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
10 * Authors: Ross Biro, <bir7@leland.Stanford.Edu>
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)); \
93 * struct sock_common - minimal network layer representation of sockets
94 * @skc_family - network address family
95 * @skc_state - Connection state
96 * @skc_reuse - %SO_REUSEADDR setting
97 * @skc_bound_dev_if - bound device index if != 0
98 * @skc_node - main hash linkage for various protocol lookup tables
99 * @skc_bind_node - bind hash linkage for various protocol lookup tables
100 * @skc_refcnt - reference count
102 * This is the minimal network layer representation of sockets, the header
103 * for struct sock and struct tcp_tw_bucket.
106 unsigned short skc_family
;
107 volatile unsigned char skc_state
;
108 unsigned char skc_reuse
;
109 int skc_bound_dev_if
;
110 struct hlist_node skc_node
;
111 struct hlist_node skc_bind_node
;
116 * struct sock - network layer representation of sockets
117 * @__sk_common - shared layout with tcp_tw_bucket
118 * @sk_shutdown - mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
119 * @sk_userlocks - %SO_SNDBUF and %SO_RCVBUF settings
120 * @sk_lock - synchronizer
121 * @sk_rcvbuf - size of receive buffer in bytes
122 * @sk_sleep - sock wait queue
123 * @sk_dst_cache - destination cache
124 * @sk_dst_lock - destination cache lock
125 * @sk_policy - flow policy
126 * @sk_rmem_alloc - receive queue bytes committed
127 * @sk_receive_queue - incoming packets
128 * @sk_wmem_alloc - transmit queue bytes committed
129 * @sk_write_queue - Packet sending queue
130 * @sk_omem_alloc - "o" is "option" or "other"
131 * @sk_wmem_queued - persistent queue size
132 * @sk_forward_alloc - space allocated forward
133 * @sk_allocation - allocation mode
134 * @sk_sndbuf - size of send buffer in bytes
135 * @sk_flags - %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
136 * @sk_no_check - %SO_NO_CHECK setting, wether or not checkup packets
137 * @sk_route_caps - route capabilities (e.g. %NETIF_F_TSO)
138 * @sk_lingertime - %SO_LINGER l_linger setting
139 * @sk_hashent - hash entry in several tables (e.g. tcp_ehash)
140 * @sk_backlog - always used with the per-socket spinlock held
141 * @sk_callback_lock - used with the callbacks in the end of this struct
142 * @sk_error_queue - rarely used
143 * @sk_prot - protocol handlers inside a network family
144 * @sk_err - last error
145 * @sk_err_soft - errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
146 * @sk_ack_backlog - current listen backlog
147 * @sk_max_ack_backlog - listen backlog set in listen()
148 * @sk_priority - %SO_PRIORITY setting
149 * @sk_type - socket type (%SOCK_STREAM, etc)
150 * @sk_protocol - which protocol this socket belongs in this network family
151 * @sk_peercred - %SO_PEERCRED setting
152 * @sk_rcvlowat - %SO_RCVLOWAT setting
153 * @sk_rcvtimeo - %SO_RCVTIMEO setting
154 * @sk_sndtimeo - %SO_SNDTIMEO setting
155 * @sk_filter - socket filtering instructions
156 * @sk_protinfo - private area, net family specific, when not using slab
157 * @sk_timer - sock cleanup timer
158 * @sk_stamp - time stamp of last packet received
159 * @sk_socket - Identd and reporting IO signals
160 * @sk_user_data - RPC layer private data
161 * @sk_sndmsg_page - cached page for sendmsg
162 * @sk_sndmsg_off - cached offset for sendmsg
163 * @sk_send_head - front of stuff to transmit
164 * @sk_write_pending - a write to stream socket waits to start
165 * @sk_state_change - callback to indicate change in the state of the sock
166 * @sk_data_ready - callback to indicate there is data to be processed
167 * @sk_write_space - callback to indicate there is bf sending space available
168 * @sk_error_report - callback to indicate errors (e.g. %MSG_ERRQUEUE)
169 * @sk_backlog_rcv - callback to process the backlog
170 * @sk_destruct - called at sock freeing time, i.e. when all refcnt == 0
174 * Now struct tcp_tw_bucket also uses sock_common, so please just
175 * don't add nothing before this first member (__sk_common) --acme
177 struct sock_common __sk_common
;
178 #define sk_family __sk_common.skc_family
179 #define sk_state __sk_common.skc_state
180 #define sk_reuse __sk_common.skc_reuse
181 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
182 #define sk_node __sk_common.skc_node
183 #define sk_bind_node __sk_common.skc_bind_node
184 #define sk_refcnt __sk_common.skc_refcnt
185 unsigned char sk_shutdown
: 2,
188 unsigned char sk_protocol
;
189 unsigned short sk_type
;
191 socket_lock_t sk_lock
;
192 wait_queue_head_t
*sk_sleep
;
193 struct dst_entry
*sk_dst_cache
;
194 struct xfrm_policy
*sk_policy
[2];
195 rwlock_t sk_dst_lock
;
196 atomic_t sk_rmem_alloc
;
197 atomic_t sk_wmem_alloc
;
198 atomic_t sk_omem_alloc
;
199 struct sk_buff_head sk_receive_queue
;
200 struct sk_buff_head sk_write_queue
;
202 int sk_forward_alloc
;
203 unsigned int sk_allocation
;
207 unsigned long sk_flags
;
208 unsigned long sk_lingertime
;
210 * The backlog queue is special, it is always used with
211 * the per-socket spinlock held and requires low latency
212 * access. Therefore we special case it's implementation.
215 struct sk_buff
*head
;
216 struct sk_buff
*tail
;
218 struct sk_buff_head sk_error_queue
;
219 struct proto
*sk_prot
;
220 rwlock_t sk_callback_lock
;
223 unsigned short sk_ack_backlog
;
224 unsigned short sk_max_ack_backlog
;
226 struct ucred sk_peercred
;
230 struct sk_filter
*sk_filter
;
232 struct timer_list sk_timer
;
233 struct timeval sk_stamp
;
234 struct socket
*sk_socket
;
236 struct page
*sk_sndmsg_page
;
237 struct sk_buff
*sk_send_head
;
239 int sk_write_pending
;
241 void (*sk_state_change
)(struct sock
*sk
);
242 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
243 void (*sk_write_space
)(struct sock
*sk
);
244 void (*sk_error_report
)(struct sock
*sk
);
245 int (*sk_backlog_rcv
)(struct sock
*sk
,
246 struct sk_buff
*skb
);
247 void (*sk_destruct
)(struct sock
*sk
);
251 * Hashed lists helper routines
253 static inline struct sock
*__sk_head(struct hlist_head
*head
)
255 return hlist_entry(head
->first
, struct sock
, sk_node
);
258 static inline struct sock
*sk_head(struct hlist_head
*head
)
260 return hlist_empty(head
) ? NULL
: __sk_head(head
);
263 static inline struct sock
*sk_next(struct sock
*sk
)
265 return sk
->sk_node
.next
?
266 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
269 static inline int sk_unhashed(struct sock
*sk
)
271 return hlist_unhashed(&sk
->sk_node
);
274 static inline int sk_hashed(struct sock
*sk
)
276 return sk
->sk_node
.pprev
!= NULL
;
279 static __inline__
void sk_node_init(struct hlist_node
*node
)
284 static __inline__
void __sk_del_node(struct sock
*sk
)
286 __hlist_del(&sk
->sk_node
);
289 static __inline__
int __sk_del_node_init(struct sock
*sk
)
293 sk_node_init(&sk
->sk_node
);
299 /* Grab socket reference count. This operation is valid only
300 when sk is ALREADY grabbed f.e. it is found in hash table
301 or a list and the lookup is made under lock preventing hash table
305 static inline void sock_hold(struct sock
*sk
)
307 atomic_inc(&sk
->sk_refcnt
);
310 /* Ungrab socket in the context, which assumes that socket refcnt
311 cannot hit zero, f.e. it is true in context of any socketcall.
313 static inline void __sock_put(struct sock
*sk
)
315 atomic_dec(&sk
->sk_refcnt
);
318 static __inline__
int sk_del_node_init(struct sock
*sk
)
320 int rc
= __sk_del_node_init(sk
);
323 /* paranoid for a while -acme */
324 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
330 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
332 hlist_add_head(&sk
->sk_node
, list
);
335 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
338 __sk_add_node(sk
, list
);
341 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
343 __hlist_del(&sk
->sk_bind_node
);
346 static __inline__
void sk_add_bind_node(struct sock
*sk
,
347 struct hlist_head
*list
)
349 hlist_add_head(&sk
->sk_bind_node
, list
);
352 #define sk_for_each(__sk, node, list) \
353 hlist_for_each_entry(__sk, node, list, sk_node)
354 #define sk_for_each_from(__sk, node) \
355 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
356 hlist_for_each_entry_from(__sk, node, sk_node)
357 #define sk_for_each_continue(__sk, node) \
358 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
359 hlist_for_each_entry_continue(__sk, node, sk_node)
360 #define sk_for_each_safe(__sk, node, tmp, list) \
361 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
362 #define sk_for_each_bound(__sk, node, list) \
363 hlist_for_each_entry(__sk, node, list, sk_bind_node)
376 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
377 SOCK_DBG
, /* %SO_DEBUG setting */
378 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
379 SOCK_NO_LARGESEND
, /* whether to sent large segments or not */
380 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
381 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
384 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
386 __set_bit(flag
, &sk
->sk_flags
);
389 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
391 __clear_bit(flag
, &sk
->sk_flags
);
394 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
396 return test_bit(flag
, &sk
->sk_flags
);
399 static inline void sk_acceptq_removed(struct sock
*sk
)
401 sk
->sk_ack_backlog
--;
404 static inline void sk_acceptq_added(struct sock
*sk
)
406 sk
->sk_ack_backlog
++;
409 static inline int sk_acceptq_is_full(struct sock
*sk
)
411 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
415 * Compute minimal free write space needed to queue new packets.
417 static inline int sk_stream_min_wspace(struct sock
*sk
)
419 return sk
->sk_wmem_queued
/ 2;
422 static inline int sk_stream_wspace(struct sock
*sk
)
424 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
427 extern void sk_stream_write_space(struct sock
*sk
);
429 static inline int sk_stream_memory_free(struct sock
*sk
)
431 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
434 extern void sk_stream_rfree(struct sk_buff
*skb
);
436 static inline void sk_stream_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
439 skb
->destructor
= sk_stream_rfree
;
440 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
441 sk
->sk_forward_alloc
-= skb
->truesize
;
444 static inline void sk_stream_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
446 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
447 sk
->sk_wmem_queued
-= skb
->truesize
;
448 sk
->sk_forward_alloc
+= skb
->truesize
;
452 /* The per-socket spinlock must be held here. */
453 #define sk_add_backlog(__sk, __skb) \
454 do { if (!(__sk)->sk_backlog.tail) { \
455 (__sk)->sk_backlog.head = \
456 (__sk)->sk_backlog.tail = (__skb); \
458 ((__sk)->sk_backlog.tail)->next = (__skb); \
459 (__sk)->sk_backlog.tail = (__skb); \
461 (__skb)->next = NULL; \
464 #define sk_wait_event(__sk, __timeo, __condition) \
466 release_sock(__sk); \
469 *(__timeo) = schedule_timeout(*(__timeo)); \
476 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
477 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
478 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
479 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
480 extern void sk_stream_kill_queues(struct sock
*sk
);
482 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
484 /* Networking protocol blocks we attach to sockets.
485 * socket layer -> transport layer interface
486 * transport -> network interface is defined by struct inet_proto
489 void (*close
)(struct sock
*sk
,
491 int (*connect
)(struct sock
*sk
,
492 struct sockaddr
*uaddr
,
494 int (*disconnect
)(struct sock
*sk
, int flags
);
496 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
498 int (*ioctl
)(struct sock
*sk
, int cmd
,
500 int (*init
)(struct sock
*sk
);
501 int (*destroy
)(struct sock
*sk
);
502 void (*shutdown
)(struct sock
*sk
, int how
);
503 int (*setsockopt
)(struct sock
*sk
, int level
,
504 int optname
, char __user
*optval
,
506 int (*getsockopt
)(struct sock
*sk
, int level
,
507 int optname
, char __user
*optval
,
509 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
510 struct msghdr
*msg
, size_t len
);
511 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
513 size_t len
, int noblock
, int flags
,
515 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
516 int offset
, size_t size
, int flags
);
517 int (*bind
)(struct sock
*sk
,
518 struct sockaddr
*uaddr
, int addr_len
);
520 int (*backlog_rcv
) (struct sock
*sk
,
521 struct sk_buff
*skb
);
523 /* Keeping track of sk's, looking them up, and port selection methods. */
524 void (*hash
)(struct sock
*sk
);
525 void (*unhash
)(struct sock
*sk
);
526 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
528 /* Memory pressure */
529 void (*enter_memory_pressure
)(void);
530 atomic_t
*memory_allocated
; /* Current allocated memory. */
531 atomic_t
*sockets_allocated
; /* Current number of sockets. */
533 * Pressure flag: try to collapse.
534 * Technical note: it is used by multiple contexts non atomically.
535 * All the sk_stream_mem_schedule() is of this nature: accounting
536 * is strict, actions are advisory and have some latency.
538 int *memory_pressure
;
545 unsigned int obj_size
;
547 struct module
*owner
;
551 struct list_head node
;
555 u8 __pad
[SMP_CACHE_BYTES
- sizeof(int)];
559 extern int proto_register(struct proto
*prot
, int alloc_slab
);
560 extern void proto_unregister(struct proto
*prot
);
562 /* Called with local bh disabled */
563 static __inline__
void sock_prot_inc_use(struct proto
*prot
)
565 prot
->stats
[smp_processor_id()].inuse
++;
568 static __inline__
void sock_prot_dec_use(struct proto
*prot
)
570 prot
->stats
[smp_processor_id()].inuse
--;
573 /* About 10 seconds */
574 #define SOCK_DESTROY_TIME (10*HZ)
576 /* Sockets 0-1023 can't be bound to unless you are superuser */
577 #define PROT_SOCK 1024
579 #define SHUTDOWN_MASK 3
580 #define RCV_SHUTDOWN 1
581 #define SEND_SHUTDOWN 2
583 #define SOCK_SNDBUF_LOCK 1
584 #define SOCK_RCVBUF_LOCK 2
585 #define SOCK_BINDADDR_LOCK 4
586 #define SOCK_BINDPORT_LOCK 8
588 /* sock_iocb: used to kick off async processing of socket ios */
590 struct list_head list
;
596 struct scm_cookie
*scm
;
597 struct msghdr
*msg
, async_msg
;
598 struct iovec async_iov
;
602 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
604 return (struct sock_iocb
*)iocb
->private;
607 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
612 struct socket_alloc
{
613 struct socket socket
;
614 struct inode vfs_inode
;
617 static inline struct socket
*SOCKET_I(struct inode
*inode
)
619 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
622 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
624 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
627 extern void __sk_stream_mem_reclaim(struct sock
*sk
);
628 extern int sk_stream_mem_schedule(struct sock
*sk
, int size
, int kind
);
630 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
632 static inline int sk_stream_pages(int amt
)
634 return (amt
+ SK_STREAM_MEM_QUANTUM
- 1) / SK_STREAM_MEM_QUANTUM
;
637 static inline void sk_stream_mem_reclaim(struct sock
*sk
)
639 if (sk
->sk_forward_alloc
>= SK_STREAM_MEM_QUANTUM
)
640 __sk_stream_mem_reclaim(sk
);
643 static inline void sk_stream_writequeue_purge(struct sock
*sk
)
647 while ((skb
= __skb_dequeue(&sk
->sk_write_queue
)) != NULL
)
648 sk_stream_free_skb(sk
, skb
);
649 sk_stream_mem_reclaim(sk
);
652 static inline int sk_stream_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
)
654 return (int)skb
->truesize
<= sk
->sk_forward_alloc
||
655 sk_stream_mem_schedule(sk
, skb
->truesize
, 1);
658 /* Used by processes to "lock" a socket state, so that
659 * interrupts and bottom half handlers won't change it
660 * from under us. It essentially blocks any incoming
661 * packets, so that we won't get any new data or any
662 * packets that change the state of the socket.
664 * While locked, BH processing will add new packets to
665 * the backlog queue. This queue is processed by the
666 * owner of the socket lock right before it is released.
668 * Since ~2.3.5 it is also exclusive sleep lock serializing
669 * accesses from user process context.
671 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
673 extern void FASTCALL(lock_sock(struct sock
*sk
));
674 extern void FASTCALL(release_sock(struct sock
*sk
));
676 /* BH context may only use the following locking interface. */
677 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
678 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
680 extern struct sock
*sk_alloc(int family
, int priority
,
681 struct proto
*prot
, int zero_it
);
682 extern void sk_free(struct sock
*sk
);
684 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
685 unsigned long size
, int force
,
687 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
688 unsigned long size
, int force
,
690 extern void sock_wfree(struct sk_buff
*skb
);
691 extern void sock_rfree(struct sk_buff
*skb
);
693 extern int sock_setsockopt(struct socket
*sock
, int level
,
694 int op
, char __user
*optval
,
697 extern int sock_getsockopt(struct socket
*sock
, int level
,
698 int op
, char __user
*optval
,
700 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
704 extern void *sock_kmalloc(struct sock
*sk
, int size
, int priority
);
705 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
706 extern void sk_send_sigurg(struct sock
*sk
);
709 * Functions to fill in entries in struct proto_ops when a protocol
710 * does not implement a particular function.
712 extern int sock_no_bind(struct socket
*,
713 struct sockaddr
*, int);
714 extern int sock_no_connect(struct socket
*,
715 struct sockaddr
*, int, int);
716 extern int sock_no_socketpair(struct socket
*,
718 extern int sock_no_accept(struct socket
*,
719 struct socket
*, int);
720 extern int sock_no_getname(struct socket
*,
721 struct sockaddr
*, int *, int);
722 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
723 struct poll_table_struct
*);
724 extern int sock_no_ioctl(struct socket
*, unsigned int,
726 extern int sock_no_listen(struct socket
*, int);
727 extern int sock_no_shutdown(struct socket
*, int);
728 extern int sock_no_getsockopt(struct socket
*, int , int,
729 char __user
*, int __user
*);
730 extern int sock_no_setsockopt(struct socket
*, int, int,
732 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
733 struct msghdr
*, size_t);
734 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
735 struct msghdr
*, size_t, int);
736 extern int sock_no_mmap(struct file
*file
,
738 struct vm_area_struct
*vma
);
739 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
741 int offset
, size_t size
,
745 * Functions to fill in entries in struct proto_ops when a protocol
746 * uses the inet style.
748 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
749 char __user
*optval
, int __user
*optlen
);
750 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
751 struct msghdr
*msg
, size_t size
, int flags
);
752 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
753 char __user
*optval
, int optlen
);
755 extern void sk_common_release(struct sock
*sk
);
758 * Default socket callbacks and setup code
761 /* Initialise core socket variables */
762 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
765 * sk_filter - run a packet through a socket filter
766 * @sk: sock associated with &sk_buff
767 * @skb: buffer to filter
768 * @needlock: set to 1 if the sock is not locked by caller.
770 * Run the filter code and then cut skb->data to correct size returned by
771 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
772 * than pkt_len we keep whole skb->data. This is the socket level
773 * wrapper to sk_run_filter. It returns 0 if the packet should
774 * be accepted or -EPERM if the packet should be tossed.
778 static inline int sk_filter(struct sock
*sk
, struct sk_buff
*skb
, int needlock
)
782 err
= security_sock_rcv_skb(sk
, skb
);
787 struct sk_filter
*filter
;
792 filter
= sk
->sk_filter
;
794 int pkt_len
= sk_run_filter(skb
, filter
->insns
,
799 skb_trim(skb
, pkt_len
);
809 * sk_filter_release: Release a socket filter
811 * @fp: filter to remove
813 * Remove a filter from a socket and release its resources.
816 static inline void sk_filter_release(struct sock
*sk
, struct sk_filter
*fp
)
818 unsigned int size
= sk_filter_len(fp
);
820 atomic_sub(size
, &sk
->sk_omem_alloc
);
822 if (atomic_dec_and_test(&fp
->refcnt
))
826 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
828 atomic_inc(&fp
->refcnt
);
829 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
833 * Socket reference counting postulates.
835 * * Each user of socket SHOULD hold a reference count.
836 * * Each access point to socket (an hash table bucket, reference from a list,
837 * running timer, skb in flight MUST hold a reference count.
838 * * When reference count hits 0, it means it will never increase back.
839 * * When reference count hits 0, it means that no references from
840 * outside exist to this socket and current process on current CPU
841 * is last user and may/should destroy this socket.
842 * * sk_free is called from any context: process, BH, IRQ. When
843 * it is called, socket has no references from outside -> sk_free
844 * may release descendant resources allocated by the socket, but
845 * to the time when it is called, socket is NOT referenced by any
846 * hash tables, lists etc.
847 * * Packets, delivered from outside (from network or from another process)
848 * and enqueued on receive/error queues SHOULD NOT grab reference count,
849 * when they sit in queue. Otherwise, packets will leak to hole, when
850 * socket is looked up by one cpu and unhasing is made by another CPU.
851 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
852 * (leak to backlog). Packet socket does all the processing inside
853 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
854 * use separate SMP lock, so that they are prone too.
857 /* Ungrab socket and destroy it, if it was the last reference. */
858 static inline void sock_put(struct sock
*sk
)
860 if (atomic_dec_and_test(&sk
->sk_refcnt
))
864 /* Detach socket from process context.
865 * Announce socket dead, detach it from wait queue and inode.
866 * Note that parent inode held reference count on this struct sock,
867 * we do not release it in this function, because protocol
868 * probably wants some additional cleanups or even continuing
869 * to work with this socket (TCP).
871 static inline void sock_orphan(struct sock
*sk
)
873 write_lock_bh(&sk
->sk_callback_lock
);
874 sock_set_flag(sk
, SOCK_DEAD
);
875 sk
->sk_socket
= NULL
;
877 write_unlock_bh(&sk
->sk_callback_lock
);
880 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
882 write_lock_bh(&sk
->sk_callback_lock
);
883 sk
->sk_sleep
= &parent
->wait
;
885 sk
->sk_socket
= parent
;
886 write_unlock_bh(&sk
->sk_callback_lock
);
889 extern int sock_i_uid(struct sock
*sk
);
890 extern unsigned long sock_i_ino(struct sock
*sk
);
892 static inline struct dst_entry
*
893 __sk_dst_get(struct sock
*sk
)
895 return sk
->sk_dst_cache
;
898 static inline struct dst_entry
*
899 sk_dst_get(struct sock
*sk
)
901 struct dst_entry
*dst
;
903 read_lock(&sk
->sk_dst_lock
);
904 dst
= sk
->sk_dst_cache
;
907 read_unlock(&sk
->sk_dst_lock
);
912 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
914 struct dst_entry
*old_dst
;
916 old_dst
= sk
->sk_dst_cache
;
917 sk
->sk_dst_cache
= dst
;
918 dst_release(old_dst
);
922 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
924 write_lock(&sk
->sk_dst_lock
);
925 __sk_dst_set(sk
, dst
);
926 write_unlock(&sk
->sk_dst_lock
);
930 __sk_dst_reset(struct sock
*sk
)
932 struct dst_entry
*old_dst
;
934 old_dst
= sk
->sk_dst_cache
;
935 sk
->sk_dst_cache
= NULL
;
936 dst_release(old_dst
);
940 sk_dst_reset(struct sock
*sk
)
942 write_lock(&sk
->sk_dst_lock
);
944 write_unlock(&sk
->sk_dst_lock
);
947 static inline struct dst_entry
*
948 __sk_dst_check(struct sock
*sk
, u32 cookie
)
950 struct dst_entry
*dst
= sk
->sk_dst_cache
;
952 if (dst
&& dst
->obsolete
&& dst
->ops
->check(dst
, cookie
) == NULL
) {
953 sk
->sk_dst_cache
= NULL
;
961 static inline struct dst_entry
*
962 sk_dst_check(struct sock
*sk
, u32 cookie
)
964 struct dst_entry
*dst
= sk_dst_get(sk
);
966 if (dst
&& dst
->obsolete
&& dst
->ops
->check(dst
, cookie
) == NULL
) {
975 static inline void sk_charge_skb(struct sock
*sk
, struct sk_buff
*skb
)
977 sk
->sk_wmem_queued
+= skb
->truesize
;
978 sk
->sk_forward_alloc
-= skb
->truesize
;
981 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
982 struct sk_buff
*skb
, struct page
*page
,
985 if (skb
->ip_summed
== CHECKSUM_NONE
) {
987 unsigned int csum
= csum_and_copy_from_user(from
,
988 page_address(page
) + off
,
992 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
993 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
997 skb
->data_len
+= copy
;
998 skb
->truesize
+= copy
;
999 sk
->sk_wmem_queued
+= copy
;
1000 sk
->sk_forward_alloc
-= copy
;
1005 * Queue a received datagram if it will fit. Stream and sequenced
1006 * protocols can't normally use this as they need to fit buffers in
1007 * and play with them.
1009 * Inlined as it's very short and called for pretty much every
1010 * packet ever received.
1013 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1017 skb
->destructor
= sock_wfree
;
1018 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1021 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1024 skb
->destructor
= sock_rfree
;
1025 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1028 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1029 unsigned long expires
);
1031 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1033 static inline int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
1038 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1039 number of warnings when compiling with -W --ANK
1041 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
1042 (unsigned)sk
->sk_rcvbuf
) {
1047 /* It would be deadlock, if sock_queue_rcv_skb is used
1048 with socket lock! We assume that users of this
1049 function are lock free.
1051 err
= sk_filter(sk
, skb
, 1);
1056 skb_set_owner_r(skb
, sk
);
1058 /* Cache the SKB length before we tack it onto the receive
1059 * queue. Once it is added it no longer belongs to us and
1060 * may be freed by other threads of control pulling packets
1065 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1067 if (!sock_flag(sk
, SOCK_DEAD
))
1068 sk
->sk_data_ready(sk
, skb_len
);
1073 static inline int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
)
1075 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1076 number of warnings when compiling with -W --ANK
1078 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
1079 (unsigned)sk
->sk_rcvbuf
)
1081 skb_set_owner_r(skb
, sk
);
1082 skb_queue_tail(&sk
->sk_error_queue
, skb
);
1083 if (!sock_flag(sk
, SOCK_DEAD
))
1084 sk
->sk_data_ready(sk
, skb
->len
);
1089 * Recover an error report and clear atomically
1092 static inline int sock_error(struct sock
*sk
)
1094 int err
= xchg(&sk
->sk_err
, 0);
1098 static inline unsigned long sock_wspace(struct sock
*sk
)
1102 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1103 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1110 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1112 if (sk
->sk_socket
&& sk
->sk_socket
->fasync_list
)
1113 sock_wake_async(sk
->sk_socket
, how
, band
);
1116 #define SOCK_MIN_SNDBUF 2048
1117 #define SOCK_MIN_RCVBUF 256
1119 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1121 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1122 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
/ 2);
1123 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1127 static inline struct sk_buff
*sk_stream_alloc_pskb(struct sock
*sk
,
1128 int size
, int mem
, int gfp
)
1130 struct sk_buff
*skb
= alloc_skb(size
+ sk
->sk_prot
->max_header
, gfp
);
1133 skb
->truesize
+= mem
;
1134 if (sk
->sk_forward_alloc
>= (int)skb
->truesize
||
1135 sk_stream_mem_schedule(sk
, skb
->truesize
, 0)) {
1136 skb_reserve(skb
, sk
->sk_prot
->max_header
);
1141 sk
->sk_prot
->enter_memory_pressure();
1142 sk_stream_moderate_sndbuf(sk
);
1147 static inline struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
,
1150 return sk_stream_alloc_pskb(sk
, size
, 0, gfp
);
1153 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1155 struct page
*page
= NULL
;
1157 if (sk
->sk_forward_alloc
>= (int)PAGE_SIZE
||
1158 sk_stream_mem_schedule(sk
, PAGE_SIZE
, 0))
1159 page
= alloc_pages(sk
->sk_allocation
, 0);
1161 sk
->sk_prot
->enter_memory_pressure();
1162 sk_stream_moderate_sndbuf(sk
);
1167 #define sk_stream_for_retrans_queue(skb, sk) \
1168 for (skb = (sk)->sk_write_queue.next; \
1169 (skb != (sk)->sk_send_head) && \
1170 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1174 * Default write policy as shown to user space via poll/select/SIGIO
1176 static inline int sock_writeable(const struct sock
*sk
)
1178 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
/ 2);
1181 static inline int gfp_any(void)
1183 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
1186 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1188 return noblock
? 0 : sk
->sk_rcvtimeo
;
1191 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1193 return noblock
? 0 : sk
->sk_sndtimeo
;
1196 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1198 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1201 /* Alas, with timeout socket operations are not restartable.
1202 * Compare this to poll().
1204 static inline int sock_intr_errno(long timeo
)
1206 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
1209 static __inline__
void
1210 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
1212 struct timeval
*stamp
= &skb
->stamp
;
1213 if (sock_flag(sk
, SOCK_RCVTSTAMP
)) {
1214 /* Race occurred between timestamp enabling and packet
1215 receiving. Fill in the current time for now. */
1216 if (stamp
->tv_sec
== 0)
1217 do_gettimeofday(stamp
);
1218 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP
, sizeof(struct timeval
),
1221 sk
->sk_stamp
= *stamp
;
1225 * sk_eat_skb - Release a skb if it is no longer needed
1226 * @sk - socket to eat this skb from
1227 * @skb - socket buffer to eat
1229 * This routine must be called with interrupts disabled or with the socket
1230 * locked so that the sk_buff queue operation is ok.
1232 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
)
1234 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1238 extern void sock_enable_timestamp(struct sock
*sk
);
1239 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
1242 * Enable debug/info messages
1246 #define NETDEBUG(x) do { } while (0)
1247 #define LIMIT_NETDEBUG(x) do {} while(0)
1249 #define NETDEBUG(x) do { x; } while (0)
1250 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1254 * Macros for sleeping on a socket. Use them like this:
1256 * SOCK_SLEEP_PRE(sk)
1259 * SOCK_SLEEP_POST(sk)
1261 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1262 * and when the last use of them in DECnet has gone, I'm intending to
1266 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1267 DECLARE_WAITQUEUE(wait, tsk); \
1268 tsk->state = TASK_INTERRUPTIBLE; \
1269 add_wait_queue((sk)->sk_sleep, &wait); \
1272 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1273 remove_wait_queue((sk)->sk_sleep, &wait); \
1277 static inline void sock_valbool_flag(struct sock
*sk
, int bit
, int valbool
)
1280 sock_set_flag(sk
, bit
);
1282 sock_reset_flag(sk
, bit
);
1285 extern __u32 sysctl_wmem_max
;
1286 extern __u32 sysctl_rmem_max
;
1289 int siocdevprivate_ioctl(unsigned int fd
, unsigned int cmd
, unsigned long arg
);
1291 static inline int siocdevprivate_ioctl(unsigned int fd
, unsigned int cmd
, unsigned long arg
)
1297 #endif /* _SOCK_H */