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/list.h>
44 #include <linux/timer.h>
45 #include <linux/cache.h>
46 #include <linux/module.h>
47 #include <linux/netdevice.h>
48 #include <linux/skbuff.h> /* struct sk_buff */
49 #include <linux/security.h>
51 #include <linux/filter.h>
53 #include <asm/atomic.h>
55 #include <net/checksum.h>
58 * This structure really needs to be cleaned up.
59 * Most of it is for TCP, and not used by any of
60 * the other protocols.
63 /* Define this to get the SOCK_DBG debugging facility. */
64 #define SOCK_DEBUGGING
66 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
67 printk(KERN_DEBUG msg); } while (0)
69 #define SOCK_DEBUG(sk, msg...) do { } while (0)
72 /* This is the per-socket lock. The spinlock provides a synchronization
73 * between user contexts and software interrupt processing, whereas the
74 * mini-semaphore synchronizes multiple users amongst themselves.
79 struct sock_iocb *owner;
83 #define sock_lock_init(__sk) \
84 do { spin_lock_init(&((__sk)->sk_lock.slock)); \
85 (__sk)->sk_lock.owner = NULL; \
86 init_waitqueue_head(&((__sk)->sk_lock.wq)); \
93 * struct sock_common - minimal network layer representation of sockets
94 * @skc_family: network address family
95 * @skc_state: Connection state
96 * @skc_reuse: %SO_REUSEADDR setting
97 * @skc_bound_dev_if: bound device index if != 0
98 * @skc_node: main hash linkage for various protocol lookup tables
99 * @skc_bind_node: bind hash linkage for various protocol lookup tables
100 * @skc_refcnt: reference count
101 * @skc_hash: hash value used with various protocol lookup tables
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 unsigned int skc_hash;
116 struct proto *skc_prot;
120 * struct sock - network layer representation of sockets
121 * @__sk_common: shared layout with inet_timewait_sock
122 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
123 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
124 * @sk_lock: synchronizer
125 * @sk_rcvbuf: size of receive buffer in bytes
126 * @sk_sleep: sock wait queue
127 * @sk_dst_cache: destination cache
128 * @sk_dst_lock: destination cache lock
129 * @sk_policy: flow policy
130 * @sk_rmem_alloc: receive queue bytes committed
131 * @sk_receive_queue: incoming packets
132 * @sk_wmem_alloc: transmit queue bytes committed
133 * @sk_write_queue: Packet sending queue
134 * @sk_async_wait_queue: DMA copied packets
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;
208 struct sk_buff_head sk_async_wait_queue;
210 int sk_forward_alloc;
215 unsigned long sk_flags;
216 unsigned long sk_lingertime;
218 * The backlog queue is special, it is always used with
219 * the per-socket spinlock held and requires low latency
220 * access. Therefore we special case it's implementation.
223 struct sk_buff *head;
224 struct sk_buff *tail;
226 struct sk_buff_head sk_error_queue;
227 struct proto *sk_prot_creator;
228 rwlock_t sk_callback_lock;
231 unsigned short sk_ack_backlog;
232 unsigned short sk_max_ack_backlog;
234 struct ucred sk_peercred;
237 struct sk_filter *sk_filter;
239 struct timer_list sk_timer;
240 struct timeval sk_stamp;
241 struct socket *sk_socket;
243 struct page *sk_sndmsg_page;
244 struct sk_buff *sk_send_head;
246 int sk_write_pending;
248 void (*sk_state_change)(struct sock *sk);
249 void (*sk_data_ready)(struct sock *sk, int bytes);
250 void (*sk_write_space)(struct sock *sk);
251 void (*sk_error_report)(struct sock *sk);
252 int (*sk_backlog_rcv)(struct sock *sk,
253 struct sk_buff *skb);
254 void (*sk_destruct)(struct sock *sk);
258 * Hashed lists helper routines
260 static inline struct sock *__sk_head(const struct hlist_head *head)
262 return hlist_entry(head->first, struct sock, sk_node);
265 static inline struct sock *sk_head(const struct hlist_head *head)
267 return hlist_empty(head) ? NULL : __sk_head(head);
270 static inline struct sock *sk_next(const struct sock *sk)
272 return sk->sk_node.next ?
273 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
276 static inline int sk_unhashed(const struct sock *sk)
278 return hlist_unhashed(&sk->sk_node);
281 static inline int sk_hashed(const struct sock *sk)
283 return !sk_unhashed(sk);
286 static __inline__ void sk_node_init(struct hlist_node *node)
291 static __inline__ void __sk_del_node(struct sock *sk)
293 __hlist_del(&sk->sk_node);
296 static __inline__ int __sk_del_node_init(struct sock *sk)
300 sk_node_init(&sk->sk_node);
306 /* Grab socket reference count. This operation is valid only
307 when sk is ALREADY grabbed f.e. it is found in hash table
308 or a list and the lookup is made under lock preventing hash table
312 static inline void sock_hold(struct sock *sk)
314 atomic_inc(&sk->sk_refcnt);
317 /* Ungrab socket in the context, which assumes that socket refcnt
318 cannot hit zero, f.e. it is true in context of any socketcall.
320 static inline void __sock_put(struct sock *sk)
322 atomic_dec(&sk->sk_refcnt);
325 static __inline__ int sk_del_node_init(struct sock *sk)
327 int rc = __sk_del_node_init(sk);
330 /* paranoid for a while -acme */
331 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
337 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
339 hlist_add_head(&sk->sk_node, list);
342 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
345 __sk_add_node(sk, list);
348 static __inline__ void __sk_del_bind_node(struct sock *sk)
350 __hlist_del(&sk->sk_bind_node);
353 static __inline__ void sk_add_bind_node(struct sock *sk,
354 struct hlist_head *list)
356 hlist_add_head(&sk->sk_bind_node, list);
359 #define sk_for_each(__sk, node, list) \
360 hlist_for_each_entry(__sk, node, list, sk_node)
361 #define sk_for_each_from(__sk, node) \
362 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
363 hlist_for_each_entry_from(__sk, node, sk_node)
364 #define sk_for_each_continue(__sk, node) \
365 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
366 hlist_for_each_entry_continue(__sk, node, sk_node)
367 #define sk_for_each_safe(__sk, node, tmp, list) \
368 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
369 #define sk_for_each_bound(__sk, node, list) \
370 hlist_for_each_entry(__sk, node, list, sk_bind_node)
383 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
384 SOCK_DBG, /* %SO_DEBUG setting */
385 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
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 skb_truesize_check(skb);
458 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
459 sk->sk_wmem_queued -= skb->truesize;
460 sk->sk_forward_alloc += skb->truesize;
464 /* The per-socket spinlock must be held here. */
465 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
467 if (!sk->sk_backlog.tail) {
468 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
470 sk->sk_backlog.tail->next = skb;
471 sk->sk_backlog.tail = skb;
476 #define sk_wait_event(__sk, __timeo, __condition) \
478 release_sock(__sk); \
481 *(__timeo) = schedule_timeout(*(__timeo)); \
488 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
489 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
490 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
491 extern int sk_stream_error(struct sock *sk, int flags, int err);
492 extern void sk_stream_kill_queues(struct sock *sk);
494 extern int sk_wait_data(struct sock *sk, long *timeo);
496 struct request_sock_ops;
497 struct timewait_sock_ops;
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 (*compat_setsockopt)(struct sock *sk,
526 int optname, char __user *optval,
528 int (*compat_getsockopt)(struct sock *sk,
530 int optname, char __user *optval,
532 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
533 struct msghdr *msg, size_t len);
534 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
536 size_t len, int noblock, int flags,
538 int (*sendpage)(struct sock *sk, struct page *page,
539 int offset, size_t size, int flags);
540 int (*bind)(struct sock *sk,
541 struct sockaddr *uaddr, int addr_len);
543 int (*backlog_rcv) (struct sock *sk,
544 struct sk_buff *skb);
546 /* Keeping track of sk's, looking them up, and port selection methods. */
547 void (*hash)(struct sock *sk);
548 void (*unhash)(struct sock *sk);
549 int (*get_port)(struct sock *sk, unsigned short snum);
551 /* Memory pressure */
552 void (*enter_memory_pressure)(void);
553 atomic_t *memory_allocated; /* Current allocated memory. */
554 atomic_t *sockets_allocated; /* Current number of sockets. */
556 * Pressure flag: try to collapse.
557 * Technical note: it is used by multiple contexts non atomically.
558 * All the sk_stream_mem_schedule() is of this nature: accounting
559 * is strict, actions are advisory and have some latency.
561 int *memory_pressure;
568 unsigned int obj_size;
570 atomic_t *orphan_count;
572 struct request_sock_ops *rsk_prot;
573 struct timewait_sock_ops *twsk_prot;
575 struct module *owner;
579 struct list_head node;
580 #ifdef SOCK_REFCNT_DEBUG
585 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
589 extern int proto_register(struct proto *prot, int alloc_slab);
590 extern void proto_unregister(struct proto *prot);
592 #ifdef SOCK_REFCNT_DEBUG
593 static inline void sk_refcnt_debug_inc(struct sock *sk)
595 atomic_inc(&sk->sk_prot->socks);
598 static inline void sk_refcnt_debug_dec(struct sock *sk)
600 atomic_dec(&sk->sk_prot->socks);
601 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
602 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
605 static inline void sk_refcnt_debug_release(const struct sock *sk)
607 if (atomic_read(&sk->sk_refcnt) != 1)
608 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
609 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
611 #else /* SOCK_REFCNT_DEBUG */
612 #define sk_refcnt_debug_inc(sk) do { } while (0)
613 #define sk_refcnt_debug_dec(sk) do { } while (0)
614 #define sk_refcnt_debug_release(sk) do { } while (0)
615 #endif /* SOCK_REFCNT_DEBUG */
617 /* Called with local bh disabled */
618 static __inline__ void sock_prot_inc_use(struct proto *prot)
620 prot->stats[smp_processor_id()].inuse++;
623 static __inline__ void sock_prot_dec_use(struct proto *prot)
625 prot->stats[smp_processor_id()].inuse--;
628 /* With per-bucket locks this operation is not-atomic, so that
629 * this version is not worse.
631 static inline void __sk_prot_rehash(struct sock *sk)
633 sk->sk_prot->unhash(sk);
634 sk->sk_prot->hash(sk);
637 /* About 10 seconds */
638 #define SOCK_DESTROY_TIME (10*HZ)
640 /* Sockets 0-1023 can't be bound to unless you are superuser */
641 #define PROT_SOCK 1024
643 #define SHUTDOWN_MASK 3
644 #define RCV_SHUTDOWN 1
645 #define SEND_SHUTDOWN 2
647 #define SOCK_SNDBUF_LOCK 1
648 #define SOCK_RCVBUF_LOCK 2
649 #define SOCK_BINDADDR_LOCK 4
650 #define SOCK_BINDPORT_LOCK 8
652 /* sock_iocb: used to kick off async processing of socket ios */
654 struct list_head list;
660 struct scm_cookie *scm;
661 struct msghdr *msg, async_msg;
662 struct iovec async_iov;
666 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
668 return (struct sock_iocb *)iocb->private;
671 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
676 struct socket_alloc {
677 struct socket socket;
678 struct inode vfs_inode;
681 static inline struct socket *SOCKET_I(struct inode *inode)
683 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
686 static inline struct inode *SOCK_INODE(struct socket *socket)
688 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
691 extern void __sk_stream_mem_reclaim(struct sock *sk);
692 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
694 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
696 static inline int sk_stream_pages(int amt)
698 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
701 static inline void sk_stream_mem_reclaim(struct sock *sk)
703 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
704 __sk_stream_mem_reclaim(sk);
707 static inline void sk_stream_writequeue_purge(struct sock *sk)
711 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
712 sk_stream_free_skb(sk, skb);
713 sk_stream_mem_reclaim(sk);
716 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
718 return (int)skb->truesize <= sk->sk_forward_alloc ||
719 sk_stream_mem_schedule(sk, skb->truesize, 1);
722 static inline int sk_stream_wmem_schedule(struct sock *sk, int size)
724 return size <= sk->sk_forward_alloc ||
725 sk_stream_mem_schedule(sk, size, 0);
728 /* Used by processes to "lock" a socket state, so that
729 * interrupts and bottom half handlers won't change it
730 * from under us. It essentially blocks any incoming
731 * packets, so that we won't get any new data or any
732 * packets that change the state of the socket.
734 * While locked, BH processing will add new packets to
735 * the backlog queue. This queue is processed by the
736 * owner of the socket lock right before it is released.
738 * Since ~2.3.5 it is also exclusive sleep lock serializing
739 * accesses from user process context.
741 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
743 extern void FASTCALL(lock_sock(struct sock *sk));
744 extern void FASTCALL(release_sock(struct sock *sk));
746 /* BH context may only use the following locking interface. */
747 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
748 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
750 extern struct sock *sk_alloc(int family,
752 struct proto *prot, int zero_it);
753 extern void sk_free(struct sock *sk);
754 extern struct sock *sk_clone(const struct sock *sk,
755 const gfp_t priority);
757 extern struct sk_buff *sock_wmalloc(struct sock *sk,
758 unsigned long size, int force,
760 extern struct sk_buff *sock_rmalloc(struct sock *sk,
761 unsigned long size, int force,
763 extern void sock_wfree(struct sk_buff *skb);
764 extern void sock_rfree(struct sk_buff *skb);
766 extern int sock_setsockopt(struct socket *sock, int level,
767 int op, char __user *optval,
770 extern int sock_getsockopt(struct socket *sock, int level,
771 int op, char __user *optval,
773 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
777 extern void *sock_kmalloc(struct sock *sk, int size,
779 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
780 extern void sk_send_sigurg(struct sock *sk);
783 * Functions to fill in entries in struct proto_ops when a protocol
784 * does not implement a particular function.
786 extern int sock_no_bind(struct socket *,
787 struct sockaddr *, int);
788 extern int sock_no_connect(struct socket *,
789 struct sockaddr *, int, int);
790 extern int sock_no_socketpair(struct socket *,
792 extern int sock_no_accept(struct socket *,
793 struct socket *, int);
794 extern int sock_no_getname(struct socket *,
795 struct sockaddr *, int *, int);
796 extern unsigned int sock_no_poll(struct file *, struct socket *,
797 struct poll_table_struct *);
798 extern int sock_no_ioctl(struct socket *, unsigned int,
800 extern int sock_no_listen(struct socket *, int);
801 extern int sock_no_shutdown(struct socket *, int);
802 extern int sock_no_getsockopt(struct socket *, int , int,
803 char __user *, int __user *);
804 extern int sock_no_setsockopt(struct socket *, int, int,
806 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
807 struct msghdr *, size_t);
808 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
809 struct msghdr *, size_t, int);
810 extern int sock_no_mmap(struct file *file,
812 struct vm_area_struct *vma);
813 extern ssize_t sock_no_sendpage(struct socket *sock,
815 int offset, size_t size,
819 * Functions to fill in entries in struct proto_ops when a protocol
820 * uses the inet style.
822 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
823 char __user *optval, int __user *optlen);
824 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
825 struct msghdr *msg, size_t size, int flags);
826 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
827 char __user *optval, int optlen);
828 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
829 int optname, char __user *optval, int __user *optlen);
830 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
831 int optname, char __user *optval, int optlen);
833 extern void sk_common_release(struct sock *sk);
836 * Default socket callbacks and setup code
839 /* Initialise core socket variables */
840 extern void sock_init_data(struct socket *sock, struct sock *sk);
843 * sk_filter - run a packet through a socket filter
844 * @sk: sock associated with &sk_buff
845 * @skb: buffer to filter
846 * @needlock: set to 1 if the sock is not locked by caller.
848 * Run the filter code and then cut skb->data to correct size returned by
849 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
850 * than pkt_len we keep whole skb->data. This is the socket level
851 * wrapper to sk_run_filter. It returns 0 if the packet should
852 * be accepted or -EPERM if the packet should be tossed.
856 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
860 err = security_sock_rcv_skb(sk, skb);
865 struct sk_filter *filter;
870 filter = sk->sk_filter;
872 unsigned int pkt_len = sk_run_filter(skb, filter->insns,
874 err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM;
884 * sk_filter_release: Release a socket filter
886 * @fp: filter to remove
888 * Remove a filter from a socket and release its resources.
891 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
893 unsigned int size = sk_filter_len(fp);
895 atomic_sub(size, &sk->sk_omem_alloc);
897 if (atomic_dec_and_test(&fp->refcnt))
901 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
903 atomic_inc(&fp->refcnt);
904 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
908 * Socket reference counting postulates.
910 * * Each user of socket SHOULD hold a reference count.
911 * * Each access point to socket (an hash table bucket, reference from a list,
912 * running timer, skb in flight MUST hold a reference count.
913 * * When reference count hits 0, it means it will never increase back.
914 * * When reference count hits 0, it means that no references from
915 * outside exist to this socket and current process on current CPU
916 * is last user and may/should destroy this socket.
917 * * sk_free is called from any context: process, BH, IRQ. When
918 * it is called, socket has no references from outside -> sk_free
919 * may release descendant resources allocated by the socket, but
920 * to the time when it is called, socket is NOT referenced by any
921 * hash tables, lists etc.
922 * * Packets, delivered from outside (from network or from another process)
923 * and enqueued on receive/error queues SHOULD NOT grab reference count,
924 * when they sit in queue. Otherwise, packets will leak to hole, when
925 * socket is looked up by one cpu and unhasing is made by another CPU.
926 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
927 * (leak to backlog). Packet socket does all the processing inside
928 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
929 * use separate SMP lock, so that they are prone too.
932 /* Ungrab socket and destroy it, if it was the last reference. */
933 static inline void sock_put(struct sock *sk)
935 if (atomic_dec_and_test(&sk->sk_refcnt))
939 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb);
941 /* Detach socket from process context.
942 * Announce socket dead, detach it from wait queue and inode.
943 * Note that parent inode held reference count on this struct sock,
944 * we do not release it in this function, because protocol
945 * probably wants some additional cleanups or even continuing
946 * to work with this socket (TCP).
948 static inline void sock_orphan(struct sock *sk)
950 write_lock_bh(&sk->sk_callback_lock);
951 sock_set_flag(sk, SOCK_DEAD);
952 sk->sk_socket = NULL;
954 write_unlock_bh(&sk->sk_callback_lock);
957 static inline void sock_graft(struct sock *sk, struct socket *parent)
959 write_lock_bh(&sk->sk_callback_lock);
960 sk->sk_sleep = &parent->wait;
962 sk->sk_socket = parent;
963 write_unlock_bh(&sk->sk_callback_lock);
966 extern int sock_i_uid(struct sock *sk);
967 extern unsigned long sock_i_ino(struct sock *sk);
969 static inline struct dst_entry *
970 __sk_dst_get(struct sock *sk)
972 return sk->sk_dst_cache;
975 static inline struct dst_entry *
976 sk_dst_get(struct sock *sk)
978 struct dst_entry *dst;
980 read_lock(&sk->sk_dst_lock);
981 dst = sk->sk_dst_cache;
984 read_unlock(&sk->sk_dst_lock);
989 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
991 struct dst_entry *old_dst;
993 old_dst = sk->sk_dst_cache;
994 sk->sk_dst_cache = dst;
995 dst_release(old_dst);
999 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1001 write_lock(&sk->sk_dst_lock);
1002 __sk_dst_set(sk, dst);
1003 write_unlock(&sk->sk_dst_lock);
1007 __sk_dst_reset(struct sock *sk)
1009 struct dst_entry *old_dst;
1011 old_dst = sk->sk_dst_cache;
1012 sk->sk_dst_cache = NULL;
1013 dst_release(old_dst);
1017 sk_dst_reset(struct sock *sk)
1019 write_lock(&sk->sk_dst_lock);
1021 write_unlock(&sk->sk_dst_lock);
1024 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1026 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1028 static inline void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1030 __sk_dst_set(sk, dst);
1031 sk->sk_route_caps = dst->dev->features;
1032 if (sk->sk_route_caps & NETIF_F_GSO)
1033 sk->sk_route_caps |= NETIF_F_TSO;
1034 if (sk->sk_route_caps & NETIF_F_TSO) {
1035 if (dst->header_len)
1036 sk->sk_route_caps &= ~NETIF_F_TSO;
1038 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
1042 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1044 sk->sk_wmem_queued += skb->truesize;
1045 sk->sk_forward_alloc -= skb->truesize;
1048 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1049 struct sk_buff *skb, struct page *page,
1052 if (skb->ip_summed == CHECKSUM_NONE) {
1054 unsigned int csum = csum_and_copy_from_user(from,
1055 page_address(page) + off,
1059 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1060 } else if (copy_from_user(page_address(page) + off, from, copy))
1064 skb->data_len += copy;
1065 skb->truesize += copy;
1066 sk->sk_wmem_queued += copy;
1067 sk->sk_forward_alloc -= copy;
1072 * Queue a received datagram if it will fit. Stream and sequenced
1073 * protocols can't normally use this as they need to fit buffers in
1074 * and play with them.
1076 * Inlined as it's very short and called for pretty much every
1077 * packet ever received.
1080 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1084 skb->destructor = sock_wfree;
1085 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1088 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1091 skb->destructor = sock_rfree;
1092 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1095 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1096 unsigned long expires);
1098 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1100 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1102 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1104 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1105 number of warnings when compiling with -W --ANK
1107 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1108 (unsigned)sk->sk_rcvbuf)
1110 skb_set_owner_r(skb, sk);
1111 skb_queue_tail(&sk->sk_error_queue, skb);
1112 if (!sock_flag(sk, SOCK_DEAD))
1113 sk->sk_data_ready(sk, skb->len);
1118 * Recover an error report and clear atomically
1121 static inline int sock_error(struct sock *sk)
1124 if (likely(!sk->sk_err))
1126 err = xchg(&sk->sk_err, 0);
1130 static inline unsigned long sock_wspace(struct sock *sk)
1134 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1135 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1142 static inline void sk_wake_async(struct sock *sk, int how, int band)
1144 if (sk->sk_socket && sk->sk_socket->fasync_list)
1145 sock_wake_async(sk->sk_socket, how, band);
1148 #define SOCK_MIN_SNDBUF 2048
1149 #define SOCK_MIN_RCVBUF 256
1151 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1153 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1154 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1155 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1159 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1163 struct sk_buff *skb;
1166 hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header);
1167 skb = alloc_skb_fclone(size + hdr_len, gfp);
1169 skb->truesize += mem;
1170 if (sk_stream_wmem_schedule(sk, skb->truesize)) {
1171 skb_reserve(skb, hdr_len);
1176 sk->sk_prot->enter_memory_pressure();
1177 sk_stream_moderate_sndbuf(sk);
1182 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1186 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1189 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1191 struct page *page = NULL;
1193 page = alloc_pages(sk->sk_allocation, 0);
1195 sk->sk_prot->enter_memory_pressure();
1196 sk_stream_moderate_sndbuf(sk);
1201 #define sk_stream_for_retrans_queue(skb, sk) \
1202 for (skb = (sk)->sk_write_queue.next; \
1203 (skb != (sk)->sk_send_head) && \
1204 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1207 /*from STCP for fast SACK Process*/
1208 #define sk_stream_for_retrans_queue_from(skb, sk) \
1209 for (; (skb != (sk)->sk_send_head) && \
1210 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1214 * Default write policy as shown to user space via poll/select/SIGIO
1216 static inline int sock_writeable(const struct sock *sk)
1218 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1221 static inline gfp_t gfp_any(void)
1223 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1226 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1228 return noblock ? 0 : sk->sk_rcvtimeo;
1231 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1233 return noblock ? 0 : sk->sk_sndtimeo;
1236 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1238 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1241 /* Alas, with timeout socket operations are not restartable.
1242 * Compare this to poll().
1244 static inline int sock_intr_errno(long timeo)
1246 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1249 static __inline__ void
1250 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1252 struct timeval stamp;
1254 skb_get_timestamp(skb, &stamp);
1255 if (sock_flag(sk, SOCK_RCVTSTAMP)) {
1256 /* Race occurred between timestamp enabling and packet
1257 receiving. Fill in the current time for now. */
1258 if (stamp.tv_sec == 0)
1259 do_gettimeofday(&stamp);
1260 skb_set_timestamp(skb, &stamp);
1261 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1264 sk->sk_stamp = stamp;
1268 * sk_eat_skb - Release a skb if it is no longer needed
1269 * @sk: socket to eat this skb from
1270 * @skb: socket buffer to eat
1271 * @copied_early: flag indicating whether DMA operations copied this data early
1273 * This routine must be called with interrupts disabled or with the socket
1274 * locked so that the sk_buff queue operation is ok.
1276 #ifdef CONFIG_NET_DMA
1277 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1279 __skb_unlink(skb, &sk->sk_receive_queue);
1283 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1286 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1288 __skb_unlink(skb, &sk->sk_receive_queue);
1293 extern void sock_enable_timestamp(struct sock *sk);
1294 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1297 * Enable debug/info messages
1300 #ifdef CONFIG_NETDEBUG
1301 #define NETDEBUG(fmt, args...) printk(fmt,##args)
1302 #define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0)
1304 #define NETDEBUG(fmt, args...) do { } while (0)
1305 #define LIMIT_NETDEBUG(fmt, args...) do { } while(0)
1309 * Macros for sleeping on a socket. Use them like this:
1311 * SOCK_SLEEP_PRE(sk)
1314 * SOCK_SLEEP_POST(sk)
1316 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1317 * and when the last use of them in DECnet has gone, I'm intending to
1321 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1322 DECLARE_WAITQUEUE(wait, tsk); \
1323 tsk->state = TASK_INTERRUPTIBLE; \
1324 add_wait_queue((sk)->sk_sleep, &wait); \
1327 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1328 remove_wait_queue((sk)->sk_sleep, &wait); \
1332 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1335 sock_set_flag(sk, bit);
1337 sock_reset_flag(sk, bit);
1340 extern __u32 sysctl_wmem_max;
1341 extern __u32 sysctl_rmem_max;
1344 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1346 static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
1352 extern void sk_init(void);
1354 #ifdef CONFIG_SYSCTL
1355 extern struct ctl_table core_table[];
1358 extern int sysctl_optmem_max;
1360 extern __u32 sysctl_wmem_default;
1361 extern __u32 sysctl_rmem_default;
1363 #endif /* _SOCK_H */