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/timer.h>
46 #include <linux/cache.h>
47 #include <linux/module.h>
48 #include <linux/lockdep.h>
49 #include <linux/netdevice.h>
50 #include <linux/pcounter.h>
51 #include <linux/skbuff.h> /* struct sk_buff */
53 #include <linux/security.h>
55 #include <linux/filter.h>
57 #include <asm/atomic.h>
59 #include <net/checksum.h>
62 * This structure really needs to be cleaned up.
63 * Most of it is for TCP, and not used by any of
64 * the other protocols.
67 /* Define this to get the SOCK_DBG debugging facility. */
68 #define SOCK_DEBUGGING
70 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
71 printk(KERN_DEBUG msg); } while (0)
73 #define SOCK_DEBUG(sk, msg...) do { } while (0)
76 /* This is the per-socket lock. The spinlock provides a synchronization
77 * between user contexts and software interrupt processing, whereas the
78 * mini-semaphore synchronizes multiple users amongst themselves.
85 * We express the mutex-alike socket_lock semantics
86 * to the lock validator by explicitly managing
87 * the slock as a lock variant (in addition to
90 #ifdef CONFIG_DEBUG_LOCK_ALLOC
91 struct lockdep_map dep_map
;
100 * struct sock_common - minimal network layer representation of sockets
101 * @skc_family: network address family
102 * @skc_state: Connection state
103 * @skc_reuse: %SO_REUSEADDR setting
104 * @skc_bound_dev_if: bound device index if != 0
105 * @skc_node: main hash linkage for various protocol lookup tables
106 * @skc_bind_node: bind hash linkage for various protocol lookup tables
107 * @skc_refcnt: reference count
108 * @skc_hash: hash value used with various protocol lookup tables
109 * @skc_prot: protocol handlers inside a network family
110 * @skc_net: reference to the network namespace of this socket
112 * This is the minimal network layer representation of sockets, the header
113 * for struct sock and struct inet_timewait_sock.
116 unsigned short skc_family
;
117 volatile unsigned char skc_state
;
118 unsigned char skc_reuse
;
119 int skc_bound_dev_if
;
120 struct hlist_node skc_node
;
121 struct hlist_node skc_bind_node
;
123 unsigned int skc_hash
;
124 struct proto
*skc_prot
;
129 * struct sock - network layer representation of sockets
130 * @__sk_common: shared layout with inet_timewait_sock
131 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
132 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
133 * @sk_lock: synchronizer
134 * @sk_rcvbuf: size of receive buffer in bytes
135 * @sk_sleep: sock wait queue
136 * @sk_dst_cache: destination cache
137 * @sk_dst_lock: destination cache lock
138 * @sk_policy: flow policy
139 * @sk_rmem_alloc: receive queue bytes committed
140 * @sk_receive_queue: incoming packets
141 * @sk_wmem_alloc: transmit queue bytes committed
142 * @sk_write_queue: Packet sending queue
143 * @sk_async_wait_queue: DMA copied packets
144 * @sk_omem_alloc: "o" is "option" or "other"
145 * @sk_wmem_queued: persistent queue size
146 * @sk_forward_alloc: space allocated forward
147 * @sk_allocation: allocation mode
148 * @sk_sndbuf: size of send buffer in bytes
149 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
150 * %SO_OOBINLINE settings
151 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
152 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
153 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
154 * @sk_lingertime: %SO_LINGER l_linger setting
155 * @sk_backlog: always used with the per-socket spinlock held
156 * @sk_callback_lock: used with the callbacks in the end of this struct
157 * @sk_error_queue: rarely used
158 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
159 * IPV6_ADDRFORM for instance)
160 * @sk_err: last error
161 * @sk_err_soft: errors that don't cause failure but are the cause of a
162 * persistent failure not just 'timed out'
163 * @sk_drops: raw drops counter
164 * @sk_ack_backlog: current listen backlog
165 * @sk_max_ack_backlog: listen backlog set in listen()
166 * @sk_priority: %SO_PRIORITY setting
167 * @sk_type: socket type (%SOCK_STREAM, etc)
168 * @sk_protocol: which protocol this socket belongs in this network family
169 * @sk_peercred: %SO_PEERCRED setting
170 * @sk_rcvlowat: %SO_RCVLOWAT setting
171 * @sk_rcvtimeo: %SO_RCVTIMEO setting
172 * @sk_sndtimeo: %SO_SNDTIMEO setting
173 * @sk_filter: socket filtering instructions
174 * @sk_protinfo: private area, net family specific, when not using slab
175 * @sk_timer: sock cleanup timer
176 * @sk_stamp: time stamp of last packet received
177 * @sk_socket: Identd and reporting IO signals
178 * @sk_user_data: RPC layer private data
179 * @sk_sndmsg_page: cached page for sendmsg
180 * @sk_sndmsg_off: cached offset for sendmsg
181 * @sk_send_head: front of stuff to transmit
182 * @sk_security: used by security modules
183 * @sk_write_pending: a write to stream socket waits to start
184 * @sk_state_change: callback to indicate change in the state of the sock
185 * @sk_data_ready: callback to indicate there is data to be processed
186 * @sk_write_space: callback to indicate there is bf sending space available
187 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
188 * @sk_backlog_rcv: callback to process the backlog
189 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
193 * Now struct inet_timewait_sock also uses sock_common, so please just
194 * don't add nothing before this first member (__sk_common) --acme
196 struct sock_common __sk_common
;
197 #define sk_family __sk_common.skc_family
198 #define sk_state __sk_common.skc_state
199 #define sk_reuse __sk_common.skc_reuse
200 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
201 #define sk_node __sk_common.skc_node
202 #define sk_bind_node __sk_common.skc_bind_node
203 #define sk_refcnt __sk_common.skc_refcnt
204 #define sk_hash __sk_common.skc_hash
205 #define sk_prot __sk_common.skc_prot
206 #define sk_net __sk_common.skc_net
207 unsigned char sk_shutdown
: 2,
210 unsigned char sk_protocol
;
211 unsigned short sk_type
;
213 socket_lock_t sk_lock
;
215 * The backlog queue is special, it is always used with
216 * the per-socket spinlock held and requires low latency
217 * access. Therefore we special case it's implementation.
220 struct sk_buff
*head
;
221 struct sk_buff
*tail
;
223 wait_queue_head_t
*sk_sleep
;
224 struct dst_entry
*sk_dst_cache
;
225 struct xfrm_policy
*sk_policy
[2];
226 rwlock_t sk_dst_lock
;
227 atomic_t sk_rmem_alloc
;
228 atomic_t sk_wmem_alloc
;
229 atomic_t sk_omem_alloc
;
231 struct sk_buff_head sk_receive_queue
;
232 struct sk_buff_head sk_write_queue
;
233 struct sk_buff_head sk_async_wait_queue
;
235 int sk_forward_alloc
;
240 unsigned long sk_flags
;
241 unsigned long sk_lingertime
;
242 struct sk_buff_head sk_error_queue
;
243 struct proto
*sk_prot_creator
;
244 rwlock_t sk_callback_lock
;
248 unsigned short sk_ack_backlog
;
249 unsigned short sk_max_ack_backlog
;
251 struct ucred sk_peercred
;
254 struct sk_filter
*sk_filter
;
256 struct timer_list sk_timer
;
258 struct socket
*sk_socket
;
260 struct page
*sk_sndmsg_page
;
261 struct sk_buff
*sk_send_head
;
263 int sk_write_pending
;
265 void (*sk_state_change
)(struct sock
*sk
);
266 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
267 void (*sk_write_space
)(struct sock
*sk
);
268 void (*sk_error_report
)(struct sock
*sk
);
269 int (*sk_backlog_rcv
)(struct sock
*sk
,
270 struct sk_buff
*skb
);
271 void (*sk_destruct
)(struct sock
*sk
);
275 * Hashed lists helper routines
277 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
279 return hlist_entry(head
->first
, struct sock
, sk_node
);
282 static inline struct sock
*sk_head(const struct hlist_head
*head
)
284 return hlist_empty(head
) ? NULL
: __sk_head(head
);
287 static inline struct sock
*sk_next(const struct sock
*sk
)
289 return sk
->sk_node
.next
?
290 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
293 static inline int sk_unhashed(const struct sock
*sk
)
295 return hlist_unhashed(&sk
->sk_node
);
298 static inline int sk_hashed(const struct sock
*sk
)
300 return !sk_unhashed(sk
);
303 static __inline__
void sk_node_init(struct hlist_node
*node
)
308 static __inline__
void __sk_del_node(struct sock
*sk
)
310 __hlist_del(&sk
->sk_node
);
313 static __inline__
int __sk_del_node_init(struct sock
*sk
)
317 sk_node_init(&sk
->sk_node
);
323 /* Grab socket reference count. This operation is valid only
324 when sk is ALREADY grabbed f.e. it is found in hash table
325 or a list and the lookup is made under lock preventing hash table
329 static inline void sock_hold(struct sock
*sk
)
331 atomic_inc(&sk
->sk_refcnt
);
334 /* Ungrab socket in the context, which assumes that socket refcnt
335 cannot hit zero, f.e. it is true in context of any socketcall.
337 static inline void __sock_put(struct sock
*sk
)
339 atomic_dec(&sk
->sk_refcnt
);
342 static __inline__
int sk_del_node_init(struct sock
*sk
)
344 int rc
= __sk_del_node_init(sk
);
347 /* paranoid for a while -acme */
348 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
354 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
356 hlist_add_head(&sk
->sk_node
, list
);
359 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
362 __sk_add_node(sk
, list
);
365 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
367 __hlist_del(&sk
->sk_bind_node
);
370 static __inline__
void sk_add_bind_node(struct sock
*sk
,
371 struct hlist_head
*list
)
373 hlist_add_head(&sk
->sk_bind_node
, list
);
376 #define sk_for_each(__sk, node, list) \
377 hlist_for_each_entry(__sk, node, list, sk_node)
378 #define sk_for_each_from(__sk, node) \
379 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
380 hlist_for_each_entry_from(__sk, node, sk_node)
381 #define sk_for_each_continue(__sk, node) \
382 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
383 hlist_for_each_entry_continue(__sk, node, sk_node)
384 #define sk_for_each_safe(__sk, node, tmp, list) \
385 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
386 #define sk_for_each_bound(__sk, node, list) \
387 hlist_for_each_entry(__sk, node, list, sk_bind_node)
400 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
401 SOCK_DBG
, /* %SO_DEBUG setting */
402 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
403 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
404 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
405 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
408 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
410 nsk
->sk_flags
= osk
->sk_flags
;
413 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
415 __set_bit(flag
, &sk
->sk_flags
);
418 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
420 __clear_bit(flag
, &sk
->sk_flags
);
423 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
425 return test_bit(flag
, &sk
->sk_flags
);
428 static inline void sk_acceptq_removed(struct sock
*sk
)
430 sk
->sk_ack_backlog
--;
433 static inline void sk_acceptq_added(struct sock
*sk
)
435 sk
->sk_ack_backlog
++;
438 static inline int sk_acceptq_is_full(struct sock
*sk
)
440 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
444 * Compute minimal free write space needed to queue new packets.
446 static inline int sk_stream_min_wspace(struct sock
*sk
)
448 return sk
->sk_wmem_queued
>> 1;
451 static inline int sk_stream_wspace(struct sock
*sk
)
453 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
456 extern void sk_stream_write_space(struct sock
*sk
);
458 static inline int sk_stream_memory_free(struct sock
*sk
)
460 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
463 extern void sk_stream_rfree(struct sk_buff
*skb
);
465 static inline void sk_stream_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
468 skb
->destructor
= sk_stream_rfree
;
469 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
470 sk
->sk_forward_alloc
-= skb
->truesize
;
473 static inline void sk_stream_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
475 skb_truesize_check(skb
);
476 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
477 sk
->sk_wmem_queued
-= skb
->truesize
;
478 sk
->sk_forward_alloc
+= skb
->truesize
;
482 /* The per-socket spinlock must be held here. */
483 static inline void sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
485 if (!sk
->sk_backlog
.tail
) {
486 sk
->sk_backlog
.head
= sk
->sk_backlog
.tail
= skb
;
488 sk
->sk_backlog
.tail
->next
= skb
;
489 sk
->sk_backlog
.tail
= skb
;
494 #define sk_wait_event(__sk, __timeo, __condition) \
496 release_sock(__sk); \
497 __rc = __condition; \
499 *(__timeo) = schedule_timeout(*(__timeo)); \
502 __rc = __condition; \
506 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
507 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
508 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
509 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
510 extern void sk_stream_kill_queues(struct sock
*sk
);
512 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
514 struct request_sock_ops
;
515 struct timewait_sock_ops
;
517 /* Networking protocol blocks we attach to sockets.
518 * socket layer -> transport layer interface
519 * transport -> network interface is defined by struct inet_proto
522 void (*close
)(struct sock
*sk
,
524 int (*connect
)(struct sock
*sk
,
525 struct sockaddr
*uaddr
,
527 int (*disconnect
)(struct sock
*sk
, int flags
);
529 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
531 int (*ioctl
)(struct sock
*sk
, int cmd
,
533 int (*init
)(struct sock
*sk
);
534 int (*destroy
)(struct sock
*sk
);
535 void (*shutdown
)(struct sock
*sk
, int how
);
536 int (*setsockopt
)(struct sock
*sk
, int level
,
537 int optname
, char __user
*optval
,
539 int (*getsockopt
)(struct sock
*sk
, int level
,
540 int optname
, char __user
*optval
,
542 int (*compat_setsockopt
)(struct sock
*sk
,
544 int optname
, char __user
*optval
,
546 int (*compat_getsockopt
)(struct sock
*sk
,
548 int optname
, char __user
*optval
,
550 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
551 struct msghdr
*msg
, size_t len
);
552 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
554 size_t len
, int noblock
, int flags
,
556 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
557 int offset
, size_t size
, int flags
);
558 int (*bind
)(struct sock
*sk
,
559 struct sockaddr
*uaddr
, int addr_len
);
561 int (*backlog_rcv
) (struct sock
*sk
,
562 struct sk_buff
*skb
);
564 /* Keeping track of sk's, looking them up, and port selection methods. */
565 void (*hash
)(struct sock
*sk
);
566 void (*unhash
)(struct sock
*sk
);
567 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
569 /* Keeping track of sockets in use */
570 struct pcounter inuse
;
572 /* Memory pressure */
573 void (*enter_memory_pressure
)(void);
574 atomic_t
*memory_allocated
; /* Current allocated memory. */
575 atomic_t
*sockets_allocated
; /* Current number of sockets. */
577 * Pressure flag: try to collapse.
578 * Technical note: it is used by multiple contexts non atomically.
579 * All the sk_stream_mem_schedule() is of this nature: accounting
580 * is strict, actions are advisory and have some latency.
582 int *memory_pressure
;
588 struct kmem_cache
*slab
;
589 unsigned int obj_size
;
591 atomic_t
*orphan_count
;
593 struct request_sock_ops
*rsk_prot
;
594 struct timewait_sock_ops
*twsk_prot
;
596 struct module
*owner
;
600 struct list_head node
;
601 #ifdef SOCK_REFCNT_DEBUG
606 #define DEFINE_PROTO_INUSE(NAME) DEFINE_PCOUNTER(NAME)
607 #define REF_PROTO_INUSE(NAME) PCOUNTER_MEMBER_INITIALIZER(NAME, .inuse)
609 extern int proto_register(struct proto
*prot
, int alloc_slab
);
610 extern void proto_unregister(struct proto
*prot
);
612 #ifdef SOCK_REFCNT_DEBUG
613 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
615 atomic_inc(&sk
->sk_prot
->socks
);
618 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
620 atomic_dec(&sk
->sk_prot
->socks
);
621 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
622 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
625 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
627 if (atomic_read(&sk
->sk_refcnt
) != 1)
628 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
629 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
631 #else /* SOCK_REFCNT_DEBUG */
632 #define sk_refcnt_debug_inc(sk) do { } while (0)
633 #define sk_refcnt_debug_dec(sk) do { } while (0)
634 #define sk_refcnt_debug_release(sk) do { } while (0)
635 #endif /* SOCK_REFCNT_DEBUG */
637 /* Called with local bh disabled */
638 static __inline__
void sock_prot_inc_use(struct proto
*prot
)
640 pcounter_add(&prot
->inuse
, 1);
643 static __inline__
void sock_prot_dec_use(struct proto
*prot
)
645 pcounter_add(&prot
->inuse
, -1);
648 static __inline__
int sock_prot_inuse(struct proto
*proto
)
650 return pcounter_getval(&proto
->inuse
);
653 /* With per-bucket locks this operation is not-atomic, so that
654 * this version is not worse.
656 static inline void __sk_prot_rehash(struct sock
*sk
)
658 sk
->sk_prot
->unhash(sk
);
659 sk
->sk_prot
->hash(sk
);
662 /* About 10 seconds */
663 #define SOCK_DESTROY_TIME (10*HZ)
665 /* Sockets 0-1023 can't be bound to unless you are superuser */
666 #define PROT_SOCK 1024
668 #define SHUTDOWN_MASK 3
669 #define RCV_SHUTDOWN 1
670 #define SEND_SHUTDOWN 2
672 #define SOCK_SNDBUF_LOCK 1
673 #define SOCK_RCVBUF_LOCK 2
674 #define SOCK_BINDADDR_LOCK 4
675 #define SOCK_BINDPORT_LOCK 8
677 /* sock_iocb: used to kick off async processing of socket ios */
679 struct list_head list
;
685 struct scm_cookie
*scm
;
686 struct msghdr
*msg
, async_msg
;
690 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
692 return (struct sock_iocb
*)iocb
->private;
695 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
700 struct socket_alloc
{
701 struct socket socket
;
702 struct inode vfs_inode
;
705 static inline struct socket
*SOCKET_I(struct inode
*inode
)
707 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
710 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
712 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
715 extern void __sk_stream_mem_reclaim(struct sock
*sk
);
716 extern int sk_stream_mem_schedule(struct sock
*sk
, int size
, int kind
);
718 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
719 #define SK_STREAM_MEM_QUANTUM_SHIFT ilog2(SK_STREAM_MEM_QUANTUM)
721 static inline int sk_stream_pages(int amt
)
723 return (amt
+ SK_STREAM_MEM_QUANTUM
- 1) >> SK_STREAM_MEM_QUANTUM_SHIFT
;
726 static inline void sk_stream_mem_reclaim(struct sock
*sk
)
728 if (sk
->sk_forward_alloc
>= SK_STREAM_MEM_QUANTUM
)
729 __sk_stream_mem_reclaim(sk
);
732 static inline int sk_stream_rmem_schedule(struct sock
*sk
, struct sk_buff
*skb
)
734 return (int)skb
->truesize
<= sk
->sk_forward_alloc
||
735 sk_stream_mem_schedule(sk
, skb
->truesize
, 1);
738 static inline int sk_stream_wmem_schedule(struct sock
*sk
, int size
)
740 return size
<= sk
->sk_forward_alloc
||
741 sk_stream_mem_schedule(sk
, size
, 0);
744 /* Used by processes to "lock" a socket state, so that
745 * interrupts and bottom half handlers won't change it
746 * from under us. It essentially blocks any incoming
747 * packets, so that we won't get any new data or any
748 * packets that change the state of the socket.
750 * While locked, BH processing will add new packets to
751 * the backlog queue. This queue is processed by the
752 * owner of the socket lock right before it is released.
754 * Since ~2.3.5 it is also exclusive sleep lock serializing
755 * accesses from user process context.
757 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
760 * Macro so as to not evaluate some arguments when
761 * lockdep is not enabled.
763 * Mark both the sk_lock and the sk_lock.slock as a
764 * per-address-family lock class.
766 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
768 sk->sk_lock.owned = 0; \
769 init_waitqueue_head(&sk->sk_lock.wq); \
770 spin_lock_init(&(sk)->sk_lock.slock); \
771 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
772 sizeof((sk)->sk_lock)); \
773 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
775 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
778 extern void lock_sock_nested(struct sock
*sk
, int subclass
);
780 static inline void lock_sock(struct sock
*sk
)
782 lock_sock_nested(sk
, 0);
785 extern void release_sock(struct sock
*sk
);
787 /* BH context may only use the following locking interface. */
788 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
789 #define bh_lock_sock_nested(__sk) \
790 spin_lock_nested(&((__sk)->sk_lock.slock), \
791 SINGLE_DEPTH_NESTING)
792 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
794 extern struct sock
*sk_alloc(struct net
*net
, int family
,
797 extern void sk_free(struct sock
*sk
);
798 extern struct sock
*sk_clone(const struct sock
*sk
,
799 const gfp_t priority
);
801 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
802 unsigned long size
, int force
,
804 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
805 unsigned long size
, int force
,
807 extern void sock_wfree(struct sk_buff
*skb
);
808 extern void sock_rfree(struct sk_buff
*skb
);
810 extern int sock_setsockopt(struct socket
*sock
, int level
,
811 int op
, char __user
*optval
,
814 extern int sock_getsockopt(struct socket
*sock
, int level
,
815 int op
, char __user
*optval
,
817 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
821 extern void *sock_kmalloc(struct sock
*sk
, int size
,
823 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
824 extern void sk_send_sigurg(struct sock
*sk
);
827 * Functions to fill in entries in struct proto_ops when a protocol
828 * does not implement a particular function.
830 extern int sock_no_bind(struct socket
*,
831 struct sockaddr
*, int);
832 extern int sock_no_connect(struct socket
*,
833 struct sockaddr
*, int, int);
834 extern int sock_no_socketpair(struct socket
*,
836 extern int sock_no_accept(struct socket
*,
837 struct socket
*, int);
838 extern int sock_no_getname(struct socket
*,
839 struct sockaddr
*, int *, int);
840 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
841 struct poll_table_struct
*);
842 extern int sock_no_ioctl(struct socket
*, unsigned int,
844 extern int sock_no_listen(struct socket
*, int);
845 extern int sock_no_shutdown(struct socket
*, int);
846 extern int sock_no_getsockopt(struct socket
*, int , int,
847 char __user
*, int __user
*);
848 extern int sock_no_setsockopt(struct socket
*, int, int,
850 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
851 struct msghdr
*, size_t);
852 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
853 struct msghdr
*, size_t, int);
854 extern int sock_no_mmap(struct file
*file
,
856 struct vm_area_struct
*vma
);
857 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
859 int offset
, size_t size
,
863 * Functions to fill in entries in struct proto_ops when a protocol
864 * uses the inet style.
866 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
867 char __user
*optval
, int __user
*optlen
);
868 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
869 struct msghdr
*msg
, size_t size
, int flags
);
870 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
871 char __user
*optval
, int optlen
);
872 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
873 int optname
, char __user
*optval
, int __user
*optlen
);
874 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
875 int optname
, char __user
*optval
, int optlen
);
877 extern void sk_common_release(struct sock
*sk
);
880 * Default socket callbacks and setup code
883 /* Initialise core socket variables */
884 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
887 * sk_filter - run a packet through a socket filter
888 * @sk: sock associated with &sk_buff
889 * @skb: buffer to filter
890 * @needlock: set to 1 if the sock is not locked by caller.
892 * Run the filter code and then cut skb->data to correct size returned by
893 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
894 * than pkt_len we keep whole skb->data. This is the socket level
895 * wrapper to sk_run_filter. It returns 0 if the packet should
896 * be accepted or -EPERM if the packet should be tossed.
900 static inline int sk_filter(struct sock
*sk
, struct sk_buff
*skb
)
903 struct sk_filter
*filter
;
905 err
= security_sock_rcv_skb(sk
, skb
);
910 filter
= rcu_dereference(sk
->sk_filter
);
912 unsigned int pkt_len
= sk_run_filter(skb
, filter
->insns
,
914 err
= pkt_len
? pskb_trim(skb
, pkt_len
) : -EPERM
;
916 rcu_read_unlock_bh();
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 sk_filter
*fp
)
931 if (atomic_dec_and_test(&fp
->refcnt
))
935 static inline void sk_filter_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
937 unsigned int size
= sk_filter_len(fp
);
939 atomic_sub(size
, &sk
->sk_omem_alloc
);
940 sk_filter_release(fp
);
943 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
945 atomic_inc(&fp
->refcnt
);
946 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
950 * Socket reference counting postulates.
952 * * Each user of socket SHOULD hold a reference count.
953 * * Each access point to socket (an hash table bucket, reference from a list,
954 * running timer, skb in flight MUST hold a reference count.
955 * * When reference count hits 0, it means it will never increase back.
956 * * When reference count hits 0, it means that no references from
957 * outside exist to this socket and current process on current CPU
958 * is last user and may/should destroy this socket.
959 * * sk_free is called from any context: process, BH, IRQ. When
960 * it is called, socket has no references from outside -> sk_free
961 * may release descendant resources allocated by the socket, but
962 * to the time when it is called, socket is NOT referenced by any
963 * hash tables, lists etc.
964 * * Packets, delivered from outside (from network or from another process)
965 * and enqueued on receive/error queues SHOULD NOT grab reference count,
966 * when they sit in queue. Otherwise, packets will leak to hole, when
967 * socket is looked up by one cpu and unhasing is made by another CPU.
968 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
969 * (leak to backlog). Packet socket does all the processing inside
970 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
971 * use separate SMP lock, so that they are prone too.
974 /* Ungrab socket and destroy it, if it was the last reference. */
975 static inline void sock_put(struct sock
*sk
)
977 if (atomic_dec_and_test(&sk
->sk_refcnt
))
981 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
984 /* Detach socket from process context.
985 * Announce socket dead, detach it from wait queue and inode.
986 * Note that parent inode held reference count on this struct sock,
987 * we do not release it in this function, because protocol
988 * probably wants some additional cleanups or even continuing
989 * to work with this socket (TCP).
991 static inline void sock_orphan(struct sock
*sk
)
993 write_lock_bh(&sk
->sk_callback_lock
);
994 sock_set_flag(sk
, SOCK_DEAD
);
995 sk
->sk_socket
= NULL
;
997 write_unlock_bh(&sk
->sk_callback_lock
);
1000 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1002 write_lock_bh(&sk
->sk_callback_lock
);
1003 sk
->sk_sleep
= &parent
->wait
;
1005 sk
->sk_socket
= parent
;
1006 security_sock_graft(sk
, parent
);
1007 write_unlock_bh(&sk
->sk_callback_lock
);
1010 extern int sock_i_uid(struct sock
*sk
);
1011 extern unsigned long sock_i_ino(struct sock
*sk
);
1013 static inline struct dst_entry
*
1014 __sk_dst_get(struct sock
*sk
)
1016 return sk
->sk_dst_cache
;
1019 static inline struct dst_entry
*
1020 sk_dst_get(struct sock
*sk
)
1022 struct dst_entry
*dst
;
1024 read_lock(&sk
->sk_dst_lock
);
1025 dst
= sk
->sk_dst_cache
;
1028 read_unlock(&sk
->sk_dst_lock
);
1033 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1035 struct dst_entry
*old_dst
;
1037 old_dst
= sk
->sk_dst_cache
;
1038 sk
->sk_dst_cache
= dst
;
1039 dst_release(old_dst
);
1043 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1045 write_lock(&sk
->sk_dst_lock
);
1046 __sk_dst_set(sk
, dst
);
1047 write_unlock(&sk
->sk_dst_lock
);
1051 __sk_dst_reset(struct sock
*sk
)
1053 struct dst_entry
*old_dst
;
1055 old_dst
= sk
->sk_dst_cache
;
1056 sk
->sk_dst_cache
= NULL
;
1057 dst_release(old_dst
);
1061 sk_dst_reset(struct sock
*sk
)
1063 write_lock(&sk
->sk_dst_lock
);
1065 write_unlock(&sk
->sk_dst_lock
);
1068 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1070 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1072 static inline int sk_can_gso(const struct sock
*sk
)
1074 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1077 extern void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1079 static inline void sk_charge_skb(struct sock
*sk
, struct sk_buff
*skb
)
1081 sk
->sk_wmem_queued
+= skb
->truesize
;
1082 sk
->sk_forward_alloc
-= skb
->truesize
;
1085 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1086 struct sk_buff
*skb
, struct page
*page
,
1089 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1091 __wsum csum
= csum_and_copy_from_user(from
,
1092 page_address(page
) + off
,
1096 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1097 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1101 skb
->data_len
+= copy
;
1102 skb
->truesize
+= copy
;
1103 sk
->sk_wmem_queued
+= copy
;
1104 sk
->sk_forward_alloc
-= copy
;
1109 * Queue a received datagram if it will fit. Stream and sequenced
1110 * protocols can't normally use this as they need to fit buffers in
1111 * and play with them.
1113 * Inlined as it's very short and called for pretty much every
1114 * packet ever received.
1117 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1121 skb
->destructor
= sock_wfree
;
1122 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1125 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1128 skb
->destructor
= sock_rfree
;
1129 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1132 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1133 unsigned long expires
);
1135 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1137 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1139 static inline int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
)
1141 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1142 number of warnings when compiling with -W --ANK
1144 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
1145 (unsigned)sk
->sk_rcvbuf
)
1147 skb_set_owner_r(skb
, sk
);
1148 skb_queue_tail(&sk
->sk_error_queue
, skb
);
1149 if (!sock_flag(sk
, SOCK_DEAD
))
1150 sk
->sk_data_ready(sk
, skb
->len
);
1155 * Recover an error report and clear atomically
1158 static inline int sock_error(struct sock
*sk
)
1161 if (likely(!sk
->sk_err
))
1163 err
= xchg(&sk
->sk_err
, 0);
1167 static inline unsigned long sock_wspace(struct sock
*sk
)
1171 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1172 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1179 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1181 if (sk
->sk_socket
&& sk
->sk_socket
->fasync_list
)
1182 sock_wake_async(sk
->sk_socket
, how
, band
);
1185 #define SOCK_MIN_SNDBUF 2048
1186 #define SOCK_MIN_RCVBUF 256
1188 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1190 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1191 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
1192 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1196 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
1198 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1200 struct page
*page
= NULL
;
1202 page
= alloc_pages(sk
->sk_allocation
, 0);
1204 sk
->sk_prot
->enter_memory_pressure();
1205 sk_stream_moderate_sndbuf(sk
);
1211 * Default write policy as shown to user space via poll/select/SIGIO
1213 static inline int sock_writeable(const struct sock
*sk
)
1215 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
1218 static inline gfp_t
gfp_any(void)
1220 return in_atomic() ? GFP_ATOMIC
: GFP_KERNEL
;
1223 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1225 return noblock
? 0 : sk
->sk_rcvtimeo
;
1228 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1230 return noblock
? 0 : sk
->sk_sndtimeo
;
1233 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1235 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1238 /* Alas, with timeout socket operations are not restartable.
1239 * Compare this to poll().
1241 static inline int sock_intr_errno(long timeo
)
1243 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
1246 extern void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
1247 struct sk_buff
*skb
);
1249 static __inline__
void
1250 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
1252 ktime_t kt
= skb
->tstamp
;
1254 if (sock_flag(sk
, SOCK_RCVTSTAMP
))
1255 __sock_recv_timestamp(msg
, sk
, skb
);
1261 * sk_eat_skb - Release a skb if it is no longer needed
1262 * @sk: socket to eat this skb from
1263 * @skb: socket buffer to eat
1264 * @copied_early: flag indicating whether DMA operations copied this data early
1266 * This routine must be called with interrupts disabled or with the socket
1267 * locked so that the sk_buff queue operation is ok.
1269 #ifdef CONFIG_NET_DMA
1270 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1272 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1276 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
1279 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1281 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1286 extern void sock_enable_timestamp(struct sock
*sk
);
1287 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
1288 extern int sock_get_timestampns(struct sock
*, struct timespec __user
*);
1291 * Enable debug/info messages
1293 extern int net_msg_warn
;
1294 #define NETDEBUG(fmt, args...) \
1295 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1297 #define LIMIT_NETDEBUG(fmt, args...) \
1298 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1301 * Macros for sleeping on a socket. Use them like this:
1303 * SOCK_SLEEP_PRE(sk)
1306 * SOCK_SLEEP_POST(sk)
1308 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1309 * and when the last use of them in DECnet has gone, I'm intending to
1313 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1314 DECLARE_WAITQUEUE(wait, tsk); \
1315 tsk->state = TASK_INTERRUPTIBLE; \
1316 add_wait_queue((sk)->sk_sleep, &wait); \
1319 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1320 remove_wait_queue((sk)->sk_sleep, &wait); \
1324 extern __u32 sysctl_wmem_max
;
1325 extern __u32 sysctl_rmem_max
;
1327 extern void sk_init(void);
1329 extern int sysctl_optmem_max
;
1331 extern __u32 sysctl_wmem_default
;
1332 extern __u32 sysctl_rmem_default
;
1334 #endif /* _SOCK_H */