2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the AF_INET socket handler.
8 * Version: @(#)sock.h 1.0.4 05/13/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
43 #include <linux/kernel.h>
44 #include <linux/list.h>
45 #include <linux/list_nulls.h>
46 #include <linux/timer.h>
47 #include <linux/cache.h>
48 #include <linux/module.h>
49 #include <linux/lockdep.h>
50 #include <linux/netdevice.h>
51 #include <linux/skbuff.h> /* struct sk_buff */
53 #include <linux/security.h>
55 #include <linux/filter.h>
56 #include <linux/rculist_nulls.h>
58 #include <asm/atomic.h>
60 #include <net/checksum.h>
63 * This structure really needs to be cleaned up.
64 * Most of it is for TCP, and not used by any of
65 * the other protocols.
68 /* Define this to get the SOCK_DBG debugging facility. */
69 #define SOCK_DEBUGGING
71 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
72 printk(KERN_DEBUG msg); } while (0)
74 /* Validate arguments and do nothing */
75 static void inline int __attribute__ ((format (printf
, 2, 3)))
76 SOCK_DEBUG(struct sock
*sk
, const char *msg
, ...)
81 /* This is the per-socket lock. The spinlock provides a synchronization
82 * between user contexts and software interrupt processing, whereas the
83 * mini-semaphore synchronizes multiple users amongst themselves.
90 * We express the mutex-alike socket_lock semantics
91 * to the lock validator by explicitly managing
92 * the slock as a lock variant (in addition to
95 #ifdef CONFIG_DEBUG_LOCK_ALLOC
96 struct lockdep_map dep_map
;
105 * struct sock_common - minimal network layer representation of sockets
106 * @skc_family: network address family
107 * @skc_state: Connection state
108 * @skc_reuse: %SO_REUSEADDR setting
109 * @skc_bound_dev_if: bound device index if != 0
110 * @skc_node: main hash linkage for various protocol lookup tables
111 * @skc_nulls_node: main hash linkage for UDP/UDP-Lite protocol
112 * @skc_bind_node: bind hash linkage for various protocol lookup tables
113 * @skc_refcnt: reference count
114 * @skc_hash: hash value used with various protocol lookup tables
115 * @skc_prot: protocol handlers inside a network family
116 * @skc_net: reference to the network namespace of this socket
118 * This is the minimal network layer representation of sockets, the header
119 * for struct sock and struct inet_timewait_sock.
122 unsigned short skc_family
;
123 volatile unsigned char skc_state
;
124 unsigned char skc_reuse
;
125 int skc_bound_dev_if
;
127 struct hlist_node skc_node
;
128 struct hlist_nulls_node skc_nulls_node
;
130 struct hlist_node skc_bind_node
;
132 unsigned int skc_hash
;
133 struct proto
*skc_prot
;
140 * struct sock - network layer representation of sockets
141 * @__sk_common: shared layout with inet_timewait_sock
142 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
143 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
144 * @sk_lock: synchronizer
145 * @sk_rcvbuf: size of receive buffer in bytes
146 * @sk_sleep: sock wait queue
147 * @sk_dst_cache: destination cache
148 * @sk_dst_lock: destination cache lock
149 * @sk_policy: flow policy
150 * @sk_rmem_alloc: receive queue bytes committed
151 * @sk_receive_queue: incoming packets
152 * @sk_wmem_alloc: transmit queue bytes committed
153 * @sk_write_queue: Packet sending queue
154 * @sk_async_wait_queue: DMA copied packets
155 * @sk_omem_alloc: "o" is "option" or "other"
156 * @sk_wmem_queued: persistent queue size
157 * @sk_forward_alloc: space allocated forward
158 * @sk_allocation: allocation mode
159 * @sk_sndbuf: size of send buffer in bytes
160 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
161 * %SO_OOBINLINE settings
162 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
163 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
164 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
165 * @sk_gso_max_size: Maximum GSO segment size to build
166 * @sk_lingertime: %SO_LINGER l_linger setting
167 * @sk_backlog: always used with the per-socket spinlock held
168 * @sk_callback_lock: used with the callbacks in the end of this struct
169 * @sk_error_queue: rarely used
170 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
171 * IPV6_ADDRFORM for instance)
172 * @sk_err: last error
173 * @sk_err_soft: errors that don't cause failure but are the cause of a
174 * persistent failure not just 'timed out'
175 * @sk_drops: raw/udp drops counter
176 * @sk_ack_backlog: current listen backlog
177 * @sk_max_ack_backlog: listen backlog set in listen()
178 * @sk_priority: %SO_PRIORITY setting
179 * @sk_type: socket type (%SOCK_STREAM, etc)
180 * @sk_protocol: which protocol this socket belongs in this network family
181 * @sk_peercred: %SO_PEERCRED setting
182 * @sk_rcvlowat: %SO_RCVLOWAT setting
183 * @sk_rcvtimeo: %SO_RCVTIMEO setting
184 * @sk_sndtimeo: %SO_SNDTIMEO setting
185 * @sk_filter: socket filtering instructions
186 * @sk_protinfo: private area, net family specific, when not using slab
187 * @sk_timer: sock cleanup timer
188 * @sk_stamp: time stamp of last packet received
189 * @sk_socket: Identd and reporting IO signals
190 * @sk_user_data: RPC layer private data
191 * @sk_sndmsg_page: cached page for sendmsg
192 * @sk_sndmsg_off: cached offset for sendmsg
193 * @sk_send_head: front of stuff to transmit
194 * @sk_security: used by security modules
195 * @sk_mark: generic packet mark
196 * @sk_write_pending: a write to stream socket waits to start
197 * @sk_state_change: callback to indicate change in the state of the sock
198 * @sk_data_ready: callback to indicate there is data to be processed
199 * @sk_write_space: callback to indicate there is bf sending space available
200 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
201 * @sk_backlog_rcv: callback to process the backlog
202 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
206 * Now struct inet_timewait_sock also uses sock_common, so please just
207 * don't add nothing before this first member (__sk_common) --acme
209 struct sock_common __sk_common
;
210 #define sk_family __sk_common.skc_family
211 #define sk_state __sk_common.skc_state
212 #define sk_reuse __sk_common.skc_reuse
213 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
214 #define sk_node __sk_common.skc_node
215 #define sk_nulls_node __sk_common.skc_nulls_node
216 #define sk_bind_node __sk_common.skc_bind_node
217 #define sk_refcnt __sk_common.skc_refcnt
218 #define sk_hash __sk_common.skc_hash
219 #define sk_prot __sk_common.skc_prot
220 #define sk_net __sk_common.skc_net
221 unsigned char sk_shutdown
: 2,
224 unsigned char sk_protocol
;
225 unsigned short sk_type
;
227 socket_lock_t sk_lock
;
229 * The backlog queue is special, it is always used with
230 * the per-socket spinlock held and requires low latency
231 * access. Therefore we special case it's implementation.
234 struct sk_buff
*head
;
235 struct sk_buff
*tail
;
237 wait_queue_head_t
*sk_sleep
;
238 struct dst_entry
*sk_dst_cache
;
240 struct xfrm_policy
*sk_policy
[2];
242 rwlock_t sk_dst_lock
;
243 atomic_t sk_rmem_alloc
;
244 atomic_t sk_wmem_alloc
;
245 atomic_t sk_omem_alloc
;
247 struct sk_buff_head sk_receive_queue
;
248 struct sk_buff_head sk_write_queue
;
249 #ifdef CONFIG_NET_DMA
250 struct sk_buff_head sk_async_wait_queue
;
253 int sk_forward_alloc
;
257 unsigned int sk_gso_max_size
;
259 unsigned long sk_flags
;
260 unsigned long sk_lingertime
;
261 struct sk_buff_head sk_error_queue
;
262 struct proto
*sk_prot_creator
;
263 rwlock_t sk_callback_lock
;
267 unsigned short sk_ack_backlog
;
268 unsigned short sk_max_ack_backlog
;
270 struct ucred sk_peercred
;
273 struct sk_filter
*sk_filter
;
275 struct timer_list sk_timer
;
277 struct socket
*sk_socket
;
279 struct page
*sk_sndmsg_page
;
280 struct sk_buff
*sk_send_head
;
282 int sk_write_pending
;
283 #ifdef CONFIG_SECURITY
287 /* XXX 4 bytes hole on 64 bit */
288 void (*sk_state_change
)(struct sock
*sk
);
289 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
290 void (*sk_write_space
)(struct sock
*sk
);
291 void (*sk_error_report
)(struct sock
*sk
);
292 int (*sk_backlog_rcv
)(struct sock
*sk
,
293 struct sk_buff
*skb
);
294 void (*sk_destruct
)(struct sock
*sk
);
298 * Hashed lists helper routines
300 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
302 return hlist_entry(head
->first
, struct sock
, sk_node
);
305 static inline struct sock
*sk_head(const struct hlist_head
*head
)
307 return hlist_empty(head
) ? NULL
: __sk_head(head
);
310 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
312 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
315 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
317 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
320 static inline struct sock
*sk_next(const struct sock
*sk
)
322 return sk
->sk_node
.next
?
323 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
326 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
328 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
329 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
330 struct sock
, sk_nulls_node
) :
334 static inline int sk_unhashed(const struct sock
*sk
)
336 return hlist_unhashed(&sk
->sk_node
);
339 static inline int sk_hashed(const struct sock
*sk
)
341 return !sk_unhashed(sk
);
344 static __inline__
void sk_node_init(struct hlist_node
*node
)
349 static __inline__
void sk_nulls_node_init(struct hlist_nulls_node
*node
)
354 static __inline__
void __sk_del_node(struct sock
*sk
)
356 __hlist_del(&sk
->sk_node
);
359 static __inline__
int __sk_del_node_init(struct sock
*sk
)
363 sk_node_init(&sk
->sk_node
);
369 /* Grab socket reference count. This operation is valid only
370 when sk is ALREADY grabbed f.e. it is found in hash table
371 or a list and the lookup is made under lock preventing hash table
375 static inline void sock_hold(struct sock
*sk
)
377 atomic_inc(&sk
->sk_refcnt
);
380 /* Ungrab socket in the context, which assumes that socket refcnt
381 cannot hit zero, f.e. it is true in context of any socketcall.
383 static inline void __sock_put(struct sock
*sk
)
385 atomic_dec(&sk
->sk_refcnt
);
388 static __inline__
int sk_del_node_init(struct sock
*sk
)
390 int rc
= __sk_del_node_init(sk
);
393 /* paranoid for a while -acme */
394 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
400 static __inline__
int __sk_nulls_del_node_init_rcu(struct sock
*sk
)
403 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
409 static __inline__
int sk_nulls_del_node_init_rcu(struct sock
*sk
)
411 int rc
= __sk_nulls_del_node_init_rcu(sk
);
414 /* paranoid for a while -acme */
415 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
421 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
423 hlist_add_head(&sk
->sk_node
, list
);
426 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
429 __sk_add_node(sk
, list
);
432 static __inline__
void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
434 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
437 static __inline__
void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
440 __sk_nulls_add_node_rcu(sk
, list
);
443 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
445 __hlist_del(&sk
->sk_bind_node
);
448 static __inline__
void sk_add_bind_node(struct sock
*sk
,
449 struct hlist_head
*list
)
451 hlist_add_head(&sk
->sk_bind_node
, list
);
454 #define sk_for_each(__sk, node, list) \
455 hlist_for_each_entry(__sk, node, list, sk_node)
456 #define sk_nulls_for_each(__sk, node, list) \
457 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
458 #define sk_nulls_for_each_rcu(__sk, node, list) \
459 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
460 #define sk_for_each_from(__sk, node) \
461 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
462 hlist_for_each_entry_from(__sk, node, sk_node)
463 #define sk_nulls_for_each_from(__sk, node) \
464 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
465 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
466 #define sk_for_each_continue(__sk, node) \
467 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
468 hlist_for_each_entry_continue(__sk, node, sk_node)
469 #define sk_for_each_safe(__sk, node, tmp, list) \
470 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
471 #define sk_for_each_bound(__sk, node, list) \
472 hlist_for_each_entry(__sk, node, list, sk_bind_node)
485 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
486 SOCK_DBG
, /* %SO_DEBUG setting */
487 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
488 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
489 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
490 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
493 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
495 nsk
->sk_flags
= osk
->sk_flags
;
498 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
500 __set_bit(flag
, &sk
->sk_flags
);
503 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
505 __clear_bit(flag
, &sk
->sk_flags
);
508 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
510 return test_bit(flag
, &sk
->sk_flags
);
513 static inline void sk_acceptq_removed(struct sock
*sk
)
515 sk
->sk_ack_backlog
--;
518 static inline void sk_acceptq_added(struct sock
*sk
)
520 sk
->sk_ack_backlog
++;
523 static inline int sk_acceptq_is_full(struct sock
*sk
)
525 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
529 * Compute minimal free write space needed to queue new packets.
531 static inline int sk_stream_min_wspace(struct sock
*sk
)
533 return sk
->sk_wmem_queued
>> 1;
536 static inline int sk_stream_wspace(struct sock
*sk
)
538 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
541 extern void sk_stream_write_space(struct sock
*sk
);
543 static inline int sk_stream_memory_free(struct sock
*sk
)
545 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
548 /* The per-socket spinlock must be held here. */
549 static inline void sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
551 if (!sk
->sk_backlog
.tail
) {
552 sk
->sk_backlog
.head
= sk
->sk_backlog
.tail
= skb
;
554 sk
->sk_backlog
.tail
->next
= skb
;
555 sk
->sk_backlog
.tail
= skb
;
560 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
562 return sk
->sk_backlog_rcv(sk
, skb
);
565 #define sk_wait_event(__sk, __timeo, __condition) \
567 release_sock(__sk); \
568 __rc = __condition; \
570 *(__timeo) = schedule_timeout(*(__timeo)); \
573 __rc = __condition; \
577 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
578 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
579 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
580 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
581 extern void sk_stream_kill_queues(struct sock
*sk
);
583 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
585 struct request_sock_ops
;
586 struct timewait_sock_ops
;
587 struct inet_hashinfo
;
590 /* Networking protocol blocks we attach to sockets.
591 * socket layer -> transport layer interface
592 * transport -> network interface is defined by struct inet_proto
595 void (*close
)(struct sock
*sk
,
597 int (*connect
)(struct sock
*sk
,
598 struct sockaddr
*uaddr
,
600 int (*disconnect
)(struct sock
*sk
, int flags
);
602 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
604 int (*ioctl
)(struct sock
*sk
, int cmd
,
606 int (*init
)(struct sock
*sk
);
607 void (*destroy
)(struct sock
*sk
);
608 void (*shutdown
)(struct sock
*sk
, int how
);
609 int (*setsockopt
)(struct sock
*sk
, int level
,
610 int optname
, char __user
*optval
,
612 int (*getsockopt
)(struct sock
*sk
, int level
,
613 int optname
, char __user
*optval
,
616 int (*compat_setsockopt
)(struct sock
*sk
,
618 int optname
, char __user
*optval
,
620 int (*compat_getsockopt
)(struct sock
*sk
,
622 int optname
, char __user
*optval
,
625 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
626 struct msghdr
*msg
, size_t len
);
627 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
629 size_t len
, int noblock
, int flags
,
631 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
632 int offset
, size_t size
, int flags
);
633 int (*bind
)(struct sock
*sk
,
634 struct sockaddr
*uaddr
, int addr_len
);
636 int (*backlog_rcv
) (struct sock
*sk
,
637 struct sk_buff
*skb
);
639 /* Keeping track of sk's, looking them up, and port selection methods. */
640 void (*hash
)(struct sock
*sk
);
641 void (*unhash
)(struct sock
*sk
);
642 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
644 /* Keeping track of sockets in use */
645 #ifdef CONFIG_PROC_FS
646 unsigned int inuse_idx
;
649 /* Memory pressure */
650 void (*enter_memory_pressure
)(struct sock
*sk
);
651 atomic_t
*memory_allocated
; /* Current allocated memory. */
652 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
654 * Pressure flag: try to collapse.
655 * Technical note: it is used by multiple contexts non atomically.
656 * All the __sk_mem_schedule() is of this nature: accounting
657 * is strict, actions are advisory and have some latency.
659 int *memory_pressure
;
665 struct kmem_cache
*slab
;
666 unsigned int obj_size
;
669 struct percpu_counter
*orphan_count
;
671 struct request_sock_ops
*rsk_prot
;
672 struct timewait_sock_ops
*twsk_prot
;
675 struct inet_hashinfo
*hashinfo
;
676 struct udp_table
*udp_table
;
677 struct raw_hashinfo
*raw_hash
;
680 struct module
*owner
;
684 struct list_head node
;
685 #ifdef SOCK_REFCNT_DEBUG
690 extern int proto_register(struct proto
*prot
, int alloc_slab
);
691 extern void proto_unregister(struct proto
*prot
);
693 #ifdef SOCK_REFCNT_DEBUG
694 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
696 atomic_inc(&sk
->sk_prot
->socks
);
699 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
701 atomic_dec(&sk
->sk_prot
->socks
);
702 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
703 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
706 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
708 if (atomic_read(&sk
->sk_refcnt
) != 1)
709 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
710 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
712 #else /* SOCK_REFCNT_DEBUG */
713 #define sk_refcnt_debug_inc(sk) do { } while (0)
714 #define sk_refcnt_debug_dec(sk) do { } while (0)
715 #define sk_refcnt_debug_release(sk) do { } while (0)
716 #endif /* SOCK_REFCNT_DEBUG */
719 #ifdef CONFIG_PROC_FS
720 /* Called with local bh disabled */
721 extern void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
722 extern int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
724 static void inline sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
731 /* With per-bucket locks this operation is not-atomic, so that
732 * this version is not worse.
734 static inline void __sk_prot_rehash(struct sock
*sk
)
736 sk
->sk_prot
->unhash(sk
);
737 sk
->sk_prot
->hash(sk
);
740 /* About 10 seconds */
741 #define SOCK_DESTROY_TIME (10*HZ)
743 /* Sockets 0-1023 can't be bound to unless you are superuser */
744 #define PROT_SOCK 1024
746 #define SHUTDOWN_MASK 3
747 #define RCV_SHUTDOWN 1
748 #define SEND_SHUTDOWN 2
750 #define SOCK_SNDBUF_LOCK 1
751 #define SOCK_RCVBUF_LOCK 2
752 #define SOCK_BINDADDR_LOCK 4
753 #define SOCK_BINDPORT_LOCK 8
755 /* sock_iocb: used to kick off async processing of socket ios */
757 struct list_head list
;
763 struct scm_cookie
*scm
;
764 struct msghdr
*msg
, async_msg
;
768 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
770 return (struct sock_iocb
*)iocb
->private;
773 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
778 struct socket_alloc
{
779 struct socket socket
;
780 struct inode vfs_inode
;
783 static inline struct socket
*SOCKET_I(struct inode
*inode
)
785 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
788 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
790 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
794 * Functions for memory accounting
796 extern int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
797 extern void __sk_mem_reclaim(struct sock
*sk
);
799 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
800 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
801 #define SK_MEM_SEND 0
802 #define SK_MEM_RECV 1
804 static inline int sk_mem_pages(int amt
)
806 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
809 static inline int sk_has_account(struct sock
*sk
)
811 /* return true if protocol supports memory accounting */
812 return !!sk
->sk_prot
->memory_allocated
;
815 static inline int sk_wmem_schedule(struct sock
*sk
, int size
)
817 if (!sk_has_account(sk
))
819 return size
<= sk
->sk_forward_alloc
||
820 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
823 static inline int sk_rmem_schedule(struct sock
*sk
, int size
)
825 if (!sk_has_account(sk
))
827 return size
<= sk
->sk_forward_alloc
||
828 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
);
831 static inline void sk_mem_reclaim(struct sock
*sk
)
833 if (!sk_has_account(sk
))
835 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
836 __sk_mem_reclaim(sk
);
839 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
841 if (!sk_has_account(sk
))
843 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
844 __sk_mem_reclaim(sk
);
847 static inline void sk_mem_charge(struct sock
*sk
, int size
)
849 if (!sk_has_account(sk
))
851 sk
->sk_forward_alloc
-= size
;
854 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
856 if (!sk_has_account(sk
))
858 sk
->sk_forward_alloc
+= size
;
861 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
863 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
864 sk
->sk_wmem_queued
-= skb
->truesize
;
865 sk_mem_uncharge(sk
, skb
->truesize
);
869 /* Used by processes to "lock" a socket state, so that
870 * interrupts and bottom half handlers won't change it
871 * from under us. It essentially blocks any incoming
872 * packets, so that we won't get any new data or any
873 * packets that change the state of the socket.
875 * While locked, BH processing will add new packets to
876 * the backlog queue. This queue is processed by the
877 * owner of the socket lock right before it is released.
879 * Since ~2.3.5 it is also exclusive sleep lock serializing
880 * accesses from user process context.
882 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
885 * Macro so as to not evaluate some arguments when
886 * lockdep is not enabled.
888 * Mark both the sk_lock and the sk_lock.slock as a
889 * per-address-family lock class.
891 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
893 sk->sk_lock.owned = 0; \
894 init_waitqueue_head(&sk->sk_lock.wq); \
895 spin_lock_init(&(sk)->sk_lock.slock); \
896 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
897 sizeof((sk)->sk_lock)); \
898 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
900 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
903 extern void lock_sock_nested(struct sock
*sk
, int subclass
);
905 static inline void lock_sock(struct sock
*sk
)
907 lock_sock_nested(sk
, 0);
910 extern void release_sock(struct sock
*sk
);
912 /* BH context may only use the following locking interface. */
913 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
914 #define bh_lock_sock_nested(__sk) \
915 spin_lock_nested(&((__sk)->sk_lock.slock), \
916 SINGLE_DEPTH_NESTING)
917 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
919 extern struct sock
*sk_alloc(struct net
*net
, int family
,
922 extern void sk_free(struct sock
*sk
);
923 extern void sk_release_kernel(struct sock
*sk
);
924 extern struct sock
*sk_clone(const struct sock
*sk
,
925 const gfp_t priority
);
927 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
928 unsigned long size
, int force
,
930 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
931 unsigned long size
, int force
,
933 extern void sock_wfree(struct sk_buff
*skb
);
934 extern void sock_rfree(struct sk_buff
*skb
);
936 extern int sock_setsockopt(struct socket
*sock
, int level
,
937 int op
, char __user
*optval
,
940 extern int sock_getsockopt(struct socket
*sock
, int level
,
941 int op
, char __user
*optval
,
943 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
947 extern void *sock_kmalloc(struct sock
*sk
, int size
,
949 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
950 extern void sk_send_sigurg(struct sock
*sk
);
953 * Functions to fill in entries in struct proto_ops when a protocol
954 * does not implement a particular function.
956 extern int sock_no_bind(struct socket
*,
957 struct sockaddr
*, int);
958 extern int sock_no_connect(struct socket
*,
959 struct sockaddr
*, int, int);
960 extern int sock_no_socketpair(struct socket
*,
962 extern int sock_no_accept(struct socket
*,
963 struct socket
*, int);
964 extern int sock_no_getname(struct socket
*,
965 struct sockaddr
*, int *, int);
966 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
967 struct poll_table_struct
*);
968 extern int sock_no_ioctl(struct socket
*, unsigned int,
970 extern int sock_no_listen(struct socket
*, int);
971 extern int sock_no_shutdown(struct socket
*, int);
972 extern int sock_no_getsockopt(struct socket
*, int , int,
973 char __user
*, int __user
*);
974 extern int sock_no_setsockopt(struct socket
*, int, int,
976 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
977 struct msghdr
*, size_t);
978 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
979 struct msghdr
*, size_t, int);
980 extern int sock_no_mmap(struct file
*file
,
982 struct vm_area_struct
*vma
);
983 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
985 int offset
, size_t size
,
989 * Functions to fill in entries in struct proto_ops when a protocol
990 * uses the inet style.
992 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
993 char __user
*optval
, int __user
*optlen
);
994 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
995 struct msghdr
*msg
, size_t size
, int flags
);
996 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
997 char __user
*optval
, int optlen
);
998 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
999 int optname
, char __user
*optval
, int __user
*optlen
);
1000 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1001 int optname
, char __user
*optval
, int optlen
);
1003 extern void sk_common_release(struct sock
*sk
);
1006 * Default socket callbacks and setup code
1009 /* Initialise core socket variables */
1010 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1013 * sk_filter_release: Release a socket filter
1014 * @fp: filter to remove
1016 * Remove a filter from a socket and release its resources.
1019 static inline void sk_filter_release(struct sk_filter
*fp
)
1021 if (atomic_dec_and_test(&fp
->refcnt
))
1025 static inline void sk_filter_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
1027 unsigned int size
= sk_filter_len(fp
);
1029 atomic_sub(size
, &sk
->sk_omem_alloc
);
1030 sk_filter_release(fp
);
1033 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
1035 atomic_inc(&fp
->refcnt
);
1036 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
1040 * Socket reference counting postulates.
1042 * * Each user of socket SHOULD hold a reference count.
1043 * * Each access point to socket (an hash table bucket, reference from a list,
1044 * running timer, skb in flight MUST hold a reference count.
1045 * * When reference count hits 0, it means it will never increase back.
1046 * * When reference count hits 0, it means that no references from
1047 * outside exist to this socket and current process on current CPU
1048 * is last user and may/should destroy this socket.
1049 * * sk_free is called from any context: process, BH, IRQ. When
1050 * it is called, socket has no references from outside -> sk_free
1051 * may release descendant resources allocated by the socket, but
1052 * to the time when it is called, socket is NOT referenced by any
1053 * hash tables, lists etc.
1054 * * Packets, delivered from outside (from network or from another process)
1055 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1056 * when they sit in queue. Otherwise, packets will leak to hole, when
1057 * socket is looked up by one cpu and unhasing is made by another CPU.
1058 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1059 * (leak to backlog). Packet socket does all the processing inside
1060 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1061 * use separate SMP lock, so that they are prone too.
1064 /* Ungrab socket and destroy it, if it was the last reference. */
1065 static inline void sock_put(struct sock
*sk
)
1067 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1071 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1074 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1076 sk
->sk_socket
= sock
;
1079 /* Detach socket from process context.
1080 * Announce socket dead, detach it from wait queue and inode.
1081 * Note that parent inode held reference count on this struct sock,
1082 * we do not release it in this function, because protocol
1083 * probably wants some additional cleanups or even continuing
1084 * to work with this socket (TCP).
1086 static inline void sock_orphan(struct sock
*sk
)
1088 write_lock_bh(&sk
->sk_callback_lock
);
1089 sock_set_flag(sk
, SOCK_DEAD
);
1090 sk_set_socket(sk
, NULL
);
1091 sk
->sk_sleep
= NULL
;
1092 write_unlock_bh(&sk
->sk_callback_lock
);
1095 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1097 write_lock_bh(&sk
->sk_callback_lock
);
1098 sk
->sk_sleep
= &parent
->wait
;
1100 sk_set_socket(sk
, parent
);
1101 security_sock_graft(sk
, parent
);
1102 write_unlock_bh(&sk
->sk_callback_lock
);
1105 extern int sock_i_uid(struct sock
*sk
);
1106 extern unsigned long sock_i_ino(struct sock
*sk
);
1108 static inline struct dst_entry
*
1109 __sk_dst_get(struct sock
*sk
)
1111 return sk
->sk_dst_cache
;
1114 static inline struct dst_entry
*
1115 sk_dst_get(struct sock
*sk
)
1117 struct dst_entry
*dst
;
1119 read_lock(&sk
->sk_dst_lock
);
1120 dst
= sk
->sk_dst_cache
;
1123 read_unlock(&sk
->sk_dst_lock
);
1128 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1130 struct dst_entry
*old_dst
;
1132 old_dst
= sk
->sk_dst_cache
;
1133 sk
->sk_dst_cache
= dst
;
1134 dst_release(old_dst
);
1138 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1140 write_lock(&sk
->sk_dst_lock
);
1141 __sk_dst_set(sk
, dst
);
1142 write_unlock(&sk
->sk_dst_lock
);
1146 __sk_dst_reset(struct sock
*sk
)
1148 struct dst_entry
*old_dst
;
1150 old_dst
= sk
->sk_dst_cache
;
1151 sk
->sk_dst_cache
= NULL
;
1152 dst_release(old_dst
);
1156 sk_dst_reset(struct sock
*sk
)
1158 write_lock(&sk
->sk_dst_lock
);
1160 write_unlock(&sk
->sk_dst_lock
);
1163 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1165 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1167 static inline int sk_can_gso(const struct sock
*sk
)
1169 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1172 extern void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1174 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1175 struct sk_buff
*skb
, struct page
*page
,
1178 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1180 __wsum csum
= csum_and_copy_from_user(from
,
1181 page_address(page
) + off
,
1185 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1186 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1190 skb
->data_len
+= copy
;
1191 skb
->truesize
+= copy
;
1192 sk
->sk_wmem_queued
+= copy
;
1193 sk_mem_charge(sk
, copy
);
1198 * Queue a received datagram if it will fit. Stream and sequenced
1199 * protocols can't normally use this as they need to fit buffers in
1200 * and play with them.
1202 * Inlined as it's very short and called for pretty much every
1203 * packet ever received.
1206 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1210 skb
->destructor
= sock_wfree
;
1211 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1214 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1217 skb
->destructor
= sock_rfree
;
1218 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1219 sk_mem_charge(sk
, skb
->truesize
);
1222 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1223 unsigned long expires
);
1225 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1227 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1229 static inline int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
)
1231 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1232 number of warnings when compiling with -W --ANK
1234 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
1235 (unsigned)sk
->sk_rcvbuf
)
1237 skb_set_owner_r(skb
, sk
);
1238 skb_queue_tail(&sk
->sk_error_queue
, skb
);
1239 if (!sock_flag(sk
, SOCK_DEAD
))
1240 sk
->sk_data_ready(sk
, skb
->len
);
1245 * Recover an error report and clear atomically
1248 static inline int sock_error(struct sock
*sk
)
1251 if (likely(!sk
->sk_err
))
1253 err
= xchg(&sk
->sk_err
, 0);
1257 static inline unsigned long sock_wspace(struct sock
*sk
)
1261 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1262 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1269 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1271 if (sk
->sk_socket
&& sk
->sk_socket
->fasync_list
)
1272 sock_wake_async(sk
->sk_socket
, how
, band
);
1275 #define SOCK_MIN_SNDBUF 2048
1276 #define SOCK_MIN_RCVBUF 256
1278 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1280 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1281 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
1282 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1286 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
1288 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1290 struct page
*page
= NULL
;
1292 page
= alloc_pages(sk
->sk_allocation
, 0);
1294 sk
->sk_prot
->enter_memory_pressure(sk
);
1295 sk_stream_moderate_sndbuf(sk
);
1301 * Default write policy as shown to user space via poll/select/SIGIO
1303 static inline int sock_writeable(const struct sock
*sk
)
1305 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
1308 static inline gfp_t
gfp_any(void)
1310 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
1313 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1315 return noblock
? 0 : sk
->sk_rcvtimeo
;
1318 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1320 return noblock
? 0 : sk
->sk_sndtimeo
;
1323 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1325 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1328 /* Alas, with timeout socket operations are not restartable.
1329 * Compare this to poll().
1331 static inline int sock_intr_errno(long timeo
)
1333 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
1336 extern void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
1337 struct sk_buff
*skb
);
1339 static __inline__
void
1340 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
1342 ktime_t kt
= skb
->tstamp
;
1344 if (sock_flag(sk
, SOCK_RCVTSTAMP
))
1345 __sock_recv_timestamp(msg
, sk
, skb
);
1351 * sk_eat_skb - Release a skb if it is no longer needed
1352 * @sk: socket to eat this skb from
1353 * @skb: socket buffer to eat
1354 * @copied_early: flag indicating whether DMA operations copied this data early
1356 * This routine must be called with interrupts disabled or with the socket
1357 * locked so that the sk_buff queue operation is ok.
1359 #ifdef CONFIG_NET_DMA
1360 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1362 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1366 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
1369 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1371 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1377 struct net
*sock_net(const struct sock
*sk
)
1379 #ifdef CONFIG_NET_NS
1387 void sock_net_set(struct sock
*sk
, struct net
*net
)
1389 #ifdef CONFIG_NET_NS
1395 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1396 * They should not hold a referrence to a namespace in order to allow
1398 * Sockets after sk_change_net should be released using sk_release_kernel
1400 static inline void sk_change_net(struct sock
*sk
, struct net
*net
)
1402 put_net(sock_net(sk
));
1403 sock_net_set(sk
, hold_net(net
));
1406 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
1408 if (unlikely(skb
->sk
)) {
1409 struct sock
*sk
= skb
->sk
;
1411 skb
->destructor
= NULL
;
1418 extern void sock_enable_timestamp(struct sock
*sk
);
1419 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
1420 extern int sock_get_timestampns(struct sock
*, struct timespec __user
*);
1423 * Enable debug/info messages
1425 extern int net_msg_warn
;
1426 #define NETDEBUG(fmt, args...) \
1427 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1429 #define LIMIT_NETDEBUG(fmt, args...) \
1430 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1432 extern __u32 sysctl_wmem_max
;
1433 extern __u32 sysctl_rmem_max
;
1435 extern void sk_init(void);
1437 extern int sysctl_optmem_max
;
1439 extern __u32 sysctl_wmem_default
;
1440 extern __u32 sysctl_rmem_default
;
1442 #endif /* _SOCK_H */