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/list.h>
44 #include <linux/timer.h>
45 #include <linux/cache.h>
46 #include <linux/module.h>
47 #include <linux/lockdep.h>
48 #include <linux/netdevice.h>
49 #include <linux/skbuff.h> /* struct sk_buff */
51 #include <linux/security.h>
53 #include <linux/filter.h>
55 #include <asm/atomic.h>
57 #include <net/checksum.h>
60 * This structure really needs to be cleaned up.
61 * Most of it is for TCP, and not used by any of
62 * the other protocols.
65 /* Define this to get the SOCK_DBG debugging facility. */
66 #define SOCK_DEBUGGING
68 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
69 printk(KERN_DEBUG msg); } while (0)
71 #define SOCK_DEBUG(sk, msg...) do { } while (0)
74 /* This is the per-socket lock. The spinlock provides a synchronization
75 * between user contexts and software interrupt processing, whereas the
76 * mini-semaphore synchronizes multiple users amongst themselves.
81 struct sock_iocb
*owner
;
84 * We express the mutex-alike socket_lock semantics
85 * to the lock validator by explicitly managing
86 * the slock as a lock variant (in addition to
89 #ifdef CONFIG_DEBUG_LOCK_ALLOC
90 struct lockdep_map dep_map
;
98 * struct sock_common - minimal network layer representation of sockets
99 * @skc_family: network address family
100 * @skc_state: Connection state
101 * @skc_reuse: %SO_REUSEADDR setting
102 * @skc_bound_dev_if: bound device index if != 0
103 * @skc_node: main hash linkage for various protocol lookup tables
104 * @skc_bind_node: bind hash linkage for various protocol lookup tables
105 * @skc_refcnt: reference count
106 * @skc_hash: hash value used with various protocol lookup tables
107 * @skc_prot: protocol handlers inside a network family
109 * This is the minimal network layer representation of sockets, the header
110 * for struct sock and struct inet_timewait_sock.
113 unsigned short skc_family
;
114 volatile unsigned char skc_state
;
115 unsigned char skc_reuse
;
116 int skc_bound_dev_if
;
117 struct hlist_node skc_node
;
118 struct hlist_node skc_bind_node
;
120 unsigned int skc_hash
;
121 struct proto
*skc_prot
;
125 * struct sock - network layer representation of sockets
126 * @__sk_common: shared layout with inet_timewait_sock
127 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
128 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
129 * @sk_lock: synchronizer
130 * @sk_rcvbuf: size of receive buffer in bytes
131 * @sk_sleep: sock wait queue
132 * @sk_dst_cache: destination cache
133 * @sk_dst_lock: destination cache lock
134 * @sk_policy: flow policy
135 * @sk_rmem_alloc: receive queue bytes committed
136 * @sk_receive_queue: incoming packets
137 * @sk_wmem_alloc: transmit queue bytes committed
138 * @sk_write_queue: Packet sending queue
139 * @sk_async_wait_queue: DMA copied packets
140 * @sk_omem_alloc: "o" is "option" or "other"
141 * @sk_wmem_queued: persistent queue size
142 * @sk_forward_alloc: space allocated forward
143 * @sk_allocation: allocation mode
144 * @sk_sndbuf: size of send buffer in bytes
145 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
146 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
147 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
148 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
149 * @sk_lingertime: %SO_LINGER l_linger setting
150 * @sk_backlog: always used with the per-socket spinlock held
151 * @sk_callback_lock: used with the callbacks in the end of this struct
152 * @sk_error_queue: rarely used
153 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
154 * @sk_err: last error
155 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
156 * @sk_ack_backlog: current listen backlog
157 * @sk_max_ack_backlog: listen backlog set in listen()
158 * @sk_priority: %SO_PRIORITY setting
159 * @sk_type: socket type (%SOCK_STREAM, etc)
160 * @sk_protocol: which protocol this socket belongs in this network family
161 * @sk_peercred: %SO_PEERCRED setting
162 * @sk_rcvlowat: %SO_RCVLOWAT setting
163 * @sk_rcvtimeo: %SO_RCVTIMEO setting
164 * @sk_sndtimeo: %SO_SNDTIMEO setting
165 * @sk_filter: socket filtering instructions
166 * @sk_protinfo: private area, net family specific, when not using slab
167 * @sk_timer: sock cleanup timer
168 * @sk_stamp: time stamp of last packet received
169 * @sk_socket: Identd and reporting IO signals
170 * @sk_user_data: RPC layer private data
171 * @sk_sndmsg_page: cached page for sendmsg
172 * @sk_sndmsg_off: cached offset for sendmsg
173 * @sk_send_head: front of stuff to transmit
174 * @sk_security: used by security modules
175 * @sk_write_pending: a write to stream socket waits to start
176 * @sk_state_change: callback to indicate change in the state of the sock
177 * @sk_data_ready: callback to indicate there is data to be processed
178 * @sk_write_space: callback to indicate there is bf sending space available
179 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
180 * @sk_backlog_rcv: callback to process the backlog
181 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
185 * Now struct inet_timewait_sock also uses sock_common, so please just
186 * don't add nothing before this first member (__sk_common) --acme
188 struct sock_common __sk_common
;
189 #define sk_family __sk_common.skc_family
190 #define sk_state __sk_common.skc_state
191 #define sk_reuse __sk_common.skc_reuse
192 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
193 #define sk_node __sk_common.skc_node
194 #define sk_bind_node __sk_common.skc_bind_node
195 #define sk_refcnt __sk_common.skc_refcnt
196 #define sk_hash __sk_common.skc_hash
197 #define sk_prot __sk_common.skc_prot
198 unsigned char sk_shutdown
: 2,
201 unsigned char sk_protocol
;
202 unsigned short sk_type
;
204 socket_lock_t sk_lock
;
205 wait_queue_head_t
*sk_sleep
;
206 struct dst_entry
*sk_dst_cache
;
207 struct xfrm_policy
*sk_policy
[2];
208 rwlock_t sk_dst_lock
;
209 atomic_t sk_rmem_alloc
;
210 atomic_t sk_wmem_alloc
;
211 atomic_t sk_omem_alloc
;
212 struct sk_buff_head sk_receive_queue
;
213 struct sk_buff_head sk_write_queue
;
214 struct sk_buff_head sk_async_wait_queue
;
216 int sk_forward_alloc
;
222 unsigned long sk_flags
;
223 unsigned long sk_lingertime
;
225 * The backlog queue is special, it is always used with
226 * the per-socket spinlock held and requires low latency
227 * access. Therefore we special case it's implementation.
230 struct sk_buff
*head
;
231 struct sk_buff
*tail
;
233 struct sk_buff_head sk_error_queue
;
234 struct proto
*sk_prot_creator
;
235 rwlock_t sk_callback_lock
;
238 unsigned short sk_ack_backlog
;
239 unsigned short sk_max_ack_backlog
;
241 struct ucred sk_peercred
;
244 struct sk_filter
*sk_filter
;
246 struct timer_list sk_timer
;
247 struct timeval sk_stamp
;
248 struct socket
*sk_socket
;
250 struct page
*sk_sndmsg_page
;
251 struct sk_buff
*sk_send_head
;
253 int sk_write_pending
;
255 void (*sk_state_change
)(struct sock
*sk
);
256 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
257 void (*sk_write_space
)(struct sock
*sk
);
258 void (*sk_error_report
)(struct sock
*sk
);
259 int (*sk_backlog_rcv
)(struct sock
*sk
,
260 struct sk_buff
*skb
);
261 void (*sk_destruct
)(struct sock
*sk
);
265 * Hashed lists helper routines
267 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
269 return hlist_entry(head
->first
, struct sock
, sk_node
);
272 static inline struct sock
*sk_head(const struct hlist_head
*head
)
274 return hlist_empty(head
) ? NULL
: __sk_head(head
);
277 static inline struct sock
*sk_next(const struct sock
*sk
)
279 return sk
->sk_node
.next
?
280 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
283 static inline int sk_unhashed(const struct sock
*sk
)
285 return hlist_unhashed(&sk
->sk_node
);
288 static inline int sk_hashed(const struct sock
*sk
)
290 return !sk_unhashed(sk
);
293 static __inline__
void sk_node_init(struct hlist_node
*node
)
298 static __inline__
void __sk_del_node(struct sock
*sk
)
300 __hlist_del(&sk
->sk_node
);
303 static __inline__
int __sk_del_node_init(struct sock
*sk
)
307 sk_node_init(&sk
->sk_node
);
313 /* Grab socket reference count. This operation is valid only
314 when sk is ALREADY grabbed f.e. it is found in hash table
315 or a list and the lookup is made under lock preventing hash table
319 static inline void sock_hold(struct sock
*sk
)
321 atomic_inc(&sk
->sk_refcnt
);
324 /* Ungrab socket in the context, which assumes that socket refcnt
325 cannot hit zero, f.e. it is true in context of any socketcall.
327 static inline void __sock_put(struct sock
*sk
)
329 atomic_dec(&sk
->sk_refcnt
);
332 static __inline__
int sk_del_node_init(struct sock
*sk
)
334 int rc
= __sk_del_node_init(sk
);
337 /* paranoid for a while -acme */
338 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
344 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
346 hlist_add_head(&sk
->sk_node
, list
);
349 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
352 __sk_add_node(sk
, list
);
355 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
357 __hlist_del(&sk
->sk_bind_node
);
360 static __inline__
void sk_add_bind_node(struct sock
*sk
,
361 struct hlist_head
*list
)
363 hlist_add_head(&sk
->sk_bind_node
, list
);
366 #define sk_for_each(__sk, node, list) \
367 hlist_for_each_entry(__sk, node, list, sk_node)
368 #define sk_for_each_from(__sk, node) \
369 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
370 hlist_for_each_entry_from(__sk, node, sk_node)
371 #define sk_for_each_continue(__sk, node) \
372 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
373 hlist_for_each_entry_continue(__sk, node, sk_node)
374 #define sk_for_each_safe(__sk, node, tmp, list) \
375 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
376 #define sk_for_each_bound(__sk, node, list) \
377 hlist_for_each_entry(__sk, node, list, sk_bind_node)
390 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
391 SOCK_DBG
, /* %SO_DEBUG setting */
392 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
393 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
394 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
397 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
399 nsk
->sk_flags
= osk
->sk_flags
;
402 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
404 __set_bit(flag
, &sk
->sk_flags
);
407 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
409 __clear_bit(flag
, &sk
->sk_flags
);
412 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
414 return test_bit(flag
, &sk
->sk_flags
);
417 static inline void sk_acceptq_removed(struct sock
*sk
)
419 sk
->sk_ack_backlog
--;
422 static inline void sk_acceptq_added(struct sock
*sk
)
424 sk
->sk_ack_backlog
++;
427 static inline int sk_acceptq_is_full(struct sock
*sk
)
429 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
433 * Compute minimal free write space needed to queue new packets.
435 static inline int sk_stream_min_wspace(struct sock
*sk
)
437 return sk
->sk_wmem_queued
/ 2;
440 static inline int sk_stream_wspace(struct sock
*sk
)
442 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
445 extern void sk_stream_write_space(struct sock
*sk
);
447 static inline int sk_stream_memory_free(struct sock
*sk
)
449 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
452 extern void sk_stream_rfree(struct sk_buff
*skb
);
454 static inline void sk_stream_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
457 skb
->destructor
= sk_stream_rfree
;
458 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
459 sk
->sk_forward_alloc
-= skb
->truesize
;
462 static inline void sk_stream_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
464 skb_truesize_check(skb
);
465 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
466 sk
->sk_wmem_queued
-= skb
->truesize
;
467 sk
->sk_forward_alloc
+= skb
->truesize
;
471 /* The per-socket spinlock must be held here. */
472 static inline void sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
474 if (!sk
->sk_backlog
.tail
) {
475 sk
->sk_backlog
.head
= sk
->sk_backlog
.tail
= skb
;
477 sk
->sk_backlog
.tail
->next
= skb
;
478 sk
->sk_backlog
.tail
= skb
;
483 #define sk_wait_event(__sk, __timeo, __condition) \
485 release_sock(__sk); \
488 *(__timeo) = schedule_timeout(*(__timeo)); \
495 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
496 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
497 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
498 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
499 extern void sk_stream_kill_queues(struct sock
*sk
);
501 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
503 struct request_sock_ops
;
504 struct timewait_sock_ops
;
506 /* Networking protocol blocks we attach to sockets.
507 * socket layer -> transport layer interface
508 * transport -> network interface is defined by struct inet_proto
511 void (*close
)(struct sock
*sk
,
513 int (*connect
)(struct sock
*sk
,
514 struct sockaddr
*uaddr
,
516 int (*disconnect
)(struct sock
*sk
, int flags
);
518 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
520 int (*ioctl
)(struct sock
*sk
, int cmd
,
522 int (*init
)(struct sock
*sk
);
523 int (*destroy
)(struct sock
*sk
);
524 void (*shutdown
)(struct sock
*sk
, int how
);
525 int (*setsockopt
)(struct sock
*sk
, int level
,
526 int optname
, char __user
*optval
,
528 int (*getsockopt
)(struct sock
*sk
, int level
,
529 int optname
, char __user
*optval
,
531 int (*compat_setsockopt
)(struct sock
*sk
,
533 int optname
, char __user
*optval
,
535 int (*compat_getsockopt
)(struct sock
*sk
,
537 int optname
, char __user
*optval
,
539 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
540 struct msghdr
*msg
, size_t len
);
541 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
543 size_t len
, int noblock
, int flags
,
545 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
546 int offset
, size_t size
, int flags
);
547 int (*bind
)(struct sock
*sk
,
548 struct sockaddr
*uaddr
, int addr_len
);
550 int (*backlog_rcv
) (struct sock
*sk
,
551 struct sk_buff
*skb
);
553 /* Keeping track of sk's, looking them up, and port selection methods. */
554 void (*hash
)(struct sock
*sk
);
555 void (*unhash
)(struct sock
*sk
);
556 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
558 /* Memory pressure */
559 void (*enter_memory_pressure
)(void);
560 atomic_t
*memory_allocated
; /* Current allocated memory. */
561 atomic_t
*sockets_allocated
; /* Current number of sockets. */
563 * Pressure flag: try to collapse.
564 * Technical note: it is used by multiple contexts non atomically.
565 * All the sk_stream_mem_schedule() is of this nature: accounting
566 * is strict, actions are advisory and have some latency.
568 int *memory_pressure
;
574 struct kmem_cache
*slab
;
575 unsigned int obj_size
;
577 atomic_t
*orphan_count
;
579 struct request_sock_ops
*rsk_prot
;
580 struct timewait_sock_ops
*twsk_prot
;
582 struct module
*owner
;
586 struct list_head node
;
587 #ifdef SOCK_REFCNT_DEBUG
592 u8 __pad
[SMP_CACHE_BYTES
- sizeof(int)];
596 extern int proto_register(struct proto
*prot
, int alloc_slab
);
597 extern void proto_unregister(struct proto
*prot
);
599 #ifdef SOCK_REFCNT_DEBUG
600 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
602 atomic_inc(&sk
->sk_prot
->socks
);
605 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
607 atomic_dec(&sk
->sk_prot
->socks
);
608 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
609 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
612 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
614 if (atomic_read(&sk
->sk_refcnt
) != 1)
615 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
616 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
618 #else /* SOCK_REFCNT_DEBUG */
619 #define sk_refcnt_debug_inc(sk) do { } while (0)
620 #define sk_refcnt_debug_dec(sk) do { } while (0)
621 #define sk_refcnt_debug_release(sk) do { } while (0)
622 #endif /* SOCK_REFCNT_DEBUG */
624 /* Called with local bh disabled */
625 static __inline__
void sock_prot_inc_use(struct proto
*prot
)
627 prot
->stats
[smp_processor_id()].inuse
++;
630 static __inline__
void sock_prot_dec_use(struct proto
*prot
)
632 prot
->stats
[smp_processor_id()].inuse
--;
635 /* With per-bucket locks this operation is not-atomic, so that
636 * this version is not worse.
638 static inline void __sk_prot_rehash(struct sock
*sk
)
640 sk
->sk_prot
->unhash(sk
);
641 sk
->sk_prot
->hash(sk
);
644 /* About 10 seconds */
645 #define SOCK_DESTROY_TIME (10*HZ)
647 /* Sockets 0-1023 can't be bound to unless you are superuser */
648 #define PROT_SOCK 1024
650 #define SHUTDOWN_MASK 3
651 #define RCV_SHUTDOWN 1
652 #define SEND_SHUTDOWN 2
654 #define SOCK_SNDBUF_LOCK 1
655 #define SOCK_RCVBUF_LOCK 2
656 #define SOCK_BINDADDR_LOCK 4
657 #define SOCK_BINDPORT_LOCK 8
659 /* sock_iocb: used to kick off async processing of socket ios */
661 struct list_head list
;
667 struct scm_cookie
*scm
;
668 struct msghdr
*msg
, async_msg
;
672 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
674 return (struct sock_iocb
*)iocb
->private;
677 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
682 struct socket_alloc
{
683 struct socket socket
;
684 struct inode vfs_inode
;
687 static inline struct socket
*SOCKET_I(struct inode
*inode
)
689 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
692 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
694 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
697 extern void __sk_stream_mem_reclaim(struct sock
*sk
);
698 extern int sk_stream_mem_schedule(struct sock
*sk
, int size
, int kind
);
700 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
702 static inline int sk_stream_pages(int amt
)
704 return (amt
+ SK_STREAM_MEM_QUANTUM
- 1) / SK_STREAM_MEM_QUANTUM
;
707 static inline void sk_stream_mem_reclaim(struct sock
*sk
)
709 if (sk
->sk_forward_alloc
>= SK_STREAM_MEM_QUANTUM
)
710 __sk_stream_mem_reclaim(sk
);
713 static inline void sk_stream_writequeue_purge(struct sock
*sk
)
717 while ((skb
= __skb_dequeue(&sk
->sk_write_queue
)) != NULL
)
718 sk_stream_free_skb(sk
, skb
);
719 sk_stream_mem_reclaim(sk
);
722 static inline int sk_stream_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
)
724 return (int)skb
->truesize
<= sk
->sk_forward_alloc
||
725 sk_stream_mem_schedule(sk
, skb
->truesize
, 1);
728 static inline int sk_stream_wmem_schedule(struct sock
*sk
, int size
)
730 return size
<= sk
->sk_forward_alloc
||
731 sk_stream_mem_schedule(sk
, size
, 0);
734 /* Used by processes to "lock" a socket state, so that
735 * interrupts and bottom half handlers won't change it
736 * from under us. It essentially blocks any incoming
737 * packets, so that we won't get any new data or any
738 * packets that change the state of the socket.
740 * While locked, BH processing will add new packets to
741 * the backlog queue. This queue is processed by the
742 * owner of the socket lock right before it is released.
744 * Since ~2.3.5 it is also exclusive sleep lock serializing
745 * accesses from user process context.
747 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
750 * Macro so as to not evaluate some arguments when
751 * lockdep is not enabled.
753 * Mark both the sk_lock and the sk_lock.slock as a
754 * per-address-family lock class.
756 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
758 sk->sk_lock.owner = NULL; \
759 init_waitqueue_head(&sk->sk_lock.wq); \
760 spin_lock_init(&(sk)->sk_lock.slock); \
761 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
762 sizeof((sk)->sk_lock)); \
763 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
765 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
768 extern void FASTCALL(lock_sock_nested(struct sock
*sk
, int subclass
));
770 static inline void lock_sock(struct sock
*sk
)
772 lock_sock_nested(sk
, 0);
775 extern void FASTCALL(release_sock(struct sock
*sk
));
777 /* BH context may only use the following locking interface. */
778 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
779 #define bh_lock_sock_nested(__sk) \
780 spin_lock_nested(&((__sk)->sk_lock.slock), \
781 SINGLE_DEPTH_NESTING)
782 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
784 extern struct sock
*sk_alloc(int family
,
786 struct proto
*prot
, int zero_it
);
787 extern void sk_free(struct sock
*sk
);
788 extern struct sock
*sk_clone(const struct sock
*sk
,
789 const gfp_t priority
);
791 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
792 unsigned long size
, int force
,
794 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
795 unsigned long size
, int force
,
797 extern void sock_wfree(struct sk_buff
*skb
);
798 extern void sock_rfree(struct sk_buff
*skb
);
800 extern int sock_setsockopt(struct socket
*sock
, int level
,
801 int op
, char __user
*optval
,
804 extern int sock_getsockopt(struct socket
*sock
, int level
,
805 int op
, char __user
*optval
,
807 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
811 extern void *sock_kmalloc(struct sock
*sk
, int size
,
813 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
814 extern void sk_send_sigurg(struct sock
*sk
);
817 * Functions to fill in entries in struct proto_ops when a protocol
818 * does not implement a particular function.
820 extern int sock_no_bind(struct socket
*,
821 struct sockaddr
*, int);
822 extern int sock_no_connect(struct socket
*,
823 struct sockaddr
*, int, int);
824 extern int sock_no_socketpair(struct socket
*,
826 extern int sock_no_accept(struct socket
*,
827 struct socket
*, int);
828 extern int sock_no_getname(struct socket
*,
829 struct sockaddr
*, int *, int);
830 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
831 struct poll_table_struct
*);
832 extern int sock_no_ioctl(struct socket
*, unsigned int,
834 extern int sock_no_listen(struct socket
*, int);
835 extern int sock_no_shutdown(struct socket
*, int);
836 extern int sock_no_getsockopt(struct socket
*, int , int,
837 char __user
*, int __user
*);
838 extern int sock_no_setsockopt(struct socket
*, int, int,
840 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
841 struct msghdr
*, size_t);
842 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
843 struct msghdr
*, size_t, int);
844 extern int sock_no_mmap(struct file
*file
,
846 struct vm_area_struct
*vma
);
847 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
849 int offset
, size_t size
,
853 * Functions to fill in entries in struct proto_ops when a protocol
854 * uses the inet style.
856 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
857 char __user
*optval
, int __user
*optlen
);
858 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
859 struct msghdr
*msg
, size_t size
, int flags
);
860 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
861 char __user
*optval
, int optlen
);
862 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
863 int optname
, char __user
*optval
, int __user
*optlen
);
864 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
865 int optname
, char __user
*optval
, int optlen
);
867 extern void sk_common_release(struct sock
*sk
);
870 * Default socket callbacks and setup code
873 /* Initialise core socket variables */
874 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
877 * sk_filter - run a packet through a socket filter
878 * @sk: sock associated with &sk_buff
879 * @skb: buffer to filter
880 * @needlock: set to 1 if the sock is not locked by caller.
882 * Run the filter code and then cut skb->data to correct size returned by
883 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
884 * than pkt_len we keep whole skb->data. This is the socket level
885 * wrapper to sk_run_filter. It returns 0 if the packet should
886 * be accepted or -EPERM if the packet should be tossed.
890 static inline int sk_filter(struct sock
*sk
, struct sk_buff
*skb
)
893 struct sk_filter
*filter
;
895 err
= security_sock_rcv_skb(sk
, skb
);
900 filter
= sk
->sk_filter
;
902 unsigned int pkt_len
= sk_run_filter(skb
, filter
->insns
,
904 err
= pkt_len
? pskb_trim(skb
, pkt_len
) : -EPERM
;
906 rcu_read_unlock_bh();
912 * sk_filter_rcu_free: Free a socket filter
913 * @rcu: rcu_head that contains the sk_filter to free
915 static inline void sk_filter_rcu_free(struct rcu_head
*rcu
)
917 struct sk_filter
*fp
= container_of(rcu
, struct sk_filter
, rcu
);
922 * sk_filter_release: Release a socket filter
924 * @fp: filter to remove
926 * Remove a filter from a socket and release its resources.
929 static inline void sk_filter_release(struct sock
*sk
, struct sk_filter
*fp
)
931 unsigned int size
= sk_filter_len(fp
);
933 atomic_sub(size
, &sk
->sk_omem_alloc
);
935 if (atomic_dec_and_test(&fp
->refcnt
))
936 call_rcu_bh(&fp
->rcu
, sk_filter_rcu_free
);
939 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
941 atomic_inc(&fp
->refcnt
);
942 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
946 * Socket reference counting postulates.
948 * * Each user of socket SHOULD hold a reference count.
949 * * Each access point to socket (an hash table bucket, reference from a list,
950 * running timer, skb in flight MUST hold a reference count.
951 * * When reference count hits 0, it means it will never increase back.
952 * * When reference count hits 0, it means that no references from
953 * outside exist to this socket and current process on current CPU
954 * is last user and may/should destroy this socket.
955 * * sk_free is called from any context: process, BH, IRQ. When
956 * it is called, socket has no references from outside -> sk_free
957 * may release descendant resources allocated by the socket, but
958 * to the time when it is called, socket is NOT referenced by any
959 * hash tables, lists etc.
960 * * Packets, delivered from outside (from network or from another process)
961 * and enqueued on receive/error queues SHOULD NOT grab reference count,
962 * when they sit in queue. Otherwise, packets will leak to hole, when
963 * socket is looked up by one cpu and unhasing is made by another CPU.
964 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
965 * (leak to backlog). Packet socket does all the processing inside
966 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
967 * use separate SMP lock, so that they are prone too.
970 /* Ungrab socket and destroy it, if it was the last reference. */
971 static inline void sock_put(struct sock
*sk
)
973 if (atomic_dec_and_test(&sk
->sk_refcnt
))
977 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
980 /* Detach socket from process context.
981 * Announce socket dead, detach it from wait queue and inode.
982 * Note that parent inode held reference count on this struct sock,
983 * we do not release it in this function, because protocol
984 * probably wants some additional cleanups or even continuing
985 * to work with this socket (TCP).
987 static inline void sock_orphan(struct sock
*sk
)
989 write_lock_bh(&sk
->sk_callback_lock
);
990 sock_set_flag(sk
, SOCK_DEAD
);
991 sk
->sk_socket
= NULL
;
993 write_unlock_bh(&sk
->sk_callback_lock
);
996 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
998 write_lock_bh(&sk
->sk_callback_lock
);
999 sk
->sk_sleep
= &parent
->wait
;
1001 sk
->sk_socket
= parent
;
1002 security_sock_graft(sk
, parent
);
1003 write_unlock_bh(&sk
->sk_callback_lock
);
1006 static inline void sock_copy(struct sock
*nsk
, const struct sock
*osk
)
1008 #ifdef CONFIG_SECURITY_NETWORK
1009 void *sptr
= nsk
->sk_security
;
1012 memcpy(nsk
, osk
, osk
->sk_prot
->obj_size
);
1013 #ifdef CONFIG_SECURITY_NETWORK
1014 nsk
->sk_security
= sptr
;
1015 security_sk_clone(osk
, nsk
);
1019 extern int sock_i_uid(struct sock
*sk
);
1020 extern unsigned long sock_i_ino(struct sock
*sk
);
1022 static inline struct dst_entry
*
1023 __sk_dst_get(struct sock
*sk
)
1025 return sk
->sk_dst_cache
;
1028 static inline struct dst_entry
*
1029 sk_dst_get(struct sock
*sk
)
1031 struct dst_entry
*dst
;
1033 read_lock(&sk
->sk_dst_lock
);
1034 dst
= sk
->sk_dst_cache
;
1037 read_unlock(&sk
->sk_dst_lock
);
1042 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1044 struct dst_entry
*old_dst
;
1046 old_dst
= sk
->sk_dst_cache
;
1047 sk
->sk_dst_cache
= dst
;
1048 dst_release(old_dst
);
1052 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1054 write_lock(&sk
->sk_dst_lock
);
1055 __sk_dst_set(sk
, dst
);
1056 write_unlock(&sk
->sk_dst_lock
);
1060 __sk_dst_reset(struct sock
*sk
)
1062 struct dst_entry
*old_dst
;
1064 old_dst
= sk
->sk_dst_cache
;
1065 sk
->sk_dst_cache
= NULL
;
1066 dst_release(old_dst
);
1070 sk_dst_reset(struct sock
*sk
)
1072 write_lock(&sk
->sk_dst_lock
);
1074 write_unlock(&sk
->sk_dst_lock
);
1077 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1079 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1081 static inline int sk_can_gso(const struct sock
*sk
)
1083 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1086 static inline void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
)
1088 __sk_dst_set(sk
, dst
);
1089 sk
->sk_route_caps
= dst
->dev
->features
;
1090 if (sk
->sk_route_caps
& NETIF_F_GSO
)
1091 sk
->sk_route_caps
|= NETIF_F_GSO_MASK
;
1092 if (sk_can_gso(sk
)) {
1093 if (dst
->header_len
)
1094 sk
->sk_route_caps
&= ~NETIF_F_GSO_MASK
;
1096 sk
->sk_route_caps
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
1100 static inline void sk_charge_skb(struct sock
*sk
, struct sk_buff
*skb
)
1102 sk
->sk_wmem_queued
+= skb
->truesize
;
1103 sk
->sk_forward_alloc
-= skb
->truesize
;
1106 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1107 struct sk_buff
*skb
, struct page
*page
,
1110 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1112 __wsum csum
= csum_and_copy_from_user(from
,
1113 page_address(page
) + off
,
1117 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1118 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1122 skb
->data_len
+= copy
;
1123 skb
->truesize
+= copy
;
1124 sk
->sk_wmem_queued
+= copy
;
1125 sk
->sk_forward_alloc
-= copy
;
1130 * Queue a received datagram if it will fit. Stream and sequenced
1131 * protocols can't normally use this as they need to fit buffers in
1132 * and play with them.
1134 * Inlined as it's very short and called for pretty much every
1135 * packet ever received.
1138 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1142 skb
->destructor
= sock_wfree
;
1143 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1146 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1149 skb
->destructor
= sock_rfree
;
1150 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1153 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1154 unsigned long expires
);
1156 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1158 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1160 static inline int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
)
1162 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1163 number of warnings when compiling with -W --ANK
1165 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
1166 (unsigned)sk
->sk_rcvbuf
)
1168 skb_set_owner_r(skb
, sk
);
1169 skb_queue_tail(&sk
->sk_error_queue
, skb
);
1170 if (!sock_flag(sk
, SOCK_DEAD
))
1171 sk
->sk_data_ready(sk
, skb
->len
);
1176 * Recover an error report and clear atomically
1179 static inline int sock_error(struct sock
*sk
)
1182 if (likely(!sk
->sk_err
))
1184 err
= xchg(&sk
->sk_err
, 0);
1188 static inline unsigned long sock_wspace(struct sock
*sk
)
1192 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1193 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1200 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1202 if (sk
->sk_socket
&& sk
->sk_socket
->fasync_list
)
1203 sock_wake_async(sk
->sk_socket
, how
, band
);
1206 #define SOCK_MIN_SNDBUF 2048
1207 #define SOCK_MIN_RCVBUF 256
1209 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1211 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1212 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
/ 2);
1213 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1217 static inline struct sk_buff
*sk_stream_alloc_pskb(struct sock
*sk
,
1221 struct sk_buff
*skb
;
1224 hdr_len
= SKB_DATA_ALIGN(sk
->sk_prot
->max_header
);
1225 skb
= alloc_skb_fclone(size
+ hdr_len
, gfp
);
1227 skb
->truesize
+= mem
;
1228 if (sk_stream_wmem_schedule(sk
, skb
->truesize
)) {
1229 skb_reserve(skb
, hdr_len
);
1234 sk
->sk_prot
->enter_memory_pressure();
1235 sk_stream_moderate_sndbuf(sk
);
1240 static inline struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
,
1244 return sk_stream_alloc_pskb(sk
, size
, 0, gfp
);
1247 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1249 struct page
*page
= NULL
;
1251 page
= alloc_pages(sk
->sk_allocation
, 0);
1253 sk
->sk_prot
->enter_memory_pressure();
1254 sk_stream_moderate_sndbuf(sk
);
1259 #define sk_stream_for_retrans_queue(skb, sk) \
1260 for (skb = (sk)->sk_write_queue.next; \
1261 (skb != (sk)->sk_send_head) && \
1262 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1265 /*from STCP for fast SACK Process*/
1266 #define sk_stream_for_retrans_queue_from(skb, sk) \
1267 for (; (skb != (sk)->sk_send_head) && \
1268 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1272 * Default write policy as shown to user space via poll/select/SIGIO
1274 static inline int sock_writeable(const struct sock
*sk
)
1276 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
/ 2);
1279 static inline gfp_t
gfp_any(void)
1281 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
1284 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1286 return noblock
? 0 : sk
->sk_rcvtimeo
;
1289 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1291 return noblock
? 0 : sk
->sk_sndtimeo
;
1294 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1296 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1299 /* Alas, with timeout socket operations are not restartable.
1300 * Compare this to poll().
1302 static inline int sock_intr_errno(long timeo
)
1304 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
1307 static __inline__
void
1308 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
1310 struct timeval stamp
;
1312 skb_get_timestamp(skb
, &stamp
);
1313 if (sock_flag(sk
, SOCK_RCVTSTAMP
)) {
1314 /* Race occurred between timestamp enabling and packet
1315 receiving. Fill in the current time for now. */
1316 if (stamp
.tv_sec
== 0)
1317 do_gettimeofday(&stamp
);
1318 skb_set_timestamp(skb
, &stamp
);
1319 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP
, sizeof(struct timeval
),
1322 sk
->sk_stamp
= stamp
;
1326 * sk_eat_skb - Release a skb if it is no longer needed
1327 * @sk: socket to eat this skb from
1328 * @skb: socket buffer to eat
1329 * @copied_early: flag indicating whether DMA operations copied this data early
1331 * This routine must be called with interrupts disabled or with the socket
1332 * locked so that the sk_buff queue operation is ok.
1334 #ifdef CONFIG_NET_DMA
1335 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1337 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1341 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
1344 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1346 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1351 extern void sock_enable_timestamp(struct sock
*sk
);
1352 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
1355 * Enable debug/info messages
1358 #ifdef CONFIG_NETDEBUG
1359 #define NETDEBUG(fmt, args...) printk(fmt,##args)
1360 #define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0)
1362 #define NETDEBUG(fmt, args...) do { } while (0)
1363 #define LIMIT_NETDEBUG(fmt, args...) do { } while(0)
1367 * Macros for sleeping on a socket. Use them like this:
1369 * SOCK_SLEEP_PRE(sk)
1372 * SOCK_SLEEP_POST(sk)
1374 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1375 * and when the last use of them in DECnet has gone, I'm intending to
1379 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1380 DECLARE_WAITQUEUE(wait, tsk); \
1381 tsk->state = TASK_INTERRUPTIBLE; \
1382 add_wait_queue((sk)->sk_sleep, &wait); \
1385 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1386 remove_wait_queue((sk)->sk_sleep, &wait); \
1390 static inline void sock_valbool_flag(struct sock
*sk
, int bit
, int valbool
)
1393 sock_set_flag(sk
, bit
);
1395 sock_reset_flag(sk
, bit
);
1398 extern __u32 sysctl_wmem_max
;
1399 extern __u32 sysctl_rmem_max
;
1402 int siocdevprivate_ioctl(unsigned int fd
, unsigned int cmd
, unsigned long arg
);
1404 static inline int siocdevprivate_ioctl(unsigned int fd
, unsigned int cmd
, unsigned long arg
)
1410 extern void sk_init(void);
1412 #ifdef CONFIG_SYSCTL
1413 extern struct ctl_table core_table
[];
1416 extern int sysctl_optmem_max
;
1418 extern __u32 sysctl_wmem_default
;
1419 extern __u32 sysctl_rmem_default
;
1421 #endif /* _SOCK_H */