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/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>
52 #include <linux/filter.h>
54 #include <asm/atomic.h>
56 #include <net/checksum.h>
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.
64 /* Define this to get the SOCK_DBG debugging facility. */
65 #define SOCK_DEBUGGING
67 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
68 printk(KERN_DEBUG msg); } while (0)
70 #define SOCK_DEBUG(sk, msg...) do { } while (0)
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.
80 struct sock_iocb
*owner
;
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)); \
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_hash: hash value used with various protocol lookup tables
103 * @skc_prot: protocol handlers inside a network family
105 * This is the minimal network layer representation of sockets, the header
106 * for struct sock and struct inet_timewait_sock.
109 unsigned short skc_family
;
110 volatile unsigned char skc_state
;
111 unsigned char skc_reuse
;
112 int skc_bound_dev_if
;
113 struct hlist_node skc_node
;
114 struct hlist_node skc_bind_node
;
116 unsigned int skc_hash
;
117 struct proto
*skc_prot
;
121 * struct sock - network layer representation of sockets
122 * @__sk_common: shared layout with inet_timewait_sock
123 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
124 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
125 * @sk_lock: synchronizer
126 * @sk_rcvbuf: size of receive buffer in bytes
127 * @sk_sleep: sock wait queue
128 * @sk_dst_cache: destination cache
129 * @sk_dst_lock: destination cache lock
130 * @sk_policy: flow policy
131 * @sk_rmem_alloc: receive queue bytes committed
132 * @sk_receive_queue: incoming packets
133 * @sk_wmem_alloc: transmit queue bytes committed
134 * @sk_write_queue: Packet sending queue
135 * @sk_async_wait_queue: DMA copied packets
136 * @sk_omem_alloc: "o" is "option" or "other"
137 * @sk_wmem_queued: persistent queue size
138 * @sk_forward_alloc: space allocated forward
139 * @sk_allocation: allocation mode
140 * @sk_sndbuf: size of send buffer in bytes
141 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
142 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
143 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
144 * @sk_lingertime: %SO_LINGER l_linger setting
145 * @sk_backlog: always used with the per-socket spinlock held
146 * @sk_callback_lock: used with the callbacks in the end of this struct
147 * @sk_error_queue: rarely used
148 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
149 * @sk_err: last error
150 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
151 * @sk_ack_backlog: current listen backlog
152 * @sk_max_ack_backlog: listen backlog set in listen()
153 * @sk_priority: %SO_PRIORITY setting
154 * @sk_type: socket type (%SOCK_STREAM, etc)
155 * @sk_protocol: which protocol this socket belongs in this network family
156 * @sk_peercred: %SO_PEERCRED setting
157 * @sk_rcvlowat: %SO_RCVLOWAT setting
158 * @sk_rcvtimeo: %SO_RCVTIMEO setting
159 * @sk_sndtimeo: %SO_SNDTIMEO setting
160 * @sk_filter: socket filtering instructions
161 * @sk_protinfo: private area, net family specific, when not using slab
162 * @sk_timer: sock cleanup timer
163 * @sk_stamp: time stamp of last packet received
164 * @sk_socket: Identd and reporting IO signals
165 * @sk_user_data: RPC layer private data
166 * @sk_sndmsg_page: cached page for sendmsg
167 * @sk_sndmsg_off: cached offset for sendmsg
168 * @sk_send_head: front of stuff to transmit
169 * @sk_security: used by security modules
170 * @sk_write_pending: a write to stream socket waits to start
171 * @sk_state_change: callback to indicate change in the state of the sock
172 * @sk_data_ready: callback to indicate there is data to be processed
173 * @sk_write_space: callback to indicate there is bf sending space available
174 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
175 * @sk_backlog_rcv: callback to process the backlog
176 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
180 * Now struct inet_timewait_sock also uses sock_common, so please just
181 * don't add nothing before this first member (__sk_common) --acme
183 struct sock_common __sk_common
;
184 #define sk_family __sk_common.skc_family
185 #define sk_state __sk_common.skc_state
186 #define sk_reuse __sk_common.skc_reuse
187 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
188 #define sk_node __sk_common.skc_node
189 #define sk_bind_node __sk_common.skc_bind_node
190 #define sk_refcnt __sk_common.skc_refcnt
191 #define sk_hash __sk_common.skc_hash
192 #define sk_prot __sk_common.skc_prot
193 unsigned char sk_shutdown
: 2,
196 unsigned char sk_protocol
;
197 unsigned short sk_type
;
199 socket_lock_t sk_lock
;
200 wait_queue_head_t
*sk_sleep
;
201 struct dst_entry
*sk_dst_cache
;
202 struct xfrm_policy
*sk_policy
[2];
203 rwlock_t sk_dst_lock
;
204 atomic_t sk_rmem_alloc
;
205 atomic_t sk_wmem_alloc
;
206 atomic_t sk_omem_alloc
;
207 struct sk_buff_head sk_receive_queue
;
208 struct sk_buff_head sk_write_queue
;
209 struct sk_buff_head sk_async_wait_queue
;
211 int sk_forward_alloc
;
216 unsigned long sk_flags
;
217 unsigned long sk_lingertime
;
219 * The backlog queue is special, it is always used with
220 * the per-socket spinlock held and requires low latency
221 * access. Therefore we special case it's implementation.
224 struct sk_buff
*head
;
225 struct sk_buff
*tail
;
227 struct sk_buff_head sk_error_queue
;
228 struct proto
*sk_prot_creator
;
229 rwlock_t sk_callback_lock
;
232 unsigned short sk_ack_backlog
;
233 unsigned short sk_max_ack_backlog
;
235 struct ucred sk_peercred
;
238 struct sk_filter
*sk_filter
;
240 struct timer_list sk_timer
;
241 struct timeval sk_stamp
;
242 struct socket
*sk_socket
;
244 struct page
*sk_sndmsg_page
;
245 struct sk_buff
*sk_send_head
;
247 int sk_write_pending
;
249 void (*sk_state_change
)(struct sock
*sk
);
250 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
251 void (*sk_write_space
)(struct sock
*sk
);
252 void (*sk_error_report
)(struct sock
*sk
);
253 int (*sk_backlog_rcv
)(struct sock
*sk
,
254 struct sk_buff
*skb
);
255 void (*sk_destruct
)(struct sock
*sk
);
259 * Hashed lists helper routines
261 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
263 return hlist_entry(head
->first
, struct sock
, sk_node
);
266 static inline struct sock
*sk_head(const struct hlist_head
*head
)
268 return hlist_empty(head
) ? NULL
: __sk_head(head
);
271 static inline struct sock
*sk_next(const struct sock
*sk
)
273 return sk
->sk_node
.next
?
274 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
277 static inline int sk_unhashed(const struct sock
*sk
)
279 return hlist_unhashed(&sk
->sk_node
);
282 static inline int sk_hashed(const struct sock
*sk
)
284 return !sk_unhashed(sk
);
287 static __inline__
void sk_node_init(struct hlist_node
*node
)
292 static __inline__
void __sk_del_node(struct sock
*sk
)
294 __hlist_del(&sk
->sk_node
);
297 static __inline__
int __sk_del_node_init(struct sock
*sk
)
301 sk_node_init(&sk
->sk_node
);
307 /* Grab socket reference count. This operation is valid only
308 when sk is ALREADY grabbed f.e. it is found in hash table
309 or a list and the lookup is made under lock preventing hash table
313 static inline void sock_hold(struct sock
*sk
)
315 atomic_inc(&sk
->sk_refcnt
);
318 /* Ungrab socket in the context, which assumes that socket refcnt
319 cannot hit zero, f.e. it is true in context of any socketcall.
321 static inline void __sock_put(struct sock
*sk
)
323 atomic_dec(&sk
->sk_refcnt
);
326 static __inline__
int sk_del_node_init(struct sock
*sk
)
328 int rc
= __sk_del_node_init(sk
);
331 /* paranoid for a while -acme */
332 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
338 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
340 hlist_add_head(&sk
->sk_node
, list
);
343 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
346 __sk_add_node(sk
, list
);
349 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
351 __hlist_del(&sk
->sk_bind_node
);
354 static __inline__
void sk_add_bind_node(struct sock
*sk
,
355 struct hlist_head
*list
)
357 hlist_add_head(&sk
->sk_bind_node
, list
);
360 #define sk_for_each(__sk, node, list) \
361 hlist_for_each_entry(__sk, node, list, sk_node)
362 #define sk_for_each_from(__sk, node) \
363 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
364 hlist_for_each_entry_from(__sk, node, sk_node)
365 #define sk_for_each_continue(__sk, node) \
366 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
367 hlist_for_each_entry_continue(__sk, node, sk_node)
368 #define sk_for_each_safe(__sk, node, tmp, list) \
369 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
370 #define sk_for_each_bound(__sk, node, list) \
371 hlist_for_each_entry(__sk, node, list, sk_bind_node)
384 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
385 SOCK_DBG
, /* %SO_DEBUG setting */
386 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
387 SOCK_NO_LARGESEND
, /* whether to sent large segments or not */
388 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
389 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
392 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
394 nsk
->sk_flags
= osk
->sk_flags
;
397 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
399 __set_bit(flag
, &sk
->sk_flags
);
402 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
404 __clear_bit(flag
, &sk
->sk_flags
);
407 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
409 return test_bit(flag
, &sk
->sk_flags
);
412 static inline void sk_acceptq_removed(struct sock
*sk
)
414 sk
->sk_ack_backlog
--;
417 static inline void sk_acceptq_added(struct sock
*sk
)
419 sk
->sk_ack_backlog
++;
422 static inline int sk_acceptq_is_full(struct sock
*sk
)
424 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
428 * Compute minimal free write space needed to queue new packets.
430 static inline int sk_stream_min_wspace(struct sock
*sk
)
432 return sk
->sk_wmem_queued
/ 2;
435 static inline int sk_stream_wspace(struct sock
*sk
)
437 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
440 extern void sk_stream_write_space(struct sock
*sk
);
442 static inline int sk_stream_memory_free(struct sock
*sk
)
444 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
447 extern void sk_stream_rfree(struct sk_buff
*skb
);
449 static inline void sk_stream_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
452 skb
->destructor
= sk_stream_rfree
;
453 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
454 sk
->sk_forward_alloc
-= skb
->truesize
;
457 static inline void sk_stream_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
459 skb_truesize_check(skb
);
460 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
461 sk
->sk_wmem_queued
-= skb
->truesize
;
462 sk
->sk_forward_alloc
+= skb
->truesize
;
466 /* The per-socket spinlock must be held here. */
467 static inline void sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
469 if (!sk
->sk_backlog
.tail
) {
470 sk
->sk_backlog
.head
= sk
->sk_backlog
.tail
= skb
;
472 sk
->sk_backlog
.tail
->next
= skb
;
473 sk
->sk_backlog
.tail
= skb
;
478 #define sk_wait_event(__sk, __timeo, __condition) \
480 release_sock(__sk); \
483 *(__timeo) = schedule_timeout(*(__timeo)); \
490 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
491 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
492 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
493 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
494 extern void sk_stream_kill_queues(struct sock
*sk
);
496 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
498 struct request_sock_ops
;
499 struct timewait_sock_ops
;
501 /* Networking protocol blocks we attach to sockets.
502 * socket layer -> transport layer interface
503 * transport -> network interface is defined by struct inet_proto
506 void (*close
)(struct sock
*sk
,
508 int (*connect
)(struct sock
*sk
,
509 struct sockaddr
*uaddr
,
511 int (*disconnect
)(struct sock
*sk
, int flags
);
513 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
515 int (*ioctl
)(struct sock
*sk
, int cmd
,
517 int (*init
)(struct sock
*sk
);
518 int (*destroy
)(struct sock
*sk
);
519 void (*shutdown
)(struct sock
*sk
, int how
);
520 int (*setsockopt
)(struct sock
*sk
, int level
,
521 int optname
, char __user
*optval
,
523 int (*getsockopt
)(struct sock
*sk
, int level
,
524 int optname
, char __user
*optval
,
526 int (*compat_setsockopt
)(struct sock
*sk
,
528 int optname
, char __user
*optval
,
530 int (*compat_getsockopt
)(struct sock
*sk
,
532 int optname
, char __user
*optval
,
534 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
535 struct msghdr
*msg
, size_t len
);
536 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
538 size_t len
, int noblock
, int flags
,
540 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
541 int offset
, size_t size
, int flags
);
542 int (*bind
)(struct sock
*sk
,
543 struct sockaddr
*uaddr
, int addr_len
);
545 int (*backlog_rcv
) (struct sock
*sk
,
546 struct sk_buff
*skb
);
548 /* Keeping track of sk's, looking them up, and port selection methods. */
549 void (*hash
)(struct sock
*sk
);
550 void (*unhash
)(struct sock
*sk
);
551 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
553 /* Memory pressure */
554 void (*enter_memory_pressure
)(void);
555 atomic_t
*memory_allocated
; /* Current allocated memory. */
556 atomic_t
*sockets_allocated
; /* Current number of sockets. */
558 * Pressure flag: try to collapse.
559 * Technical note: it is used by multiple contexts non atomically.
560 * All the sk_stream_mem_schedule() is of this nature: accounting
561 * is strict, actions are advisory and have some latency.
563 int *memory_pressure
;
570 unsigned int obj_size
;
572 atomic_t
*orphan_count
;
574 struct request_sock_ops
*rsk_prot
;
575 struct timewait_sock_ops
*twsk_prot
;
577 struct module
*owner
;
581 struct list_head node
;
582 #ifdef SOCK_REFCNT_DEBUG
587 u8 __pad
[SMP_CACHE_BYTES
- sizeof(int)];
591 extern int proto_register(struct proto
*prot
, int alloc_slab
);
592 extern void proto_unregister(struct proto
*prot
);
594 #ifdef SOCK_REFCNT_DEBUG
595 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
597 atomic_inc(&sk
->sk_prot
->socks
);
600 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
602 atomic_dec(&sk
->sk_prot
->socks
);
603 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
604 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
607 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
609 if (atomic_read(&sk
->sk_refcnt
) != 1)
610 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
611 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
613 #else /* SOCK_REFCNT_DEBUG */
614 #define sk_refcnt_debug_inc(sk) do { } while (0)
615 #define sk_refcnt_debug_dec(sk) do { } while (0)
616 #define sk_refcnt_debug_release(sk) do { } while (0)
617 #endif /* SOCK_REFCNT_DEBUG */
619 /* Called with local bh disabled */
620 static __inline__
void sock_prot_inc_use(struct proto
*prot
)
622 prot
->stats
[smp_processor_id()].inuse
++;
625 static __inline__
void sock_prot_dec_use(struct proto
*prot
)
627 prot
->stats
[smp_processor_id()].inuse
--;
630 /* With per-bucket locks this operation is not-atomic, so that
631 * this version is not worse.
633 static inline void __sk_prot_rehash(struct sock
*sk
)
635 sk
->sk_prot
->unhash(sk
);
636 sk
->sk_prot
->hash(sk
);
639 /* About 10 seconds */
640 #define SOCK_DESTROY_TIME (10*HZ)
642 /* Sockets 0-1023 can't be bound to unless you are superuser */
643 #define PROT_SOCK 1024
645 #define SHUTDOWN_MASK 3
646 #define RCV_SHUTDOWN 1
647 #define SEND_SHUTDOWN 2
649 #define SOCK_SNDBUF_LOCK 1
650 #define SOCK_RCVBUF_LOCK 2
651 #define SOCK_BINDADDR_LOCK 4
652 #define SOCK_BINDPORT_LOCK 8
654 /* sock_iocb: used to kick off async processing of socket ios */
656 struct list_head list
;
662 struct scm_cookie
*scm
;
663 struct msghdr
*msg
, async_msg
;
664 struct iovec async_iov
;
668 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
670 return (struct sock_iocb
*)iocb
->private;
673 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
678 struct socket_alloc
{
679 struct socket socket
;
680 struct inode vfs_inode
;
683 static inline struct socket
*SOCKET_I(struct inode
*inode
)
685 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
688 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
690 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
693 extern void __sk_stream_mem_reclaim(struct sock
*sk
);
694 extern int sk_stream_mem_schedule(struct sock
*sk
, int size
, int kind
);
696 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
698 static inline int sk_stream_pages(int amt
)
700 return (amt
+ SK_STREAM_MEM_QUANTUM
- 1) / SK_STREAM_MEM_QUANTUM
;
703 static inline void sk_stream_mem_reclaim(struct sock
*sk
)
705 if (sk
->sk_forward_alloc
>= SK_STREAM_MEM_QUANTUM
)
706 __sk_stream_mem_reclaim(sk
);
709 static inline void sk_stream_writequeue_purge(struct sock
*sk
)
713 while ((skb
= __skb_dequeue(&sk
->sk_write_queue
)) != NULL
)
714 sk_stream_free_skb(sk
, skb
);
715 sk_stream_mem_reclaim(sk
);
718 static inline int sk_stream_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
)
720 return (int)skb
->truesize
<= sk
->sk_forward_alloc
||
721 sk_stream_mem_schedule(sk
, skb
->truesize
, 1);
724 static inline int sk_stream_wmem_schedule(struct sock
*sk
, int size
)
726 return size
<= sk
->sk_forward_alloc
||
727 sk_stream_mem_schedule(sk
, size
, 0);
730 /* Used by processes to "lock" a socket state, so that
731 * interrupts and bottom half handlers won't change it
732 * from under us. It essentially blocks any incoming
733 * packets, so that we won't get any new data or any
734 * packets that change the state of the socket.
736 * While locked, BH processing will add new packets to
737 * the backlog queue. This queue is processed by the
738 * owner of the socket lock right before it is released.
740 * Since ~2.3.5 it is also exclusive sleep lock serializing
741 * accesses from user process context.
743 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
745 extern void FASTCALL(lock_sock(struct sock
*sk
));
746 extern void FASTCALL(release_sock(struct sock
*sk
));
748 /* BH context may only use the following locking interface. */
749 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
750 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
752 extern struct sock
*sk_alloc(int family
,
754 struct proto
*prot
, int zero_it
);
755 extern void sk_free(struct sock
*sk
);
756 extern struct sock
*sk_clone(const struct sock
*sk
,
757 const gfp_t priority
);
759 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
760 unsigned long size
, int force
,
762 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
763 unsigned long size
, int force
,
765 extern void sock_wfree(struct sk_buff
*skb
);
766 extern void sock_rfree(struct sk_buff
*skb
);
768 extern int sock_setsockopt(struct socket
*sock
, int level
,
769 int op
, char __user
*optval
,
772 extern int sock_getsockopt(struct socket
*sock
, int level
,
773 int op
, char __user
*optval
,
775 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
779 extern void *sock_kmalloc(struct sock
*sk
, int size
,
781 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
782 extern void sk_send_sigurg(struct sock
*sk
);
785 * Functions to fill in entries in struct proto_ops when a protocol
786 * does not implement a particular function.
788 extern int sock_no_bind(struct socket
*,
789 struct sockaddr
*, int);
790 extern int sock_no_connect(struct socket
*,
791 struct sockaddr
*, int, int);
792 extern int sock_no_socketpair(struct socket
*,
794 extern int sock_no_accept(struct socket
*,
795 struct socket
*, int);
796 extern int sock_no_getname(struct socket
*,
797 struct sockaddr
*, int *, int);
798 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
799 struct poll_table_struct
*);
800 extern int sock_no_ioctl(struct socket
*, unsigned int,
802 extern int sock_no_listen(struct socket
*, int);
803 extern int sock_no_shutdown(struct socket
*, int);
804 extern int sock_no_getsockopt(struct socket
*, int , int,
805 char __user
*, int __user
*);
806 extern int sock_no_setsockopt(struct socket
*, int, int,
808 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
809 struct msghdr
*, size_t);
810 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
811 struct msghdr
*, size_t, int);
812 extern int sock_no_mmap(struct file
*file
,
814 struct vm_area_struct
*vma
);
815 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
817 int offset
, size_t size
,
821 * Functions to fill in entries in struct proto_ops when a protocol
822 * uses the inet style.
824 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
825 char __user
*optval
, int __user
*optlen
);
826 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
827 struct msghdr
*msg
, size_t size
, int flags
);
828 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
829 char __user
*optval
, int optlen
);
830 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
831 int optname
, char __user
*optval
, int __user
*optlen
);
832 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
833 int optname
, char __user
*optval
, int optlen
);
835 extern void sk_common_release(struct sock
*sk
);
838 * Default socket callbacks and setup code
841 /* Initialise core socket variables */
842 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
845 * sk_filter - run a packet through a socket filter
846 * @sk: sock associated with &sk_buff
847 * @skb: buffer to filter
848 * @needlock: set to 1 if the sock is not locked by caller.
850 * Run the filter code and then cut skb->data to correct size returned by
851 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
852 * than pkt_len we keep whole skb->data. This is the socket level
853 * wrapper to sk_run_filter. It returns 0 if the packet should
854 * be accepted or -EPERM if the packet should be tossed.
858 static inline int sk_filter(struct sock
*sk
, struct sk_buff
*skb
, int needlock
)
862 err
= security_sock_rcv_skb(sk
, skb
);
867 struct sk_filter
*filter
;
872 filter
= sk
->sk_filter
;
874 unsigned int pkt_len
= sk_run_filter(skb
, filter
->insns
,
879 skb_trim(skb
, pkt_len
);
889 * sk_filter_release: Release a socket filter
891 * @fp: filter to remove
893 * Remove a filter from a socket and release its resources.
896 static inline void sk_filter_release(struct sock
*sk
, struct sk_filter
*fp
)
898 unsigned int size
= sk_filter_len(fp
);
900 atomic_sub(size
, &sk
->sk_omem_alloc
);
902 if (atomic_dec_and_test(&fp
->refcnt
))
906 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
908 atomic_inc(&fp
->refcnt
);
909 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
913 * Socket reference counting postulates.
915 * * Each user of socket SHOULD hold a reference count.
916 * * Each access point to socket (an hash table bucket, reference from a list,
917 * running timer, skb in flight MUST hold a reference count.
918 * * When reference count hits 0, it means it will never increase back.
919 * * When reference count hits 0, it means that no references from
920 * outside exist to this socket and current process on current CPU
921 * is last user and may/should destroy this socket.
922 * * sk_free is called from any context: process, BH, IRQ. When
923 * it is called, socket has no references from outside -> sk_free
924 * may release descendant resources allocated by the socket, but
925 * to the time when it is called, socket is NOT referenced by any
926 * hash tables, lists etc.
927 * * Packets, delivered from outside (from network or from another process)
928 * and enqueued on receive/error queues SHOULD NOT grab reference count,
929 * when they sit in queue. Otherwise, packets will leak to hole, when
930 * socket is looked up by one cpu and unhasing is made by another CPU.
931 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
932 * (leak to backlog). Packet socket does all the processing inside
933 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
934 * use separate SMP lock, so that they are prone too.
937 /* Ungrab socket and destroy it, if it was the last reference. */
938 static inline void sock_put(struct sock
*sk
)
940 if (atomic_dec_and_test(&sk
->sk_refcnt
))
944 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
);
946 /* Detach socket from process context.
947 * Announce socket dead, detach it from wait queue and inode.
948 * Note that parent inode held reference count on this struct sock,
949 * we do not release it in this function, because protocol
950 * probably wants some additional cleanups or even continuing
951 * to work with this socket (TCP).
953 static inline void sock_orphan(struct sock
*sk
)
955 write_lock_bh(&sk
->sk_callback_lock
);
956 sock_set_flag(sk
, SOCK_DEAD
);
957 sk
->sk_socket
= NULL
;
959 write_unlock_bh(&sk
->sk_callback_lock
);
962 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
964 write_lock_bh(&sk
->sk_callback_lock
);
965 sk
->sk_sleep
= &parent
->wait
;
967 sk
->sk_socket
= parent
;
968 write_unlock_bh(&sk
->sk_callback_lock
);
971 extern int sock_i_uid(struct sock
*sk
);
972 extern unsigned long sock_i_ino(struct sock
*sk
);
974 static inline struct dst_entry
*
975 __sk_dst_get(struct sock
*sk
)
977 return sk
->sk_dst_cache
;
980 static inline struct dst_entry
*
981 sk_dst_get(struct sock
*sk
)
983 struct dst_entry
*dst
;
985 read_lock(&sk
->sk_dst_lock
);
986 dst
= sk
->sk_dst_cache
;
989 read_unlock(&sk
->sk_dst_lock
);
994 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
996 struct dst_entry
*old_dst
;
998 old_dst
= sk
->sk_dst_cache
;
999 sk
->sk_dst_cache
= dst
;
1000 dst_release(old_dst
);
1004 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1006 write_lock(&sk
->sk_dst_lock
);
1007 __sk_dst_set(sk
, dst
);
1008 write_unlock(&sk
->sk_dst_lock
);
1012 __sk_dst_reset(struct sock
*sk
)
1014 struct dst_entry
*old_dst
;
1016 old_dst
= sk
->sk_dst_cache
;
1017 sk
->sk_dst_cache
= NULL
;
1018 dst_release(old_dst
);
1022 sk_dst_reset(struct sock
*sk
)
1024 write_lock(&sk
->sk_dst_lock
);
1026 write_unlock(&sk
->sk_dst_lock
);
1029 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1031 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1033 static inline void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
)
1035 __sk_dst_set(sk
, dst
);
1036 sk
->sk_route_caps
= dst
->dev
->features
;
1037 if (sk
->sk_route_caps
& NETIF_F_TSO
) {
1038 if (sock_flag(sk
, SOCK_NO_LARGESEND
) || dst
->header_len
)
1039 sk
->sk_route_caps
&= ~NETIF_F_TSO
;
1043 static inline void sk_charge_skb(struct sock
*sk
, struct sk_buff
*skb
)
1045 sk
->sk_wmem_queued
+= skb
->truesize
;
1046 sk
->sk_forward_alloc
-= skb
->truesize
;
1049 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1050 struct sk_buff
*skb
, struct page
*page
,
1053 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1055 unsigned int csum
= csum_and_copy_from_user(from
,
1056 page_address(page
) + off
,
1060 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1061 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1065 skb
->data_len
+= copy
;
1066 skb
->truesize
+= copy
;
1067 sk
->sk_wmem_queued
+= copy
;
1068 sk
->sk_forward_alloc
-= copy
;
1073 * Queue a received datagram if it will fit. Stream and sequenced
1074 * protocols can't normally use this as they need to fit buffers in
1075 * and play with them.
1077 * Inlined as it's very short and called for pretty much every
1078 * packet ever received.
1081 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1085 skb
->destructor
= sock_wfree
;
1086 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1089 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1092 skb
->destructor
= sock_rfree
;
1093 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1096 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1097 unsigned long expires
);
1099 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1101 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1103 static inline int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
)
1105 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1106 number of warnings when compiling with -W --ANK
1108 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
1109 (unsigned)sk
->sk_rcvbuf
)
1111 skb_set_owner_r(skb
, sk
);
1112 skb_queue_tail(&sk
->sk_error_queue
, skb
);
1113 if (!sock_flag(sk
, SOCK_DEAD
))
1114 sk
->sk_data_ready(sk
, skb
->len
);
1119 * Recover an error report and clear atomically
1122 static inline int sock_error(struct sock
*sk
)
1125 if (likely(!sk
->sk_err
))
1127 err
= xchg(&sk
->sk_err
, 0);
1131 static inline unsigned long sock_wspace(struct sock
*sk
)
1135 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1136 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1143 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1145 if (sk
->sk_socket
&& sk
->sk_socket
->fasync_list
)
1146 sock_wake_async(sk
->sk_socket
, how
, band
);
1149 #define SOCK_MIN_SNDBUF 2048
1150 #define SOCK_MIN_RCVBUF 256
1152 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1154 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1155 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
/ 2);
1156 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1160 static inline struct sk_buff
*sk_stream_alloc_pskb(struct sock
*sk
,
1164 struct sk_buff
*skb
;
1167 hdr_len
= SKB_DATA_ALIGN(sk
->sk_prot
->max_header
);
1168 skb
= alloc_skb_fclone(size
+ hdr_len
, gfp
);
1170 skb
->truesize
+= mem
;
1171 if (sk_stream_wmem_schedule(sk
, skb
->truesize
)) {
1172 skb_reserve(skb
, hdr_len
);
1177 sk
->sk_prot
->enter_memory_pressure();
1178 sk_stream_moderate_sndbuf(sk
);
1183 static inline struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
,
1187 return sk_stream_alloc_pskb(sk
, size
, 0, gfp
);
1190 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1192 struct page
*page
= NULL
;
1194 page
= alloc_pages(sk
->sk_allocation
, 0);
1196 sk
->sk_prot
->enter_memory_pressure();
1197 sk_stream_moderate_sndbuf(sk
);
1202 #define sk_stream_for_retrans_queue(skb, sk) \
1203 for (skb = (sk)->sk_write_queue.next; \
1204 (skb != (sk)->sk_send_head) && \
1205 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1208 /*from STCP for fast SACK Process*/
1209 #define sk_stream_for_retrans_queue_from(skb, sk) \
1210 for (; (skb != (sk)->sk_send_head) && \
1211 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1215 * Default write policy as shown to user space via poll/select/SIGIO
1217 static inline int sock_writeable(const struct sock
*sk
)
1219 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
/ 2);
1222 static inline gfp_t
gfp_any(void)
1224 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
1227 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1229 return noblock
? 0 : sk
->sk_rcvtimeo
;
1232 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1234 return noblock
? 0 : sk
->sk_sndtimeo
;
1237 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1239 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1242 /* Alas, with timeout socket operations are not restartable.
1243 * Compare this to poll().
1245 static inline int sock_intr_errno(long timeo
)
1247 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
1250 static __inline__
void
1251 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
1253 struct timeval stamp
;
1255 skb_get_timestamp(skb
, &stamp
);
1256 if (sock_flag(sk
, SOCK_RCVTSTAMP
)) {
1257 /* Race occurred between timestamp enabling and packet
1258 receiving. Fill in the current time for now. */
1259 if (stamp
.tv_sec
== 0)
1260 do_gettimeofday(&stamp
);
1261 skb_set_timestamp(skb
, &stamp
);
1262 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP
, sizeof(struct timeval
),
1265 sk
->sk_stamp
= stamp
;
1269 * sk_eat_skb - Release a skb if it is no longer needed
1270 * @sk: socket to eat this skb from
1271 * @skb: socket buffer to eat
1273 * This routine must be called with interrupts disabled or with the socket
1274 * locked so that the sk_buff queue operation is ok.
1276 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
1278 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1282 extern void sock_enable_timestamp(struct sock
*sk
);
1283 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
1286 * Enable debug/info messages
1289 #ifdef CONFIG_NETDEBUG
1290 #define NETDEBUG(fmt, args...) printk(fmt,##args)
1291 #define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0)
1293 #define NETDEBUG(fmt, args...) do { } while (0)
1294 #define LIMIT_NETDEBUG(fmt, args...) do { } while(0)
1298 * Macros for sleeping on a socket. Use them like this:
1300 * SOCK_SLEEP_PRE(sk)
1303 * SOCK_SLEEP_POST(sk)
1305 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1306 * and when the last use of them in DECnet has gone, I'm intending to
1310 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1311 DECLARE_WAITQUEUE(wait, tsk); \
1312 tsk->state = TASK_INTERRUPTIBLE; \
1313 add_wait_queue((sk)->sk_sleep, &wait); \
1316 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1317 remove_wait_queue((sk)->sk_sleep, &wait); \
1321 static inline void sock_valbool_flag(struct sock
*sk
, int bit
, int valbool
)
1324 sock_set_flag(sk
, bit
);
1326 sock_reset_flag(sk
, bit
);
1329 extern __u32 sysctl_wmem_max
;
1330 extern __u32 sysctl_rmem_max
;
1333 int siocdevprivate_ioctl(unsigned int fd
, unsigned int cmd
, unsigned long arg
);
1335 static inline int siocdevprivate_ioctl(unsigned int fd
, unsigned int cmd
, unsigned long arg
)
1341 extern void sk_init(void);
1343 #ifdef CONFIG_SYSCTL
1344 extern struct ctl_table core_table
[];
1347 extern int sysctl_optmem_max
;
1349 extern __u32 sysctl_wmem_default
;
1350 extern __u32 sysctl_rmem_default
;
1352 #endif /* _SOCK_H */