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.
6 * Definitions for the AF_INET socket handler.
8 * Version: @(#)sock.h 1.0.4 05/13/93
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>
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
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
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.
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 */
53 #include <linux/security.h>
55 #include <linux/filter.h>
56 #include <linux/rculist_nulls.h>
58 #include <asm/atomic.h>
60 #include <net/checksum.h>
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.
68 /* Define this to get the SOCK_DBG debugging facility. */
69 #define SOCK_DEBUGGING
71 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
72 printk(KERN_DEBUG msg); } while (0)
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
, ...)
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.
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
95 #ifdef CONFIG_DEBUG_LOCK_ALLOC
96 struct lockdep_map dep_map
;
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
118 * This is the minimal network layer representation of sockets, the header
119 * for struct sock and struct inet_timewait_sock.
122 unsigned short skc_family
;
123 volatile unsigned char skc_state
;
124 unsigned char skc_reuse
;
125 int skc_bound_dev_if
;
127 struct hlist_node skc_node
;
128 struct hlist_nulls_node skc_nulls_node
;
130 struct hlist_node skc_bind_node
;
132 unsigned int skc_hash
;
133 struct proto
*skc_prot
;
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
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
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,
225 kmemcheck_bitfield_end(flags
);
226 unsigned char sk_protocol
;
227 unsigned short sk_type
;
229 socket_lock_t sk_lock
;
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.
236 struct sk_buff
*head
;
237 struct sk_buff
*tail
;
239 wait_queue_head_t
*sk_sleep
;
240 struct dst_entry
*sk_dst_cache
;
242 struct xfrm_policy
*sk_policy
[2];
244 rwlock_t sk_dst_lock
;
245 atomic_t sk_rmem_alloc
;
246 atomic_t sk_wmem_alloc
;
247 atomic_t sk_omem_alloc
;
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
;
255 int sk_forward_alloc
;
259 unsigned int sk_gso_max_size
;
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
;
269 unsigned short sk_ack_backlog
;
270 unsigned short sk_max_ack_backlog
;
272 struct ucred sk_peercred
;
275 struct sk_filter
*sk_filter
;
277 struct timer_list sk_timer
;
279 struct socket
*sk_socket
;
281 struct page
*sk_sndmsg_page
;
282 struct sk_buff
*sk_send_head
;
284 int sk_write_pending
;
285 #ifdef CONFIG_SECURITY
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
);
300 * Hashed lists helper routines
302 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
304 return hlist_entry(head
->first
, struct sock
, sk_node
);
307 static inline struct sock
*sk_head(const struct hlist_head
*head
)
309 return hlist_empty(head
) ? NULL
: __sk_head(head
);
312 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
314 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
317 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
319 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
322 static inline struct sock
*sk_next(const struct sock
*sk
)
324 return sk
->sk_node
.next
?
325 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
328 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
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
) :
336 static inline int sk_unhashed(const struct sock
*sk
)
338 return hlist_unhashed(&sk
->sk_node
);
341 static inline int sk_hashed(const struct sock
*sk
)
343 return !sk_unhashed(sk
);
346 static __inline__
void sk_node_init(struct hlist_node
*node
)
351 static __inline__
void sk_nulls_node_init(struct hlist_nulls_node
*node
)
356 static __inline__
void __sk_del_node(struct sock
*sk
)
358 __hlist_del(&sk
->sk_node
);
361 static __inline__
int __sk_del_node_init(struct sock
*sk
)
365 sk_node_init(&sk
->sk_node
);
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
377 static inline void sock_hold(struct sock
*sk
)
379 atomic_inc(&sk
->sk_refcnt
);
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.
385 static inline void __sock_put(struct sock
*sk
)
387 atomic_dec(&sk
->sk_refcnt
);
390 static __inline__
int sk_del_node_init(struct sock
*sk
)
392 int rc
= __sk_del_node_init(sk
);
395 /* paranoid for a while -acme */
396 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
402 static __inline__
int __sk_nulls_del_node_init_rcu(struct sock
*sk
)
405 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
411 static __inline__
int sk_nulls_del_node_init_rcu(struct sock
*sk
)
413 int rc
= __sk_nulls_del_node_init_rcu(sk
);
416 /* paranoid for a while -acme */
417 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
423 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
425 hlist_add_head(&sk
->sk_node
, list
);
428 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
431 __sk_add_node(sk
, list
);
434 static __inline__
void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
436 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
439 static __inline__
void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
442 __sk_nulls_add_node_rcu(sk
, list
);
445 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
447 __hlist_del(&sk
->sk_bind_node
);
450 static __inline__
void sk_add_bind_node(struct sock
*sk
,
451 struct hlist_head
*list
)
453 hlist_add_head(&sk
->sk_bind_node
, list
);
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)
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 */
502 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
504 nsk
->sk_flags
= osk
->sk_flags
;
507 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
509 __set_bit(flag
, &sk
->sk_flags
);
512 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
514 __clear_bit(flag
, &sk
->sk_flags
);
517 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
519 return test_bit(flag
, &sk
->sk_flags
);
522 static inline void sk_acceptq_removed(struct sock
*sk
)
524 sk
->sk_ack_backlog
--;
527 static inline void sk_acceptq_added(struct sock
*sk
)
529 sk
->sk_ack_backlog
++;
532 static inline int sk_acceptq_is_full(struct sock
*sk
)
534 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
538 * Compute minimal free write space needed to queue new packets.
540 static inline int sk_stream_min_wspace(struct sock
*sk
)
542 return sk
->sk_wmem_queued
>> 1;
545 static inline int sk_stream_wspace(struct sock
*sk
)
547 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
550 extern void sk_stream_write_space(struct sock
*sk
);
552 static inline int sk_stream_memory_free(struct sock
*sk
)
554 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
557 /* The per-socket spinlock must be held here. */
558 static inline void sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
560 if (!sk
->sk_backlog
.tail
) {
561 sk
->sk_backlog
.head
= sk
->sk_backlog
.tail
= skb
;
563 sk
->sk_backlog
.tail
->next
= skb
;
564 sk
->sk_backlog
.tail
= skb
;
569 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
571 return sk
->sk_backlog_rcv(sk
, skb
);
574 #define sk_wait_event(__sk, __timeo, __condition) \
576 release_sock(__sk); \
577 __rc = __condition; \
579 *(__timeo) = schedule_timeout(*(__timeo)); \
582 __rc = __condition; \
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
);
592 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
594 struct request_sock_ops
;
595 struct timewait_sock_ops
;
596 struct inet_hashinfo
;
599 /* Networking protocol blocks we attach to sockets.
600 * socket layer -> transport layer interface
601 * transport -> network interface is defined by struct inet_proto
604 void (*close
)(struct sock
*sk
,
606 int (*connect
)(struct sock
*sk
,
607 struct sockaddr
*uaddr
,
609 int (*disconnect
)(struct sock
*sk
, int flags
);
611 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
613 int (*ioctl
)(struct sock
*sk
, int cmd
,
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
,
621 int (*getsockopt
)(struct sock
*sk
, int level
,
622 int optname
, char __user
*optval
,
625 int (*compat_setsockopt
)(struct sock
*sk
,
627 int optname
, char __user
*optval
,
629 int (*compat_getsockopt
)(struct sock
*sk
,
631 int optname
, char __user
*optval
,
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
,
638 size_t len
, int noblock
, int flags
,
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
);
645 int (*backlog_rcv
) (struct sock
*sk
,
646 struct sk_buff
*skb
);
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
);
653 /* Keeping track of sockets in use */
654 #ifdef CONFIG_PROC_FS
655 unsigned int inuse_idx
;
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. */
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.
668 int *memory_pressure
;
674 struct kmem_cache
*slab
;
675 unsigned int obj_size
;
678 struct percpu_counter
*orphan_count
;
680 struct request_sock_ops
*rsk_prot
;
681 struct timewait_sock_ops
*twsk_prot
;
684 struct inet_hashinfo
*hashinfo
;
685 struct udp_table
*udp_table
;
686 struct raw_hashinfo
*raw_hash
;
689 struct module
*owner
;
693 struct list_head node
;
694 #ifdef SOCK_REFCNT_DEBUG
699 extern int proto_register(struct proto
*prot
, int alloc_slab
);
700 extern void proto_unregister(struct proto
*prot
);
702 #ifdef SOCK_REFCNT_DEBUG
703 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
705 atomic_inc(&sk
->sk_prot
->socks
);
708 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
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
));
715 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
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
));
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 */
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
);
733 static void inline sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
740 /* With per-bucket locks this operation is not-atomic, so that
741 * this version is not worse.
743 static inline void __sk_prot_rehash(struct sock
*sk
)
745 sk
->sk_prot
->unhash(sk
);
746 sk
->sk_prot
->hash(sk
);
749 /* About 10 seconds */
750 #define SOCK_DESTROY_TIME (10*HZ)
752 /* Sockets 0-1023 can't be bound to unless you are superuser */
753 #define PROT_SOCK 1024
755 #define SHUTDOWN_MASK 3
756 #define RCV_SHUTDOWN 1
757 #define SEND_SHUTDOWN 2
759 #define SOCK_SNDBUF_LOCK 1
760 #define SOCK_RCVBUF_LOCK 2
761 #define SOCK_BINDADDR_LOCK 4
762 #define SOCK_BINDPORT_LOCK 8
764 /* sock_iocb: used to kick off async processing of socket ios */
766 struct list_head list
;
772 struct scm_cookie
*scm
;
773 struct msghdr
*msg
, async_msg
;
777 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
779 return (struct sock_iocb
*)iocb
->private;
782 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
787 struct socket_alloc
{
788 struct socket socket
;
789 struct inode vfs_inode
;
792 static inline struct socket
*SOCKET_I(struct inode
*inode
)
794 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
797 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
799 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
803 * Functions for memory accounting
805 extern int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
806 extern void __sk_mem_reclaim(struct sock
*sk
);
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
813 static inline int sk_mem_pages(int amt
)
815 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
818 static inline int sk_has_account(struct sock
*sk
)
820 /* return true if protocol supports memory accounting */
821 return !!sk
->sk_prot
->memory_allocated
;
824 static inline int sk_wmem_schedule(struct sock
*sk
, int size
)
826 if (!sk_has_account(sk
))
828 return size
<= sk
->sk_forward_alloc
||
829 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
832 static inline int sk_rmem_schedule(struct sock
*sk
, int size
)
834 if (!sk_has_account(sk
))
836 return size
<= sk
->sk_forward_alloc
||
837 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
);
840 static inline void sk_mem_reclaim(struct sock
*sk
)
842 if (!sk_has_account(sk
))
844 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
845 __sk_mem_reclaim(sk
);
848 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
850 if (!sk_has_account(sk
))
852 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
853 __sk_mem_reclaim(sk
);
856 static inline void sk_mem_charge(struct sock
*sk
, int size
)
858 if (!sk_has_account(sk
))
860 sk
->sk_forward_alloc
-= size
;
863 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
865 if (!sk_has_account(sk
))
867 sk
->sk_forward_alloc
+= size
;
870 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
872 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
873 sk
->sk_wmem_queued
-= skb
->truesize
;
874 sk_mem_uncharge(sk
, skb
->truesize
);
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.
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.
888 * Since ~2.3.5 it is also exclusive sleep lock serializing
889 * accesses from user process context.
891 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
894 * Macro so as to not evaluate some arguments when
895 * lockdep is not enabled.
897 * Mark both the sk_lock and the sk_lock.slock as a
898 * per-address-family lock class.
900 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
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, \
909 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
912 extern void lock_sock_nested(struct sock
*sk
, int subclass
);
914 static inline void lock_sock(struct sock
*sk
)
916 lock_sock_nested(sk
, 0);
919 extern void release_sock(struct sock
*sk
);
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))
928 extern struct sock
*sk_alloc(struct net
*net
, int family
,
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
);
936 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
937 unsigned long size
, int force
,
939 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
940 unsigned long size
, int force
,
942 extern void sock_wfree(struct sk_buff
*skb
);
943 extern void sock_rfree(struct sk_buff
*skb
);
945 extern int sock_setsockopt(struct socket
*sock
, int level
,
946 int op
, char __user
*optval
,
949 extern int sock_getsockopt(struct socket
*sock
, int level
,
950 int op
, char __user
*optval
,
952 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
956 extern struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
,
957 unsigned long header_len
,
958 unsigned long data_len
,
961 extern void *sock_kmalloc(struct sock
*sk
, int size
,
963 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
964 extern void sk_send_sigurg(struct sock
*sk
);
967 * Functions to fill in entries in struct proto_ops when a protocol
968 * does not implement a particular function.
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
*,
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,
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,
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
,
996 struct vm_area_struct
*vma
);
997 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
999 int offset
, size_t size
,
1003 * Functions to fill in entries in struct proto_ops when a protocol
1004 * uses the inet style.
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
);
1017 extern void sk_common_release(struct sock
*sk
);
1020 * Default socket callbacks and setup code
1023 /* Initialise core socket variables */
1024 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1027 * sk_filter_release: Release a socket filter
1028 * @fp: filter to remove
1030 * Remove a filter from a socket and release its resources.
1033 static inline void sk_filter_release(struct sk_filter
*fp
)
1035 if (atomic_dec_and_test(&fp
->refcnt
))
1039 static inline void sk_filter_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
1041 unsigned int size
= sk_filter_len(fp
);
1043 atomic_sub(size
, &sk
->sk_omem_alloc
);
1044 sk_filter_release(fp
);
1047 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
1049 atomic_inc(&fp
->refcnt
);
1050 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
1054 * Socket reference counting postulates.
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.
1078 /* Ungrab socket and destroy it, if it was the last reference. */
1079 static inline void sock_put(struct sock
*sk
)
1081 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1085 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1088 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1090 sk
->sk_socket
= sock
;
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).
1100 static inline void sock_orphan(struct sock
*sk
)
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
);
1109 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1111 write_lock_bh(&sk
->sk_callback_lock
);
1112 sk
->sk_sleep
= &parent
->wait
;
1114 sk_set_socket(sk
, parent
);
1115 security_sock_graft(sk
, parent
);
1116 write_unlock_bh(&sk
->sk_callback_lock
);
1119 extern int sock_i_uid(struct sock
*sk
);
1120 extern unsigned long sock_i_ino(struct sock
*sk
);
1122 static inline struct dst_entry
*
1123 __sk_dst_get(struct sock
*sk
)
1125 return sk
->sk_dst_cache
;
1128 static inline struct dst_entry
*
1129 sk_dst_get(struct sock
*sk
)
1131 struct dst_entry
*dst
;
1133 read_lock(&sk
->sk_dst_lock
);
1134 dst
= sk
->sk_dst_cache
;
1137 read_unlock(&sk
->sk_dst_lock
);
1142 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1144 struct dst_entry
*old_dst
;
1146 old_dst
= sk
->sk_dst_cache
;
1147 sk
->sk_dst_cache
= dst
;
1148 dst_release(old_dst
);
1152 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1154 write_lock(&sk
->sk_dst_lock
);
1155 __sk_dst_set(sk
, dst
);
1156 write_unlock(&sk
->sk_dst_lock
);
1160 __sk_dst_reset(struct sock
*sk
)
1162 struct dst_entry
*old_dst
;
1164 old_dst
= sk
->sk_dst_cache
;
1165 sk
->sk_dst_cache
= NULL
;
1166 dst_release(old_dst
);
1170 sk_dst_reset(struct sock
*sk
)
1172 write_lock(&sk
->sk_dst_lock
);
1174 write_unlock(&sk
->sk_dst_lock
);
1177 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1179 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1181 static inline int sk_can_gso(const struct sock
*sk
)
1183 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1186 extern void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1188 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1189 struct sk_buff
*skb
, struct page
*page
,
1192 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1194 __wsum csum
= csum_and_copy_from_user(from
,
1195 page_address(page
) + off
,
1199 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1200 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1204 skb
->data_len
+= copy
;
1205 skb
->truesize
+= copy
;
1206 sk
->sk_wmem_queued
+= copy
;
1207 sk_mem_charge(sk
, copy
);
1212 * sk_wmem_alloc_get - returns write allocations
1215 * Returns sk_wmem_alloc minus initial offset of one
1217 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1219 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1223 * sk_rmem_alloc_get - returns read allocations
1226 * Returns sk_rmem_alloc
1228 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1230 return atomic_read(&sk
->sk_rmem_alloc
);
1234 * sk_has_allocations - check if allocations are outstanding
1237 * Returns true if socket has write or read allocations
1239 static inline int sk_has_allocations(const struct sock
*sk
)
1241 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1245 * Queue a received datagram if it will fit. Stream and sequenced
1246 * protocols can't normally use this as they need to fit buffers in
1247 * and play with them.
1249 * Inlined as it's very short and called for pretty much every
1250 * packet ever received.
1253 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1256 skb
->destructor
= sock_wfree
;
1258 * We used to take a refcount on sk, but following operation
1259 * is enough to guarantee sk_free() wont free this sock until
1260 * all in-flight packets are completed
1262 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1265 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1268 skb
->destructor
= sock_rfree
;
1269 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1270 sk_mem_charge(sk
, skb
->truesize
);
1273 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1274 unsigned long expires
);
1276 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1278 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1280 static inline int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
)
1282 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1283 number of warnings when compiling with -W --ANK
1285 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
1286 (unsigned)sk
->sk_rcvbuf
)
1288 skb_set_owner_r(skb
, sk
);
1289 skb_queue_tail(&sk
->sk_error_queue
, skb
);
1290 if (!sock_flag(sk
, SOCK_DEAD
))
1291 sk
->sk_data_ready(sk
, skb
->len
);
1296 * Recover an error report and clear atomically
1299 static inline int sock_error(struct sock
*sk
)
1302 if (likely(!sk
->sk_err
))
1304 err
= xchg(&sk
->sk_err
, 0);
1308 static inline unsigned long sock_wspace(struct sock
*sk
)
1312 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1313 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1320 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1322 if (sk
->sk_socket
&& sk
->sk_socket
->fasync_list
)
1323 sock_wake_async(sk
->sk_socket
, how
, band
);
1326 #define SOCK_MIN_SNDBUF 2048
1327 #define SOCK_MIN_RCVBUF 256
1329 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1331 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1332 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
1333 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1337 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
1339 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1341 struct page
*page
= NULL
;
1343 page
= alloc_pages(sk
->sk_allocation
, 0);
1345 sk
->sk_prot
->enter_memory_pressure(sk
);
1346 sk_stream_moderate_sndbuf(sk
);
1352 * Default write policy as shown to user space via poll/select/SIGIO
1354 static inline int sock_writeable(const struct sock
*sk
)
1356 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
1359 static inline gfp_t
gfp_any(void)
1361 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
1364 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1366 return noblock
? 0 : sk
->sk_rcvtimeo
;
1369 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1371 return noblock
? 0 : sk
->sk_sndtimeo
;
1374 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1376 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1379 /* Alas, with timeout socket operations are not restartable.
1380 * Compare this to poll().
1382 static inline int sock_intr_errno(long timeo
)
1384 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
1387 extern void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
1388 struct sk_buff
*skb
);
1390 static __inline__
void
1391 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
1393 ktime_t kt
= skb
->tstamp
;
1394 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
1397 * generate control messages if
1398 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1399 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1400 * - software time stamp available and wanted
1401 * (SOCK_TIMESTAMPING_SOFTWARE)
1402 * - hardware time stamps available and wanted
1403 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1404 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1406 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
1407 sock_flag(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
) ||
1408 (kt
.tv64
&& sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) ||
1409 (hwtstamps
->hwtstamp
.tv64
&&
1410 sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
)) ||
1411 (hwtstamps
->syststamp
.tv64
&&
1412 sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
)))
1413 __sock_recv_timestamp(msg
, sk
, skb
);
1419 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1420 * @msg: outgoing packet
1421 * @sk: socket sending this packet
1422 * @shtx: filled with instructions for time stamping
1424 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1425 * parameters are invalid.
1427 extern int sock_tx_timestamp(struct msghdr
*msg
,
1429 union skb_shared_tx
*shtx
);
1433 * sk_eat_skb - Release a skb if it is no longer needed
1434 * @sk: socket to eat this skb from
1435 * @skb: socket buffer to eat
1436 * @copied_early: flag indicating whether DMA operations copied this data early
1438 * This routine must be called with interrupts disabled or with the socket
1439 * locked so that the sk_buff queue operation is ok.
1441 #ifdef CONFIG_NET_DMA
1442 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1444 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1448 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
1451 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1453 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1459 struct net
*sock_net(const struct sock
*sk
)
1461 #ifdef CONFIG_NET_NS
1469 void sock_net_set(struct sock
*sk
, struct net
*net
)
1471 #ifdef CONFIG_NET_NS
1477 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1478 * They should not hold a referrence to a namespace in order to allow
1480 * Sockets after sk_change_net should be released using sk_release_kernel
1482 static inline void sk_change_net(struct sock
*sk
, struct net
*net
)
1484 put_net(sock_net(sk
));
1485 sock_net_set(sk
, hold_net(net
));
1488 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
1490 if (unlikely(skb
->sk
)) {
1491 struct sock
*sk
= skb
->sk
;
1493 skb
->destructor
= NULL
;
1500 extern void sock_enable_timestamp(struct sock
*sk
, int flag
);
1501 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
1502 extern int sock_get_timestampns(struct sock
*, struct timespec __user
*);
1505 * Enable debug/info messages
1507 extern int net_msg_warn
;
1508 #define NETDEBUG(fmt, args...) \
1509 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1511 #define LIMIT_NETDEBUG(fmt, args...) \
1512 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1514 extern __u32 sysctl_wmem_max
;
1515 extern __u32 sysctl_rmem_max
;
1517 extern void sk_init(void);
1519 extern int sysctl_optmem_max
;
1521 extern __u32 sysctl_wmem_default
;
1522 extern __u32 sysctl_rmem_default
;
1524 #endif /* _SOCK_H */