818 
819 extern void sk_common_release(struct sock *sk);
820 
821 /*
822  *	Default socket callbacks and setup code
823  */
824 
825 /* Initialise core socket variables */
826 extern void sock_init_data(struct socket *sock, struct sock *sk);
827 
828 /**
829  *	sk_filter - run a packet through a socket filter
830  *	@sk: sock associated with &sk_buff
831  *	@skb: buffer to filter
832  *	@needlock: set to 1 if the sock is not locked by caller.
833  *
834  * Run the filter code and then cut skb->data to correct size returned by
835  * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
836  * than pkt_len we keep whole skb->data. This is the socket level
837  * wrapper to sk_run_filter. It returns 0 if the packet should
838  * be accepted or -EPERM if the packet should be tossed.
839  *
840  */
841 
842 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
843 {
844 	int err;
845 
846 	err = security_sock_rcv_skb(sk, skb);
847 	if (err)
848 		return err;
849 
850 	if (sk->sk_filter) {
851 		struct sk_filter *filter;
852 
853 		if (needlock)
854 			bh_lock_sock(sk);
855 
856 		filter = sk->sk_filter;
857 		if (filter) {
858 			int pkt_len = sk_run_filter(skb, filter->insns,
859 						    filter->len);
860 			if (!pkt_len)
861 				err = -EPERM;
862 			else
863 				skb_trim(skb, pkt_len);
864 		}
865 
866 		if (needlock)
867 			bh_unlock_sock(sk);
868 	}
869 	return err;
870 }
871 
872 /**
873  *	sk_filter_release: Release a socket filter
874  *	@sk: socket
875  *	@fp: filter to remove
876  *
877  *	Remove a filter from a socket and release its resources.
878  */
879 
880 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
881 {
882 	unsigned int size = sk_filter_len(fp);
883 
884 	atomic_sub(size, &sk->sk_omem_alloc);
885 
886 	if (atomic_dec_and_test(&fp->refcnt))
887 		kfree(fp);
888 }
889 
890 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
891 {
892 	atomic_inc(&fp->refcnt);
893 	atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
894 }
895 
896 /*
897  * Socket reference counting postulates.
898  *
899  * * Each user of socket SHOULD hold a reference count.
900  * * Each access point to socket (an hash table bucket, reference from a list,
901  *   running timer, skb in flight MUST hold a reference count.
902  * * When reference count hits 0, it means it will never increase back.
903  * * When reference count hits 0, it means that no references from
904  *   outside exist to this socket and current process on current CPU
905  *   is last user and may/should destroy this socket.
906  * * sk_free is called from any context: process, BH, IRQ. When
907  *   it is called, socket has no references from outside -> sk_free
908  *   may release descendant resources allocated by the socket, but
909  *   to the time when it is called, socket is NOT referenced by any
910  *   hash tables, lists etc.
911  * * Packets, delivered from outside (from network or from another process)
912  *   and enqueued on receive/error queues SHOULD NOT grab reference count,
913  *   when they sit in queue. Otherwise, packets will leak to hole, when
914  *   socket is looked up by one cpu and unhasing is made by another CPU.
915  *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
916  *   (leak to backlog). Packet socket does all the processing inside
917  *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
918  *   use separate SMP lock, so that they are prone too.
919  */
920 
921 /* Ungrab socket and destroy it, if it was the last reference. */
922 static inline void sock_put(struct sock *sk)
923 {
924 	if (atomic_dec_and_test(&sk->sk_refcnt))
925 		sk_free(sk);
926 }
927 
928 /* Detach socket from process context.
929  * Announce socket dead, detach it from wait queue and inode.
930  * Note that parent inode held reference count on this struct sock,
931  * we do not release it in this function, because protocol
932  * probably wants some additional cleanups or even continuing
933  * to work with this socket (TCP).
934  */
935 static inline void sock_orphan(struct sock *sk)
936 {
937 	write_lock_bh(&sk->sk_callback_lock);
938 	sock_set_flag(sk, SOCK_DEAD);
939 	sk->sk_socket = NULL;
940 	sk->sk_sleep  = NULL;
941 	write_unlock_bh(&sk->sk_callback_lock);
942 }
943 
944 static inline void sock_graft(struct sock *sk, struct socket *parent)
945 {
946 	write_lock_bh(&sk->sk_callback_lock);
947 	sk->sk_sleep = &parent->wait;
948 	parent->sk = sk;
949 	sk->sk_socket = parent;
950 	write_unlock_bh(&sk->sk_callback_lock);
951 }
952 
953 extern int sock_i_uid(struct sock *sk);
954 extern unsigned long sock_i_ino(struct sock *sk);
955 
956 static inline struct dst_entry *
957 __sk_dst_get(struct sock *sk)
958 {
959 	return sk->sk_dst_cache;
960 }
961 
962 static inline struct dst_entry *
963 sk_dst_get(struct sock *sk)
964 {
965 	struct dst_entry *dst;
966 
967 	read_lock(&sk->sk_dst_lock);
968 	dst = sk->sk_dst_cache;
969 	if (dst)
970 		dst_hold(dst);
971 	read_unlock(&sk->sk_dst_lock);
972 	return dst;
973 }
974 
975 static inline void
976 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
977 {
978 	struct dst_entry *old_dst;
979 
980 	old_dst = sk->sk_dst_cache;
981 	sk->sk_dst_cache = dst;
982 	dst_release(old_dst);
983 }
984 
985 static inline void
986 sk_dst_set(struct sock *sk, struct dst_entry *dst)
987 {
988 	write_lock(&sk->sk_dst_lock);
989 	__sk_dst_set(sk, dst);
990 	write_unlock(&sk->sk_dst_lock);
991 }
992 
993 static inline void
994 __sk_dst_reset(struct sock *sk)
995 {
996 	struct dst_entry *old_dst;
997 
998 	old_dst = sk->sk_dst_cache;
999 	sk->sk_dst_cache = NULL;
1000 	dst_release(old_dst);
1001 }
1002 
1003 static inline void
1004 sk_dst_reset(struct sock *sk)
1005 {
1006 	write_lock(&sk->sk_dst_lock);
1007 	__sk_dst_reset(sk);
1008 	write_unlock(&sk->sk_dst_lock);
1009 }
1010 
1011 static inline struct dst_entry *
1012 __sk_dst_check(struct sock *sk, u32 cookie)
1013 {
1014 	struct dst_entry *dst = sk->sk_dst_cache;
1015 
1016 	if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1017 		sk->sk_dst_cache = NULL;
1018 		dst_release(dst);
1019 		return NULL;
1020 	}
1021 
1022 	return dst;
1023 }
1024 
1025 static inline struct dst_entry *
1026 sk_dst_check(struct sock *sk, u32 cookie)
1027 {
1028 	struct dst_entry *dst = sk_dst_get(sk);
1029 
1030 	if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1031 		sk_dst_reset(sk);
1032 		dst_release(dst);
1033 		return NULL;
1034 	}
1035 
1036 	return dst;
1037 }
1038 
1039 static inline void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1040 {
1041 	__sk_dst_set(sk, dst);
1042 	sk->sk_route_caps = dst->dev->features;
1043 	if (sk->sk_route_caps & NETIF_F_TSO) {
1044 		if (sock_flag(sk, SOCK_NO_LARGESEND) || dst->header_len)
1045 			sk->sk_route_caps &= ~NETIF_F_TSO;
1046 	}
1047 }
1048 
1049 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1050 {
1051 	sk->sk_wmem_queued   += skb->truesize;
1052 	sk->sk_forward_alloc -= skb->truesize;
1053 }
1054 
1055 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1056 				   struct sk_buff *skb, struct page *page,
1057 				   int off, int copy)
1058 {
1059 	if (skb->ip_summed == CHECKSUM_NONE) {
1060 		int err = 0;
1061 		unsigned int csum = csum_and_copy_from_user(from,
1062 						     page_address(page) + off,
1063 							    copy, 0, &err);
1064 		if (err)
1065 			return err;
1066 		skb->csum = csum_block_add(skb->csum, csum, skb->len);
1067 	} else if (copy_from_user(page_address(page) + off, from, copy))
1068 		return -EFAULT;
1069 
1070 	skb->len	     += copy;
1071 	skb->data_len	     += copy;
1072 	skb->truesize	     += copy;
1073 	sk->sk_wmem_queued   += copy;
1074 	sk->sk_forward_alloc -= copy;
1075 	return 0;
1076 }
1077 
1078 /*
1079  * 	Queue a received datagram if it will fit. Stream and sequenced
1080  *	protocols can't normally use this as they need to fit buffers in
1081  *	and play with them.
1082  *
1083  * 	Inlined as it's very short and called for pretty much every
1084  *	packet ever received.
1085  */
1086 
1087 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1088 {
1089 	sock_hold(sk);
1090 	skb->sk = sk;
1091 	skb->destructor = sock_wfree;
1092 	atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1093 }
1094 
1095 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1096 {
1097 	skb->sk = sk;
1098 	skb->destructor = sock_rfree;
1099 	atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1100 }
1101 
1102 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1103 			   unsigned long expires);
1104 
1105 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1106 
1107 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1108 {
1109 	int err = 0;
1110 	int skb_len;
1111 
1112 	/* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1113 	   number of warnings when compiling with -W --ANK
1114 	 */
1115 	if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1116 	    (unsigned)sk->sk_rcvbuf) {
1117 		err = -ENOMEM;
1118 		goto out;
1119 	}
1120 
1121 	/* It would be deadlock, if sock_queue_rcv_skb is used
1122 	   with socket lock! We assume that users of this
1123 	   function are lock free.
1124 	*/
1125 	err = sk_filter(sk, skb, 1);
1126 	if (err)
1127 		goto out;
1128 
1129 	skb->dev = NULL;
1130 	skb_set_owner_r(skb, sk);
1131 
1132 	/* Cache the SKB length before we tack it onto the receive
1133 	 * queue.  Once it is added it no longer belongs to us and
1134 	 * may be freed by other threads of control pulling packets
1135 	 * from the queue.
1136 	 */
1137 	skb_len = skb->len;
1138 
1139 	skb_queue_tail(&sk->sk_receive_queue, skb);
1140 
1141 	if (!sock_flag(sk, SOCK_DEAD))
1142 		sk->sk_data_ready(sk, skb_len);
1143 out:
1144 	return err;
1145 }
1146 
1147 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1148 {
1149 	/* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1150 	   number of warnings when compiling with -W --ANK
1151 	 */
1152 	if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1153 	    (unsigned)sk->sk_rcvbuf)
1154 		return -ENOMEM;
1155 	skb_set_owner_r(skb, sk);
1156 	skb_queue_tail(&sk->sk_error_queue, skb);
1157 	if (!sock_flag(sk, SOCK_DEAD))
1158 		sk->sk_data_ready(sk, skb->len);
1159 	return 0;
1160 }
1161 
1162 /*
1163  *	Recover an error report and clear atomically
1164  */
1165 
1166 static inline int sock_error(struct sock *sk)
1167 {
1168 	int err = xchg(&sk->sk_err, 0);
1169 	return -err;
1170 }
1171 
1172 static inline unsigned long sock_wspace(struct sock *sk)
1173 {
1174 	int amt = 0;
1175 
1176 	if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1177 		amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1178 		if (amt < 0)
1179 			amt = 0;
1180 	}
1181 	return amt;
1182 }
1183 
1184 static inline void sk_wake_async(struct sock *sk, int how, int band)
1185 {
1186 	if (sk->sk_socket && sk->sk_socket->fasync_list)
1187 		sock_wake_async(sk->sk_socket, how, band);
1188 }
1189 
1190 #define SOCK_MIN_SNDBUF 2048
1191 #define SOCK_MIN_RCVBUF 256
1192 
1193 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1194 {
1195 	if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1196 		sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1197 		sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1198 	}
1199 }
1200 
1201 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1202 						   int size, int mem,
1203 						   unsigned int __nocast gfp)
1204 {
1205 	struct sk_buff *skb;
1206 	int hdr_len;
1207 
1208 	hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header);
1209 	skb = alloc_skb_fclone(size + hdr_len, gfp);
1210 	if (skb) {
1211 		skb->truesize += mem;
1212 		if (sk_stream_wmem_schedule(sk, skb->truesize)) {
1213 			skb_reserve(skb, hdr_len);
1214 			return skb;
1215 		}
1216 		__kfree_skb(skb);
1217 	} else {
1218 		sk->sk_prot->enter_memory_pressure();
1219 		sk_stream_moderate_sndbuf(sk);
1220 	}
1221 	return NULL;
1222 }
1223 
1224 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1225 						  int size,
1226 						  unsigned int __nocast gfp)
1227 {
1228 	return sk_stream_alloc_pskb(sk, size, 0, gfp);
1229 }
1230 
1231 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1232 {
1233 	struct page *page = NULL;
1234 
1235 	page = alloc_pages(sk->sk_allocation, 0);
1236 	if (!page) {
1237 		sk->sk_prot->enter_memory_pressure();
1238 		sk_stream_moderate_sndbuf(sk);
1239 	}
1240 	return page;
1241 }
1242 
1243 #define sk_stream_for_retrans_queue(skb, sk)				\
1244 		for (skb = (sk)->sk_write_queue.next;			\
1245 		     (skb != (sk)->sk_send_head) &&			\
1246 		     (skb != (struct sk_buff *)&(sk)->sk_write_queue);	\
1247 		     skb = skb->next)
1248 
1249 /*
1250  *	Default write policy as shown to user space via poll/select/SIGIO
1251  */
1252 static inline int sock_writeable(const struct sock *sk)
1253 {
1254 	return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1255 }
1256 
1257 static inline unsigned int __nocast gfp_any(void)
1258 {
1259 	return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1260 }
1261 
1262 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1263 {
1264 	return noblock ? 0 : sk->sk_rcvtimeo;
1265 }
1266 
1267 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1268 {
1269 	return noblock ? 0 : sk->sk_sndtimeo;
1270 }
1271 
1272 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1273 {
1274 	return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1275 }
1276 
1277 /* Alas, with timeout socket operations are not restartable.
1278  * Compare this to poll().
1279  */
1280 static inline int sock_intr_errno(long timeo)
1281 {
1282 	return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1283 }
1284 
1285 static __inline__ void
1286 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1287 {
1288 	struct timeval stamp;
1289 
1290 	skb_get_timestamp(skb, &stamp);
1291 	if (sock_flag(sk, SOCK_RCVTSTAMP)) {
1292 		/* Race occurred between timestamp enabling and packet
1293 		   receiving.  Fill in the current time for now. */
1294 		if (stamp.tv_sec == 0)
1295 			do_gettimeofday(&stamp);
1296 		skb_set_timestamp(skb, &stamp);
1297 		put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1298 			 &stamp);
1299 	} else
1300 		sk->sk_stamp = stamp;
1301 }
1302 
1303 /**
1304  * sk_eat_skb - Release a skb if it is no longer needed
1305  * @sk: socket to eat this skb from
1306  * @skb: socket buffer to eat
1307  *
1308  * This routine must be called with interrupts disabled or with the socket
1309  * locked so that the sk_buff queue operation is ok.
1310 */
1311 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1312 {
1313 	__skb_unlink(skb, &sk->sk_receive_queue);
1314 	__kfree_skb(skb);
1315 }
1316 
1317 extern void sock_enable_timestamp(struct sock *sk);
1318 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1319 
1320 /*
1321  *	Enable debug/info messages
1322  */
1323 
1324 #if 0
1325 #define NETDEBUG(fmt, args...)	do { } while (0)
1326 #define LIMIT_NETDEBUG(fmt, args...) do { } while(0)
1327 #else
1328 #define NETDEBUG(fmt, args...)	printk(fmt,##args)
1329 #define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0)
1330 #endif
1331 
1332 /*
1333  * Macros for sleeping on a socket. Use them like this:
1334  *
1335  * SOCK_SLEEP_PRE(sk)
1336  * if (condition)
1337  * 	schedule();
1338  * SOCK_SLEEP_POST(sk)
1339  *
1340  * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1341  * and when the last use of them in DECnet has gone, I'm intending to
1342  * remove them.
1343  */
1344 
1345 #define SOCK_SLEEP_PRE(sk) 	{ struct task_struct *tsk = current; \
1346 				DECLARE_WAITQUEUE(wait, tsk); \
1347 				tsk->state = TASK_INTERRUPTIBLE; \
1348 				add_wait_queue((sk)->sk_sleep, &wait); \
1349 				release_sock(sk);
1350 
1351 #define SOCK_SLEEP_POST(sk)	tsk->state = TASK_RUNNING; \
1352 				remove_wait_queue((sk)->sk_sleep, &wait); \
1353 				lock_sock(sk); \
1354 				}
1355 
1356 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1357 {
1358 	if (valbool)
1359 		sock_set_flag(sk, bit);
1360 	else
1361 		sock_reset_flag(sk, bit);
1362 }
1363 
1364 extern __u32 sysctl_wmem_max;
1365 extern __u32 sysctl_rmem_max;
1366 
1367 #ifdef CONFIG_NET
1368 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1369 #else
1370 static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
1371 {
1372 	return -ENODEV;
1373 }
1374 #endif
1375 
1376 extern void sk_init(void);
1377 
1378 #ifdef CONFIG_SYSCTL
1379 extern struct ctl_table core_table[];
1380 #endif
1381 
1382 extern int sysctl_optmem_max;
1383 
1384 extern __u32 sysctl_wmem_default;
1385 extern __u32 sysctl_rmem_default;
1386 
1387 #endif	/* _SOCK_H */
1388