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_omem_alloc: "o" is "option" or "other"
136 * @sk_wmem_queued: persistent queue size
137 * @sk_forward_alloc: space allocated forward
138 * @sk_allocation: allocation mode
139 * @sk_sndbuf: size of send buffer in bytes
140 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
141 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
142 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
143 * @sk_lingertime: %SO_LINGER l_linger setting
144 * @sk_backlog: always used with the per-socket spinlock held
145 * @sk_callback_lock: used with the callbacks in the end of this struct
146 * @sk_error_queue: rarely used
147 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
148 * @sk_err: last error
149 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
150 * @sk_ack_backlog: current listen backlog
151 * @sk_max_ack_backlog: listen backlog set in listen()
152 * @sk_priority: %SO_PRIORITY setting
153 * @sk_type: socket type (%SOCK_STREAM, etc)
154 * @sk_protocol: which protocol this socket belongs in this network family
155 * @sk_peercred: %SO_PEERCRED setting
156 * @sk_rcvlowat: %SO_RCVLOWAT setting
157 * @sk_rcvtimeo: %SO_RCVTIMEO setting
158 * @sk_sndtimeo: %SO_SNDTIMEO setting
159 * @sk_filter: socket filtering instructions
160 * @sk_protinfo: private area, net family specific, when not using slab
161 * @sk_timer: sock cleanup timer
162 * @sk_stamp: time stamp of last packet received
163 * @sk_socket: Identd and reporting IO signals
164 * @sk_user_data: RPC layer private data
165 * @sk_sndmsg_page: cached page for sendmsg
166 * @sk_sndmsg_off: cached offset for sendmsg
167 * @sk_send_head: front of stuff to transmit
168 * @sk_security: used by security modules
169 * @sk_write_pending: a write to stream socket waits to start
170 * @sk_state_change: callback to indicate change in the state of the sock
171 * @sk_data_ready: callback to indicate there is data to be processed
172 * @sk_write_space: callback to indicate there is bf sending space available
173 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
174 * @sk_backlog_rcv: callback to process the backlog
175 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
179 * Now struct inet_timewait_sock also uses sock_common, so please just
180 * don't add nothing before this first member (__sk_common) --acme
182 struct sock_common __sk_common
;
183 #define sk_family __sk_common.skc_family
184 #define sk_state __sk_common.skc_state
185 #define sk_reuse __sk_common.skc_reuse
186 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
187 #define sk_node __sk_common.skc_node
188 #define sk_bind_node __sk_common.skc_bind_node
189 #define sk_refcnt __sk_common.skc_refcnt
190 #define sk_hash __sk_common.skc_hash
191 #define sk_prot __sk_common.skc_prot
192 unsigned char sk_shutdown
: 2,
195 unsigned char sk_protocol
;
196 unsigned short sk_type
;
198 socket_lock_t sk_lock
;
199 wait_queue_head_t
*sk_sleep
;
200 struct dst_entry
*sk_dst_cache
;
201 struct xfrm_policy
*sk_policy
[2];
202 rwlock_t sk_dst_lock
;
203 atomic_t sk_rmem_alloc
;
204 atomic_t sk_wmem_alloc
;
205 atomic_t sk_omem_alloc
;
206 struct sk_buff_head sk_receive_queue
;
207 struct sk_buff_head sk_write_queue
;
209 int sk_forward_alloc
;
213 unsigned long sk_flags
;
214 unsigned long sk_lingertime
;
216 * The backlog queue is special, it is always used with
217 * the per-socket spinlock held and requires low latency
218 * access. Therefore we special case it's implementation.
221 struct sk_buff
*head
;
222 struct sk_buff
*tail
;
224 struct sk_buff_head sk_error_queue
;
225 struct proto
*sk_prot_creator
;
226 rwlock_t sk_callback_lock
;
229 unsigned short sk_ack_backlog
;
230 unsigned short sk_max_ack_backlog
;
232 struct ucred sk_peercred
;
236 struct sk_filter
*sk_filter
;
238 struct timer_list sk_timer
;
239 struct timeval sk_stamp
;
240 struct socket
*sk_socket
;
242 struct page
*sk_sndmsg_page
;
243 struct sk_buff
*sk_send_head
;
245 int sk_write_pending
;
247 void (*sk_state_change
)(struct sock
*sk
);
248 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
249 void (*sk_write_space
)(struct sock
*sk
);
250 void (*sk_error_report
)(struct sock
*sk
);
251 int (*sk_backlog_rcv
)(struct sock
*sk
,
252 struct sk_buff
*skb
);
253 void (*sk_destruct
)(struct sock
*sk
);
257 * Hashed lists helper routines
259 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
261 return hlist_entry(head
->first
, struct sock
, sk_node
);
264 static inline struct sock
*sk_head(const struct hlist_head
*head
)
266 return hlist_empty(head
) ? NULL
: __sk_head(head
);
269 static inline struct sock
*sk_next(const struct sock
*sk
)
271 return sk
->sk_node
.next
?
272 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
275 static inline int sk_unhashed(const struct sock
*sk
)
277 return hlist_unhashed(&sk
->sk_node
);
280 static inline int sk_hashed(const struct sock
*sk
)
282 return sk
->sk_node
.pprev
!= NULL
;
285 static __inline__
void sk_node_init(struct hlist_node
*node
)
290 static __inline__
void __sk_del_node(struct sock
*sk
)
292 __hlist_del(&sk
->sk_node
);
295 static __inline__
int __sk_del_node_init(struct sock
*sk
)
299 sk_node_init(&sk
->sk_node
);
305 /* Grab socket reference count. This operation is valid only
306 when sk is ALREADY grabbed f.e. it is found in hash table
307 or a list and the lookup is made under lock preventing hash table
311 static inline void sock_hold(struct sock
*sk
)
313 atomic_inc(&sk
->sk_refcnt
);
316 /* Ungrab socket in the context, which assumes that socket refcnt
317 cannot hit zero, f.e. it is true in context of any socketcall.
319 static inline void __sock_put(struct sock
*sk
)
321 atomic_dec(&sk
->sk_refcnt
);
324 static __inline__
int sk_del_node_init(struct sock
*sk
)
326 int rc
= __sk_del_node_init(sk
);
329 /* paranoid for a while -acme */
330 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
336 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
338 hlist_add_head(&sk
->sk_node
, list
);
341 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
344 __sk_add_node(sk
, list
);
347 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
349 __hlist_del(&sk
->sk_bind_node
);
352 static __inline__
void sk_add_bind_node(struct sock
*sk
,
353 struct hlist_head
*list
)
355 hlist_add_head(&sk
->sk_bind_node
, list
);
358 #define sk_for_each(__sk, node, list) \
359 hlist_for_each_entry(__sk, node, list, sk_node)
360 #define sk_for_each_from(__sk, node) \
361 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
362 hlist_for_each_entry_from(__sk, node, sk_node)
363 #define sk_for_each_continue(__sk, node) \
364 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
365 hlist_for_each_entry_continue(__sk, node, sk_node)
366 #define sk_for_each_safe(__sk, node, tmp, list) \
367 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
368 #define sk_for_each_bound(__sk, node, list) \
369 hlist_for_each_entry(__sk, node, list, sk_bind_node)
382 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
383 SOCK_DBG
, /* %SO_DEBUG setting */
384 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
385 SOCK_NO_LARGESEND
, /* whether to sent large segments or not */
386 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
387 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
390 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
392 nsk
->sk_flags
= osk
->sk_flags
;
395 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
397 __set_bit(flag
, &sk
->sk_flags
);
400 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
402 __clear_bit(flag
, &sk
->sk_flags
);
405 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
407 return test_bit(flag
, &sk
->sk_flags
);
410 static inline void sk_acceptq_removed(struct sock
*sk
)
412 sk
->sk_ack_backlog
--;
415 static inline void sk_acceptq_added(struct sock
*sk
)
417 sk
->sk_ack_backlog
++;
420 static inline int sk_acceptq_is_full(struct sock
*sk
)
422 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
426 * Compute minimal free write space needed to queue new packets.
428 static inline int sk_stream_min_wspace(struct sock
*sk
)
430 return sk
->sk_wmem_queued
/ 2;
433 static inline int sk_stream_wspace(struct sock
*sk
)
435 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
438 extern void sk_stream_write_space(struct sock
*sk
);
440 static inline int sk_stream_memory_free(struct sock
*sk
)
442 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
445 extern void sk_stream_rfree(struct sk_buff
*skb
);
447 static inline void sk_stream_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
450 skb
->destructor
= sk_stream_rfree
;
451 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
452 sk
->sk_forward_alloc
-= skb
->truesize
;
455 static inline void sk_stream_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
457 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
458 sk
->sk_wmem_queued
-= skb
->truesize
;
459 sk
->sk_forward_alloc
+= skb
->truesize
;
463 /* The per-socket spinlock must be held here. */
464 static inline void sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
466 if (!sk
->sk_backlog
.tail
) {
467 sk
->sk_backlog
.head
= sk
->sk_backlog
.tail
= skb
;
469 sk
->sk_backlog
.tail
->next
= skb
;
470 sk
->sk_backlog
.tail
= skb
;
475 #define sk_wait_event(__sk, __timeo, __condition) \
477 release_sock(__sk); \
480 *(__timeo) = schedule_timeout(*(__timeo)); \
487 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
488 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
489 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
490 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
491 extern void sk_stream_kill_queues(struct sock
*sk
);
493 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
495 struct request_sock_ops
;
496 struct timewait_sock_ops
;
498 /* Networking protocol blocks we attach to sockets.
499 * socket layer -> transport layer interface
500 * transport -> network interface is defined by struct inet_proto
503 void (*close
)(struct sock
*sk
,
505 int (*connect
)(struct sock
*sk
,
506 struct sockaddr
*uaddr
,
508 int (*disconnect
)(struct sock
*sk
, int flags
);
510 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
512 int (*ioctl
)(struct sock
*sk
, int cmd
,
514 int (*init
)(struct sock
*sk
);
515 int (*destroy
)(struct sock
*sk
);
516 void (*shutdown
)(struct sock
*sk
, int how
);
517 int (*setsockopt
)(struct sock
*sk
, int level
,
518 int optname
, char __user
*optval
,
520 int (*getsockopt
)(struct sock
*sk
, int level
,
521 int optname
, char __user
*optval
,
523 int (*compat_setsockopt
)(struct sock
*sk
,
525 int optname
, char __user
*optval
,
527 int (*compat_getsockopt
)(struct sock
*sk
,
529 int optname
, char __user
*optval
,
531 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
532 struct msghdr
*msg
, size_t len
);
533 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
535 size_t len
, int noblock
, int flags
,
537 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
538 int offset
, size_t size
, int flags
);
539 int (*bind
)(struct sock
*sk
,
540 struct sockaddr
*uaddr
, int addr_len
);
542 int (*backlog_rcv
) (struct sock
*sk
,
543 struct sk_buff
*skb
);
545 /* Keeping track of sk's, looking them up, and port selection methods. */
546 void (*hash
)(struct sock
*sk
);
547 void (*unhash
)(struct sock
*sk
);
548 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
550 /* Memory pressure */
551 void (*enter_memory_pressure
)(void);
552 atomic_t
*memory_allocated
; /* Current allocated memory. */
553 atomic_t
*sockets_allocated
; /* Current number of sockets. */
555 * Pressure flag: try to collapse.
556 * Technical note: it is used by multiple contexts non atomically.
557 * All the sk_stream_mem_schedule() is of this nature: accounting
558 * is strict, actions are advisory and have some latency.
560 int *memory_pressure
;
567 unsigned int obj_size
;
569 atomic_t
*orphan_count
;
571 struct request_sock_ops
*rsk_prot
;
572 struct timewait_sock_ops
*twsk_prot
;
574 struct module
*owner
;
578 struct list_head node
;
579 #ifdef SOCK_REFCNT_DEBUG
584 u8 __pad
[SMP_CACHE_BYTES
- sizeof(int)];
588 extern int proto_register(struct proto
*prot
, int alloc_slab
);
589 extern void proto_unregister(struct proto
*prot
);
591 #ifdef SOCK_REFCNT_DEBUG
592 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
594 atomic_inc(&sk
->sk_prot
->socks
);
597 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
599 atomic_dec(&sk
->sk_prot
->socks
);
600 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
601 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
604 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
606 if (atomic_read(&sk
->sk_refcnt
) != 1)
607 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
608 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
610 #else /* SOCK_REFCNT_DEBUG */
611 #define sk_refcnt_debug_inc(sk) do { } while (0)
612 #define sk_refcnt_debug_dec(sk) do { } while (0)
613 #define sk_refcnt_debug_release(sk) do { } while (0)
614 #endif /* SOCK_REFCNT_DEBUG */
616 /* Called with local bh disabled */
617 static __inline__
void sock_prot_inc_use(struct proto
*prot
)
619 prot
->stats
[smp_processor_id()].inuse
++;
622 static __inline__
void sock_prot_dec_use(struct proto
*prot
)
624 prot
->stats
[smp_processor_id()].inuse
--;
627 /* With per-bucket locks this operation is not-atomic, so that
628 * this version is not worse.
630 static inline void __sk_prot_rehash(struct sock
*sk
)
632 sk
->sk_prot
->unhash(sk
);
633 sk
->sk_prot
->hash(sk
);
636 /* About 10 seconds */
637 #define SOCK_DESTROY_TIME (10*HZ)
639 /* Sockets 0-1023 can't be bound to unless you are superuser */
640 #define PROT_SOCK 1024
642 #define SHUTDOWN_MASK 3
643 #define RCV_SHUTDOWN 1
644 #define SEND_SHUTDOWN 2
646 #define SOCK_SNDBUF_LOCK 1
647 #define SOCK_RCVBUF_LOCK 2
648 #define SOCK_BINDADDR_LOCK 4
649 #define SOCK_BINDPORT_LOCK 8
651 /* sock_iocb: used to kick off async processing of socket ios */
653 struct list_head list
;
659 struct scm_cookie
*scm
;
660 struct msghdr
*msg
, async_msg
;
661 struct iovec async_iov
;
665 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
667 return (struct sock_iocb
*)iocb
->private;
670 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
675 struct socket_alloc
{
676 struct socket socket
;
677 struct inode vfs_inode
;
680 static inline struct socket
*SOCKET_I(struct inode
*inode
)
682 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
685 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
687 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
690 extern void __sk_stream_mem_reclaim(struct sock
*sk
);
691 extern int sk_stream_mem_schedule(struct sock
*sk
, int size
, int kind
);
693 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
695 static inline int sk_stream_pages(int amt
)
697 return (amt
+ SK_STREAM_MEM_QUANTUM
- 1) / SK_STREAM_MEM_QUANTUM
;
700 static inline void sk_stream_mem_reclaim(struct sock
*sk
)
702 if (sk
->sk_forward_alloc
>= SK_STREAM_MEM_QUANTUM
)
703 __sk_stream_mem_reclaim(sk
);
706 static inline void sk_stream_writequeue_purge(struct sock
*sk
)
710 while ((skb
= __skb_dequeue(&sk
->sk_write_queue
)) != NULL
)
711 sk_stream_free_skb(sk
, skb
);
712 sk_stream_mem_reclaim(sk
);
715 static inline int sk_stream_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
)
717 return (int)skb
->truesize
<= sk
->sk_forward_alloc
||
718 sk_stream_mem_schedule(sk
, skb
->truesize
, 1);
721 static inline int sk_stream_wmem_schedule(struct sock
*sk
, int size
)
723 return size
<= sk
->sk_forward_alloc
||
724 sk_stream_mem_schedule(sk
, size
, 0);
727 /* Used by processes to "lock" a socket state, so that
728 * interrupts and bottom half handlers won't change it
729 * from under us. It essentially blocks any incoming
730 * packets, so that we won't get any new data or any
731 * packets that change the state of the socket.
733 * While locked, BH processing will add new packets to
734 * the backlog queue. This queue is processed by the
735 * owner of the socket lock right before it is released.
737 * Since ~2.3.5 it is also exclusive sleep lock serializing
738 * accesses from user process context.
740 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
742 extern void FASTCALL(lock_sock(struct sock
*sk
));
743 extern void FASTCALL(release_sock(struct sock
*sk
));
745 /* BH context may only use the following locking interface. */
746 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
747 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
749 extern struct sock
*sk_alloc(int family
,
751 struct proto
*prot
, int zero_it
);
752 extern void sk_free(struct sock
*sk
);
753 extern struct sock
*sk_clone(const struct sock
*sk
,
754 const gfp_t priority
);
756 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
757 unsigned long size
, int force
,
759 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
760 unsigned long size
, int force
,
762 extern void sock_wfree(struct sk_buff
*skb
);
763 extern void sock_rfree(struct sk_buff
*skb
);
765 extern int sock_setsockopt(struct socket
*sock
, int level
,
766 int op
, char __user
*optval
,
769 extern int sock_getsockopt(struct socket
*sock
, int level
,
770 int op
, char __user
*optval
,
772 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
776 extern void *sock_kmalloc(struct sock
*sk
, int size
,
778 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
779 extern void sk_send_sigurg(struct sock
*sk
);
782 * Functions to fill in entries in struct proto_ops when a protocol
783 * does not implement a particular function.
785 extern int sock_no_bind(struct socket
*,
786 struct sockaddr
*, int);
787 extern int sock_no_connect(struct socket
*,
788 struct sockaddr
*, int, int);
789 extern int sock_no_socketpair(struct socket
*,
791 extern int sock_no_accept(struct socket
*,
792 struct socket
*, int);
793 extern int sock_no_getname(struct socket
*,
794 struct sockaddr
*, int *, int);
795 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
796 struct poll_table_struct
*);
797 extern int sock_no_ioctl(struct socket
*, unsigned int,
799 extern int sock_no_listen(struct socket
*, int);
800 extern int sock_no_shutdown(struct socket
*, int);
801 extern int sock_no_getsockopt(struct socket
*, int , int,
802 char __user
*, int __user
*);
803 extern int sock_no_setsockopt(struct socket
*, int, int,
805 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
806 struct msghdr
*, size_t);
807 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
808 struct msghdr
*, size_t, int);
809 extern int sock_no_mmap(struct file
*file
,
811 struct vm_area_struct
*vma
);
812 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
814 int offset
, size_t size
,
818 * Functions to fill in entries in struct proto_ops when a protocol
819 * uses the inet style.
821 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
822 char __user
*optval
, int __user
*optlen
);
823 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
824 struct msghdr
*msg
, size_t size
, int flags
);
825 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
826 char __user
*optval
, int optlen
);
827 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
828 int optname
, char __user
*optval
, int __user
*optlen
);
829 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
830 int optname
, char __user
*optval
, int optlen
);
832 extern void sk_common_release(struct sock
*sk
);
835 * Default socket callbacks and setup code
838 /* Initialise core socket variables */
839 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
842 * sk_filter - run a packet through a socket filter
843 * @sk: sock associated with &sk_buff
844 * @skb: buffer to filter
845 * @needlock: set to 1 if the sock is not locked by caller.
847 * Run the filter code and then cut skb->data to correct size returned by
848 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
849 * than pkt_len we keep whole skb->data. This is the socket level
850 * wrapper to sk_run_filter. It returns 0 if the packet should
851 * be accepted or -EPERM if the packet should be tossed.
855 static inline int sk_filter(struct sock
*sk
, struct sk_buff
*skb
, int needlock
)
859 err
= security_sock_rcv_skb(sk
, skb
);
864 struct sk_filter
*filter
;
869 filter
= sk
->sk_filter
;
871 unsigned int pkt_len
= sk_run_filter(skb
, filter
->insns
,
876 skb_trim(skb
, pkt_len
);
886 * sk_filter_release: Release a socket filter
888 * @fp: filter to remove
890 * Remove a filter from a socket and release its resources.
893 static inline void sk_filter_release(struct sock
*sk
, struct sk_filter
*fp
)
895 unsigned int size
= sk_filter_len(fp
);
897 atomic_sub(size
, &sk
->sk_omem_alloc
);
899 if (atomic_dec_and_test(&fp
->refcnt
))
903 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
905 atomic_inc(&fp
->refcnt
);
906 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
910 * Socket reference counting postulates.
912 * * Each user of socket SHOULD hold a reference count.
913 * * Each access point to socket (an hash table bucket, reference from a list,
914 * running timer, skb in flight MUST hold a reference count.
915 * * When reference count hits 0, it means it will never increase back.
916 * * When reference count hits 0, it means that no references from
917 * outside exist to this socket and current process on current CPU
918 * is last user and may/should destroy this socket.
919 * * sk_free is called from any context: process, BH, IRQ. When
920 * it is called, socket has no references from outside -> sk_free
921 * may release descendant resources allocated by the socket, but
922 * to the time when it is called, socket is NOT referenced by any
923 * hash tables, lists etc.
924 * * Packets, delivered from outside (from network or from another process)
925 * and enqueued on receive/error queues SHOULD NOT grab reference count,
926 * when they sit in queue. Otherwise, packets will leak to hole, when
927 * socket is looked up by one cpu and unhasing is made by another CPU.
928 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
929 * (leak to backlog). Packet socket does all the processing inside
930 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
931 * use separate SMP lock, so that they are prone too.
934 /* Ungrab socket and destroy it, if it was the last reference. */
935 static inline void sock_put(struct sock
*sk
)
937 if (atomic_dec_and_test(&sk
->sk_refcnt
))
941 static inline int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
)
943 int rc
= NET_RX_SUCCESS
;
945 if (sk_filter(sk
, skb
, 0))
946 goto discard_and_relse
;
951 if (!sock_owned_by_user(sk
))
952 rc
= sk
->sk_backlog_rcv(sk
, skb
);
954 sk_add_backlog(sk
, skb
);
964 /* Detach socket from process context.
965 * Announce socket dead, detach it from wait queue and inode.
966 * Note that parent inode held reference count on this struct sock,
967 * we do not release it in this function, because protocol
968 * probably wants some additional cleanups or even continuing
969 * to work with this socket (TCP).
971 static inline void sock_orphan(struct sock
*sk
)
973 write_lock_bh(&sk
->sk_callback_lock
);
974 sock_set_flag(sk
, SOCK_DEAD
);
975 sk
->sk_socket
= NULL
;
977 write_unlock_bh(&sk
->sk_callback_lock
);
980 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
982 write_lock_bh(&sk
->sk_callback_lock
);
983 sk
->sk_sleep
= &parent
->wait
;
985 sk
->sk_socket
= parent
;
986 write_unlock_bh(&sk
->sk_callback_lock
);
989 extern int sock_i_uid(struct sock
*sk
);
990 extern unsigned long sock_i_ino(struct sock
*sk
);
992 static inline struct dst_entry
*
993 __sk_dst_get(struct sock
*sk
)
995 return sk
->sk_dst_cache
;
998 static inline struct dst_entry
*
999 sk_dst_get(struct sock
*sk
)
1001 struct dst_entry
*dst
;
1003 read_lock(&sk
->sk_dst_lock
);
1004 dst
= sk
->sk_dst_cache
;
1007 read_unlock(&sk
->sk_dst_lock
);
1012 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1014 struct dst_entry
*old_dst
;
1016 old_dst
= sk
->sk_dst_cache
;
1017 sk
->sk_dst_cache
= dst
;
1018 dst_release(old_dst
);
1022 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1024 write_lock(&sk
->sk_dst_lock
);
1025 __sk_dst_set(sk
, dst
);
1026 write_unlock(&sk
->sk_dst_lock
);
1030 __sk_dst_reset(struct sock
*sk
)
1032 struct dst_entry
*old_dst
;
1034 old_dst
= sk
->sk_dst_cache
;
1035 sk
->sk_dst_cache
= NULL
;
1036 dst_release(old_dst
);
1040 sk_dst_reset(struct sock
*sk
)
1042 write_lock(&sk
->sk_dst_lock
);
1044 write_unlock(&sk
->sk_dst_lock
);
1047 static inline struct dst_entry
*
1048 __sk_dst_check(struct sock
*sk
, u32 cookie
)
1050 struct dst_entry
*dst
= sk
->sk_dst_cache
;
1052 if (dst
&& dst
->obsolete
&& dst
->ops
->check(dst
, cookie
) == NULL
) {
1053 sk
->sk_dst_cache
= NULL
;
1061 static inline struct dst_entry
*
1062 sk_dst_check(struct sock
*sk
, u32 cookie
)
1064 struct dst_entry
*dst
= sk_dst_get(sk
);
1066 if (dst
&& dst
->obsolete
&& dst
->ops
->check(dst
, cookie
) == NULL
) {
1075 static inline void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
)
1077 __sk_dst_set(sk
, dst
);
1078 sk
->sk_route_caps
= dst
->dev
->features
;
1079 if (sk
->sk_route_caps
& NETIF_F_TSO
) {
1080 if (sock_flag(sk
, SOCK_NO_LARGESEND
) || dst
->header_len
)
1081 sk
->sk_route_caps
&= ~NETIF_F_TSO
;
1085 static inline void sk_charge_skb(struct sock
*sk
, struct sk_buff
*skb
)
1087 sk
->sk_wmem_queued
+= skb
->truesize
;
1088 sk
->sk_forward_alloc
-= skb
->truesize
;
1091 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1092 struct sk_buff
*skb
, struct page
*page
,
1095 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1097 unsigned int csum
= csum_and_copy_from_user(from
,
1098 page_address(page
) + off
,
1102 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1103 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1107 skb
->data_len
+= copy
;
1108 skb
->truesize
+= copy
;
1109 sk
->sk_wmem_queued
+= copy
;
1110 sk
->sk_forward_alloc
-= copy
;
1115 * Queue a received datagram if it will fit. Stream and sequenced
1116 * protocols can't normally use this as they need to fit buffers in
1117 * and play with them.
1119 * Inlined as it's very short and called for pretty much every
1120 * packet ever received.
1123 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1127 skb
->destructor
= sock_wfree
;
1128 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1131 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1134 skb
->destructor
= sock_rfree
;
1135 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1138 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1139 unsigned long expires
);
1141 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1143 static inline int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
1148 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1149 number of warnings when compiling with -W --ANK
1151 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
1152 (unsigned)sk
->sk_rcvbuf
) {
1157 /* It would be deadlock, if sock_queue_rcv_skb is used
1158 with socket lock! We assume that users of this
1159 function are lock free.
1161 err
= sk_filter(sk
, skb
, 1);
1166 skb_set_owner_r(skb
, sk
);
1168 /* Cache the SKB length before we tack it onto the receive
1169 * queue. Once it is added it no longer belongs to us and
1170 * may be freed by other threads of control pulling packets
1175 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1177 if (!sock_flag(sk
, SOCK_DEAD
))
1178 sk
->sk_data_ready(sk
, skb_len
);
1183 static inline int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
)
1185 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1186 number of warnings when compiling with -W --ANK
1188 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
1189 (unsigned)sk
->sk_rcvbuf
)
1191 skb_set_owner_r(skb
, sk
);
1192 skb_queue_tail(&sk
->sk_error_queue
, skb
);
1193 if (!sock_flag(sk
, SOCK_DEAD
))
1194 sk
->sk_data_ready(sk
, skb
->len
);
1199 * Recover an error report and clear atomically
1202 static inline int sock_error(struct sock
*sk
)
1205 if (likely(!sk
->sk_err
))
1207 err
= xchg(&sk
->sk_err
, 0);
1211 static inline unsigned long sock_wspace(struct sock
*sk
)
1215 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1216 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1223 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1225 if (sk
->sk_socket
&& sk
->sk_socket
->fasync_list
)
1226 sock_wake_async(sk
->sk_socket
, how
, band
);
1229 #define SOCK_MIN_SNDBUF 2048
1230 #define SOCK_MIN_RCVBUF 256
1232 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1234 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1235 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
/ 2);
1236 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1240 static inline struct sk_buff
*sk_stream_alloc_pskb(struct sock
*sk
,
1244 struct sk_buff
*skb
;
1247 hdr_len
= SKB_DATA_ALIGN(sk
->sk_prot
->max_header
);
1248 skb
= alloc_skb_fclone(size
+ hdr_len
, gfp
);
1250 skb
->truesize
+= mem
;
1251 if (sk_stream_wmem_schedule(sk
, skb
->truesize
)) {
1252 skb_reserve(skb
, hdr_len
);
1257 sk
->sk_prot
->enter_memory_pressure();
1258 sk_stream_moderate_sndbuf(sk
);
1263 static inline struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
,
1267 return sk_stream_alloc_pskb(sk
, size
, 0, gfp
);
1270 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1272 struct page
*page
= NULL
;
1274 page
= alloc_pages(sk
->sk_allocation
, 0);
1276 sk
->sk_prot
->enter_memory_pressure();
1277 sk_stream_moderate_sndbuf(sk
);
1282 #define sk_stream_for_retrans_queue(skb, sk) \
1283 for (skb = (sk)->sk_write_queue.next; \
1284 (skb != (sk)->sk_send_head) && \
1285 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1288 /*from STCP for fast SACK Process*/
1289 #define sk_stream_for_retrans_queue_from(skb, sk) \
1290 for (; (skb != (sk)->sk_send_head) && \
1291 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1295 * Default write policy as shown to user space via poll/select/SIGIO
1297 static inline int sock_writeable(const struct sock
*sk
)
1299 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
/ 2);
1302 static inline gfp_t
gfp_any(void)
1304 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
1307 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1309 return noblock
? 0 : sk
->sk_rcvtimeo
;
1312 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1314 return noblock
? 0 : sk
->sk_sndtimeo
;
1317 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1319 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1322 /* Alas, with timeout socket operations are not restartable.
1323 * Compare this to poll().
1325 static inline int sock_intr_errno(long timeo
)
1327 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
1330 static __inline__
void
1331 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
1333 struct timeval stamp
;
1335 skb_get_timestamp(skb
, &stamp
);
1336 if (sock_flag(sk
, SOCK_RCVTSTAMP
)) {
1337 /* Race occurred between timestamp enabling and packet
1338 receiving. Fill in the current time for now. */
1339 if (stamp
.tv_sec
== 0)
1340 do_gettimeofday(&stamp
);
1341 skb_set_timestamp(skb
, &stamp
);
1342 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP
, sizeof(struct timeval
),
1345 sk
->sk_stamp
= stamp
;
1349 * sk_eat_skb - Release a skb if it is no longer needed
1350 * @sk: socket to eat this skb from
1351 * @skb: socket buffer to eat
1353 * This routine must be called with interrupts disabled or with the socket
1354 * locked so that the sk_buff queue operation is ok.
1356 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
1358 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1362 extern void sock_enable_timestamp(struct sock
*sk
);
1363 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
1366 * Enable debug/info messages
1369 #ifdef CONFIG_NETDEBUG
1370 #define NETDEBUG(fmt, args...) printk(fmt,##args)
1371 #define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0)
1373 #define NETDEBUG(fmt, args...) do { } while (0)
1374 #define LIMIT_NETDEBUG(fmt, args...) do { } while(0)
1378 * Macros for sleeping on a socket. Use them like this:
1380 * SOCK_SLEEP_PRE(sk)
1383 * SOCK_SLEEP_POST(sk)
1385 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1386 * and when the last use of them in DECnet has gone, I'm intending to
1390 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1391 DECLARE_WAITQUEUE(wait, tsk); \
1392 tsk->state = TASK_INTERRUPTIBLE; \
1393 add_wait_queue((sk)->sk_sleep, &wait); \
1396 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1397 remove_wait_queue((sk)->sk_sleep, &wait); \
1401 static inline void sock_valbool_flag(struct sock
*sk
, int bit
, int valbool
)
1404 sock_set_flag(sk
, bit
);
1406 sock_reset_flag(sk
, bit
);
1409 extern __u32 sysctl_wmem_max
;
1410 extern __u32 sysctl_rmem_max
;
1413 int siocdevprivate_ioctl(unsigned int fd
, unsigned int cmd
, unsigned long arg
);
1415 static inline int siocdevprivate_ioctl(unsigned int fd
, unsigned int cmd
, unsigned long arg
)
1421 extern void sk_init(void);
1423 #ifdef CONFIG_SYSCTL
1424 extern struct ctl_table core_table
[];
1427 extern int sysctl_optmem_max
;
1429 extern __u32 sysctl_wmem_default
;
1430 extern __u32 sysctl_rmem_default
;
1432 #endif /* _SOCK_H */