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/skbuff.h> /* struct sk_buff */
52 #include <linux/security.h>
54 #include <linux/filter.h>
56 #include <asm/atomic.h>
58 #include <net/checksum.h>
61 * This structure really needs to be cleaned up.
62 * Most of it is for TCP, and not used by any of
63 * the other protocols.
66 /* Define this to get the SOCK_DBG debugging facility. */
67 #define SOCK_DEBUGGING
69 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
70 printk(KERN_DEBUG msg); } while (0)
72 /* Validate arguments and do nothing */
73 static void inline int __attribute__ ((format (printf, 2, 3)))
74 SOCK_DEBUG(struct sock *sk, const char *msg, ...)
79 /* This is the per-socket lock. The spinlock provides a synchronization
80 * between user contexts and software interrupt processing, whereas the
81 * mini-semaphore synchronizes multiple users amongst themselves.
88 * We express the mutex-alike socket_lock semantics
89 * to the lock validator by explicitly managing
90 * the slock as a lock variant (in addition to
93 #ifdef CONFIG_DEBUG_LOCK_ALLOC
94 struct lockdep_map dep_map;
103 * struct sock_common - minimal network layer representation of sockets
104 * @skc_family: network address family
105 * @skc_state: Connection state
106 * @skc_reuse: %SO_REUSEADDR setting
107 * @skc_bound_dev_if: bound device index if != 0
108 * @skc_node: main hash linkage for various protocol lookup tables
109 * @skc_bind_node: bind hash linkage for various protocol lookup tables
110 * @skc_refcnt: reference count
111 * @skc_hash: hash value used with various protocol lookup tables
112 * @skc_prot: protocol handlers inside a network family
113 * @skc_net: reference to the network namespace of this socket
115 * This is the minimal network layer representation of sockets, the header
116 * for struct sock and struct inet_timewait_sock.
119 unsigned short skc_family;
120 volatile unsigned char skc_state;
121 unsigned char skc_reuse;
122 int skc_bound_dev_if;
123 struct hlist_node skc_node;
124 struct hlist_node skc_bind_node;
126 unsigned int skc_hash;
127 struct proto *skc_prot;
134 * struct sock - network layer representation of sockets
135 * @__sk_common: shared layout with inet_timewait_sock
136 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
137 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
138 * @sk_lock: synchronizer
139 * @sk_rcvbuf: size of receive buffer in bytes
140 * @sk_sleep: sock wait queue
141 * @sk_dst_cache: destination cache
142 * @sk_dst_lock: destination cache lock
143 * @sk_policy: flow policy
144 * @sk_rmem_alloc: receive queue bytes committed
145 * @sk_receive_queue: incoming packets
146 * @sk_wmem_alloc: transmit queue bytes committed
147 * @sk_write_queue: Packet sending queue
148 * @sk_async_wait_queue: DMA copied packets
149 * @sk_omem_alloc: "o" is "option" or "other"
150 * @sk_wmem_queued: persistent queue size
151 * @sk_forward_alloc: space allocated forward
152 * @sk_allocation: allocation mode
153 * @sk_sndbuf: size of send buffer in bytes
154 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
155 * %SO_OOBINLINE settings
156 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
157 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
158 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
159 * @sk_gso_max_size: Maximum GSO segment size to build
160 * @sk_lingertime: %SO_LINGER l_linger setting
161 * @sk_backlog: always used with the per-socket spinlock held
162 * @sk_callback_lock: used with the callbacks in the end of this struct
163 * @sk_error_queue: rarely used
164 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
165 * IPV6_ADDRFORM for instance)
166 * @sk_err: last error
167 * @sk_err_soft: errors that don't cause failure but are the cause of a
168 * persistent failure not just 'timed out'
169 * @sk_drops: raw/udp drops counter
170 * @sk_ack_backlog: current listen backlog
171 * @sk_max_ack_backlog: listen backlog set in listen()
172 * @sk_priority: %SO_PRIORITY setting
173 * @sk_type: socket type (%SOCK_STREAM, etc)
174 * @sk_protocol: which protocol this socket belongs in this network family
175 * @sk_peercred: %SO_PEERCRED setting
176 * @sk_rcvlowat: %SO_RCVLOWAT setting
177 * @sk_rcvtimeo: %SO_RCVTIMEO setting
178 * @sk_sndtimeo: %SO_SNDTIMEO setting
179 * @sk_filter: socket filtering instructions
180 * @sk_protinfo: private area, net family specific, when not using slab
181 * @sk_timer: sock cleanup timer
182 * @sk_stamp: time stamp of last packet received
183 * @sk_socket: Identd and reporting IO signals
184 * @sk_user_data: RPC layer private data
185 * @sk_sndmsg_page: cached page for sendmsg
186 * @sk_sndmsg_off: cached offset for sendmsg
187 * @sk_send_head: front of stuff to transmit
188 * @sk_security: used by security modules
189 * @sk_mark: generic packet mark
190 * @sk_write_pending: a write to stream socket waits to start
191 * @sk_state_change: callback to indicate change in the state of the sock
192 * @sk_data_ready: callback to indicate there is data to be processed
193 * @sk_write_space: callback to indicate there is bf sending space available
194 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
195 * @sk_backlog_rcv: callback to process the backlog
196 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
200 * Now struct inet_timewait_sock also uses sock_common, so please just
201 * don't add nothing before this first member (__sk_common) --acme
203 struct sock_common __sk_common;
204 #define sk_family __sk_common.skc_family
205 #define sk_state __sk_common.skc_state
206 #define sk_reuse __sk_common.skc_reuse
207 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
208 #define sk_node __sk_common.skc_node
209 #define sk_bind_node __sk_common.skc_bind_node
210 #define sk_refcnt __sk_common.skc_refcnt
211 #define sk_hash __sk_common.skc_hash
212 #define sk_prot __sk_common.skc_prot
213 #define sk_net __sk_common.skc_net
214 unsigned char sk_shutdown : 2,
217 unsigned char sk_protocol;
218 unsigned short sk_type;
220 socket_lock_t sk_lock;
222 * The backlog queue is special, it is always used with
223 * the per-socket spinlock held and requires low latency
224 * access. Therefore we special case it's implementation.
227 struct sk_buff *head;
228 struct sk_buff *tail;
230 wait_queue_head_t *sk_sleep;
231 struct dst_entry *sk_dst_cache;
232 struct xfrm_policy *sk_policy[2];
233 rwlock_t sk_dst_lock;
234 atomic_t sk_rmem_alloc;
235 atomic_t sk_wmem_alloc;
236 atomic_t sk_omem_alloc;
238 struct sk_buff_head sk_receive_queue;
239 struct sk_buff_head sk_write_queue;
240 struct sk_buff_head sk_async_wait_queue;
242 int sk_forward_alloc;
246 unsigned int sk_gso_max_size;
248 unsigned long sk_flags;
249 unsigned long sk_lingertime;
250 struct sk_buff_head sk_error_queue;
251 struct proto *sk_prot_creator;
252 rwlock_t sk_callback_lock;
256 unsigned short sk_ack_backlog;
257 unsigned short sk_max_ack_backlog;
259 struct ucred sk_peercred;
262 struct sk_filter *sk_filter;
264 struct timer_list sk_timer;
266 struct socket *sk_socket;
268 struct page *sk_sndmsg_page;
269 struct sk_buff *sk_send_head;
271 int sk_write_pending;
274 /* XXX 4 bytes hole on 64 bit */
275 void (*sk_state_change)(struct sock *sk);
276 void (*sk_data_ready)(struct sock *sk, int bytes);
277 void (*sk_write_space)(struct sock *sk);
278 void (*sk_error_report)(struct sock *sk);
279 int (*sk_backlog_rcv)(struct sock *sk,
280 struct sk_buff *skb);
281 void (*sk_destruct)(struct sock *sk);
285 * Hashed lists helper routines
287 static inline struct sock *__sk_head(const struct hlist_head *head)
289 return hlist_entry(head->first, struct sock, sk_node);
292 static inline struct sock *sk_head(const struct hlist_head *head)
294 return hlist_empty(head) ? NULL : __sk_head(head);
297 static inline struct sock *sk_next(const struct sock *sk)
299 return sk->sk_node.next ?
300 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
303 static inline int sk_unhashed(const struct sock *sk)
305 return hlist_unhashed(&sk->sk_node);
308 static inline int sk_hashed(const struct sock *sk)
310 return !sk_unhashed(sk);
313 static __inline__ void sk_node_init(struct hlist_node *node)
318 static __inline__ void __sk_del_node(struct sock *sk)
320 __hlist_del(&sk->sk_node);
323 static __inline__ int __sk_del_node_init(struct sock *sk)
327 sk_node_init(&sk->sk_node);
333 /* Grab socket reference count. This operation is valid only
334 when sk is ALREADY grabbed f.e. it is found in hash table
335 or a list and the lookup is made under lock preventing hash table
339 static inline void sock_hold(struct sock *sk)
341 atomic_inc(&sk->sk_refcnt);
344 /* Ungrab socket in the context, which assumes that socket refcnt
345 cannot hit zero, f.e. it is true in context of any socketcall.
347 static inline void __sock_put(struct sock *sk)
349 atomic_dec(&sk->sk_refcnt);
352 static __inline__ int sk_del_node_init(struct sock *sk)
354 int rc = __sk_del_node_init(sk);
357 /* paranoid for a while -acme */
358 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
364 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
366 hlist_add_head(&sk->sk_node, list);
369 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
372 __sk_add_node(sk, list);
375 static __inline__ void __sk_del_bind_node(struct sock *sk)
377 __hlist_del(&sk->sk_bind_node);
380 static __inline__ void sk_add_bind_node(struct sock *sk,
381 struct hlist_head *list)
383 hlist_add_head(&sk->sk_bind_node, list);
386 #define sk_for_each(__sk, node, list) \
387 hlist_for_each_entry(__sk, node, list, sk_node)
388 #define sk_for_each_from(__sk, node) \
389 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
390 hlist_for_each_entry_from(__sk, node, sk_node)
391 #define sk_for_each_continue(__sk, node) \
392 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
393 hlist_for_each_entry_continue(__sk, node, sk_node)
394 #define sk_for_each_safe(__sk, node, tmp, list) \
395 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
396 #define sk_for_each_bound(__sk, node, list) \
397 hlist_for_each_entry(__sk, node, list, sk_bind_node)
410 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
411 SOCK_DBG, /* %SO_DEBUG setting */
412 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
413 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
414 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
415 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
418 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
420 nsk->sk_flags = osk->sk_flags;
423 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
425 __set_bit(flag, &sk->sk_flags);
428 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
430 __clear_bit(flag, &sk->sk_flags);
433 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
435 return test_bit(flag, &sk->sk_flags);
438 static inline void sk_acceptq_removed(struct sock *sk)
440 sk->sk_ack_backlog--;
443 static inline void sk_acceptq_added(struct sock *sk)
445 sk->sk_ack_backlog++;
448 static inline int sk_acceptq_is_full(struct sock *sk)
450 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
454 * Compute minimal free write space needed to queue new packets.
456 static inline int sk_stream_min_wspace(struct sock *sk)
458 return sk->sk_wmem_queued >> 1;
461 static inline int sk_stream_wspace(struct sock *sk)
463 return sk->sk_sndbuf - sk->sk_wmem_queued;
466 extern void sk_stream_write_space(struct sock *sk);
468 static inline int sk_stream_memory_free(struct sock *sk)
470 return sk->sk_wmem_queued < sk->sk_sndbuf;
473 /* The per-socket spinlock must be held here. */
474 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
476 if (!sk->sk_backlog.tail) {
477 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
479 sk->sk_backlog.tail->next = skb;
480 sk->sk_backlog.tail = skb;
485 #define sk_wait_event(__sk, __timeo, __condition) \
487 release_sock(__sk); \
488 __rc = __condition; \
490 *(__timeo) = schedule_timeout(*(__timeo)); \
493 __rc = __condition; \
497 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
498 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
499 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
500 extern int sk_stream_error(struct sock *sk, int flags, int err);
501 extern void sk_stream_kill_queues(struct sock *sk);
503 extern int sk_wait_data(struct sock *sk, long *timeo);
505 struct request_sock_ops;
506 struct timewait_sock_ops;
507 struct inet_hashinfo;
510 /* Networking protocol blocks we attach to sockets.
511 * socket layer -> transport layer interface
512 * transport -> network interface is defined by struct inet_proto
515 void (*close)(struct sock *sk,
517 int (*connect)(struct sock *sk,
518 struct sockaddr *uaddr,
520 int (*disconnect)(struct sock *sk, int flags);
522 struct sock * (*accept) (struct sock *sk, int flags, int *err);
524 int (*ioctl)(struct sock *sk, int cmd,
526 int (*init)(struct sock *sk);
527 void (*destroy)(struct sock *sk);
528 void (*shutdown)(struct sock *sk, int how);
529 int (*setsockopt)(struct sock *sk, int level,
530 int optname, char __user *optval,
532 int (*getsockopt)(struct sock *sk, int level,
533 int optname, char __user *optval,
535 int (*compat_setsockopt)(struct sock *sk,
537 int optname, char __user *optval,
539 int (*compat_getsockopt)(struct sock *sk,
541 int optname, char __user *optval,
543 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
544 struct msghdr *msg, size_t len);
545 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
547 size_t len, int noblock, int flags,
549 int (*sendpage)(struct sock *sk, struct page *page,
550 int offset, size_t size, int flags);
551 int (*bind)(struct sock *sk,
552 struct sockaddr *uaddr, int addr_len);
554 int (*backlog_rcv) (struct sock *sk,
555 struct sk_buff *skb);
557 /* Keeping track of sk's, looking them up, and port selection methods. */
558 void (*hash)(struct sock *sk);
559 void (*unhash)(struct sock *sk);
560 int (*get_port)(struct sock *sk, unsigned short snum);
562 /* Keeping track of sockets in use */
563 #ifdef CONFIG_PROC_FS
564 unsigned int inuse_idx;
567 /* Memory pressure */
568 void (*enter_memory_pressure)(struct sock *sk);
569 atomic_t *memory_allocated; /* Current allocated memory. */
570 atomic_t *sockets_allocated; /* Current number of sockets. */
572 * Pressure flag: try to collapse.
573 * Technical note: it is used by multiple contexts non atomically.
574 * All the __sk_mem_schedule() is of this nature: accounting
575 * is strict, actions are advisory and have some latency.
577 int *memory_pressure;
583 struct kmem_cache *slab;
584 unsigned int obj_size;
586 atomic_t *orphan_count;
588 struct request_sock_ops *rsk_prot;
589 struct timewait_sock_ops *twsk_prot;
592 struct inet_hashinfo *hashinfo;
593 struct hlist_head *udp_hash;
594 struct raw_hashinfo *raw_hash;
597 struct module *owner;
601 struct list_head node;
602 #ifdef SOCK_REFCNT_DEBUG
607 extern int proto_register(struct proto *prot, int alloc_slab);
608 extern void proto_unregister(struct proto *prot);
610 #ifdef SOCK_REFCNT_DEBUG
611 static inline void sk_refcnt_debug_inc(struct sock *sk)
613 atomic_inc(&sk->sk_prot->socks);
616 static inline void sk_refcnt_debug_dec(struct sock *sk)
618 atomic_dec(&sk->sk_prot->socks);
619 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
620 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
623 static inline void sk_refcnt_debug_release(const struct sock *sk)
625 if (atomic_read(&sk->sk_refcnt) != 1)
626 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
627 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
629 #else /* SOCK_REFCNT_DEBUG */
630 #define sk_refcnt_debug_inc(sk) do { } while (0)
631 #define sk_refcnt_debug_dec(sk) do { } while (0)
632 #define sk_refcnt_debug_release(sk) do { } while (0)
633 #endif /* SOCK_REFCNT_DEBUG */
636 #ifdef CONFIG_PROC_FS
637 /* Called with local bh disabled */
638 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
639 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
641 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
648 /* With per-bucket locks this operation is not-atomic, so that
649 * this version is not worse.
651 static inline void __sk_prot_rehash(struct sock *sk)
653 sk->sk_prot->unhash(sk);
654 sk->sk_prot->hash(sk);
657 /* About 10 seconds */
658 #define SOCK_DESTROY_TIME (10*HZ)
660 /* Sockets 0-1023 can't be bound to unless you are superuser */
661 #define PROT_SOCK 1024
663 #define SHUTDOWN_MASK 3
664 #define RCV_SHUTDOWN 1
665 #define SEND_SHUTDOWN 2
667 #define SOCK_SNDBUF_LOCK 1
668 #define SOCK_RCVBUF_LOCK 2
669 #define SOCK_BINDADDR_LOCK 4
670 #define SOCK_BINDPORT_LOCK 8
672 /* sock_iocb: used to kick off async processing of socket ios */
674 struct list_head list;
680 struct scm_cookie *scm;
681 struct msghdr *msg, async_msg;
685 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
687 return (struct sock_iocb *)iocb->private;
690 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
695 struct socket_alloc {
696 struct socket socket;
697 struct inode vfs_inode;
700 static inline struct socket *SOCKET_I(struct inode *inode)
702 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
705 static inline struct inode *SOCK_INODE(struct socket *socket)
707 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
711 * Functions for memory accounting
713 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
714 extern void __sk_mem_reclaim(struct sock *sk);
716 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
717 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
718 #define SK_MEM_SEND 0
719 #define SK_MEM_RECV 1
721 static inline int sk_mem_pages(int amt)
723 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
726 static inline int sk_has_account(struct sock *sk)
728 /* return true if protocol supports memory accounting */
729 return !!sk->sk_prot->memory_allocated;
732 static inline int sk_wmem_schedule(struct sock *sk, int size)
734 if (!sk_has_account(sk))
736 return size <= sk->sk_forward_alloc ||
737 __sk_mem_schedule(sk, size, SK_MEM_SEND);
740 static inline int sk_rmem_schedule(struct sock *sk, int size)
742 if (!sk_has_account(sk))
744 return size <= sk->sk_forward_alloc ||
745 __sk_mem_schedule(sk, size, SK_MEM_RECV);
748 static inline void sk_mem_reclaim(struct sock *sk)
750 if (!sk_has_account(sk))
752 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
753 __sk_mem_reclaim(sk);
756 static inline void sk_mem_reclaim_partial(struct sock *sk)
758 if (!sk_has_account(sk))
760 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
761 __sk_mem_reclaim(sk);
764 static inline void sk_mem_charge(struct sock *sk, int size)
766 if (!sk_has_account(sk))
768 sk->sk_forward_alloc -= size;
771 static inline void sk_mem_uncharge(struct sock *sk, int size)
773 if (!sk_has_account(sk))
775 sk->sk_forward_alloc += size;
778 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
780 skb_truesize_check(skb);
781 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
782 sk->sk_wmem_queued -= skb->truesize;
783 sk_mem_uncharge(sk, skb->truesize);
787 /* Used by processes to "lock" a socket state, so that
788 * interrupts and bottom half handlers won't change it
789 * from under us. It essentially blocks any incoming
790 * packets, so that we won't get any new data or any
791 * packets that change the state of the socket.
793 * While locked, BH processing will add new packets to
794 * the backlog queue. This queue is processed by the
795 * owner of the socket lock right before it is released.
797 * Since ~2.3.5 it is also exclusive sleep lock serializing
798 * accesses from user process context.
800 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
803 * Macro so as to not evaluate some arguments when
804 * lockdep is not enabled.
806 * Mark both the sk_lock and the sk_lock.slock as a
807 * per-address-family lock class.
809 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
811 sk->sk_lock.owned = 0; \
812 init_waitqueue_head(&sk->sk_lock.wq); \
813 spin_lock_init(&(sk)->sk_lock.slock); \
814 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
815 sizeof((sk)->sk_lock)); \
816 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
818 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
821 extern void lock_sock_nested(struct sock *sk, int subclass);
823 static inline void lock_sock(struct sock *sk)
825 lock_sock_nested(sk, 0);
828 extern void release_sock(struct sock *sk);
830 /* BH context may only use the following locking interface. */
831 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
832 #define bh_lock_sock_nested(__sk) \
833 spin_lock_nested(&((__sk)->sk_lock.slock), \
834 SINGLE_DEPTH_NESTING)
835 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
837 extern struct sock *sk_alloc(struct net *net, int family,
840 extern void sk_free(struct sock *sk);
841 extern void sk_release_kernel(struct sock *sk);
842 extern struct sock *sk_clone(const struct sock *sk,
843 const gfp_t priority);
845 extern struct sk_buff *sock_wmalloc(struct sock *sk,
846 unsigned long size, int force,
848 extern struct sk_buff *sock_rmalloc(struct sock *sk,
849 unsigned long size, int force,
851 extern void sock_wfree(struct sk_buff *skb);
852 extern void sock_rfree(struct sk_buff *skb);
854 extern int sock_setsockopt(struct socket *sock, int level,
855 int op, char __user *optval,
858 extern int sock_getsockopt(struct socket *sock, int level,
859 int op, char __user *optval,
861 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
865 extern void *sock_kmalloc(struct sock *sk, int size,
867 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
868 extern void sk_send_sigurg(struct sock *sk);
871 * Functions to fill in entries in struct proto_ops when a protocol
872 * does not implement a particular function.
874 extern int sock_no_bind(struct socket *,
875 struct sockaddr *, int);
876 extern int sock_no_connect(struct socket *,
877 struct sockaddr *, int, int);
878 extern int sock_no_socketpair(struct socket *,
880 extern int sock_no_accept(struct socket *,
881 struct socket *, int);
882 extern int sock_no_getname(struct socket *,
883 struct sockaddr *, int *, int);
884 extern unsigned int sock_no_poll(struct file *, struct socket *,
885 struct poll_table_struct *);
886 extern int sock_no_ioctl(struct socket *, unsigned int,
888 extern int sock_no_listen(struct socket *, int);
889 extern int sock_no_shutdown(struct socket *, int);
890 extern int sock_no_getsockopt(struct socket *, int , int,
891 char __user *, int __user *);
892 extern int sock_no_setsockopt(struct socket *, int, int,
894 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
895 struct msghdr *, size_t);
896 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
897 struct msghdr *, size_t, int);
898 extern int sock_no_mmap(struct file *file,
900 struct vm_area_struct *vma);
901 extern ssize_t sock_no_sendpage(struct socket *sock,
903 int offset, size_t size,
907 * Functions to fill in entries in struct proto_ops when a protocol
908 * uses the inet style.
910 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
911 char __user *optval, int __user *optlen);
912 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
913 struct msghdr *msg, size_t size, int flags);
914 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
915 char __user *optval, int optlen);
916 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
917 int optname, char __user *optval, int __user *optlen);
918 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
919 int optname, char __user *optval, int optlen);
921 extern void sk_common_release(struct sock *sk);
924 * Default socket callbacks and setup code
927 /* Initialise core socket variables */
928 extern void sock_init_data(struct socket *sock, struct sock *sk);
931 * sk_filter_release: Release a socket filter
933 * @fp: filter to remove
935 * Remove a filter from a socket and release its resources.
938 static inline void sk_filter_release(struct sk_filter *fp)
940 if (atomic_dec_and_test(&fp->refcnt))
944 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
946 unsigned int size = sk_filter_len(fp);
948 atomic_sub(size, &sk->sk_omem_alloc);
949 sk_filter_release(fp);
952 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
954 atomic_inc(&fp->refcnt);
955 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
959 * Socket reference counting postulates.
961 * * Each user of socket SHOULD hold a reference count.
962 * * Each access point to socket (an hash table bucket, reference from a list,
963 * running timer, skb in flight MUST hold a reference count.
964 * * When reference count hits 0, it means it will never increase back.
965 * * When reference count hits 0, it means that no references from
966 * outside exist to this socket and current process on current CPU
967 * is last user and may/should destroy this socket.
968 * * sk_free is called from any context: process, BH, IRQ. When
969 * it is called, socket has no references from outside -> sk_free
970 * may release descendant resources allocated by the socket, but
971 * to the time when it is called, socket is NOT referenced by any
972 * hash tables, lists etc.
973 * * Packets, delivered from outside (from network or from another process)
974 * and enqueued on receive/error queues SHOULD NOT grab reference count,
975 * when they sit in queue. Otherwise, packets will leak to hole, when
976 * socket is looked up by one cpu and unhasing is made by another CPU.
977 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
978 * (leak to backlog). Packet socket does all the processing inside
979 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
980 * use separate SMP lock, so that they are prone too.
983 /* Ungrab socket and destroy it, if it was the last reference. */
984 static inline void sock_put(struct sock *sk)
986 if (atomic_dec_and_test(&sk->sk_refcnt))
990 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
993 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
995 sk->sk_socket = sock;
998 /* Detach socket from process context.
999 * Announce socket dead, detach it from wait queue and inode.
1000 * Note that parent inode held reference count on this struct sock,
1001 * we do not release it in this function, because protocol
1002 * probably wants some additional cleanups or even continuing
1003 * to work with this socket (TCP).
1005 static inline void sock_orphan(struct sock *sk)
1007 write_lock_bh(&sk->sk_callback_lock);
1008 sock_set_flag(sk, SOCK_DEAD);
1009 sk_set_socket(sk, NULL);
1010 sk->sk_sleep = NULL;
1011 write_unlock_bh(&sk->sk_callback_lock);
1014 static inline void sock_graft(struct sock *sk, struct socket *parent)
1016 write_lock_bh(&sk->sk_callback_lock);
1017 sk->sk_sleep = &parent->wait;
1019 sk_set_socket(sk, parent);
1020 security_sock_graft(sk, parent);
1021 write_unlock_bh(&sk->sk_callback_lock);
1024 extern int sock_i_uid(struct sock *sk);
1025 extern unsigned long sock_i_ino(struct sock *sk);
1027 static inline struct dst_entry *
1028 __sk_dst_get(struct sock *sk)
1030 return sk->sk_dst_cache;
1033 static inline struct dst_entry *
1034 sk_dst_get(struct sock *sk)
1036 struct dst_entry *dst;
1038 read_lock(&sk->sk_dst_lock);
1039 dst = sk->sk_dst_cache;
1042 read_unlock(&sk->sk_dst_lock);
1047 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1049 struct dst_entry *old_dst;
1051 old_dst = sk->sk_dst_cache;
1052 sk->sk_dst_cache = dst;
1053 dst_release(old_dst);
1057 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1059 write_lock(&sk->sk_dst_lock);
1060 __sk_dst_set(sk, dst);
1061 write_unlock(&sk->sk_dst_lock);
1065 __sk_dst_reset(struct sock *sk)
1067 struct dst_entry *old_dst;
1069 old_dst = sk->sk_dst_cache;
1070 sk->sk_dst_cache = NULL;
1071 dst_release(old_dst);
1075 sk_dst_reset(struct sock *sk)
1077 write_lock(&sk->sk_dst_lock);
1079 write_unlock(&sk->sk_dst_lock);
1082 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1084 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1086 static inline int sk_can_gso(const struct sock *sk)
1088 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1091 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1093 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1094 struct sk_buff *skb, struct page *page,
1097 if (skb->ip_summed == CHECKSUM_NONE) {
1099 __wsum csum = csum_and_copy_from_user(from,
1100 page_address(page) + off,
1104 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1105 } else if (copy_from_user(page_address(page) + off, from, copy))
1109 skb->data_len += copy;
1110 skb->truesize += copy;
1111 sk->sk_wmem_queued += copy;
1112 sk_mem_charge(sk, copy);
1117 * Queue a received datagram if it will fit. Stream and sequenced
1118 * protocols can't normally use this as they need to fit buffers in
1119 * and play with them.
1121 * Inlined as it's very short and called for pretty much every
1122 * packet ever received.
1125 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1129 skb->destructor = sock_wfree;
1130 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1133 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1136 skb->destructor = sock_rfree;
1137 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1138 sk_mem_charge(sk, skb->truesize);
1141 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1142 unsigned long expires);
1144 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1146 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1148 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1150 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1151 number of warnings when compiling with -W --ANK
1153 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1154 (unsigned)sk->sk_rcvbuf)
1156 skb_set_owner_r(skb, sk);
1157 skb_queue_tail(&sk->sk_error_queue, skb);
1158 if (!sock_flag(sk, SOCK_DEAD))
1159 sk->sk_data_ready(sk, skb->len);
1164 * Recover an error report and clear atomically
1167 static inline int sock_error(struct sock *sk)
1170 if (likely(!sk->sk_err))
1172 err = xchg(&sk->sk_err, 0);
1176 static inline unsigned long sock_wspace(struct sock *sk)
1180 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1181 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1188 static inline void sk_wake_async(struct sock *sk, int how, int band)
1190 if (sk->sk_socket && sk->sk_socket->fasync_list)
1191 sock_wake_async(sk->sk_socket, how, band);
1194 #define SOCK_MIN_SNDBUF 2048
1195 #define SOCK_MIN_RCVBUF 256
1197 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1199 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1200 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1201 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1205 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1207 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1209 struct page *page = NULL;
1211 page = alloc_pages(sk->sk_allocation, 0);
1213 sk->sk_prot->enter_memory_pressure(sk);
1214 sk_stream_moderate_sndbuf(sk);
1220 * Default write policy as shown to user space via poll/select/SIGIO
1222 static inline int sock_writeable(const struct sock *sk)
1224 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1227 static inline gfp_t gfp_any(void)
1229 return in_atomic() ? GFP_ATOMIC : GFP_KERNEL;
1232 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1234 return noblock ? 0 : sk->sk_rcvtimeo;
1237 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1239 return noblock ? 0 : sk->sk_sndtimeo;
1242 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1244 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1247 /* Alas, with timeout socket operations are not restartable.
1248 * Compare this to poll().
1250 static inline int sock_intr_errno(long timeo)
1252 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1255 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1256 struct sk_buff *skb);
1258 static __inline__ void
1259 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1261 ktime_t kt = skb->tstamp;
1263 if (sock_flag(sk, SOCK_RCVTSTAMP))
1264 __sock_recv_timestamp(msg, sk, skb);
1270 * sk_eat_skb - Release a skb if it is no longer needed
1271 * @sk: socket to eat this skb from
1272 * @skb: socket buffer to eat
1273 * @copied_early: flag indicating whether DMA operations copied this data early
1275 * This routine must be called with interrupts disabled or with the socket
1276 * locked so that the sk_buff queue operation is ok.
1278 #ifdef CONFIG_NET_DMA
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);
1285 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1288 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1290 __skb_unlink(skb, &sk->sk_receive_queue);
1296 struct net *sock_net(const struct sock *sk)
1298 #ifdef CONFIG_NET_NS
1306 void sock_net_set(struct sock *sk, struct net *net)
1308 #ifdef CONFIG_NET_NS
1314 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1315 * They should not hold a referrence to a namespace in order to allow
1317 * Sockets after sk_change_net should be released using sk_release_kernel
1319 static inline void sk_change_net(struct sock *sk, struct net *net)
1321 put_net(sock_net(sk));
1322 sock_net_set(sk, hold_net(net));
1325 extern void sock_enable_timestamp(struct sock *sk);
1326 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1327 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1330 * Enable debug/info messages
1332 extern int net_msg_warn;
1333 #define NETDEBUG(fmt, args...) \
1334 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1336 #define LIMIT_NETDEBUG(fmt, args...) \
1337 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1339 extern __u32 sysctl_wmem_max;
1340 extern __u32 sysctl_rmem_max;
1342 extern void sk_init(void);
1344 extern int sysctl_optmem_max;
1346 extern __u32 sysctl_wmem_default;
1347 extern __u32 sysctl_rmem_default;
1349 #endif /* _SOCK_H */