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>
54 #include <linux/slab.h>
56 #include <linux/filter.h>
57 #include <linux/rculist_nulls.h>
58 #include <linux/poll.h>
60 #include <linux/atomic.h>
62 #include <net/checksum.h>
65 * This structure really needs to be cleaned up.
66 * Most of it is for TCP, and not used by any of
67 * the other protocols.
70 /* Define this to get the SOCK_DBG debugging facility. */
71 #define SOCK_DEBUGGING
73 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
74 printk(KERN_DEBUG msg); } while (0)
76 /* Validate arguments and do nothing */
77 static inline void __attribute__ ((format (printf
, 2, 3)))
78 SOCK_DEBUG(struct sock
*sk
, const char *msg
, ...)
83 /* This is the per-socket lock. The spinlock provides a synchronization
84 * between user contexts and software interrupt processing, whereas the
85 * mini-semaphore synchronizes multiple users amongst themselves.
92 * We express the mutex-alike socket_lock semantics
93 * to the lock validator by explicitly managing
94 * the slock as a lock variant (in addition to
97 #ifdef CONFIG_DEBUG_LOCK_ALLOC
98 struct lockdep_map dep_map
;
107 * struct sock_common - minimal network layer representation of sockets
108 * @skc_daddr: Foreign IPv4 addr
109 * @skc_rcv_saddr: Bound local IPv4 addr
110 * @skc_hash: hash value used with various protocol lookup tables
111 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
112 * @skc_family: network address family
113 * @skc_state: Connection state
114 * @skc_reuse: %SO_REUSEADDR setting
115 * @skc_bound_dev_if: bound device index if != 0
116 * @skc_bind_node: bind hash linkage for various protocol lookup tables
117 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
118 * @skc_prot: protocol handlers inside a network family
119 * @skc_net: reference to the network namespace of this socket
120 * @skc_node: main hash linkage for various protocol lookup tables
121 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
122 * @skc_tx_queue_mapping: tx queue number for this connection
123 * @skc_refcnt: reference count
125 * This is the minimal network layer representation of sockets, the header
126 * for struct sock and struct inet_timewait_sock.
129 /* skc_daddr and skc_rcv_saddr must be grouped :
130 * cf INET_MATCH() and INET_TW_MATCH()
133 __be32 skc_rcv_saddr
;
136 unsigned int skc_hash
;
137 __u16 skc_u16hashes
[2];
139 unsigned short skc_family
;
140 volatile unsigned char skc_state
;
141 unsigned char skc_reuse
;
142 int skc_bound_dev_if
;
144 struct hlist_node skc_bind_node
;
145 struct hlist_nulls_node skc_portaddr_node
;
147 struct proto
*skc_prot
;
152 * fields between dontcopy_begin/dontcopy_end
153 * are not copied in sock_copy()
156 int skc_dontcopy_begin
[0];
159 struct hlist_node skc_node
;
160 struct hlist_nulls_node skc_nulls_node
;
162 int skc_tx_queue_mapping
;
165 int skc_dontcopy_end
[0];
170 * struct sock - network layer representation of sockets
171 * @__sk_common: shared layout with inet_timewait_sock
172 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
173 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
174 * @sk_lock: synchronizer
175 * @sk_rcvbuf: size of receive buffer in bytes
176 * @sk_wq: sock wait queue and async head
177 * @sk_dst_cache: destination cache
178 * @sk_dst_lock: destination cache lock
179 * @sk_policy: flow policy
180 * @sk_rmem_alloc: receive queue bytes committed
181 * @sk_receive_queue: incoming packets
182 * @sk_wmem_alloc: transmit queue bytes committed
183 * @sk_write_queue: Packet sending queue
184 * @sk_async_wait_queue: DMA copied packets
185 * @sk_omem_alloc: "o" is "option" or "other"
186 * @sk_wmem_queued: persistent queue size
187 * @sk_forward_alloc: space allocated forward
188 * @sk_allocation: allocation mode
189 * @sk_sndbuf: size of send buffer in bytes
190 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
191 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
192 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
193 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
194 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
195 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
196 * @sk_gso_max_size: Maximum GSO segment size to build
197 * @sk_lingertime: %SO_LINGER l_linger setting
198 * @sk_backlog: always used with the per-socket spinlock held
199 * @sk_callback_lock: used with the callbacks in the end of this struct
200 * @sk_error_queue: rarely used
201 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
202 * IPV6_ADDRFORM for instance)
203 * @sk_err: last error
204 * @sk_err_soft: errors that don't cause failure but are the cause of a
205 * persistent failure not just 'timed out'
206 * @sk_drops: raw/udp drops counter
207 * @sk_ack_backlog: current listen backlog
208 * @sk_max_ack_backlog: listen backlog set in listen()
209 * @sk_priority: %SO_PRIORITY setting
210 * @sk_type: socket type (%SOCK_STREAM, etc)
211 * @sk_protocol: which protocol this socket belongs in this network family
212 * @sk_peer_pid: &struct pid for this socket's peer
213 * @sk_peer_cred: %SO_PEERCRED setting
214 * @sk_rcvlowat: %SO_RCVLOWAT setting
215 * @sk_rcvtimeo: %SO_RCVTIMEO setting
216 * @sk_sndtimeo: %SO_SNDTIMEO setting
217 * @sk_rxhash: flow hash received from netif layer
218 * @sk_filter: socket filtering instructions
219 * @sk_protinfo: private area, net family specific, when not using slab
220 * @sk_timer: sock cleanup timer
221 * @sk_stamp: time stamp of last packet received
222 * @sk_socket: Identd and reporting IO signals
223 * @sk_user_data: RPC layer private data
224 * @sk_sndmsg_page: cached page for sendmsg
225 * @sk_sndmsg_off: cached offset for sendmsg
226 * @sk_send_head: front of stuff to transmit
227 * @sk_security: used by security modules
228 * @sk_mark: generic packet mark
229 * @sk_classid: this socket's cgroup classid
230 * @sk_write_pending: a write to stream socket waits to start
231 * @sk_state_change: callback to indicate change in the state of the sock
232 * @sk_data_ready: callback to indicate there is data to be processed
233 * @sk_write_space: callback to indicate there is bf sending space available
234 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
235 * @sk_backlog_rcv: callback to process the backlog
236 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
240 * Now struct inet_timewait_sock also uses sock_common, so please just
241 * don't add nothing before this first member (__sk_common) --acme
243 struct sock_common __sk_common
;
244 #define sk_node __sk_common.skc_node
245 #define sk_nulls_node __sk_common.skc_nulls_node
246 #define sk_refcnt __sk_common.skc_refcnt
247 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
249 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
250 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
251 #define sk_hash __sk_common.skc_hash
252 #define sk_family __sk_common.skc_family
253 #define sk_state __sk_common.skc_state
254 #define sk_reuse __sk_common.skc_reuse
255 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
256 #define sk_bind_node __sk_common.skc_bind_node
257 #define sk_prot __sk_common.skc_prot
258 #define sk_net __sk_common.skc_net
259 socket_lock_t sk_lock
;
260 struct sk_buff_head sk_receive_queue
;
262 * The backlog queue is special, it is always used with
263 * the per-socket spinlock held and requires low latency
264 * access. Therefore we special case it's implementation.
265 * Note : rmem_alloc is in this structure to fill a hole
266 * on 64bit arches, not because its logically part of
272 struct sk_buff
*head
;
273 struct sk_buff
*tail
;
275 #define sk_rmem_alloc sk_backlog.rmem_alloc
276 int sk_forward_alloc
;
283 struct sk_filter __rcu
*sk_filter
;
284 struct socket_wq
*sk_wq
;
286 #ifdef CONFIG_NET_DMA
287 struct sk_buff_head sk_async_wait_queue
;
291 struct xfrm_policy
*sk_policy
[2];
293 unsigned long sk_flags
;
294 struct dst_entry
*sk_dst_cache
;
295 spinlock_t sk_dst_lock
;
296 atomic_t sk_wmem_alloc
;
297 atomic_t sk_omem_alloc
;
299 struct sk_buff_head sk_write_queue
;
300 kmemcheck_bitfield_begin(flags
);
301 unsigned int sk_shutdown
: 2,
306 kmemcheck_bitfield_end(flags
);
312 unsigned int sk_gso_max_size
;
314 unsigned long sk_lingertime
;
315 struct sk_buff_head sk_error_queue
;
316 struct proto
*sk_prot_creator
;
317 rwlock_t sk_callback_lock
;
320 unsigned short sk_ack_backlog
;
321 unsigned short sk_max_ack_backlog
;
323 struct pid
*sk_peer_pid
;
324 const struct cred
*sk_peer_cred
;
328 struct timer_list sk_timer
;
330 struct socket
*sk_socket
;
332 struct page
*sk_sndmsg_page
;
333 struct sk_buff
*sk_send_head
;
335 int sk_write_pending
;
336 #ifdef CONFIG_SECURITY
341 void (*sk_state_change
)(struct sock
*sk
);
342 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
343 void (*sk_write_space
)(struct sock
*sk
);
344 void (*sk_error_report
)(struct sock
*sk
);
345 int (*sk_backlog_rcv
)(struct sock
*sk
,
346 struct sk_buff
*skb
);
347 void (*sk_destruct
)(struct sock
*sk
);
351 * Hashed lists helper routines
353 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
355 return hlist_entry(node
, struct sock
, sk_node
);
358 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
360 return hlist_entry(head
->first
, struct sock
, sk_node
);
363 static inline struct sock
*sk_head(const struct hlist_head
*head
)
365 return hlist_empty(head
) ? NULL
: __sk_head(head
);
368 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
370 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
373 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
375 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
378 static inline struct sock
*sk_next(const struct sock
*sk
)
380 return sk
->sk_node
.next
?
381 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
384 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
386 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
387 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
388 struct sock
, sk_nulls_node
) :
392 static inline int sk_unhashed(const struct sock
*sk
)
394 return hlist_unhashed(&sk
->sk_node
);
397 static inline int sk_hashed(const struct sock
*sk
)
399 return !sk_unhashed(sk
);
402 static __inline__
void sk_node_init(struct hlist_node
*node
)
407 static __inline__
void sk_nulls_node_init(struct hlist_nulls_node
*node
)
412 static __inline__
void __sk_del_node(struct sock
*sk
)
414 __hlist_del(&sk
->sk_node
);
417 /* NB: equivalent to hlist_del_init_rcu */
418 static __inline__
int __sk_del_node_init(struct sock
*sk
)
422 sk_node_init(&sk
->sk_node
);
428 /* Grab socket reference count. This operation is valid only
429 when sk is ALREADY grabbed f.e. it is found in hash table
430 or a list and the lookup is made under lock preventing hash table
434 static inline void sock_hold(struct sock
*sk
)
436 atomic_inc(&sk
->sk_refcnt
);
439 /* Ungrab socket in the context, which assumes that socket refcnt
440 cannot hit zero, f.e. it is true in context of any socketcall.
442 static inline void __sock_put(struct sock
*sk
)
444 atomic_dec(&sk
->sk_refcnt
);
447 static __inline__
int sk_del_node_init(struct sock
*sk
)
449 int rc
= __sk_del_node_init(sk
);
452 /* paranoid for a while -acme */
453 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
458 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
460 static __inline__
int __sk_nulls_del_node_init_rcu(struct sock
*sk
)
463 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
469 static __inline__
int sk_nulls_del_node_init_rcu(struct sock
*sk
)
471 int rc
= __sk_nulls_del_node_init_rcu(sk
);
474 /* paranoid for a while -acme */
475 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
481 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
483 hlist_add_head(&sk
->sk_node
, list
);
486 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
489 __sk_add_node(sk
, list
);
492 static __inline__
void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
495 hlist_add_head_rcu(&sk
->sk_node
, list
);
498 static __inline__
void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
500 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
503 static __inline__
void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
506 __sk_nulls_add_node_rcu(sk
, list
);
509 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
511 __hlist_del(&sk
->sk_bind_node
);
514 static __inline__
void sk_add_bind_node(struct sock
*sk
,
515 struct hlist_head
*list
)
517 hlist_add_head(&sk
->sk_bind_node
, list
);
520 #define sk_for_each(__sk, node, list) \
521 hlist_for_each_entry(__sk, node, list, sk_node)
522 #define sk_for_each_rcu(__sk, node, list) \
523 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
524 #define sk_nulls_for_each(__sk, node, list) \
525 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
526 #define sk_nulls_for_each_rcu(__sk, node, list) \
527 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
528 #define sk_for_each_from(__sk, node) \
529 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
530 hlist_for_each_entry_from(__sk, node, sk_node)
531 #define sk_nulls_for_each_from(__sk, node) \
532 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
533 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
534 #define sk_for_each_safe(__sk, node, tmp, list) \
535 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
536 #define sk_for_each_bound(__sk, node, list) \
537 hlist_for_each_entry(__sk, node, list, sk_bind_node)
550 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
551 SOCK_DBG
, /* %SO_DEBUG setting */
552 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
553 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
554 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
555 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
556 SOCK_TIMESTAMPING_TX_HARDWARE
, /* %SOF_TIMESTAMPING_TX_HARDWARE */
557 SOCK_TIMESTAMPING_TX_SOFTWARE
, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
558 SOCK_TIMESTAMPING_RX_HARDWARE
, /* %SOF_TIMESTAMPING_RX_HARDWARE */
559 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
560 SOCK_TIMESTAMPING_SOFTWARE
, /* %SOF_TIMESTAMPING_SOFTWARE */
561 SOCK_TIMESTAMPING_RAW_HARDWARE
, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
562 SOCK_TIMESTAMPING_SYS_HARDWARE
, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
563 SOCK_FASYNC
, /* fasync() active */
567 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
569 nsk
->sk_flags
= osk
->sk_flags
;
572 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
574 __set_bit(flag
, &sk
->sk_flags
);
577 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
579 __clear_bit(flag
, &sk
->sk_flags
);
582 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
584 return test_bit(flag
, &sk
->sk_flags
);
587 static inline void sk_acceptq_removed(struct sock
*sk
)
589 sk
->sk_ack_backlog
--;
592 static inline void sk_acceptq_added(struct sock
*sk
)
594 sk
->sk_ack_backlog
++;
597 static inline int sk_acceptq_is_full(struct sock
*sk
)
599 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
603 * Compute minimal free write space needed to queue new packets.
605 static inline int sk_stream_min_wspace(struct sock
*sk
)
607 return sk
->sk_wmem_queued
>> 1;
610 static inline int sk_stream_wspace(struct sock
*sk
)
612 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
615 extern void sk_stream_write_space(struct sock
*sk
);
617 static inline int sk_stream_memory_free(struct sock
*sk
)
619 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
622 /* OOB backlog add */
623 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
625 /* dont let skb dst not refcounted, we are going to leave rcu lock */
628 if (!sk
->sk_backlog
.tail
)
629 sk
->sk_backlog
.head
= skb
;
631 sk
->sk_backlog
.tail
->next
= skb
;
633 sk
->sk_backlog
.tail
= skb
;
638 * Take into account size of receive queue and backlog queue
640 static inline bool sk_rcvqueues_full(const struct sock
*sk
, const struct sk_buff
*skb
)
642 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
644 return qsize
+ skb
->truesize
> sk
->sk_rcvbuf
;
647 /* The per-socket spinlock must be held here. */
648 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
650 if (sk_rcvqueues_full(sk
, skb
))
653 __sk_add_backlog(sk
, skb
);
654 sk
->sk_backlog
.len
+= skb
->truesize
;
658 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
660 return sk
->sk_backlog_rcv(sk
, skb
);
663 static inline void sock_rps_record_flow(const struct sock
*sk
)
666 struct rps_sock_flow_table
*sock_flow_table
;
669 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
670 rps_record_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
675 static inline void sock_rps_reset_flow(const struct sock
*sk
)
678 struct rps_sock_flow_table
*sock_flow_table
;
681 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
682 rps_reset_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
687 static inline void sock_rps_save_rxhash(struct sock
*sk
, u32 rxhash
)
690 if (unlikely(sk
->sk_rxhash
!= rxhash
)) {
691 sock_rps_reset_flow(sk
);
692 sk
->sk_rxhash
= rxhash
;
697 #define sk_wait_event(__sk, __timeo, __condition) \
699 release_sock(__sk); \
700 __rc = __condition; \
702 *(__timeo) = schedule_timeout(*(__timeo)); \
705 __rc = __condition; \
709 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
710 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
711 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
712 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
713 extern void sk_stream_kill_queues(struct sock
*sk
);
715 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
717 struct request_sock_ops
;
718 struct timewait_sock_ops
;
719 struct inet_hashinfo
;
722 /* Networking protocol blocks we attach to sockets.
723 * socket layer -> transport layer interface
724 * transport -> network interface is defined by struct inet_proto
727 void (*close
)(struct sock
*sk
,
729 int (*connect
)(struct sock
*sk
,
730 struct sockaddr
*uaddr
,
732 int (*disconnect
)(struct sock
*sk
, int flags
);
734 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
736 int (*ioctl
)(struct sock
*sk
, int cmd
,
738 int (*init
)(struct sock
*sk
);
739 void (*destroy
)(struct sock
*sk
);
740 void (*shutdown
)(struct sock
*sk
, int how
);
741 int (*setsockopt
)(struct sock
*sk
, int level
,
742 int optname
, char __user
*optval
,
743 unsigned int optlen
);
744 int (*getsockopt
)(struct sock
*sk
, int level
,
745 int optname
, char __user
*optval
,
748 int (*compat_setsockopt
)(struct sock
*sk
,
750 int optname
, char __user
*optval
,
751 unsigned int optlen
);
752 int (*compat_getsockopt
)(struct sock
*sk
,
754 int optname
, char __user
*optval
,
757 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
758 struct msghdr
*msg
, size_t len
);
759 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
761 size_t len
, int noblock
, int flags
,
763 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
764 int offset
, size_t size
, int flags
);
765 int (*bind
)(struct sock
*sk
,
766 struct sockaddr
*uaddr
, int addr_len
);
768 int (*backlog_rcv
) (struct sock
*sk
,
769 struct sk_buff
*skb
);
771 /* Keeping track of sk's, looking them up, and port selection methods. */
772 void (*hash
)(struct sock
*sk
);
773 void (*unhash
)(struct sock
*sk
);
774 void (*rehash
)(struct sock
*sk
);
775 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
776 void (*clear_sk
)(struct sock
*sk
, int size
);
778 /* Keeping track of sockets in use */
779 #ifdef CONFIG_PROC_FS
780 unsigned int inuse_idx
;
783 /* Memory pressure */
784 void (*enter_memory_pressure
)(struct sock
*sk
);
785 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
786 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
788 * Pressure flag: try to collapse.
789 * Technical note: it is used by multiple contexts non atomically.
790 * All the __sk_mem_schedule() is of this nature: accounting
791 * is strict, actions are advisory and have some latency.
793 int *memory_pressure
;
800 struct kmem_cache
*slab
;
801 unsigned int obj_size
;
804 struct percpu_counter
*orphan_count
;
806 struct request_sock_ops
*rsk_prot
;
807 struct timewait_sock_ops
*twsk_prot
;
810 struct inet_hashinfo
*hashinfo
;
811 struct udp_table
*udp_table
;
812 struct raw_hashinfo
*raw_hash
;
815 struct module
*owner
;
819 struct list_head node
;
820 #ifdef SOCK_REFCNT_DEBUG
825 extern int proto_register(struct proto
*prot
, int alloc_slab
);
826 extern void proto_unregister(struct proto
*prot
);
828 #ifdef SOCK_REFCNT_DEBUG
829 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
831 atomic_inc(&sk
->sk_prot
->socks
);
834 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
836 atomic_dec(&sk
->sk_prot
->socks
);
837 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
838 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
841 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
843 if (atomic_read(&sk
->sk_refcnt
) != 1)
844 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
845 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
847 #else /* SOCK_REFCNT_DEBUG */
848 #define sk_refcnt_debug_inc(sk) do { } while (0)
849 #define sk_refcnt_debug_dec(sk) do { } while (0)
850 #define sk_refcnt_debug_release(sk) do { } while (0)
851 #endif /* SOCK_REFCNT_DEBUG */
854 #ifdef CONFIG_PROC_FS
855 /* Called with local bh disabled */
856 extern void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
857 extern int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
859 static void inline sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
866 /* With per-bucket locks this operation is not-atomic, so that
867 * this version is not worse.
869 static inline void __sk_prot_rehash(struct sock
*sk
)
871 sk
->sk_prot
->unhash(sk
);
872 sk
->sk_prot
->hash(sk
);
875 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
877 /* About 10 seconds */
878 #define SOCK_DESTROY_TIME (10*HZ)
880 /* Sockets 0-1023 can't be bound to unless you are superuser */
881 #define PROT_SOCK 1024
883 #define SHUTDOWN_MASK 3
884 #define RCV_SHUTDOWN 1
885 #define SEND_SHUTDOWN 2
887 #define SOCK_SNDBUF_LOCK 1
888 #define SOCK_RCVBUF_LOCK 2
889 #define SOCK_BINDADDR_LOCK 4
890 #define SOCK_BINDPORT_LOCK 8
892 /* sock_iocb: used to kick off async processing of socket ios */
894 struct list_head list
;
900 struct scm_cookie
*scm
;
901 struct msghdr
*msg
, async_msg
;
905 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
907 return (struct sock_iocb
*)iocb
->private;
910 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
915 struct socket_alloc
{
916 struct socket socket
;
917 struct inode vfs_inode
;
920 static inline struct socket
*SOCKET_I(struct inode
*inode
)
922 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
925 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
927 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
931 * Functions for memory accounting
933 extern int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
934 extern void __sk_mem_reclaim(struct sock
*sk
);
936 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
937 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
938 #define SK_MEM_SEND 0
939 #define SK_MEM_RECV 1
941 static inline int sk_mem_pages(int amt
)
943 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
946 static inline int sk_has_account(struct sock
*sk
)
948 /* return true if protocol supports memory accounting */
949 return !!sk
->sk_prot
->memory_allocated
;
952 static inline int sk_wmem_schedule(struct sock
*sk
, int size
)
954 if (!sk_has_account(sk
))
956 return size
<= sk
->sk_forward_alloc
||
957 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
960 static inline int sk_rmem_schedule(struct sock
*sk
, int size
)
962 if (!sk_has_account(sk
))
964 return size
<= sk
->sk_forward_alloc
||
965 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
);
968 static inline void sk_mem_reclaim(struct sock
*sk
)
970 if (!sk_has_account(sk
))
972 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
973 __sk_mem_reclaim(sk
);
976 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
978 if (!sk_has_account(sk
))
980 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
981 __sk_mem_reclaim(sk
);
984 static inline void sk_mem_charge(struct sock
*sk
, int size
)
986 if (!sk_has_account(sk
))
988 sk
->sk_forward_alloc
-= size
;
991 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
993 if (!sk_has_account(sk
))
995 sk
->sk_forward_alloc
+= size
;
998 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1000 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1001 sk
->sk_wmem_queued
-= skb
->truesize
;
1002 sk_mem_uncharge(sk
, skb
->truesize
);
1006 /* Used by processes to "lock" a socket state, so that
1007 * interrupts and bottom half handlers won't change it
1008 * from under us. It essentially blocks any incoming
1009 * packets, so that we won't get any new data or any
1010 * packets that change the state of the socket.
1012 * While locked, BH processing will add new packets to
1013 * the backlog queue. This queue is processed by the
1014 * owner of the socket lock right before it is released.
1016 * Since ~2.3.5 it is also exclusive sleep lock serializing
1017 * accesses from user process context.
1019 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1022 * Macro so as to not evaluate some arguments when
1023 * lockdep is not enabled.
1025 * Mark both the sk_lock and the sk_lock.slock as a
1026 * per-address-family lock class.
1028 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1030 sk->sk_lock.owned = 0; \
1031 init_waitqueue_head(&sk->sk_lock.wq); \
1032 spin_lock_init(&(sk)->sk_lock.slock); \
1033 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1034 sizeof((sk)->sk_lock)); \
1035 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1037 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1040 extern void lock_sock_nested(struct sock
*sk
, int subclass
);
1042 static inline void lock_sock(struct sock
*sk
)
1044 lock_sock_nested(sk
, 0);
1047 extern void release_sock(struct sock
*sk
);
1049 /* BH context may only use the following locking interface. */
1050 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1051 #define bh_lock_sock_nested(__sk) \
1052 spin_lock_nested(&((__sk)->sk_lock.slock), \
1053 SINGLE_DEPTH_NESTING)
1054 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1056 extern bool lock_sock_fast(struct sock
*sk
);
1058 * unlock_sock_fast - complement of lock_sock_fast
1062 * fast unlock socket for user context.
1063 * If slow mode is on, we call regular release_sock()
1065 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1070 spin_unlock_bh(&sk
->sk_lock
.slock
);
1074 extern struct sock
*sk_alloc(struct net
*net
, int family
,
1076 struct proto
*prot
);
1077 extern void sk_free(struct sock
*sk
);
1078 extern void sk_release_kernel(struct sock
*sk
);
1079 extern struct sock
*sk_clone(const struct sock
*sk
,
1080 const gfp_t priority
);
1082 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
1083 unsigned long size
, int force
,
1085 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
1086 unsigned long size
, int force
,
1088 extern void sock_wfree(struct sk_buff
*skb
);
1089 extern void sock_rfree(struct sk_buff
*skb
);
1091 extern int sock_setsockopt(struct socket
*sock
, int level
,
1092 int op
, char __user
*optval
,
1093 unsigned int optlen
);
1095 extern int sock_getsockopt(struct socket
*sock
, int level
,
1096 int op
, char __user
*optval
,
1097 int __user
*optlen
);
1098 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
1102 extern struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
,
1103 unsigned long header_len
,
1104 unsigned long data_len
,
1107 extern void *sock_kmalloc(struct sock
*sk
, int size
,
1109 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1110 extern void sk_send_sigurg(struct sock
*sk
);
1112 #ifdef CONFIG_CGROUPS
1113 extern void sock_update_classid(struct sock
*sk
);
1115 static inline void sock_update_classid(struct sock
*sk
)
1121 * Functions to fill in entries in struct proto_ops when a protocol
1122 * does not implement a particular function.
1124 extern int sock_no_bind(struct socket
*,
1125 struct sockaddr
*, int);
1126 extern int sock_no_connect(struct socket
*,
1127 struct sockaddr
*, int, int);
1128 extern int sock_no_socketpair(struct socket
*,
1130 extern int sock_no_accept(struct socket
*,
1131 struct socket
*, int);
1132 extern int sock_no_getname(struct socket
*,
1133 struct sockaddr
*, int *, int);
1134 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
1135 struct poll_table_struct
*);
1136 extern int sock_no_ioctl(struct socket
*, unsigned int,
1138 extern int sock_no_listen(struct socket
*, int);
1139 extern int sock_no_shutdown(struct socket
*, int);
1140 extern int sock_no_getsockopt(struct socket
*, int , int,
1141 char __user
*, int __user
*);
1142 extern int sock_no_setsockopt(struct socket
*, int, int,
1143 char __user
*, unsigned int);
1144 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
1145 struct msghdr
*, size_t);
1146 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
1147 struct msghdr
*, size_t, int);
1148 extern int sock_no_mmap(struct file
*file
,
1149 struct socket
*sock
,
1150 struct vm_area_struct
*vma
);
1151 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
1153 int offset
, size_t size
,
1157 * Functions to fill in entries in struct proto_ops when a protocol
1158 * uses the inet style.
1160 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1161 char __user
*optval
, int __user
*optlen
);
1162 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1163 struct msghdr
*msg
, size_t size
, int flags
);
1164 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1165 char __user
*optval
, unsigned int optlen
);
1166 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1167 int optname
, char __user
*optval
, int __user
*optlen
);
1168 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1169 int optname
, char __user
*optval
, unsigned int optlen
);
1171 extern void sk_common_release(struct sock
*sk
);
1174 * Default socket callbacks and setup code
1177 /* Initialise core socket variables */
1178 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1180 extern void sk_filter_release_rcu(struct rcu_head
*rcu
);
1183 * sk_filter_release - release a socket filter
1184 * @fp: filter to remove
1186 * Remove a filter from a socket and release its resources.
1189 static inline void sk_filter_release(struct sk_filter
*fp
)
1191 if (atomic_dec_and_test(&fp
->refcnt
))
1192 call_rcu_bh(&fp
->rcu
, sk_filter_release_rcu
);
1195 static inline void sk_filter_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
1197 unsigned int size
= sk_filter_len(fp
);
1199 atomic_sub(size
, &sk
->sk_omem_alloc
);
1200 sk_filter_release(fp
);
1203 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
1205 atomic_inc(&fp
->refcnt
);
1206 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
1210 * Socket reference counting postulates.
1212 * * Each user of socket SHOULD hold a reference count.
1213 * * Each access point to socket (an hash table bucket, reference from a list,
1214 * running timer, skb in flight MUST hold a reference count.
1215 * * When reference count hits 0, it means it will never increase back.
1216 * * When reference count hits 0, it means that no references from
1217 * outside exist to this socket and current process on current CPU
1218 * is last user and may/should destroy this socket.
1219 * * sk_free is called from any context: process, BH, IRQ. When
1220 * it is called, socket has no references from outside -> sk_free
1221 * may release descendant resources allocated by the socket, but
1222 * to the time when it is called, socket is NOT referenced by any
1223 * hash tables, lists etc.
1224 * * Packets, delivered from outside (from network or from another process)
1225 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1226 * when they sit in queue. Otherwise, packets will leak to hole, when
1227 * socket is looked up by one cpu and unhasing is made by another CPU.
1228 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1229 * (leak to backlog). Packet socket does all the processing inside
1230 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1231 * use separate SMP lock, so that they are prone too.
1234 /* Ungrab socket and destroy it, if it was the last reference. */
1235 static inline void sock_put(struct sock
*sk
)
1237 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1241 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1244 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1246 sk
->sk_tx_queue_mapping
= tx_queue
;
1249 static inline void sk_tx_queue_clear(struct sock
*sk
)
1251 sk
->sk_tx_queue_mapping
= -1;
1254 static inline int sk_tx_queue_get(const struct sock
*sk
)
1256 return sk
? sk
->sk_tx_queue_mapping
: -1;
1259 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1261 sk_tx_queue_clear(sk
);
1262 sk
->sk_socket
= sock
;
1265 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1267 return &sk
->sk_wq
->wait
;
1269 /* Detach socket from process context.
1270 * Announce socket dead, detach it from wait queue and inode.
1271 * Note that parent inode held reference count on this struct sock,
1272 * we do not release it in this function, because protocol
1273 * probably wants some additional cleanups or even continuing
1274 * to work with this socket (TCP).
1276 static inline void sock_orphan(struct sock
*sk
)
1278 write_lock_bh(&sk
->sk_callback_lock
);
1279 sock_set_flag(sk
, SOCK_DEAD
);
1280 sk_set_socket(sk
, NULL
);
1282 write_unlock_bh(&sk
->sk_callback_lock
);
1285 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1287 write_lock_bh(&sk
->sk_callback_lock
);
1288 rcu_assign_pointer(sk
->sk_wq
, parent
->wq
);
1290 sk_set_socket(sk
, parent
);
1291 security_sock_graft(sk
, parent
);
1292 write_unlock_bh(&sk
->sk_callback_lock
);
1295 extern int sock_i_uid(struct sock
*sk
);
1296 extern unsigned long sock_i_ino(struct sock
*sk
);
1298 static inline struct dst_entry
*
1299 __sk_dst_get(struct sock
*sk
)
1301 return rcu_dereference_check(sk
->sk_dst_cache
, rcu_read_lock_held() ||
1302 sock_owned_by_user(sk
) ||
1303 lockdep_is_held(&sk
->sk_lock
.slock
));
1306 static inline struct dst_entry
*
1307 sk_dst_get(struct sock
*sk
)
1309 struct dst_entry
*dst
;
1312 dst
= rcu_dereference(sk
->sk_dst_cache
);
1319 extern void sk_reset_txq(struct sock
*sk
);
1321 static inline void dst_negative_advice(struct sock
*sk
)
1323 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1325 if (dst
&& dst
->ops
->negative_advice
) {
1326 ndst
= dst
->ops
->negative_advice(dst
);
1329 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1336 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1338 struct dst_entry
*old_dst
;
1340 sk_tx_queue_clear(sk
);
1342 * This can be called while sk is owned by the caller only,
1343 * with no state that can be checked in a rcu_dereference_check() cond
1345 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1346 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1347 dst_release(old_dst
);
1351 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1353 spin_lock(&sk
->sk_dst_lock
);
1354 __sk_dst_set(sk
, dst
);
1355 spin_unlock(&sk
->sk_dst_lock
);
1359 __sk_dst_reset(struct sock
*sk
)
1361 __sk_dst_set(sk
, NULL
);
1365 sk_dst_reset(struct sock
*sk
)
1367 spin_lock(&sk
->sk_dst_lock
);
1369 spin_unlock(&sk
->sk_dst_lock
);
1372 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1374 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1376 static inline int sk_can_gso(const struct sock
*sk
)
1378 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1381 extern void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1383 static inline void sk_nocaps_add(struct sock
*sk
, int flags
)
1385 sk
->sk_route_nocaps
|= flags
;
1386 sk
->sk_route_caps
&= ~flags
;
1389 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1390 struct sk_buff
*skb
, struct page
*page
,
1393 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1395 __wsum csum
= csum_and_copy_from_user(from
,
1396 page_address(page
) + off
,
1400 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1401 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1405 skb
->data_len
+= copy
;
1406 skb
->truesize
+= copy
;
1407 sk
->sk_wmem_queued
+= copy
;
1408 sk_mem_charge(sk
, copy
);
1413 * sk_wmem_alloc_get - returns write allocations
1416 * Returns sk_wmem_alloc minus initial offset of one
1418 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1420 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1424 * sk_rmem_alloc_get - returns read allocations
1427 * Returns sk_rmem_alloc
1429 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1431 return atomic_read(&sk
->sk_rmem_alloc
);
1435 * sk_has_allocations - check if allocations are outstanding
1438 * Returns true if socket has write or read allocations
1440 static inline int sk_has_allocations(const struct sock
*sk
)
1442 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1446 * wq_has_sleeper - check if there are any waiting processes
1447 * @wq: struct socket_wq
1449 * Returns true if socket_wq has waiting processes
1451 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1452 * barrier call. They were added due to the race found within the tcp code.
1454 * Consider following tcp code paths:
1458 * sys_select receive packet
1460 * __add_wait_queue update tp->rcv_nxt
1462 * tp->rcv_nxt check sock_def_readable
1464 * schedule rcu_read_lock();
1465 * wq = rcu_dereference(sk->sk_wq);
1466 * if (wq && waitqueue_active(&wq->wait))
1467 * wake_up_interruptible(&wq->wait)
1471 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1472 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1473 * could then endup calling schedule and sleep forever if there are no more
1474 * data on the socket.
1477 static inline bool wq_has_sleeper(struct socket_wq
*wq
)
1481 * We need to be sure we are in sync with the
1482 * add_wait_queue modifications to the wait queue.
1484 * This memory barrier is paired in the sock_poll_wait.
1487 return wq
&& waitqueue_active(&wq
->wait
);
1491 * sock_poll_wait - place memory barrier behind the poll_wait call.
1493 * @wait_address: socket wait queue
1496 * See the comments in the wq_has_sleeper function.
1498 static inline void sock_poll_wait(struct file
*filp
,
1499 wait_queue_head_t
*wait_address
, poll_table
*p
)
1501 if (p
&& wait_address
) {
1502 poll_wait(filp
, wait_address
, p
);
1504 * We need to be sure we are in sync with the
1505 * socket flags modification.
1507 * This memory barrier is paired in the wq_has_sleeper.
1514 * Queue a received datagram if it will fit. Stream and sequenced
1515 * protocols can't normally use this as they need to fit buffers in
1516 * and play with them.
1518 * Inlined as it's very short and called for pretty much every
1519 * packet ever received.
1522 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1526 skb
->destructor
= sock_wfree
;
1528 * We used to take a refcount on sk, but following operation
1529 * is enough to guarantee sk_free() wont free this sock until
1530 * all in-flight packets are completed
1532 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1535 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1539 skb
->destructor
= sock_rfree
;
1540 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1541 sk_mem_charge(sk
, skb
->truesize
);
1544 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1545 unsigned long expires
);
1547 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1549 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1551 extern int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
1554 * Recover an error report and clear atomically
1557 static inline int sock_error(struct sock
*sk
)
1560 if (likely(!sk
->sk_err
))
1562 err
= xchg(&sk
->sk_err
, 0);
1566 static inline unsigned long sock_wspace(struct sock
*sk
)
1570 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1571 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1578 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1580 if (sock_flag(sk
, SOCK_FASYNC
))
1581 sock_wake_async(sk
->sk_socket
, how
, band
);
1584 #define SOCK_MIN_SNDBUF 2048
1586 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1587 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1589 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1591 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1593 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1594 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
1595 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1599 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
1601 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1603 struct page
*page
= NULL
;
1605 page
= alloc_pages(sk
->sk_allocation
, 0);
1607 sk
->sk_prot
->enter_memory_pressure(sk
);
1608 sk_stream_moderate_sndbuf(sk
);
1614 * Default write policy as shown to user space via poll/select/SIGIO
1616 static inline int sock_writeable(const struct sock
*sk
)
1618 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
1621 static inline gfp_t
gfp_any(void)
1623 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
1626 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1628 return noblock
? 0 : sk
->sk_rcvtimeo
;
1631 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1633 return noblock
? 0 : sk
->sk_sndtimeo
;
1636 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1638 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1641 /* Alas, with timeout socket operations are not restartable.
1642 * Compare this to poll().
1644 static inline int sock_intr_errno(long timeo
)
1646 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
1649 extern void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
1650 struct sk_buff
*skb
);
1652 static __inline__
void
1653 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
1655 ktime_t kt
= skb
->tstamp
;
1656 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
1659 * generate control messages if
1660 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1661 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1662 * - software time stamp available and wanted
1663 * (SOCK_TIMESTAMPING_SOFTWARE)
1664 * - hardware time stamps available and wanted
1665 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1666 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1668 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
1669 sock_flag(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
) ||
1670 (kt
.tv64
&& sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) ||
1671 (hwtstamps
->hwtstamp
.tv64
&&
1672 sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
)) ||
1673 (hwtstamps
->syststamp
.tv64
&&
1674 sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
)))
1675 __sock_recv_timestamp(msg
, sk
, skb
);
1680 extern void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
1681 struct sk_buff
*skb
);
1683 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
1684 struct sk_buff
*skb
)
1686 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
1687 (1UL << SOCK_RCVTSTAMP) | \
1688 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
1689 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
1690 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
1691 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
1693 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
)
1694 __sock_recv_ts_and_drops(msg
, sk
, skb
);
1696 sk
->sk_stamp
= skb
->tstamp
;
1700 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1701 * @sk: socket sending this packet
1702 * @tx_flags: filled with instructions for time stamping
1704 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1705 * parameters are invalid.
1707 extern int sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
);
1710 * sk_eat_skb - Release a skb if it is no longer needed
1711 * @sk: socket to eat this skb from
1712 * @skb: socket buffer to eat
1713 * @copied_early: flag indicating whether DMA operations copied this data early
1715 * This routine must be called with interrupts disabled or with the socket
1716 * locked so that the sk_buff queue operation is ok.
1718 #ifdef CONFIG_NET_DMA
1719 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1721 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1725 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
1728 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1730 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1736 struct net
*sock_net(const struct sock
*sk
)
1738 return read_pnet(&sk
->sk_net
);
1742 void sock_net_set(struct sock
*sk
, struct net
*net
)
1744 write_pnet(&sk
->sk_net
, net
);
1748 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1749 * They should not hold a referrence to a namespace in order to allow
1751 * Sockets after sk_change_net should be released using sk_release_kernel
1753 static inline void sk_change_net(struct sock
*sk
, struct net
*net
)
1755 put_net(sock_net(sk
));
1756 sock_net_set(sk
, hold_net(net
));
1759 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
1761 if (unlikely(skb
->sk
)) {
1762 struct sock
*sk
= skb
->sk
;
1764 skb
->destructor
= NULL
;
1771 extern void sock_enable_timestamp(struct sock
*sk
, int flag
);
1772 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
1773 extern int sock_get_timestampns(struct sock
*, struct timespec __user
*);
1776 * Enable debug/info messages
1778 extern int net_msg_warn
;
1779 #define NETDEBUG(fmt, args...) \
1780 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1782 #define LIMIT_NETDEBUG(fmt, args...) \
1783 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1785 extern __u32 sysctl_wmem_max
;
1786 extern __u32 sysctl_rmem_max
;
1788 extern void sk_init(void);
1790 extern int sysctl_optmem_max
;
1792 extern __u32 sysctl_wmem_default
;
1793 extern __u32 sysctl_rmem_default
;
1795 #endif /* _SOCK_H */