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/config.h>
44 #include <linux/list.h>
45 #include <linux/timer.h>
46 #include <linux/cache.h>
47 #include <linux/module.h>
48 #include <linux/netdevice.h>
49 #include <linux/skbuff.h> /* struct sk_buff */
50 #include <linux/security.h>
52 #include <linux/filter.h>
54 #include <asm/atomic.h>
56 #include <net/checksum.h>
59 * This structure really needs to be cleaned up.
60 * Most of it is for TCP, and not used by any of
61 * the other protocols.
64 /* Define this to get the SOCK_DBG debugging facility. */
65 #define SOCK_DEBUGGING
67 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
68 printk(KERN_DEBUG msg); } while (0)
70 #define SOCK_DEBUG(sk, msg...) do { } while (0)
73 /* This is the per-socket lock. The spinlock provides a synchronization
74 * between user contexts and software interrupt processing, whereas the
75 * mini-semaphore synchronizes multiple users amongst themselves.
80 struct sock_iocb *owner;
84 #define sock_lock_init(__sk) \
85 do { spin_lock_init(&((__sk)->sk_lock.slock)); \
86 (__sk)->sk_lock.owner = NULL; \
87 init_waitqueue_head(&((__sk)->sk_lock.wq)); \
94 * struct sock_common - minimal network layer representation of sockets
95 * @skc_family: network address family
96 * @skc_state: Connection state
97 * @skc_reuse: %SO_REUSEADDR setting
98 * @skc_bound_dev_if: bound device index if != 0
99 * @skc_node: main hash linkage for various protocol lookup tables
100 * @skc_bind_node: bind hash linkage for various protocol lookup tables
101 * @skc_refcnt: reference count
102 * @skc_hash: hash value used with various protocol lookup tables
103 * @skc_prot: protocol handlers inside a network family
105 * This is the minimal network layer representation of sockets, the header
106 * for struct sock and struct inet_timewait_sock.
109 unsigned short skc_family;
110 volatile unsigned char skc_state;
111 unsigned char skc_reuse;
112 int skc_bound_dev_if;
113 struct hlist_node skc_node;
114 struct hlist_node skc_bind_node;
116 unsigned int skc_hash;
117 struct proto *skc_prot;
121 * struct sock - network layer representation of sockets
122 * @__sk_common: shared layout with inet_timewait_sock
123 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
124 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
125 * @sk_lock: synchronizer
126 * @sk_rcvbuf: size of receive buffer in bytes
127 * @sk_sleep: sock wait queue
128 * @sk_dst_cache: destination cache
129 * @sk_dst_lock: destination cache lock
130 * @sk_policy: flow policy
131 * @sk_rmem_alloc: receive queue bytes committed
132 * @sk_receive_queue: incoming packets
133 * @sk_wmem_alloc: transmit queue bytes committed
134 * @sk_write_queue: Packet sending queue
135 * @sk_omem_alloc: "o" is "option" or "other"
136 * @sk_wmem_queued: persistent queue size
137 * @sk_forward_alloc: space allocated forward
138 * @sk_allocation: allocation mode
139 * @sk_sndbuf: size of send buffer in bytes
140 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
141 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
142 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
143 * @sk_lingertime: %SO_LINGER l_linger setting
144 * @sk_backlog: always used with the per-socket spinlock held
145 * @sk_callback_lock: used with the callbacks in the end of this struct
146 * @sk_error_queue: rarely used
147 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
148 * @sk_err: last error
149 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
150 * @sk_ack_backlog: current listen backlog
151 * @sk_max_ack_backlog: listen backlog set in listen()
152 * @sk_priority: %SO_PRIORITY setting
153 * @sk_type: socket type (%SOCK_STREAM, etc)
154 * @sk_protocol: which protocol this socket belongs in this network family
155 * @sk_peercred: %SO_PEERCRED setting
156 * @sk_rcvlowat: %SO_RCVLOWAT setting
157 * @sk_rcvtimeo: %SO_RCVTIMEO setting
158 * @sk_sndtimeo: %SO_SNDTIMEO setting
159 * @sk_filter: socket filtering instructions
160 * @sk_protinfo: private area, net family specific, when not using slab
161 * @sk_timer: sock cleanup timer
162 * @sk_stamp: time stamp of last packet received
163 * @sk_socket: Identd and reporting IO signals
164 * @sk_user_data: RPC layer private data
165 * @sk_sndmsg_page: cached page for sendmsg
166 * @sk_sndmsg_off: cached offset for sendmsg
167 * @sk_send_head: front of stuff to transmit
168 * @sk_security: used by security modules
169 * @sk_write_pending: a write to stream socket waits to start
170 * @sk_state_change: callback to indicate change in the state of the sock
171 * @sk_data_ready: callback to indicate there is data to be processed
172 * @sk_write_space: callback to indicate there is bf sending space available
173 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
174 * @sk_backlog_rcv: callback to process the backlog
175 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
179 * Now struct inet_timewait_sock also uses sock_common, so please just
180 * don't add nothing before this first member (__sk_common) --acme
182 struct sock_common __sk_common;
183 #define sk_family __sk_common.skc_family
184 #define sk_state __sk_common.skc_state
185 #define sk_reuse __sk_common.skc_reuse
186 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
187 #define sk_node __sk_common.skc_node
188 #define sk_bind_node __sk_common.skc_bind_node
189 #define sk_refcnt __sk_common.skc_refcnt
190 #define sk_hash __sk_common.skc_hash
191 #define sk_prot __sk_common.skc_prot
192 unsigned char sk_shutdown : 2,
195 unsigned char sk_protocol;
196 unsigned short sk_type;
198 socket_lock_t sk_lock;
199 wait_queue_head_t *sk_sleep;
200 struct dst_entry *sk_dst_cache;
201 struct xfrm_policy *sk_policy[2];
202 rwlock_t sk_dst_lock;
203 atomic_t sk_rmem_alloc;
204 atomic_t sk_wmem_alloc;
205 atomic_t sk_omem_alloc;
206 struct sk_buff_head sk_receive_queue;
207 struct sk_buff_head sk_write_queue;
209 int sk_forward_alloc;
213 unsigned long sk_flags;
214 unsigned long sk_lingertime;
216 * The backlog queue is special, it is always used with
217 * the per-socket spinlock held and requires low latency
218 * access. Therefore we special case it's implementation.
221 struct sk_buff *head;
222 struct sk_buff *tail;
224 struct sk_buff_head sk_error_queue;
225 struct proto *sk_prot_creator;
226 rwlock_t sk_callback_lock;
229 unsigned short sk_ack_backlog;
230 unsigned short sk_max_ack_backlog;
232 struct ucred sk_peercred;
236 struct sk_filter *sk_filter;
238 struct timer_list sk_timer;
239 struct timeval sk_stamp;
240 struct socket *sk_socket;
242 struct page *sk_sndmsg_page;
243 struct sk_buff *sk_send_head;
245 int sk_write_pending;
247 void (*sk_state_change)(struct sock *sk);
248 void (*sk_data_ready)(struct sock *sk, int bytes);
249 void (*sk_write_space)(struct sock *sk);
250 void (*sk_error_report)(struct sock *sk);
251 int (*sk_backlog_rcv)(struct sock *sk,
252 struct sk_buff *skb);
253 void (*sk_destruct)(struct sock *sk);
257 * Hashed lists helper routines
259 static inline struct sock *__sk_head(const struct hlist_head *head)
261 return hlist_entry(head->first, struct sock, sk_node);
264 static inline struct sock *sk_head(const struct hlist_head *head)
266 return hlist_empty(head) ? NULL : __sk_head(head);
269 static inline struct sock *sk_next(const struct sock *sk)
271 return sk->sk_node.next ?
272 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
275 static inline int sk_unhashed(const struct sock *sk)
277 return hlist_unhashed(&sk->sk_node);
280 static inline int sk_hashed(const struct sock *sk)
282 return sk->sk_node.pprev != NULL;
285 static __inline__ void sk_node_init(struct hlist_node *node)
290 static __inline__ void __sk_del_node(struct sock *sk)
292 __hlist_del(&sk->sk_node);
295 static __inline__ int __sk_del_node_init(struct sock *sk)
299 sk_node_init(&sk->sk_node);
305 /* Grab socket reference count. This operation is valid only
306 when sk is ALREADY grabbed f.e. it is found in hash table
307 or a list and the lookup is made under lock preventing hash table
311 static inline void sock_hold(struct sock *sk)
313 atomic_inc(&sk->sk_refcnt);
316 /* Ungrab socket in the context, which assumes that socket refcnt
317 cannot hit zero, f.e. it is true in context of any socketcall.
319 static inline void __sock_put(struct sock *sk)
321 atomic_dec(&sk->sk_refcnt);
324 static __inline__ int sk_del_node_init(struct sock *sk)
326 int rc = __sk_del_node_init(sk);
329 /* paranoid for a while -acme */
330 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
336 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
338 hlist_add_head(&sk->sk_node, list);
341 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
344 __sk_add_node(sk, list);
347 static __inline__ void __sk_del_bind_node(struct sock *sk)
349 __hlist_del(&sk->sk_bind_node);
352 static __inline__ void sk_add_bind_node(struct sock *sk,
353 struct hlist_head *list)
355 hlist_add_head(&sk->sk_bind_node, list);
358 #define sk_for_each(__sk, node, list) \
359 hlist_for_each_entry(__sk, node, list, sk_node)
360 #define sk_for_each_from(__sk, node) \
361 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
362 hlist_for_each_entry_from(__sk, node, sk_node)
363 #define sk_for_each_continue(__sk, node) \
364 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
365 hlist_for_each_entry_continue(__sk, node, sk_node)
366 #define sk_for_each_safe(__sk, node, tmp, list) \
367 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
368 #define sk_for_each_bound(__sk, node, list) \
369 hlist_for_each_entry(__sk, node, list, sk_bind_node)
382 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
383 SOCK_DBG, /* %SO_DEBUG setting */
384 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
385 SOCK_NO_LARGESEND, /* whether to sent large segments or not */
386 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
387 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
390 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
392 nsk->sk_flags = osk->sk_flags;
395 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
397 __set_bit(flag, &sk->sk_flags);
400 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
402 __clear_bit(flag, &sk->sk_flags);
405 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
407 return test_bit(flag, &sk->sk_flags);
410 static inline void sk_acceptq_removed(struct sock *sk)
412 sk->sk_ack_backlog--;
415 static inline void sk_acceptq_added(struct sock *sk)
417 sk->sk_ack_backlog++;
420 static inline int sk_acceptq_is_full(struct sock *sk)
422 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
426 * Compute minimal free write space needed to queue new packets.
428 static inline int sk_stream_min_wspace(struct sock *sk)
430 return sk->sk_wmem_queued / 2;
433 static inline int sk_stream_wspace(struct sock *sk)
435 return sk->sk_sndbuf - sk->sk_wmem_queued;
438 extern void sk_stream_write_space(struct sock *sk);
440 static inline int sk_stream_memory_free(struct sock *sk)
442 return sk->sk_wmem_queued < sk->sk_sndbuf;
445 extern void sk_stream_rfree(struct sk_buff *skb);
447 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
450 skb->destructor = sk_stream_rfree;
451 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
452 sk->sk_forward_alloc -= skb->truesize;
455 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
457 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
458 sk->sk_wmem_queued -= skb->truesize;
459 sk->sk_forward_alloc += skb->truesize;
463 /* The per-socket spinlock must be held here. */
464 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
466 if (!sk->sk_backlog.tail) {
467 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
469 sk->sk_backlog.tail->next = skb;
470 sk->sk_backlog.tail = skb;
475 #define sk_wait_event(__sk, __timeo, __condition) \
477 release_sock(__sk); \
480 *(__timeo) = schedule_timeout(*(__timeo)); \
487 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
488 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
489 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
490 extern int sk_stream_error(struct sock *sk, int flags, int err);
491 extern void sk_stream_kill_queues(struct sock *sk);
493 extern int sk_wait_data(struct sock *sk, long *timeo);
495 struct request_sock_ops;
496 struct timewait_sock_ops;
498 /* Networking protocol blocks we attach to sockets.
499 * socket layer -> transport layer interface
500 * transport -> network interface is defined by struct inet_proto
503 void (*close)(struct sock *sk,
505 int (*connect)(struct sock *sk,
506 struct sockaddr *uaddr,
508 int (*disconnect)(struct sock *sk, int flags);
510 struct sock * (*accept) (struct sock *sk, int flags, int *err);
512 int (*ioctl)(struct sock *sk, int cmd,
514 int (*init)(struct sock *sk);
515 int (*destroy)(struct sock *sk);
516 void (*shutdown)(struct sock *sk, int how);
517 int (*setsockopt)(struct sock *sk, int level,
518 int optname, char __user *optval,
520 int (*getsockopt)(struct sock *sk, int level,
521 int optname, char __user *optval,
523 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
524 struct msghdr *msg, size_t len);
525 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
527 size_t len, int noblock, int flags,
529 int (*sendpage)(struct sock *sk, struct page *page,
530 int offset, size_t size, int flags);
531 int (*bind)(struct sock *sk,
532 struct sockaddr *uaddr, int addr_len);
534 int (*backlog_rcv) (struct sock *sk,
535 struct sk_buff *skb);
537 /* Keeping track of sk's, looking them up, and port selection methods. */
538 void (*hash)(struct sock *sk);
539 void (*unhash)(struct sock *sk);
540 int (*get_port)(struct sock *sk, unsigned short snum);
542 /* Memory pressure */
543 void (*enter_memory_pressure)(void);
544 atomic_t *memory_allocated; /* Current allocated memory. */
545 atomic_t *sockets_allocated; /* Current number of sockets. */
547 * Pressure flag: try to collapse.
548 * Technical note: it is used by multiple contexts non atomically.
549 * All the sk_stream_mem_schedule() is of this nature: accounting
550 * is strict, actions are advisory and have some latency.
552 int *memory_pressure;
559 unsigned int obj_size;
561 atomic_t *orphan_count;
563 struct request_sock_ops *rsk_prot;
564 struct timewait_sock_ops *twsk_prot;
566 struct module *owner;
570 struct list_head node;
571 #ifdef SOCK_REFCNT_DEBUG
576 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
580 extern int proto_register(struct proto *prot, int alloc_slab);
581 extern void proto_unregister(struct proto *prot);
583 #ifdef SOCK_REFCNT_DEBUG
584 static inline void sk_refcnt_debug_inc(struct sock *sk)
586 atomic_inc(&sk->sk_prot->socks);
589 static inline void sk_refcnt_debug_dec(struct sock *sk)
591 atomic_dec(&sk->sk_prot->socks);
592 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
593 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
596 static inline void sk_refcnt_debug_release(const struct sock *sk)
598 if (atomic_read(&sk->sk_refcnt) != 1)
599 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
600 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
602 #else /* SOCK_REFCNT_DEBUG */
603 #define sk_refcnt_debug_inc(sk) do { } while (0)
604 #define sk_refcnt_debug_dec(sk) do { } while (0)
605 #define sk_refcnt_debug_release(sk) do { } while (0)
606 #endif /* SOCK_REFCNT_DEBUG */
608 /* Called with local bh disabled */
609 static __inline__ void sock_prot_inc_use(struct proto *prot)
611 prot->stats[smp_processor_id()].inuse++;
614 static __inline__ void sock_prot_dec_use(struct proto *prot)
616 prot->stats[smp_processor_id()].inuse--;
619 /* With per-bucket locks this operation is not-atomic, so that
620 * this version is not worse.
622 static inline void __sk_prot_rehash(struct sock *sk)
624 sk->sk_prot->unhash(sk);
625 sk->sk_prot->hash(sk);
628 /* About 10 seconds */
629 #define SOCK_DESTROY_TIME (10*HZ)
631 /* Sockets 0-1023 can't be bound to unless you are superuser */
632 #define PROT_SOCK 1024
634 #define SHUTDOWN_MASK 3
635 #define RCV_SHUTDOWN 1
636 #define SEND_SHUTDOWN 2
638 #define SOCK_SNDBUF_LOCK 1
639 #define SOCK_RCVBUF_LOCK 2
640 #define SOCK_BINDADDR_LOCK 4
641 #define SOCK_BINDPORT_LOCK 8
643 /* sock_iocb: used to kick off async processing of socket ios */
645 struct list_head list;
651 struct scm_cookie *scm;
652 struct msghdr *msg, async_msg;
653 struct iovec async_iov;
657 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
659 return (struct sock_iocb *)iocb->private;
662 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
667 struct socket_alloc {
668 struct socket socket;
669 struct inode vfs_inode;
672 static inline struct socket *SOCKET_I(struct inode *inode)
674 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
677 static inline struct inode *SOCK_INODE(struct socket *socket)
679 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
682 extern void __sk_stream_mem_reclaim(struct sock *sk);
683 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
685 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
687 static inline int sk_stream_pages(int amt)
689 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
692 static inline void sk_stream_mem_reclaim(struct sock *sk)
694 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
695 __sk_stream_mem_reclaim(sk);
698 static inline void sk_stream_writequeue_purge(struct sock *sk)
702 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
703 sk_stream_free_skb(sk, skb);
704 sk_stream_mem_reclaim(sk);
707 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
709 return (int)skb->truesize <= sk->sk_forward_alloc ||
710 sk_stream_mem_schedule(sk, skb->truesize, 1);
713 static inline int sk_stream_wmem_schedule(struct sock *sk, int size)
715 return size <= sk->sk_forward_alloc ||
716 sk_stream_mem_schedule(sk, size, 0);
719 /* Used by processes to "lock" a socket state, so that
720 * interrupts and bottom half handlers won't change it
721 * from under us. It essentially blocks any incoming
722 * packets, so that we won't get any new data or any
723 * packets that change the state of the socket.
725 * While locked, BH processing will add new packets to
726 * the backlog queue. This queue is processed by the
727 * owner of the socket lock right before it is released.
729 * Since ~2.3.5 it is also exclusive sleep lock serializing
730 * accesses from user process context.
732 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
734 extern void FASTCALL(lock_sock(struct sock *sk));
735 extern void FASTCALL(release_sock(struct sock *sk));
737 /* BH context may only use the following locking interface. */
738 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
739 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
741 extern struct sock *sk_alloc(int family,
743 struct proto *prot, int zero_it);
744 extern void sk_free(struct sock *sk);
745 extern struct sock *sk_clone(const struct sock *sk,
746 const gfp_t priority);
748 extern struct sk_buff *sock_wmalloc(struct sock *sk,
749 unsigned long size, int force,
751 extern struct sk_buff *sock_rmalloc(struct sock *sk,
752 unsigned long size, int force,
754 extern void sock_wfree(struct sk_buff *skb);
755 extern void sock_rfree(struct sk_buff *skb);
757 extern int sock_setsockopt(struct socket *sock, int level,
758 int op, char __user *optval,
761 extern int sock_getsockopt(struct socket *sock, int level,
762 int op, char __user *optval,
764 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
768 extern void *sock_kmalloc(struct sock *sk, int size,
770 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
771 extern void sk_send_sigurg(struct sock *sk);
774 * Functions to fill in entries in struct proto_ops when a protocol
775 * does not implement a particular function.
777 extern int sock_no_bind(struct socket *,
778 struct sockaddr *, int);
779 extern int sock_no_connect(struct socket *,
780 struct sockaddr *, int, int);
781 extern int sock_no_socketpair(struct socket *,
783 extern int sock_no_accept(struct socket *,
784 struct socket *, int);
785 extern int sock_no_getname(struct socket *,
786 struct sockaddr *, int *, int);
787 extern unsigned int sock_no_poll(struct file *, struct socket *,
788 struct poll_table_struct *);
789 extern int sock_no_ioctl(struct socket *, unsigned int,
791 extern int sock_no_listen(struct socket *, int);
792 extern int sock_no_shutdown(struct socket *, int);
793 extern int sock_no_getsockopt(struct socket *, int , int,
794 char __user *, int __user *);
795 extern int sock_no_setsockopt(struct socket *, int, int,
797 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
798 struct msghdr *, size_t);
799 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
800 struct msghdr *, size_t, int);
801 extern int sock_no_mmap(struct file *file,
803 struct vm_area_struct *vma);
804 extern ssize_t sock_no_sendpage(struct socket *sock,
806 int offset, size_t size,
810 * Functions to fill in entries in struct proto_ops when a protocol
811 * uses the inet style.
813 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
814 char __user *optval, int __user *optlen);
815 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
816 struct msghdr *msg, size_t size, int flags);
817 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
818 char __user *optval, int optlen);
820 extern void sk_common_release(struct sock *sk);
823 * Default socket callbacks and setup code
826 /* Initialise core socket variables */
827 extern void sock_init_data(struct socket *sock, struct sock *sk);
830 * sk_filter - run a packet through a socket filter
831 * @sk: sock associated with &sk_buff
832 * @skb: buffer to filter
833 * @needlock: set to 1 if the sock is not locked by caller.
835 * Run the filter code and then cut skb->data to correct size returned by
836 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
837 * than pkt_len we keep whole skb->data. This is the socket level
838 * wrapper to sk_run_filter. It returns 0 if the packet should
839 * be accepted or -EPERM if the packet should be tossed.
843 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
847 err = security_sock_rcv_skb(sk, skb);
852 struct sk_filter *filter;
857 filter = sk->sk_filter;
859 unsigned int pkt_len = sk_run_filter(skb, filter->insns,
864 skb_trim(skb, pkt_len);
874 * sk_filter_release: Release a socket filter
876 * @fp: filter to remove
878 * Remove a filter from a socket and release its resources.
881 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
883 unsigned int size = sk_filter_len(fp);
885 atomic_sub(size, &sk->sk_omem_alloc);
887 if (atomic_dec_and_test(&fp->refcnt))
891 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
893 atomic_inc(&fp->refcnt);
894 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
898 * Socket reference counting postulates.
900 * * Each user of socket SHOULD hold a reference count.
901 * * Each access point to socket (an hash table bucket, reference from a list,
902 * running timer, skb in flight MUST hold a reference count.
903 * * When reference count hits 0, it means it will never increase back.
904 * * When reference count hits 0, it means that no references from
905 * outside exist to this socket and current process on current CPU
906 * is last user and may/should destroy this socket.
907 * * sk_free is called from any context: process, BH, IRQ. When
908 * it is called, socket has no references from outside -> sk_free
909 * may release descendant resources allocated by the socket, but
910 * to the time when it is called, socket is NOT referenced by any
911 * hash tables, lists etc.
912 * * Packets, delivered from outside (from network or from another process)
913 * and enqueued on receive/error queues SHOULD NOT grab reference count,
914 * when they sit in queue. Otherwise, packets will leak to hole, when
915 * socket is looked up by one cpu and unhasing is made by another CPU.
916 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
917 * (leak to backlog). Packet socket does all the processing inside
918 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
919 * use separate SMP lock, so that they are prone too.
922 /* Ungrab socket and destroy it, if it was the last reference. */
923 static inline void sock_put(struct sock *sk)
925 if (atomic_dec_and_test(&sk->sk_refcnt))
929 static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb)
931 int rc = NET_RX_SUCCESS;
933 if (sk_filter(sk, skb, 0))
934 goto discard_and_relse;
939 if (!sock_owned_by_user(sk))
940 rc = sk->sk_backlog_rcv(sk, skb);
942 sk_add_backlog(sk, skb);
952 /* Detach socket from process context.
953 * Announce socket dead, detach it from wait queue and inode.
954 * Note that parent inode held reference count on this struct sock,
955 * we do not release it in this function, because protocol
956 * probably wants some additional cleanups or even continuing
957 * to work with this socket (TCP).
959 static inline void sock_orphan(struct sock *sk)
961 write_lock_bh(&sk->sk_callback_lock);
962 sock_set_flag(sk, SOCK_DEAD);
963 sk->sk_socket = NULL;
965 write_unlock_bh(&sk->sk_callback_lock);
968 static inline void sock_graft(struct sock *sk, struct socket *parent)
970 write_lock_bh(&sk->sk_callback_lock);
971 sk->sk_sleep = &parent->wait;
973 sk->sk_socket = parent;
974 write_unlock_bh(&sk->sk_callback_lock);
977 extern int sock_i_uid(struct sock *sk);
978 extern unsigned long sock_i_ino(struct sock *sk);
980 static inline struct dst_entry *
981 __sk_dst_get(struct sock *sk)
983 return sk->sk_dst_cache;
986 static inline struct dst_entry *
987 sk_dst_get(struct sock *sk)
989 struct dst_entry *dst;
991 read_lock(&sk->sk_dst_lock);
992 dst = sk->sk_dst_cache;
995 read_unlock(&sk->sk_dst_lock);
1000 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1002 struct dst_entry *old_dst;
1004 old_dst = sk->sk_dst_cache;
1005 sk->sk_dst_cache = dst;
1006 dst_release(old_dst);
1010 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1012 write_lock(&sk->sk_dst_lock);
1013 __sk_dst_set(sk, dst);
1014 write_unlock(&sk->sk_dst_lock);
1018 __sk_dst_reset(struct sock *sk)
1020 struct dst_entry *old_dst;
1022 old_dst = sk->sk_dst_cache;
1023 sk->sk_dst_cache = NULL;
1024 dst_release(old_dst);
1028 sk_dst_reset(struct sock *sk)
1030 write_lock(&sk->sk_dst_lock);
1032 write_unlock(&sk->sk_dst_lock);
1035 static inline struct dst_entry *
1036 __sk_dst_check(struct sock *sk, u32 cookie)
1038 struct dst_entry *dst = sk->sk_dst_cache;
1040 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1041 sk->sk_dst_cache = NULL;
1049 static inline struct dst_entry *
1050 sk_dst_check(struct sock *sk, u32 cookie)
1052 struct dst_entry *dst = sk_dst_get(sk);
1054 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1063 static inline void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1065 __sk_dst_set(sk, dst);
1066 sk->sk_route_caps = dst->dev->features;
1067 if (sk->sk_route_caps & NETIF_F_TSO) {
1068 if (sock_flag(sk, SOCK_NO_LARGESEND) || dst->header_len)
1069 sk->sk_route_caps &= ~NETIF_F_TSO;
1073 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1075 sk->sk_wmem_queued += skb->truesize;
1076 sk->sk_forward_alloc -= skb->truesize;
1079 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1080 struct sk_buff *skb, struct page *page,
1083 if (skb->ip_summed == CHECKSUM_NONE) {
1085 unsigned int csum = csum_and_copy_from_user(from,
1086 page_address(page) + off,
1090 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1091 } else if (copy_from_user(page_address(page) + off, from, copy))
1095 skb->data_len += copy;
1096 skb->truesize += copy;
1097 sk->sk_wmem_queued += copy;
1098 sk->sk_forward_alloc -= copy;
1103 * Queue a received datagram if it will fit. Stream and sequenced
1104 * protocols can't normally use this as they need to fit buffers in
1105 * and play with them.
1107 * Inlined as it's very short and called for pretty much every
1108 * packet ever received.
1111 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1115 skb->destructor = sock_wfree;
1116 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1119 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1122 skb->destructor = sock_rfree;
1123 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1126 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1127 unsigned long expires);
1129 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1131 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1136 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1137 number of warnings when compiling with -W --ANK
1139 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1140 (unsigned)sk->sk_rcvbuf) {
1145 /* It would be deadlock, if sock_queue_rcv_skb is used
1146 with socket lock! We assume that users of this
1147 function are lock free.
1149 err = sk_filter(sk, skb, 1);
1154 skb_set_owner_r(skb, sk);
1156 /* Cache the SKB length before we tack it onto the receive
1157 * queue. Once it is added it no longer belongs to us and
1158 * may be freed by other threads of control pulling packets
1163 skb_queue_tail(&sk->sk_receive_queue, skb);
1165 if (!sock_flag(sk, SOCK_DEAD))
1166 sk->sk_data_ready(sk, skb_len);
1171 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1173 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1174 number of warnings when compiling with -W --ANK
1176 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1177 (unsigned)sk->sk_rcvbuf)
1179 skb_set_owner_r(skb, sk);
1180 skb_queue_tail(&sk->sk_error_queue, skb);
1181 if (!sock_flag(sk, SOCK_DEAD))
1182 sk->sk_data_ready(sk, skb->len);
1187 * Recover an error report and clear atomically
1190 static inline int sock_error(struct sock *sk)
1193 if (likely(!sk->sk_err))
1195 err = xchg(&sk->sk_err, 0);
1199 static inline unsigned long sock_wspace(struct sock *sk)
1203 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1204 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1211 static inline void sk_wake_async(struct sock *sk, int how, int band)
1213 if (sk->sk_socket && sk->sk_socket->fasync_list)
1214 sock_wake_async(sk->sk_socket, how, band);
1217 #define SOCK_MIN_SNDBUF 2048
1218 #define SOCK_MIN_RCVBUF 256
1220 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1222 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1223 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1224 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1228 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1232 struct sk_buff *skb;
1235 hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header);
1236 skb = alloc_skb_fclone(size + hdr_len, gfp);
1238 skb->truesize += mem;
1239 if (sk_stream_wmem_schedule(sk, skb->truesize)) {
1240 skb_reserve(skb, hdr_len);
1245 sk->sk_prot->enter_memory_pressure();
1246 sk_stream_moderate_sndbuf(sk);
1251 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1255 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1258 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1260 struct page *page = NULL;
1262 page = alloc_pages(sk->sk_allocation, 0);
1264 sk->sk_prot->enter_memory_pressure();
1265 sk_stream_moderate_sndbuf(sk);
1270 #define sk_stream_for_retrans_queue(skb, sk) \
1271 for (skb = (sk)->sk_write_queue.next; \
1272 (skb != (sk)->sk_send_head) && \
1273 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1276 /*from STCP for fast SACK Process*/
1277 #define sk_stream_for_retrans_queue_from(skb, sk) \
1278 for (; (skb != (sk)->sk_send_head) && \
1279 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1283 * Default write policy as shown to user space via poll/select/SIGIO
1285 static inline int sock_writeable(const struct sock *sk)
1287 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1290 static inline gfp_t gfp_any(void)
1292 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1295 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1297 return noblock ? 0 : sk->sk_rcvtimeo;
1300 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1302 return noblock ? 0 : sk->sk_sndtimeo;
1305 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1307 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1310 /* Alas, with timeout socket operations are not restartable.
1311 * Compare this to poll().
1313 static inline int sock_intr_errno(long timeo)
1315 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1318 static __inline__ void
1319 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1321 struct timeval stamp;
1323 skb_get_timestamp(skb, &stamp);
1324 if (sock_flag(sk, SOCK_RCVTSTAMP)) {
1325 /* Race occurred between timestamp enabling and packet
1326 receiving. Fill in the current time for now. */
1327 if (stamp.tv_sec == 0)
1328 do_gettimeofday(&stamp);
1329 skb_set_timestamp(skb, &stamp);
1330 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1333 sk->sk_stamp = stamp;
1337 * sk_eat_skb - Release a skb if it is no longer needed
1338 * @sk: socket to eat this skb from
1339 * @skb: socket buffer to eat
1341 * This routine must be called with interrupts disabled or with the socket
1342 * locked so that the sk_buff queue operation is ok.
1344 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1346 __skb_unlink(skb, &sk->sk_receive_queue);
1350 extern void sock_enable_timestamp(struct sock *sk);
1351 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1354 * Enable debug/info messages
1357 #ifdef CONFIG_NETDEBUG
1358 #define NETDEBUG(fmt, args...) printk(fmt,##args)
1359 #define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0)
1361 #define NETDEBUG(fmt, args...) do { } while (0)
1362 #define LIMIT_NETDEBUG(fmt, args...) do { } while(0)
1366 * Macros for sleeping on a socket. Use them like this:
1368 * SOCK_SLEEP_PRE(sk)
1371 * SOCK_SLEEP_POST(sk)
1373 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1374 * and when the last use of them in DECnet has gone, I'm intending to
1378 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1379 DECLARE_WAITQUEUE(wait, tsk); \
1380 tsk->state = TASK_INTERRUPTIBLE; \
1381 add_wait_queue((sk)->sk_sleep, &wait); \
1384 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1385 remove_wait_queue((sk)->sk_sleep, &wait); \
1389 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1392 sock_set_flag(sk, bit);
1394 sock_reset_flag(sk, bit);
1397 extern __u32 sysctl_wmem_max;
1398 extern __u32 sysctl_rmem_max;
1401 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1403 static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
1409 extern void sk_init(void);
1411 #ifdef CONFIG_SYSCTL
1412 extern struct ctl_table core_table[];
1415 extern int sysctl_optmem_max;
1417 extern __u32 sysctl_wmem_default;
1418 extern __u32 sysctl_rmem_default;
1420 #endif /* _SOCK_H */