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_prot: protocol handlers inside a network family
104 * This is the minimal network layer representation of sockets, the header
105 * for struct sock and struct inet_timewait_sock.
108 unsigned short skc_family;
109 volatile unsigned char skc_state;
110 unsigned char skc_reuse;
111 int skc_bound_dev_if;
112 struct hlist_node skc_node;
113 struct hlist_node skc_bind_node;
115 struct proto *skc_prot;
119 * struct sock - network layer representation of sockets
120 * @__sk_common: shared layout with inet_timewait_sock
121 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
122 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
123 * @sk_lock: synchronizer
124 * @sk_rcvbuf: size of receive buffer in bytes
125 * @sk_sleep: sock wait queue
126 * @sk_dst_cache: destination cache
127 * @sk_dst_lock: destination cache lock
128 * @sk_policy: flow policy
129 * @sk_rmem_alloc: receive queue bytes committed
130 * @sk_receive_queue: incoming packets
131 * @sk_wmem_alloc: transmit queue bytes committed
132 * @sk_write_queue: Packet sending queue
133 * @sk_omem_alloc: "o" is "option" or "other"
134 * @sk_wmem_queued: persistent queue size
135 * @sk_forward_alloc: space allocated forward
136 * @sk_allocation: allocation mode
137 * @sk_sndbuf: size of send buffer in bytes
138 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
139 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
140 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
141 * @sk_lingertime: %SO_LINGER l_linger setting
142 * @sk_hashent: hash entry in several tables (e.g. inet_hashinfo.ehash)
143 * @sk_backlog: always used with the per-socket spinlock held
144 * @sk_callback_lock: used with the callbacks in the end of this struct
145 * @sk_error_queue: rarely used
146 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
147 * @sk_err: last error
148 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
149 * @sk_ack_backlog: current listen backlog
150 * @sk_max_ack_backlog: listen backlog set in listen()
151 * @sk_priority: %SO_PRIORITY setting
152 * @sk_type: socket type (%SOCK_STREAM, etc)
153 * @sk_protocol: which protocol this socket belongs in this network family
154 * @sk_peercred: %SO_PEERCRED setting
155 * @sk_rcvlowat: %SO_RCVLOWAT setting
156 * @sk_rcvtimeo: %SO_RCVTIMEO setting
157 * @sk_sndtimeo: %SO_SNDTIMEO setting
158 * @sk_filter: socket filtering instructions
159 * @sk_protinfo: private area, net family specific, when not using slab
160 * @sk_timer: sock cleanup timer
161 * @sk_stamp: time stamp of last packet received
162 * @sk_socket: Identd and reporting IO signals
163 * @sk_user_data: RPC layer private data
164 * @sk_sndmsg_page: cached page for sendmsg
165 * @sk_sndmsg_off: cached offset for sendmsg
166 * @sk_send_head: front of stuff to transmit
167 * @sk_security: used by security modules
168 * @sk_write_pending: a write to stream socket waits to start
169 * @sk_state_change: callback to indicate change in the state of the sock
170 * @sk_data_ready: callback to indicate there is data to be processed
171 * @sk_write_space: callback to indicate there is bf sending space available
172 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
173 * @sk_backlog_rcv: callback to process the backlog
174 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
178 * Now struct inet_timewait_sock also uses sock_common, so please just
179 * don't add nothing before this first member (__sk_common) --acme
181 struct sock_common __sk_common;
182 #define sk_family __sk_common.skc_family
183 #define sk_state __sk_common.skc_state
184 #define sk_reuse __sk_common.skc_reuse
185 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
186 #define sk_node __sk_common.skc_node
187 #define sk_bind_node __sk_common.skc_bind_node
188 #define sk_refcnt __sk_common.skc_refcnt
189 #define sk_prot __sk_common.skc_prot
190 unsigned char sk_shutdown : 2,
193 unsigned char sk_protocol;
194 unsigned short sk_type;
196 socket_lock_t sk_lock;
197 wait_queue_head_t *sk_sleep;
198 struct dst_entry *sk_dst_cache;
199 struct xfrm_policy *sk_policy[2];
200 rwlock_t sk_dst_lock;
201 atomic_t sk_rmem_alloc;
202 atomic_t sk_wmem_alloc;
203 atomic_t sk_omem_alloc;
204 struct sk_buff_head sk_receive_queue;
205 struct sk_buff_head sk_write_queue;
207 int sk_forward_alloc;
208 unsigned int sk_allocation;
212 unsigned long sk_flags;
213 unsigned long sk_lingertime;
215 * The backlog queue is special, it is always used with
216 * the per-socket spinlock held and requires low latency
217 * access. Therefore we special case it's implementation.
220 struct sk_buff *head;
221 struct sk_buff *tail;
223 struct sk_buff_head sk_error_queue;
224 struct proto *sk_prot_creator;
225 rwlock_t sk_callback_lock;
228 unsigned short sk_ack_backlog;
229 unsigned short sk_max_ack_backlog;
231 struct ucred sk_peercred;
235 struct sk_filter *sk_filter;
237 struct timer_list sk_timer;
238 struct timeval sk_stamp;
239 struct socket *sk_socket;
241 struct page *sk_sndmsg_page;
242 struct sk_buff *sk_send_head;
244 int sk_write_pending;
246 void (*sk_state_change)(struct sock *sk);
247 void (*sk_data_ready)(struct sock *sk, int bytes);
248 void (*sk_write_space)(struct sock *sk);
249 void (*sk_error_report)(struct sock *sk);
250 int (*sk_backlog_rcv)(struct sock *sk,
251 struct sk_buff *skb);
252 void (*sk_destruct)(struct sock *sk);
256 * Hashed lists helper routines
258 static inline struct sock *__sk_head(const struct hlist_head *head)
260 return hlist_entry(head->first, struct sock, sk_node);
263 static inline struct sock *sk_head(const struct hlist_head *head)
265 return hlist_empty(head) ? NULL : __sk_head(head);
268 static inline struct sock *sk_next(const struct sock *sk)
270 return sk->sk_node.next ?
271 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
274 static inline int sk_unhashed(const struct sock *sk)
276 return hlist_unhashed(&sk->sk_node);
279 static inline int sk_hashed(const struct sock *sk)
281 return sk->sk_node.pprev != NULL;
284 static __inline__ void sk_node_init(struct hlist_node *node)
289 static __inline__ void __sk_del_node(struct sock *sk)
291 __hlist_del(&sk->sk_node);
294 static __inline__ int __sk_del_node_init(struct sock *sk)
298 sk_node_init(&sk->sk_node);
304 /* Grab socket reference count. This operation is valid only
305 when sk is ALREADY grabbed f.e. it is found in hash table
306 or a list and the lookup is made under lock preventing hash table
310 static inline void sock_hold(struct sock *sk)
312 atomic_inc(&sk->sk_refcnt);
315 /* Ungrab socket in the context, which assumes that socket refcnt
316 cannot hit zero, f.e. it is true in context of any socketcall.
318 static inline void __sock_put(struct sock *sk)
320 atomic_dec(&sk->sk_refcnt);
323 static __inline__ int sk_del_node_init(struct sock *sk)
325 int rc = __sk_del_node_init(sk);
328 /* paranoid for a while -acme */
329 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
335 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
337 hlist_add_head(&sk->sk_node, list);
340 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
343 __sk_add_node(sk, list);
346 static __inline__ void __sk_del_bind_node(struct sock *sk)
348 __hlist_del(&sk->sk_bind_node);
351 static __inline__ void sk_add_bind_node(struct sock *sk,
352 struct hlist_head *list)
354 hlist_add_head(&sk->sk_bind_node, list);
357 #define sk_for_each(__sk, node, list) \
358 hlist_for_each_entry(__sk, node, list, sk_node)
359 #define sk_for_each_from(__sk, node) \
360 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
361 hlist_for_each_entry_from(__sk, node, sk_node)
362 #define sk_for_each_continue(__sk, node) \
363 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
364 hlist_for_each_entry_continue(__sk, node, sk_node)
365 #define sk_for_each_safe(__sk, node, tmp, list) \
366 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
367 #define sk_for_each_bound(__sk, node, list) \
368 hlist_for_each_entry(__sk, node, list, sk_bind_node)
381 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
382 SOCK_DBG, /* %SO_DEBUG setting */
383 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
384 SOCK_NO_LARGESEND, /* whether to sent large segments or not */
385 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
386 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
389 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
391 nsk->sk_flags = osk->sk_flags;
394 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
396 __set_bit(flag, &sk->sk_flags);
399 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
401 __clear_bit(flag, &sk->sk_flags);
404 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
406 return test_bit(flag, &sk->sk_flags);
409 static inline void sk_acceptq_removed(struct sock *sk)
411 sk->sk_ack_backlog--;
414 static inline void sk_acceptq_added(struct sock *sk)
416 sk->sk_ack_backlog++;
419 static inline int sk_acceptq_is_full(struct sock *sk)
421 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
425 * Compute minimal free write space needed to queue new packets.
427 static inline int sk_stream_min_wspace(struct sock *sk)
429 return sk->sk_wmem_queued / 2;
432 static inline int sk_stream_wspace(struct sock *sk)
434 return sk->sk_sndbuf - sk->sk_wmem_queued;
437 extern void sk_stream_write_space(struct sock *sk);
439 static inline int sk_stream_memory_free(struct sock *sk)
441 return sk->sk_wmem_queued < sk->sk_sndbuf;
444 extern void sk_stream_rfree(struct sk_buff *skb);
446 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
449 skb->destructor = sk_stream_rfree;
450 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
451 sk->sk_forward_alloc -= skb->truesize;
454 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
456 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
457 sk->sk_wmem_queued -= skb->truesize;
458 sk->sk_forward_alloc += skb->truesize;
462 /* The per-socket spinlock must be held here. */
463 #define sk_add_backlog(__sk, __skb) \
464 do { if (!(__sk)->sk_backlog.tail) { \
465 (__sk)->sk_backlog.head = \
466 (__sk)->sk_backlog.tail = (__skb); \
468 ((__sk)->sk_backlog.tail)->next = (__skb); \
469 (__sk)->sk_backlog.tail = (__skb); \
471 (__skb)->next = NULL; \
474 #define sk_wait_event(__sk, __timeo, __condition) \
476 release_sock(__sk); \
479 *(__timeo) = schedule_timeout(*(__timeo)); \
486 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
487 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
488 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
489 extern int sk_stream_error(struct sock *sk, int flags, int err);
490 extern void sk_stream_kill_queues(struct sock *sk);
492 extern int sk_wait_data(struct sock *sk, long *timeo);
494 struct request_sock_ops;
496 /* Here is the right place to enable sock refcounting debugging */
497 //#define SOCK_REFCNT_DEBUG
499 /* Networking protocol blocks we attach to sockets.
500 * socket layer -> transport layer interface
501 * transport -> network interface is defined by struct inet_proto
504 void (*close)(struct sock *sk,
506 int (*connect)(struct sock *sk,
507 struct sockaddr *uaddr,
509 int (*disconnect)(struct sock *sk, int flags);
511 struct sock * (*accept) (struct sock *sk, int flags, int *err);
513 int (*ioctl)(struct sock *sk, int cmd,
515 int (*init)(struct sock *sk);
516 int (*destroy)(struct sock *sk);
517 void (*shutdown)(struct sock *sk, int how);
518 int (*setsockopt)(struct sock *sk, int level,
519 int optname, char __user *optval,
521 int (*getsockopt)(struct sock *sk, int level,
522 int optname, char __user *optval,
524 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
525 struct msghdr *msg, size_t len);
526 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
528 size_t len, int noblock, int flags,
530 int (*sendpage)(struct sock *sk, struct page *page,
531 int offset, size_t size, int flags);
532 int (*bind)(struct sock *sk,
533 struct sockaddr *uaddr, int addr_len);
535 int (*backlog_rcv) (struct sock *sk,
536 struct sk_buff *skb);
538 /* Keeping track of sk's, looking them up, and port selection methods. */
539 void (*hash)(struct sock *sk);
540 void (*unhash)(struct sock *sk);
541 int (*get_port)(struct sock *sk, unsigned short snum);
543 /* Memory pressure */
544 void (*enter_memory_pressure)(void);
545 atomic_t *memory_allocated; /* Current allocated memory. */
546 atomic_t *sockets_allocated; /* Current number of sockets. */
548 * Pressure flag: try to collapse.
549 * Technical note: it is used by multiple contexts non atomically.
550 * All the sk_stream_mem_schedule() is of this nature: accounting
551 * is strict, actions are advisory and have some latency.
553 int *memory_pressure;
560 unsigned int obj_size;
562 kmem_cache_t *twsk_slab;
563 unsigned int twsk_obj_size;
565 struct request_sock_ops *rsk_prot;
567 struct module *owner;
571 struct list_head node;
572 #ifdef SOCK_REFCNT_DEBUG
577 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
581 extern int proto_register(struct proto *prot, int alloc_slab);
582 extern void proto_unregister(struct proto *prot);
584 #ifdef SOCK_REFCNT_DEBUG
585 static inline void sk_refcnt_debug_inc(struct sock *sk)
587 atomic_inc(&sk->sk_prot->socks);
590 static inline void sk_refcnt_debug_dec(struct sock *sk)
592 atomic_dec(&sk->sk_prot->socks);
593 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
594 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
597 static inline void sk_refcnt_debug_release(const struct sock *sk)
599 if (atomic_read(&sk->sk_refcnt) != 1)
600 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
601 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
603 #else /* SOCK_REFCNT_DEBUG */
604 #define sk_refcnt_debug_inc(sk) do { } while (0)
605 #define sk_refcnt_debug_dec(sk) do { } while (0)
606 #define sk_refcnt_debug_release(sk) do { } while (0)
607 #endif /* SOCK_REFCNT_DEBUG */
609 /* Called with local bh disabled */
610 static __inline__ void sock_prot_inc_use(struct proto *prot)
612 prot->stats[smp_processor_id()].inuse++;
615 static __inline__ void sock_prot_dec_use(struct proto *prot)
617 prot->stats[smp_processor_id()].inuse--;
620 /* With per-bucket locks this operation is not-atomic, so that
621 * this version is not worse.
623 static inline void __sk_prot_rehash(struct sock *sk)
625 sk->sk_prot->unhash(sk);
626 sk->sk_prot->hash(sk);
629 /* About 10 seconds */
630 #define SOCK_DESTROY_TIME (10*HZ)
632 /* Sockets 0-1023 can't be bound to unless you are superuser */
633 #define PROT_SOCK 1024
635 #define SHUTDOWN_MASK 3
636 #define RCV_SHUTDOWN 1
637 #define SEND_SHUTDOWN 2
639 #define SOCK_SNDBUF_LOCK 1
640 #define SOCK_RCVBUF_LOCK 2
641 #define SOCK_BINDADDR_LOCK 4
642 #define SOCK_BINDPORT_LOCK 8
644 /* sock_iocb: used to kick off async processing of socket ios */
646 struct list_head list;
652 struct scm_cookie *scm;
653 struct msghdr *msg, async_msg;
654 struct iovec async_iov;
658 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
660 return (struct sock_iocb *)iocb->private;
663 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
668 struct socket_alloc {
669 struct socket socket;
670 struct inode vfs_inode;
673 static inline struct socket *SOCKET_I(struct inode *inode)
675 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
678 static inline struct inode *SOCK_INODE(struct socket *socket)
680 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
683 extern void __sk_stream_mem_reclaim(struct sock *sk);
684 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
686 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
688 static inline int sk_stream_pages(int amt)
690 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
693 static inline void sk_stream_mem_reclaim(struct sock *sk)
695 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
696 __sk_stream_mem_reclaim(sk);
699 static inline void sk_stream_writequeue_purge(struct sock *sk)
703 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
704 sk_stream_free_skb(sk, skb);
705 sk_stream_mem_reclaim(sk);
708 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
710 return (int)skb->truesize <= sk->sk_forward_alloc ||
711 sk_stream_mem_schedule(sk, skb->truesize, 1);
714 /* Used by processes to "lock" a socket state, so that
715 * interrupts and bottom half handlers won't change it
716 * from under us. It essentially blocks any incoming
717 * packets, so that we won't get any new data or any
718 * packets that change the state of the socket.
720 * While locked, BH processing will add new packets to
721 * the backlog queue. This queue is processed by the
722 * owner of the socket lock right before it is released.
724 * Since ~2.3.5 it is also exclusive sleep lock serializing
725 * accesses from user process context.
727 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
729 extern void FASTCALL(lock_sock(struct sock *sk));
730 extern void FASTCALL(release_sock(struct sock *sk));
732 /* BH context may only use the following locking interface. */
733 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
734 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
736 extern struct sock *sk_alloc(int family,
737 unsigned int __nocast priority,
738 struct proto *prot, int zero_it);
739 extern void sk_free(struct sock *sk);
740 extern struct sock *sk_clone(const struct sock *sk,
741 const unsigned int __nocast priority);
743 extern struct sk_buff *sock_wmalloc(struct sock *sk,
744 unsigned long size, int force,
745 unsigned int __nocast priority);
746 extern struct sk_buff *sock_rmalloc(struct sock *sk,
747 unsigned long size, int force,
748 unsigned int __nocast priority);
749 extern void sock_wfree(struct sk_buff *skb);
750 extern void sock_rfree(struct sk_buff *skb);
752 extern int sock_setsockopt(struct socket *sock, int level,
753 int op, char __user *optval,
756 extern int sock_getsockopt(struct socket *sock, int level,
757 int op, char __user *optval,
759 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
763 extern void *sock_kmalloc(struct sock *sk, int size,
764 unsigned int __nocast priority);
765 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
766 extern void sk_send_sigurg(struct sock *sk);
769 * Functions to fill in entries in struct proto_ops when a protocol
770 * does not implement a particular function.
772 extern int sock_no_bind(struct socket *,
773 struct sockaddr *, int);
774 extern int sock_no_connect(struct socket *,
775 struct sockaddr *, int, int);
776 extern int sock_no_socketpair(struct socket *,
778 extern int sock_no_accept(struct socket *,
779 struct socket *, int);
780 extern int sock_no_getname(struct socket *,
781 struct sockaddr *, int *, int);
782 extern unsigned int sock_no_poll(struct file *, struct socket *,
783 struct poll_table_struct *);
784 extern int sock_no_ioctl(struct socket *, unsigned int,
786 extern int sock_no_listen(struct socket *, int);
787 extern int sock_no_shutdown(struct socket *, int);
788 extern int sock_no_getsockopt(struct socket *, int , int,
789 char __user *, int __user *);
790 extern int sock_no_setsockopt(struct socket *, int, int,
792 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
793 struct msghdr *, size_t);
794 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
795 struct msghdr *, size_t, int);
796 extern int sock_no_mmap(struct file *file,
798 struct vm_area_struct *vma);
799 extern ssize_t sock_no_sendpage(struct socket *sock,
801 int offset, size_t size,
805 * Functions to fill in entries in struct proto_ops when a protocol
806 * uses the inet style.
808 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
809 char __user *optval, int __user *optlen);
810 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
811 struct msghdr *msg, size_t size, int flags);
812 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
813 char __user *optval, int optlen);
815 extern void sk_common_release(struct sock *sk);
818 * Default socket callbacks and setup code
821 /* Initialise core socket variables */
822 extern void sock_init_data(struct socket *sock, struct sock *sk);
825 * sk_filter - run a packet through a socket filter
826 * @sk: sock associated with &sk_buff
827 * @skb: buffer to filter
828 * @needlock: set to 1 if the sock is not locked by caller.
830 * Run the filter code and then cut skb->data to correct size returned by
831 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
832 * than pkt_len we keep whole skb->data. This is the socket level
833 * wrapper to sk_run_filter. It returns 0 if the packet should
834 * be accepted or -EPERM if the packet should be tossed.
838 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
842 err = security_sock_rcv_skb(sk, skb);
847 struct sk_filter *filter;
852 filter = sk->sk_filter;
854 int pkt_len = sk_run_filter(skb, filter->insns,
859 skb_trim(skb, pkt_len);
869 * sk_filter_release: Release a socket filter
871 * @fp: filter to remove
873 * Remove a filter from a socket and release its resources.
876 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
878 unsigned int size = sk_filter_len(fp);
880 atomic_sub(size, &sk->sk_omem_alloc);
882 if (atomic_dec_and_test(&fp->refcnt))
886 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
888 atomic_inc(&fp->refcnt);
889 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
893 * Socket reference counting postulates.
895 * * Each user of socket SHOULD hold a reference count.
896 * * Each access point to socket (an hash table bucket, reference from a list,
897 * running timer, skb in flight MUST hold a reference count.
898 * * When reference count hits 0, it means it will never increase back.
899 * * When reference count hits 0, it means that no references from
900 * outside exist to this socket and current process on current CPU
901 * is last user and may/should destroy this socket.
902 * * sk_free is called from any context: process, BH, IRQ. When
903 * it is called, socket has no references from outside -> sk_free
904 * may release descendant resources allocated by the socket, but
905 * to the time when it is called, socket is NOT referenced by any
906 * hash tables, lists etc.
907 * * Packets, delivered from outside (from network or from another process)
908 * and enqueued on receive/error queues SHOULD NOT grab reference count,
909 * when they sit in queue. Otherwise, packets will leak to hole, when
910 * socket is looked up by one cpu and unhasing is made by another CPU.
911 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
912 * (leak to backlog). Packet socket does all the processing inside
913 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
914 * use separate SMP lock, so that they are prone too.
917 /* Ungrab socket and destroy it, if it was the last reference. */
918 static inline void sock_put(struct sock *sk)
920 if (atomic_dec_and_test(&sk->sk_refcnt))
924 /* Detach socket from process context.
925 * Announce socket dead, detach it from wait queue and inode.
926 * Note that parent inode held reference count on this struct sock,
927 * we do not release it in this function, because protocol
928 * probably wants some additional cleanups or even continuing
929 * to work with this socket (TCP).
931 static inline void sock_orphan(struct sock *sk)
933 write_lock_bh(&sk->sk_callback_lock);
934 sock_set_flag(sk, SOCK_DEAD);
935 sk->sk_socket = NULL;
937 write_unlock_bh(&sk->sk_callback_lock);
940 static inline void sock_graft(struct sock *sk, struct socket *parent)
942 write_lock_bh(&sk->sk_callback_lock);
943 sk->sk_sleep = &parent->wait;
945 sk->sk_socket = parent;
946 write_unlock_bh(&sk->sk_callback_lock);
949 extern int sock_i_uid(struct sock *sk);
950 extern unsigned long sock_i_ino(struct sock *sk);
952 static inline struct dst_entry *
953 __sk_dst_get(struct sock *sk)
955 return sk->sk_dst_cache;
958 static inline struct dst_entry *
959 sk_dst_get(struct sock *sk)
961 struct dst_entry *dst;
963 read_lock(&sk->sk_dst_lock);
964 dst = sk->sk_dst_cache;
967 read_unlock(&sk->sk_dst_lock);
972 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
974 struct dst_entry *old_dst;
976 old_dst = sk->sk_dst_cache;
977 sk->sk_dst_cache = dst;
978 dst_release(old_dst);
982 sk_dst_set(struct sock *sk, struct dst_entry *dst)
984 write_lock(&sk->sk_dst_lock);
985 __sk_dst_set(sk, dst);
986 write_unlock(&sk->sk_dst_lock);
990 __sk_dst_reset(struct sock *sk)
992 struct dst_entry *old_dst;
994 old_dst = sk->sk_dst_cache;
995 sk->sk_dst_cache = NULL;
996 dst_release(old_dst);
1000 sk_dst_reset(struct sock *sk)
1002 write_lock(&sk->sk_dst_lock);
1004 write_unlock(&sk->sk_dst_lock);
1007 static inline struct dst_entry *
1008 __sk_dst_check(struct sock *sk, u32 cookie)
1010 struct dst_entry *dst = sk->sk_dst_cache;
1012 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1013 sk->sk_dst_cache = NULL;
1021 static inline struct dst_entry *
1022 sk_dst_check(struct sock *sk, u32 cookie)
1024 struct dst_entry *dst = sk_dst_get(sk);
1026 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1035 static inline void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1037 __sk_dst_set(sk, dst);
1038 sk->sk_route_caps = dst->dev->features;
1039 if (sk->sk_route_caps & NETIF_F_TSO) {
1040 if (sock_flag(sk, SOCK_NO_LARGESEND) || dst->header_len)
1041 sk->sk_route_caps &= ~NETIF_F_TSO;
1045 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1047 sk->sk_wmem_queued += skb->truesize;
1048 sk->sk_forward_alloc -= skb->truesize;
1051 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1052 struct sk_buff *skb, struct page *page,
1055 if (skb->ip_summed == CHECKSUM_NONE) {
1057 unsigned int csum = csum_and_copy_from_user(from,
1058 page_address(page) + off,
1062 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1063 } else if (copy_from_user(page_address(page) + off, from, copy))
1067 skb->data_len += copy;
1068 skb->truesize += copy;
1069 sk->sk_wmem_queued += copy;
1070 sk->sk_forward_alloc -= copy;
1075 * Queue a received datagram if it will fit. Stream and sequenced
1076 * protocols can't normally use this as they need to fit buffers in
1077 * and play with them.
1079 * Inlined as it's very short and called for pretty much every
1080 * packet ever received.
1083 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1087 skb->destructor = sock_wfree;
1088 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1091 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1094 skb->destructor = sock_rfree;
1095 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1098 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1099 unsigned long expires);
1101 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1103 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1108 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1109 number of warnings when compiling with -W --ANK
1111 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1112 (unsigned)sk->sk_rcvbuf) {
1117 /* It would be deadlock, if sock_queue_rcv_skb is used
1118 with socket lock! We assume that users of this
1119 function are lock free.
1121 err = sk_filter(sk, skb, 1);
1126 skb_set_owner_r(skb, sk);
1128 /* Cache the SKB length before we tack it onto the receive
1129 * queue. Once it is added it no longer belongs to us and
1130 * may be freed by other threads of control pulling packets
1135 skb_queue_tail(&sk->sk_receive_queue, skb);
1137 if (!sock_flag(sk, SOCK_DEAD))
1138 sk->sk_data_ready(sk, skb_len);
1143 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1145 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1146 number of warnings when compiling with -W --ANK
1148 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1149 (unsigned)sk->sk_rcvbuf)
1151 skb_set_owner_r(skb, sk);
1152 skb_queue_tail(&sk->sk_error_queue, skb);
1153 if (!sock_flag(sk, SOCK_DEAD))
1154 sk->sk_data_ready(sk, skb->len);
1159 * Recover an error report and clear atomically
1162 static inline int sock_error(struct sock *sk)
1164 int err = xchg(&sk->sk_err, 0);
1168 static inline unsigned long sock_wspace(struct sock *sk)
1172 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1173 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1180 static inline void sk_wake_async(struct sock *sk, int how, int band)
1182 if (sk->sk_socket && sk->sk_socket->fasync_list)
1183 sock_wake_async(sk->sk_socket, how, band);
1186 #define SOCK_MIN_SNDBUF 2048
1187 #define SOCK_MIN_RCVBUF 256
1189 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1191 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1192 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1193 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1197 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1199 unsigned int __nocast gfp)
1201 struct sk_buff *skb;
1204 hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header);
1205 skb = alloc_skb(size + hdr_len, gfp);
1207 skb->truesize += mem;
1208 if (sk->sk_forward_alloc >= (int)skb->truesize ||
1209 sk_stream_mem_schedule(sk, skb->truesize, 0)) {
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,
1223 unsigned int __nocast gfp)
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 if (sk->sk_forward_alloc >= (int)PAGE_SIZE ||
1233 sk_stream_mem_schedule(sk, PAGE_SIZE, 0))
1234 page = alloc_pages(sk->sk_allocation, 0);
1236 sk->sk_prot->enter_memory_pressure();
1237 sk_stream_moderate_sndbuf(sk);
1242 #define sk_stream_for_retrans_queue(skb, sk) \
1243 for (skb = (sk)->sk_write_queue.next; \
1244 (skb != (sk)->sk_send_head) && \
1245 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1249 * Default write policy as shown to user space via poll/select/SIGIO
1251 static inline int sock_writeable(const struct sock *sk)
1253 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1256 static inline unsigned int __nocast gfp_any(void)
1258 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1261 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1263 return noblock ? 0 : sk->sk_rcvtimeo;
1266 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1268 return noblock ? 0 : sk->sk_sndtimeo;
1271 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1273 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1276 /* Alas, with timeout socket operations are not restartable.
1277 * Compare this to poll().
1279 static inline int sock_intr_errno(long timeo)
1281 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1284 static __inline__ void
1285 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1287 struct timeval *stamp = &skb->stamp;
1288 if (sock_flag(sk, SOCK_RCVTSTAMP)) {
1289 /* Race occurred between timestamp enabling and packet
1290 receiving. Fill in the current time for now. */
1291 if (stamp->tv_sec == 0)
1292 do_gettimeofday(stamp);
1293 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1296 sk->sk_stamp = *stamp;
1300 * sk_eat_skb - Release a skb if it is no longer needed
1301 * @sk: socket to eat this skb from
1302 * @skb: socket buffer to eat
1304 * This routine must be called with interrupts disabled or with the socket
1305 * locked so that the sk_buff queue operation is ok.
1307 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1309 __skb_unlink(skb, &sk->sk_receive_queue);
1313 extern void sock_enable_timestamp(struct sock *sk);
1314 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1317 * Enable debug/info messages
1321 #define NETDEBUG(x) do { } while (0)
1322 #define LIMIT_NETDEBUG(x) do {} while(0)
1324 #define NETDEBUG(x) do { x; } while (0)
1325 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1329 * Macros for sleeping on a socket. Use them like this:
1331 * SOCK_SLEEP_PRE(sk)
1334 * SOCK_SLEEP_POST(sk)
1336 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1337 * and when the last use of them in DECnet has gone, I'm intending to
1341 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1342 DECLARE_WAITQUEUE(wait, tsk); \
1343 tsk->state = TASK_INTERRUPTIBLE; \
1344 add_wait_queue((sk)->sk_sleep, &wait); \
1347 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1348 remove_wait_queue((sk)->sk_sleep, &wait); \
1352 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1355 sock_set_flag(sk, bit);
1357 sock_reset_flag(sk, bit);
1360 extern __u32 sysctl_wmem_max;
1361 extern __u32 sysctl_rmem_max;
1364 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1366 static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
1372 #endif /* _SOCK_H */