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 #define SOCK_DEBUG(sk, msg...) do { } while (0)
75 /* This is the per-socket lock. The spinlock provides a synchronization
76 * between user contexts and software interrupt processing, whereas the
77 * mini-semaphore synchronizes multiple users amongst themselves.
82 struct sock_iocb *owner;
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;
99 * struct sock_common - minimal network layer representation of sockets
100 * @skc_family: network address family
101 * @skc_state: Connection state
102 * @skc_reuse: %SO_REUSEADDR setting
103 * @skc_bound_dev_if: bound device index if != 0
104 * @skc_node: main hash linkage for various protocol lookup tables
105 * @skc_bind_node: bind hash linkage for various protocol lookup tables
106 * @skc_refcnt: reference count
107 * @skc_hash: hash value used with various protocol lookup tables
108 * @skc_prot: protocol handlers inside a network family
110 * This is the minimal network layer representation of sockets, the header
111 * for struct sock and struct inet_timewait_sock.
114 unsigned short skc_family;
115 volatile unsigned char skc_state;
116 unsigned char skc_reuse;
117 int skc_bound_dev_if;
118 struct hlist_node skc_node;
119 struct hlist_node skc_bind_node;
121 unsigned int skc_hash;
122 struct proto *skc_prot;
126 * struct sock - network layer representation of sockets
127 * @__sk_common: shared layout with inet_timewait_sock
128 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
129 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
130 * @sk_lock: synchronizer
131 * @sk_rcvbuf: size of receive buffer in bytes
132 * @sk_sleep: sock wait queue
133 * @sk_dst_cache: destination cache
134 * @sk_dst_lock: destination cache lock
135 * @sk_policy: flow policy
136 * @sk_rmem_alloc: receive queue bytes committed
137 * @sk_receive_queue: incoming packets
138 * @sk_wmem_alloc: transmit queue bytes committed
139 * @sk_write_queue: Packet sending queue
140 * @sk_async_wait_queue: DMA copied packets
141 * @sk_omem_alloc: "o" is "option" or "other"
142 * @sk_wmem_queued: persistent queue size
143 * @sk_forward_alloc: space allocated forward
144 * @sk_allocation: allocation mode
145 * @sk_sndbuf: size of send buffer in bytes
146 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
147 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
148 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
149 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
150 * @sk_lingertime: %SO_LINGER l_linger setting
151 * @sk_backlog: always used with the per-socket spinlock held
152 * @sk_callback_lock: used with the callbacks in the end of this struct
153 * @sk_error_queue: rarely used
154 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
155 * @sk_err: last error
156 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
157 * @sk_ack_backlog: current listen backlog
158 * @sk_max_ack_backlog: listen backlog set in listen()
159 * @sk_priority: %SO_PRIORITY setting
160 * @sk_type: socket type (%SOCK_STREAM, etc)
161 * @sk_protocol: which protocol this socket belongs in this network family
162 * @sk_peercred: %SO_PEERCRED setting
163 * @sk_rcvlowat: %SO_RCVLOWAT setting
164 * @sk_rcvtimeo: %SO_RCVTIMEO setting
165 * @sk_sndtimeo: %SO_SNDTIMEO setting
166 * @sk_filter: socket filtering instructions
167 * @sk_protinfo: private area, net family specific, when not using slab
168 * @sk_timer: sock cleanup timer
169 * @sk_stamp: time stamp of last packet received
170 * @sk_socket: Identd and reporting IO signals
171 * @sk_user_data: RPC layer private data
172 * @sk_sndmsg_page: cached page for sendmsg
173 * @sk_sndmsg_off: cached offset for sendmsg
174 * @sk_send_head: front of stuff to transmit
175 * @sk_security: used by security modules
176 * @sk_write_pending: a write to stream socket waits to start
177 * @sk_state_change: callback to indicate change in the state of the sock
178 * @sk_data_ready: callback to indicate there is data to be processed
179 * @sk_write_space: callback to indicate there is bf sending space available
180 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
181 * @sk_backlog_rcv: callback to process the backlog
182 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
186 * Now struct inet_timewait_sock also uses sock_common, so please just
187 * don't add nothing before this first member (__sk_common) --acme
189 struct sock_common __sk_common;
190 #define sk_family __sk_common.skc_family
191 #define sk_state __sk_common.skc_state
192 #define sk_reuse __sk_common.skc_reuse
193 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
194 #define sk_node __sk_common.skc_node
195 #define sk_bind_node __sk_common.skc_bind_node
196 #define sk_refcnt __sk_common.skc_refcnt
197 #define sk_hash __sk_common.skc_hash
198 #define sk_prot __sk_common.skc_prot
199 unsigned char sk_shutdown : 2,
202 unsigned char sk_protocol;
203 unsigned short sk_type;
205 socket_lock_t sk_lock;
207 * The backlog queue is special, it is always used with
208 * the per-socket spinlock held and requires low latency
209 * access. Therefore we special case it's implementation.
212 struct sk_buff *head;
213 struct sk_buff *tail;
215 wait_queue_head_t *sk_sleep;
216 struct dst_entry *sk_dst_cache;
217 struct xfrm_policy *sk_policy[2];
218 rwlock_t sk_dst_lock;
219 atomic_t sk_rmem_alloc;
220 atomic_t sk_wmem_alloc;
221 atomic_t sk_omem_alloc;
223 struct sk_buff_head sk_receive_queue;
224 struct sk_buff_head sk_write_queue;
225 struct sk_buff_head sk_async_wait_queue;
227 int sk_forward_alloc;
232 unsigned long sk_flags;
233 unsigned long sk_lingertime;
234 struct sk_buff_head sk_error_queue;
235 struct proto *sk_prot_creator;
236 rwlock_t sk_callback_lock;
239 unsigned short sk_ack_backlog;
240 unsigned short sk_max_ack_backlog;
242 struct ucred sk_peercred;
245 struct sk_filter *sk_filter;
247 struct timer_list sk_timer;
249 struct socket *sk_socket;
251 struct page *sk_sndmsg_page;
252 struct sk_buff *sk_send_head;
254 int sk_write_pending;
256 void (*sk_state_change)(struct sock *sk);
257 void (*sk_data_ready)(struct sock *sk, int bytes);
258 void (*sk_write_space)(struct sock *sk);
259 void (*sk_error_report)(struct sock *sk);
260 int (*sk_backlog_rcv)(struct sock *sk,
261 struct sk_buff *skb);
262 void (*sk_destruct)(struct sock *sk);
266 * Hashed lists helper routines
268 static inline struct sock *__sk_head(const struct hlist_head *head)
270 return hlist_entry(head->first, struct sock, sk_node);
273 static inline struct sock *sk_head(const struct hlist_head *head)
275 return hlist_empty(head) ? NULL : __sk_head(head);
278 static inline struct sock *sk_next(const struct sock *sk)
280 return sk->sk_node.next ?
281 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
284 static inline int sk_unhashed(const struct sock *sk)
286 return hlist_unhashed(&sk->sk_node);
289 static inline int sk_hashed(const struct sock *sk)
291 return !sk_unhashed(sk);
294 static __inline__ void sk_node_init(struct hlist_node *node)
299 static __inline__ void __sk_del_node(struct sock *sk)
301 __hlist_del(&sk->sk_node);
304 static __inline__ int __sk_del_node_init(struct sock *sk)
308 sk_node_init(&sk->sk_node);
314 /* Grab socket reference count. This operation is valid only
315 when sk is ALREADY grabbed f.e. it is found in hash table
316 or a list and the lookup is made under lock preventing hash table
320 static inline void sock_hold(struct sock *sk)
322 atomic_inc(&sk->sk_refcnt);
325 /* Ungrab socket in the context, which assumes that socket refcnt
326 cannot hit zero, f.e. it is true in context of any socketcall.
328 static inline void __sock_put(struct sock *sk)
330 atomic_dec(&sk->sk_refcnt);
333 static __inline__ int sk_del_node_init(struct sock *sk)
335 int rc = __sk_del_node_init(sk);
338 /* paranoid for a while -acme */
339 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
345 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
347 hlist_add_head(&sk->sk_node, list);
350 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
353 __sk_add_node(sk, list);
356 static __inline__ void __sk_del_bind_node(struct sock *sk)
358 __hlist_del(&sk->sk_bind_node);
361 static __inline__ void sk_add_bind_node(struct sock *sk,
362 struct hlist_head *list)
364 hlist_add_head(&sk->sk_bind_node, list);
367 #define sk_for_each(__sk, node, list) \
368 hlist_for_each_entry(__sk, node, list, sk_node)
369 #define sk_for_each_from(__sk, node) \
370 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
371 hlist_for_each_entry_from(__sk, node, sk_node)
372 #define sk_for_each_continue(__sk, node) \
373 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
374 hlist_for_each_entry_continue(__sk, node, sk_node)
375 #define sk_for_each_safe(__sk, node, tmp, list) \
376 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
377 #define sk_for_each_bound(__sk, node, list) \
378 hlist_for_each_entry(__sk, node, list, sk_bind_node)
391 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
392 SOCK_DBG, /* %SO_DEBUG setting */
393 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
394 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
395 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
396 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
399 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
401 nsk->sk_flags = osk->sk_flags;
404 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
406 __set_bit(flag, &sk->sk_flags);
409 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
411 __clear_bit(flag, &sk->sk_flags);
414 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
416 return test_bit(flag, &sk->sk_flags);
419 static inline void sk_acceptq_removed(struct sock *sk)
421 sk->sk_ack_backlog--;
424 static inline void sk_acceptq_added(struct sock *sk)
426 sk->sk_ack_backlog++;
429 static inline int sk_acceptq_is_full(struct sock *sk)
431 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
435 * Compute minimal free write space needed to queue new packets.
437 static inline int sk_stream_min_wspace(struct sock *sk)
439 return sk->sk_wmem_queued / 2;
442 static inline int sk_stream_wspace(struct sock *sk)
444 return sk->sk_sndbuf - sk->sk_wmem_queued;
447 extern void sk_stream_write_space(struct sock *sk);
449 static inline int sk_stream_memory_free(struct sock *sk)
451 return sk->sk_wmem_queued < sk->sk_sndbuf;
454 extern void sk_stream_rfree(struct sk_buff *skb);
456 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
459 skb->destructor = sk_stream_rfree;
460 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
461 sk->sk_forward_alloc -= skb->truesize;
464 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
466 skb_truesize_check(skb);
467 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
468 sk->sk_wmem_queued -= skb->truesize;
469 sk->sk_forward_alloc += skb->truesize;
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); \
490 *(__timeo) = schedule_timeout(*(__timeo)); \
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;
508 /* Networking protocol blocks we attach to sockets.
509 * socket layer -> transport layer interface
510 * transport -> network interface is defined by struct inet_proto
513 void (*close)(struct sock *sk,
515 int (*connect)(struct sock *sk,
516 struct sockaddr *uaddr,
518 int (*disconnect)(struct sock *sk, int flags);
520 struct sock * (*accept) (struct sock *sk, int flags, int *err);
522 int (*ioctl)(struct sock *sk, int cmd,
524 int (*init)(struct sock *sk);
525 int (*destroy)(struct sock *sk);
526 void (*shutdown)(struct sock *sk, int how);
527 int (*setsockopt)(struct sock *sk, int level,
528 int optname, char __user *optval,
530 int (*getsockopt)(struct sock *sk, int level,
531 int optname, char __user *optval,
533 int (*compat_setsockopt)(struct sock *sk,
535 int optname, char __user *optval,
537 int (*compat_getsockopt)(struct sock *sk,
539 int optname, char __user *optval,
541 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
542 struct msghdr *msg, size_t len);
543 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
545 size_t len, int noblock, int flags,
547 int (*sendpage)(struct sock *sk, struct page *page,
548 int offset, size_t size, int flags);
549 int (*bind)(struct sock *sk,
550 struct sockaddr *uaddr, int addr_len);
552 int (*backlog_rcv) (struct sock *sk,
553 struct sk_buff *skb);
555 /* Keeping track of sk's, looking them up, and port selection methods. */
556 void (*hash)(struct sock *sk);
557 void (*unhash)(struct sock *sk);
558 int (*get_port)(struct sock *sk, unsigned short snum);
560 /* Memory pressure */
561 void (*enter_memory_pressure)(void);
562 atomic_t *memory_allocated; /* Current allocated memory. */
563 atomic_t *sockets_allocated; /* Current number of sockets. */
565 * Pressure flag: try to collapse.
566 * Technical note: it is used by multiple contexts non atomically.
567 * All the sk_stream_mem_schedule() is of this nature: accounting
568 * is strict, actions are advisory and have some latency.
570 int *memory_pressure;
576 struct kmem_cache *slab;
577 unsigned int obj_size;
579 atomic_t *orphan_count;
581 struct request_sock_ops *rsk_prot;
582 struct timewait_sock_ops *twsk_prot;
584 struct module *owner;
588 struct list_head node;
589 #ifdef SOCK_REFCNT_DEBUG
594 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
598 extern int proto_register(struct proto *prot, int alloc_slab);
599 extern void proto_unregister(struct proto *prot);
601 #ifdef SOCK_REFCNT_DEBUG
602 static inline void sk_refcnt_debug_inc(struct sock *sk)
604 atomic_inc(&sk->sk_prot->socks);
607 static inline void sk_refcnt_debug_dec(struct sock *sk)
609 atomic_dec(&sk->sk_prot->socks);
610 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
611 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
614 static inline void sk_refcnt_debug_release(const struct sock *sk)
616 if (atomic_read(&sk->sk_refcnt) != 1)
617 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
618 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
620 #else /* SOCK_REFCNT_DEBUG */
621 #define sk_refcnt_debug_inc(sk) do { } while (0)
622 #define sk_refcnt_debug_dec(sk) do { } while (0)
623 #define sk_refcnt_debug_release(sk) do { } while (0)
624 #endif /* SOCK_REFCNT_DEBUG */
626 /* Called with local bh disabled */
627 static __inline__ void sock_prot_inc_use(struct proto *prot)
629 prot->stats[smp_processor_id()].inuse++;
632 static __inline__ void sock_prot_dec_use(struct proto *prot)
634 prot->stats[smp_processor_id()].inuse--;
637 /* With per-bucket locks this operation is not-atomic, so that
638 * this version is not worse.
640 static inline void __sk_prot_rehash(struct sock *sk)
642 sk->sk_prot->unhash(sk);
643 sk->sk_prot->hash(sk);
646 /* About 10 seconds */
647 #define SOCK_DESTROY_TIME (10*HZ)
649 /* Sockets 0-1023 can't be bound to unless you are superuser */
650 #define PROT_SOCK 1024
652 #define SHUTDOWN_MASK 3
653 #define RCV_SHUTDOWN 1
654 #define SEND_SHUTDOWN 2
656 #define SOCK_SNDBUF_LOCK 1
657 #define SOCK_RCVBUF_LOCK 2
658 #define SOCK_BINDADDR_LOCK 4
659 #define SOCK_BINDPORT_LOCK 8
661 /* sock_iocb: used to kick off async processing of socket ios */
663 struct list_head list;
669 struct scm_cookie *scm;
670 struct msghdr *msg, async_msg;
674 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
676 return (struct sock_iocb *)iocb->private;
679 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
684 struct socket_alloc {
685 struct socket socket;
686 struct inode vfs_inode;
689 static inline struct socket *SOCKET_I(struct inode *inode)
691 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
694 static inline struct inode *SOCK_INODE(struct socket *socket)
696 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
699 extern void __sk_stream_mem_reclaim(struct sock *sk);
700 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
702 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
704 static inline int sk_stream_pages(int amt)
706 return DIV_ROUND_UP(amt, SK_STREAM_MEM_QUANTUM);
709 static inline void sk_stream_mem_reclaim(struct sock *sk)
711 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
712 __sk_stream_mem_reclaim(sk);
715 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
717 return (int)skb->truesize <= sk->sk_forward_alloc ||
718 sk_stream_mem_schedule(sk, skb->truesize, 1);
721 static inline int sk_stream_wmem_schedule(struct sock *sk, int size)
723 return size <= sk->sk_forward_alloc ||
724 sk_stream_mem_schedule(sk, size, 0);
727 /* Used by processes to "lock" a socket state, so that
728 * interrupts and bottom half handlers won't change it
729 * from under us. It essentially blocks any incoming
730 * packets, so that we won't get any new data or any
731 * packets that change the state of the socket.
733 * While locked, BH processing will add new packets to
734 * the backlog queue. This queue is processed by the
735 * owner of the socket lock right before it is released.
737 * Since ~2.3.5 it is also exclusive sleep lock serializing
738 * accesses from user process context.
740 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
743 * Macro so as to not evaluate some arguments when
744 * lockdep is not enabled.
746 * Mark both the sk_lock and the sk_lock.slock as a
747 * per-address-family lock class.
749 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
751 sk->sk_lock.owner = NULL; \
752 init_waitqueue_head(&sk->sk_lock.wq); \
753 spin_lock_init(&(sk)->sk_lock.slock); \
754 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
755 sizeof((sk)->sk_lock)); \
756 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
758 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
761 extern void FASTCALL(lock_sock_nested(struct sock *sk, int subclass));
763 static inline void lock_sock(struct sock *sk)
765 lock_sock_nested(sk, 0);
768 extern void FASTCALL(release_sock(struct sock *sk));
770 /* BH context may only use the following locking interface. */
771 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
772 #define bh_lock_sock_nested(__sk) \
773 spin_lock_nested(&((__sk)->sk_lock.slock), \
774 SINGLE_DEPTH_NESTING)
775 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
777 extern struct sock *sk_alloc(int family,
779 struct proto *prot, int zero_it);
780 extern void sk_free(struct sock *sk);
781 extern struct sock *sk_clone(const struct sock *sk,
782 const gfp_t priority);
784 extern struct sk_buff *sock_wmalloc(struct sock *sk,
785 unsigned long size, int force,
787 extern struct sk_buff *sock_rmalloc(struct sock *sk,
788 unsigned long size, int force,
790 extern void sock_wfree(struct sk_buff *skb);
791 extern void sock_rfree(struct sk_buff *skb);
793 extern int sock_setsockopt(struct socket *sock, int level,
794 int op, char __user *optval,
797 extern int sock_getsockopt(struct socket *sock, int level,
798 int op, char __user *optval,
800 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
804 extern void *sock_kmalloc(struct sock *sk, int size,
806 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
807 extern void sk_send_sigurg(struct sock *sk);
810 * Functions to fill in entries in struct proto_ops when a protocol
811 * does not implement a particular function.
813 extern int sock_no_bind(struct socket *,
814 struct sockaddr *, int);
815 extern int sock_no_connect(struct socket *,
816 struct sockaddr *, int, int);
817 extern int sock_no_socketpair(struct socket *,
819 extern int sock_no_accept(struct socket *,
820 struct socket *, int);
821 extern int sock_no_getname(struct socket *,
822 struct sockaddr *, int *, int);
823 extern unsigned int sock_no_poll(struct file *, struct socket *,
824 struct poll_table_struct *);
825 extern int sock_no_ioctl(struct socket *, unsigned int,
827 extern int sock_no_listen(struct socket *, int);
828 extern int sock_no_shutdown(struct socket *, int);
829 extern int sock_no_getsockopt(struct socket *, int , int,
830 char __user *, int __user *);
831 extern int sock_no_setsockopt(struct socket *, int, int,
833 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
834 struct msghdr *, size_t);
835 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
836 struct msghdr *, size_t, int);
837 extern int sock_no_mmap(struct file *file,
839 struct vm_area_struct *vma);
840 extern ssize_t sock_no_sendpage(struct socket *sock,
842 int offset, size_t size,
846 * Functions to fill in entries in struct proto_ops when a protocol
847 * uses the inet style.
849 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
850 char __user *optval, int __user *optlen);
851 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
852 struct msghdr *msg, size_t size, int flags);
853 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
854 char __user *optval, int optlen);
855 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
856 int optname, char __user *optval, int __user *optlen);
857 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
858 int optname, char __user *optval, int optlen);
860 extern void sk_common_release(struct sock *sk);
863 * Default socket callbacks and setup code
866 /* Initialise core socket variables */
867 extern void sock_init_data(struct socket *sock, struct sock *sk);
870 * sk_filter - run a packet through a socket filter
871 * @sk: sock associated with &sk_buff
872 * @skb: buffer to filter
873 * @needlock: set to 1 if the sock is not locked by caller.
875 * Run the filter code and then cut skb->data to correct size returned by
876 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
877 * than pkt_len we keep whole skb->data. This is the socket level
878 * wrapper to sk_run_filter. It returns 0 if the packet should
879 * be accepted or -EPERM if the packet should be tossed.
883 static inline int sk_filter(struct sock *sk, struct sk_buff *skb)
886 struct sk_filter *filter;
888 err = security_sock_rcv_skb(sk, skb);
893 filter = sk->sk_filter;
895 unsigned int pkt_len = sk_run_filter(skb, filter->insns,
897 err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM;
899 rcu_read_unlock_bh();
905 * sk_filter_rcu_free: Free a socket filter
906 * @rcu: rcu_head that contains the sk_filter to free
908 static inline void sk_filter_rcu_free(struct rcu_head *rcu)
910 struct sk_filter *fp = container_of(rcu, struct sk_filter, rcu);
915 * sk_filter_release: Release a socket filter
917 * @fp: filter to remove
919 * Remove a filter from a socket and release its resources.
922 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
924 unsigned int size = sk_filter_len(fp);
926 atomic_sub(size, &sk->sk_omem_alloc);
928 if (atomic_dec_and_test(&fp->refcnt))
929 call_rcu_bh(&fp->rcu, sk_filter_rcu_free);
932 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
934 atomic_inc(&fp->refcnt);
935 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
939 * Socket reference counting postulates.
941 * * Each user of socket SHOULD hold a reference count.
942 * * Each access point to socket (an hash table bucket, reference from a list,
943 * running timer, skb in flight MUST hold a reference count.
944 * * When reference count hits 0, it means it will never increase back.
945 * * When reference count hits 0, it means that no references from
946 * outside exist to this socket and current process on current CPU
947 * is last user and may/should destroy this socket.
948 * * sk_free is called from any context: process, BH, IRQ. When
949 * it is called, socket has no references from outside -> sk_free
950 * may release descendant resources allocated by the socket, but
951 * to the time when it is called, socket is NOT referenced by any
952 * hash tables, lists etc.
953 * * Packets, delivered from outside (from network or from another process)
954 * and enqueued on receive/error queues SHOULD NOT grab reference count,
955 * when they sit in queue. Otherwise, packets will leak to hole, when
956 * socket is looked up by one cpu and unhasing is made by another CPU.
957 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
958 * (leak to backlog). Packet socket does all the processing inside
959 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
960 * use separate SMP lock, so that they are prone too.
963 /* Ungrab socket and destroy it, if it was the last reference. */
964 static inline void sock_put(struct sock *sk)
966 if (atomic_dec_and_test(&sk->sk_refcnt))
970 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
973 /* Detach socket from process context.
974 * Announce socket dead, detach it from wait queue and inode.
975 * Note that parent inode held reference count on this struct sock,
976 * we do not release it in this function, because protocol
977 * probably wants some additional cleanups or even continuing
978 * to work with this socket (TCP).
980 static inline void sock_orphan(struct sock *sk)
982 write_lock_bh(&sk->sk_callback_lock);
983 sock_set_flag(sk, SOCK_DEAD);
984 sk->sk_socket = NULL;
986 write_unlock_bh(&sk->sk_callback_lock);
989 static inline void sock_graft(struct sock *sk, struct socket *parent)
991 write_lock_bh(&sk->sk_callback_lock);
992 sk->sk_sleep = &parent->wait;
994 sk->sk_socket = parent;
995 security_sock_graft(sk, parent);
996 write_unlock_bh(&sk->sk_callback_lock);
999 static inline void sock_copy(struct sock *nsk, const struct sock *osk)
1001 #ifdef CONFIG_SECURITY_NETWORK
1002 void *sptr = nsk->sk_security;
1005 memcpy(nsk, osk, osk->sk_prot->obj_size);
1006 #ifdef CONFIG_SECURITY_NETWORK
1007 nsk->sk_security = sptr;
1008 security_sk_clone(osk, nsk);
1012 extern int sock_i_uid(struct sock *sk);
1013 extern unsigned long sock_i_ino(struct sock *sk);
1015 static inline struct dst_entry *
1016 __sk_dst_get(struct sock *sk)
1018 return sk->sk_dst_cache;
1021 static inline struct dst_entry *
1022 sk_dst_get(struct sock *sk)
1024 struct dst_entry *dst;
1026 read_lock(&sk->sk_dst_lock);
1027 dst = sk->sk_dst_cache;
1030 read_unlock(&sk->sk_dst_lock);
1035 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1037 struct dst_entry *old_dst;
1039 old_dst = sk->sk_dst_cache;
1040 sk->sk_dst_cache = dst;
1041 dst_release(old_dst);
1045 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1047 write_lock(&sk->sk_dst_lock);
1048 __sk_dst_set(sk, dst);
1049 write_unlock(&sk->sk_dst_lock);
1053 __sk_dst_reset(struct sock *sk)
1055 struct dst_entry *old_dst;
1057 old_dst = sk->sk_dst_cache;
1058 sk->sk_dst_cache = NULL;
1059 dst_release(old_dst);
1063 sk_dst_reset(struct sock *sk)
1065 write_lock(&sk->sk_dst_lock);
1067 write_unlock(&sk->sk_dst_lock);
1070 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1072 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1074 static inline int sk_can_gso(const struct sock *sk)
1076 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1079 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1081 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1083 sk->sk_wmem_queued += skb->truesize;
1084 sk->sk_forward_alloc -= skb->truesize;
1087 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1088 struct sk_buff *skb, struct page *page,
1091 if (skb->ip_summed == CHECKSUM_NONE) {
1093 __wsum csum = csum_and_copy_from_user(from,
1094 page_address(page) + off,
1098 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1099 } else if (copy_from_user(page_address(page) + off, from, copy))
1103 skb->data_len += copy;
1104 skb->truesize += copy;
1105 sk->sk_wmem_queued += copy;
1106 sk->sk_forward_alloc -= copy;
1111 * Queue a received datagram if it will fit. Stream and sequenced
1112 * protocols can't normally use this as they need to fit buffers in
1113 * and play with them.
1115 * Inlined as it's very short and called for pretty much every
1116 * packet ever received.
1119 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1123 skb->destructor = sock_wfree;
1124 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1127 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1130 skb->destructor = sock_rfree;
1131 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1134 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1135 unsigned long expires);
1137 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1139 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1141 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1143 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1144 number of warnings when compiling with -W --ANK
1146 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1147 (unsigned)sk->sk_rcvbuf)
1149 skb_set_owner_r(skb, sk);
1150 skb_queue_tail(&sk->sk_error_queue, skb);
1151 if (!sock_flag(sk, SOCK_DEAD))
1152 sk->sk_data_ready(sk, skb->len);
1157 * Recover an error report and clear atomically
1160 static inline int sock_error(struct sock *sk)
1163 if (likely(!sk->sk_err))
1165 err = xchg(&sk->sk_err, 0);
1169 static inline unsigned long sock_wspace(struct sock *sk)
1173 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1174 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1181 static inline void sk_wake_async(struct sock *sk, int how, int band)
1183 if (sk->sk_socket && sk->sk_socket->fasync_list)
1184 sock_wake_async(sk->sk_socket, how, band);
1187 #define SOCK_MIN_SNDBUF 2048
1188 #define SOCK_MIN_RCVBUF 256
1190 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1192 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1193 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1194 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1198 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1202 struct sk_buff *skb;
1205 hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header);
1206 skb = alloc_skb_fclone(size + hdr_len, gfp);
1208 skb->truesize += mem;
1209 if (sk_stream_wmem_schedule(sk, skb->truesize)) {
1210 skb_reserve(skb, hdr_len);
1215 sk->sk_prot->enter_memory_pressure();
1216 sk_stream_moderate_sndbuf(sk);
1221 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1225 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1228 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1230 struct page *page = NULL;
1232 page = alloc_pages(sk->sk_allocation, 0);
1234 sk->sk_prot->enter_memory_pressure();
1235 sk_stream_moderate_sndbuf(sk);
1241 * Default write policy as shown to user space via poll/select/SIGIO
1243 static inline int sock_writeable(const struct sock *sk)
1245 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1248 static inline gfp_t gfp_any(void)
1250 return in_atomic() ? GFP_ATOMIC : GFP_KERNEL;
1253 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1255 return noblock ? 0 : sk->sk_rcvtimeo;
1258 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1260 return noblock ? 0 : sk->sk_sndtimeo;
1263 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1265 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1268 /* Alas, with timeout socket operations are not restartable.
1269 * Compare this to poll().
1271 static inline int sock_intr_errno(long timeo)
1273 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1276 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1277 struct sk_buff *skb);
1279 static __inline__ void
1280 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1282 ktime_t kt = skb->tstamp;
1284 if (sock_flag(sk, SOCK_RCVTSTAMP))
1285 __sock_recv_timestamp(msg, sk, skb);
1291 * sk_eat_skb - Release a skb if it is no longer needed
1292 * @sk: socket to eat this skb from
1293 * @skb: socket buffer to eat
1294 * @copied_early: flag indicating whether DMA operations copied this data early
1296 * This routine must be called with interrupts disabled or with the socket
1297 * locked so that the sk_buff queue operation is ok.
1299 #ifdef CONFIG_NET_DMA
1300 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1302 __skb_unlink(skb, &sk->sk_receive_queue);
1306 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1309 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1311 __skb_unlink(skb, &sk->sk_receive_queue);
1316 extern void sock_enable_timestamp(struct sock *sk);
1317 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1318 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1321 * Enable debug/info messages
1323 extern int net_msg_warn;
1324 #define NETDEBUG(fmt, args...) \
1325 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1327 #define LIMIT_NETDEBUG(fmt, args...) \
1328 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1331 * Macros for sleeping on a socket. Use them like this:
1333 * SOCK_SLEEP_PRE(sk)
1336 * SOCK_SLEEP_POST(sk)
1338 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1339 * and when the last use of them in DECnet has gone, I'm intending to
1343 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1344 DECLARE_WAITQUEUE(wait, tsk); \
1345 tsk->state = TASK_INTERRUPTIBLE; \
1346 add_wait_queue((sk)->sk_sleep, &wait); \
1349 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1350 remove_wait_queue((sk)->sk_sleep, &wait); \
1354 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1357 sock_set_flag(sk, bit);
1359 sock_reset_flag(sk, bit);
1362 extern __u32 sysctl_wmem_max;
1363 extern __u32 sysctl_rmem_max;
1365 extern void sk_init(void);
1367 #ifdef CONFIG_SYSCTL
1368 extern struct ctl_table core_table[];
1371 extern int sysctl_optmem_max;
1373 extern __u32 sysctl_wmem_default;
1374 extern __u32 sysctl_rmem_default;
1376 #endif /* _SOCK_H */