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;
119 struct vx_info *skc_vx_info;
121 struct nx_info *skc_nx_info;
125 * struct sock - network layer representation of sockets
126 * @__sk_common: shared layout with inet_timewait_sock
127 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
128 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
129 * @sk_lock: synchronizer
130 * @sk_rcvbuf: size of receive buffer in bytes
131 * @sk_sleep: sock wait queue
132 * @sk_dst_cache: destination cache
133 * @sk_dst_lock: destination cache lock
134 * @sk_policy: flow policy
135 * @sk_rmem_alloc: receive queue bytes committed
136 * @sk_receive_queue: incoming packets
137 * @sk_wmem_alloc: transmit queue bytes committed
138 * @sk_write_queue: Packet sending queue
139 * @sk_omem_alloc: "o" is "option" or "other"
140 * @sk_wmem_queued: persistent queue size
141 * @sk_forward_alloc: space allocated forward
142 * @sk_allocation: allocation mode
143 * @sk_sndbuf: size of send buffer in bytes
144 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
145 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
146 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
147 * @sk_lingertime: %SO_LINGER l_linger setting
148 * @sk_backlog: always used with the per-socket spinlock held
149 * @sk_callback_lock: used with the callbacks in the end of this struct
150 * @sk_error_queue: rarely used
151 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
152 * @sk_err: last error
153 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
154 * @sk_ack_backlog: current listen backlog
155 * @sk_max_ack_backlog: listen backlog set in listen()
156 * @sk_priority: %SO_PRIORITY setting
157 * @sk_type: socket type (%SOCK_STREAM, etc)
158 * @sk_protocol: which protocol this socket belongs in this network family
159 * @sk_peercred: %SO_PEERCRED setting
160 * @sk_rcvlowat: %SO_RCVLOWAT setting
161 * @sk_rcvtimeo: %SO_RCVTIMEO setting
162 * @sk_sndtimeo: %SO_SNDTIMEO setting
163 * @sk_filter: socket filtering instructions
164 * @sk_protinfo: private area, net family specific, when not using slab
165 * @sk_timer: sock cleanup timer
166 * @sk_stamp: time stamp of last packet received
167 * @sk_socket: Identd and reporting IO signals
168 * @sk_user_data: RPC and Tux layer private data
169 * @sk_sndmsg_page: cached page for sendmsg
170 * @sk_sndmsg_off: cached offset for sendmsg
171 * @sk_send_head: front of stuff to transmit
172 * @sk_security: used by security modules
173 * @sk_write_pending: a write to stream socket waits to start
174 * @sk_state_change: callback to indicate change in the state of the sock
175 * @sk_data_ready: callback to indicate there is data to be processed
176 * @sk_write_space: callback to indicate there is bf sending space available
177 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
178 * @sk_create_child - callback to get new socket events
179 * @sk_backlog_rcv: callback to process the backlog
180 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
184 * Now struct inet_timewait_sock also uses sock_common, so please just
185 * don't add nothing before this first member (__sk_common) --acme
187 struct sock_common __sk_common;
188 #define sk_family __sk_common.skc_family
189 #define sk_state __sk_common.skc_state
190 #define sk_reuse __sk_common.skc_reuse
191 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
192 #define sk_node __sk_common.skc_node
193 #define sk_bind_node __sk_common.skc_bind_node
194 #define sk_refcnt __sk_common.skc_refcnt
195 #define sk_hash __sk_common.skc_hash
196 #define sk_prot __sk_common.skc_prot
197 #define sk_xid __sk_common.skc_xid
198 #define sk_vx_info __sk_common.skc_vx_info
199 #define sk_nid __sk_common.skc_nid
200 #define sk_nx_info __sk_common.skc_nx_info
201 unsigned char sk_shutdown : 2,
204 unsigned char sk_protocol;
205 unsigned short sk_type;
207 socket_lock_t sk_lock;
208 wait_queue_head_t *sk_sleep;
209 struct dst_entry *sk_dst_cache;
210 struct xfrm_policy *sk_policy[2];
211 rwlock_t sk_dst_lock;
212 atomic_t sk_rmem_alloc;
213 atomic_t sk_wmem_alloc;
214 atomic_t sk_omem_alloc;
215 struct sk_buff_head sk_receive_queue;
216 struct sk_buff_head sk_write_queue;
218 int sk_forward_alloc;
223 unsigned long sk_flags;
224 unsigned long sk_lingertime;
226 * The backlog queue is special, it is always used with
227 * the per-socket spinlock held and requires low latency
228 * access. Therefore we special case it's implementation.
231 struct sk_buff *head;
232 struct sk_buff *tail;
234 struct sk_buff_head sk_error_queue;
235 struct proto *sk_prot_creator;
236 rwlock_t sk_callback_lock;
239 unsigned short sk_ack_backlog;
240 unsigned short sk_max_ack_backlog;
242 struct ucred sk_peercred;
245 struct sk_filter *sk_filter;
247 struct timer_list sk_timer;
248 struct timeval sk_stamp;
249 struct socket *sk_socket;
251 struct page *sk_sndmsg_page;
252 struct sk_buff *sk_send_head;
254 int sk_write_pending;
256 void (*sk_state_change)(struct sock *sk);
257 void (*sk_data_ready)(struct sock *sk, int bytes);
258 void (*sk_write_space)(struct sock *sk);
259 void (*sk_error_report)(struct sock *sk);
260 int (*sk_backlog_rcv)(struct sock *sk,
261 struct sk_buff *skb);
262 void (*sk_create_child)(struct sock *sk, struct sock *newsk);
263 void (*sk_destruct)(struct sock *sk);
267 * Hashed lists helper routines
269 static inline struct sock *__sk_head(const struct hlist_head *head)
271 return hlist_entry(head->first, struct sock, sk_node);
274 static inline struct sock *sk_head(const struct hlist_head *head)
276 return hlist_empty(head) ? NULL : __sk_head(head);
279 static inline struct sock *sk_next(const struct sock *sk)
281 return sk->sk_node.next ?
282 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
285 static inline int sk_unhashed(const struct sock *sk)
287 return hlist_unhashed(&sk->sk_node);
290 static inline int sk_hashed(const struct sock *sk)
292 return !sk_unhashed(sk);
295 static __inline__ void sk_node_init(struct hlist_node *node)
300 static __inline__ void __sk_del_node(struct sock *sk)
302 __hlist_del(&sk->sk_node);
305 static __inline__ int __sk_del_node_init(struct sock *sk)
309 sk_node_init(&sk->sk_node);
315 /* Grab socket reference count. This operation is valid only
316 when sk is ALREADY grabbed f.e. it is found in hash table
317 or a list and the lookup is made under lock preventing hash table
321 static inline void sock_hold(struct sock *sk)
323 atomic_inc(&sk->sk_refcnt);
326 /* Ungrab socket in the context, which assumes that socket refcnt
327 cannot hit zero, f.e. it is true in context of any socketcall.
329 static inline void __sock_put(struct sock *sk)
331 atomic_dec(&sk->sk_refcnt);
334 static __inline__ int sk_del_node_init(struct sock *sk)
336 int rc = __sk_del_node_init(sk);
339 /* paranoid for a while -acme */
340 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
346 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
348 hlist_add_head(&sk->sk_node, list);
351 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
354 __sk_add_node(sk, list);
357 static __inline__ void __sk_del_bind_node(struct sock *sk)
359 __hlist_del(&sk->sk_bind_node);
362 static __inline__ void sk_add_bind_node(struct sock *sk,
363 struct hlist_head *list)
365 hlist_add_head(&sk->sk_bind_node, list);
368 #define sk_for_each(__sk, node, list) \
369 hlist_for_each_entry(__sk, node, list, sk_node)
370 #define sk_for_each_from(__sk, node) \
371 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
372 hlist_for_each_entry_from(__sk, node, sk_node)
373 #define sk_for_each_continue(__sk, node) \
374 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
375 hlist_for_each_entry_continue(__sk, node, sk_node)
376 #define sk_for_each_safe(__sk, node, tmp, list) \
377 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
378 #define sk_for_each_bound(__sk, node, list) \
379 hlist_for_each_entry(__sk, node, list, sk_bind_node)
392 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
393 SOCK_DBG, /* %SO_DEBUG setting */
394 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
395 SOCK_NO_LARGESEND, /* whether to sent large segments or not */
396 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
397 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
400 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
402 nsk->sk_flags = osk->sk_flags;
405 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
407 __set_bit(flag, &sk->sk_flags);
410 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
412 __clear_bit(flag, &sk->sk_flags);
415 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
417 return test_bit(flag, &sk->sk_flags);
420 static inline void sk_acceptq_removed(struct sock *sk)
422 sk->sk_ack_backlog--;
425 static inline void sk_acceptq_added(struct sock *sk)
427 sk->sk_ack_backlog++;
430 static inline int sk_acceptq_is_full(struct sock *sk)
432 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
436 * Compute minimal free write space needed to queue new packets.
438 static inline int sk_stream_min_wspace(struct sock *sk)
440 return sk->sk_wmem_queued / 2;
443 static inline int sk_stream_wspace(struct sock *sk)
445 return sk->sk_sndbuf - sk->sk_wmem_queued;
448 extern void sk_stream_write_space(struct sock *sk);
450 static inline int sk_stream_memory_free(struct sock *sk)
452 return sk->sk_wmem_queued < sk->sk_sndbuf;
455 extern void sk_stream_rfree(struct sk_buff *skb);
457 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
460 skb->destructor = sk_stream_rfree;
461 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
462 sk->sk_forward_alloc -= skb->truesize;
465 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
467 skb_truesize_check(skb);
468 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
469 sk->sk_wmem_queued -= skb->truesize;
470 sk->sk_forward_alloc += skb->truesize;
474 /* The per-socket spinlock must be held here. */
475 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
477 if (!sk->sk_backlog.tail) {
478 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
480 sk->sk_backlog.tail->next = skb;
481 sk->sk_backlog.tail = skb;
486 #define sk_wait_event(__sk, __timeo, __condition) \
488 release_sock(__sk); \
491 *(__timeo) = schedule_timeout(*(__timeo)); \
498 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
499 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
500 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
501 extern int sk_stream_error(struct sock *sk, int flags, int err);
502 extern void sk_stream_kill_queues(struct sock *sk);
504 extern int sk_wait_data(struct sock *sk, long *timeo);
506 struct request_sock_ops;
507 struct timewait_sock_ops;
509 /* Networking protocol blocks we attach to sockets.
510 * socket layer -> transport layer interface
511 * transport -> network interface is defined by struct inet_proto
514 void (*close)(struct sock *sk,
516 int (*connect)(struct sock *sk,
517 struct sockaddr *uaddr,
519 int (*disconnect)(struct sock *sk, int flags);
521 struct sock * (*accept) (struct sock *sk, int flags, int *err);
523 int (*ioctl)(struct sock *sk, int cmd,
525 int (*init)(struct sock *sk);
526 int (*destroy)(struct sock *sk);
527 void (*shutdown)(struct sock *sk, int how);
528 int (*setsockopt)(struct sock *sk, int level,
529 int optname, char __user *optval,
531 int (*getsockopt)(struct sock *sk, int level,
532 int optname, char __user *optval,
534 int (*compat_setsockopt)(struct sock *sk,
536 int optname, char __user *optval,
538 int (*compat_getsockopt)(struct sock *sk,
540 int optname, char __user *optval,
542 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
543 struct msghdr *msg, size_t len);
544 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
546 size_t len, int noblock, int flags,
548 int (*sendpage)(struct sock *sk, struct page *page,
549 int offset, size_t size, int flags);
550 int (*bind)(struct sock *sk,
551 struct sockaddr *uaddr, int addr_len);
553 int (*backlog_rcv) (struct sock *sk,
554 struct sk_buff *skb);
556 /* Keeping track of sk's, looking them up, and port selection methods. */
557 void (*hash)(struct sock *sk);
558 void (*unhash)(struct sock *sk);
559 int (*get_port)(struct sock *sk, unsigned short snum);
561 /* Memory pressure */
562 void (*enter_memory_pressure)(void);
563 atomic_t *memory_allocated; /* Current allocated memory. */
564 atomic_t *sockets_allocated; /* Current number of sockets. */
566 * Pressure flag: try to collapse.
567 * Technical note: it is used by multiple contexts non atomically.
568 * All the sk_stream_mem_schedule() is of this nature: accounting
569 * is strict, actions are advisory and have some latency.
571 int *memory_pressure;
578 unsigned int obj_size;
580 atomic_t *orphan_count;
582 struct request_sock_ops *rsk_prot;
583 struct timewait_sock_ops *twsk_prot;
585 struct module *owner;
589 struct list_head node;
590 #ifdef SOCK_REFCNT_DEBUG
595 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
599 extern int proto_register(struct proto *prot, int alloc_slab);
600 extern void proto_unregister(struct proto *prot);
602 #ifdef SOCK_REFCNT_DEBUG
603 static inline void sk_refcnt_debug_inc(struct sock *sk)
605 atomic_inc(&sk->sk_prot->socks);
608 static inline void sk_refcnt_debug_dec(struct sock *sk)
610 atomic_dec(&sk->sk_prot->socks);
611 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
612 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
615 static inline void sk_refcnt_debug_release(const struct sock *sk)
617 if (atomic_read(&sk->sk_refcnt) != 1)
618 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
619 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
621 #else /* SOCK_REFCNT_DEBUG */
622 #define sk_refcnt_debug_inc(sk) do { } while (0)
623 #define sk_refcnt_debug_dec(sk) do { } while (0)
624 #define sk_refcnt_debug_release(sk) do { } while (0)
625 #endif /* SOCK_REFCNT_DEBUG */
627 /* Called with local bh disabled */
628 static __inline__ void sock_prot_inc_use(struct proto *prot)
630 prot->stats[smp_processor_id()].inuse++;
633 static __inline__ void sock_prot_dec_use(struct proto *prot)
635 prot->stats[smp_processor_id()].inuse--;
638 /* With per-bucket locks this operation is not-atomic, so that
639 * this version is not worse.
641 static inline void __sk_prot_rehash(struct sock *sk)
643 sk->sk_prot->unhash(sk);
644 sk->sk_prot->hash(sk);
647 /* About 10 seconds */
648 #define SOCK_DESTROY_TIME (10*HZ)
650 /* Sockets 0-1023 can't be bound to unless you are superuser */
651 #define PROT_SOCK 1024
653 #define SHUTDOWN_MASK 3
654 #define RCV_SHUTDOWN 1
655 #define SEND_SHUTDOWN 2
657 #define SOCK_SNDBUF_LOCK 1
658 #define SOCK_RCVBUF_LOCK 2
659 #define SOCK_BINDADDR_LOCK 4
660 #define SOCK_BINDPORT_LOCK 8
662 /* sock_iocb: used to kick off async processing of socket ios */
664 struct list_head list;
670 struct scm_cookie *scm;
671 struct msghdr *msg, async_msg;
672 struct iovec async_iov;
676 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
678 return (struct sock_iocb *)iocb->private;
681 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
686 struct socket_alloc {
687 struct socket socket;
688 struct inode vfs_inode;
691 static inline struct socket *SOCKET_I(struct inode *inode)
693 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
696 static inline struct inode *SOCK_INODE(struct socket *socket)
698 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
701 extern void __sk_stream_mem_reclaim(struct sock *sk);
702 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
704 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
706 static inline int sk_stream_pages(int amt)
708 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
711 static inline void sk_stream_mem_reclaim(struct sock *sk)
713 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
714 __sk_stream_mem_reclaim(sk);
717 static inline void sk_stream_writequeue_purge(struct sock *sk)
721 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
722 sk_stream_free_skb(sk, skb);
723 sk_stream_mem_reclaim(sk);
726 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
728 return (int)skb->truesize <= sk->sk_forward_alloc ||
729 sk_stream_mem_schedule(sk, skb->truesize, 1);
732 static inline int sk_stream_wmem_schedule(struct sock *sk, int size)
734 return size <= sk->sk_forward_alloc ||
735 sk_stream_mem_schedule(sk, size, 0);
738 /* Used by processes to "lock" a socket state, so that
739 * interrupts and bottom half handlers won't change it
740 * from under us. It essentially blocks any incoming
741 * packets, so that we won't get any new data or any
742 * packets that change the state of the socket.
744 * While locked, BH processing will add new packets to
745 * the backlog queue. This queue is processed by the
746 * owner of the socket lock right before it is released.
748 * Since ~2.3.5 it is also exclusive sleep lock serializing
749 * accesses from user process context.
751 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
753 extern void FASTCALL(lock_sock(struct sock *sk));
754 extern void FASTCALL(release_sock(struct sock *sk));
756 /* BH context may only use the following locking interface. */
757 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
758 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
760 extern struct sock *sk_alloc(int family,
762 struct proto *prot, int zero_it);
763 extern void sk_free(struct sock *sk);
764 extern struct sock *sk_clone(struct sock *sk,
765 const gfp_t priority);
767 extern struct sk_buff *sock_wmalloc(struct sock *sk,
768 unsigned long size, int force,
770 extern struct sk_buff *sock_rmalloc(struct sock *sk,
771 unsigned long size, int force,
773 extern void sock_wfree(struct sk_buff *skb);
774 extern void sock_rfree(struct sk_buff *skb);
776 extern int sock_setsockopt(struct socket *sock, int level,
777 int op, char __user *optval,
780 extern int sock_getsockopt(struct socket *sock, int level,
781 int op, char __user *optval,
783 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
787 extern void *sock_kmalloc(struct sock *sk, int size,
789 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
790 extern void sk_send_sigurg(struct sock *sk);
793 * Functions to fill in entries in struct proto_ops when a protocol
794 * does not implement a particular function.
796 extern int sock_no_bind(struct socket *,
797 struct sockaddr *, int);
798 extern int sock_no_connect(struct socket *,
799 struct sockaddr *, int, int);
800 extern int sock_no_socketpair(struct socket *,
802 extern int sock_no_accept(struct socket *,
803 struct socket *, int);
804 extern int sock_no_getname(struct socket *,
805 struct sockaddr *, int *, int);
806 extern unsigned int sock_no_poll(struct file *, struct socket *,
807 struct poll_table_struct *);
808 extern int sock_no_ioctl(struct socket *, unsigned int,
810 extern int sock_no_listen(struct socket *, int);
811 extern int sock_no_shutdown(struct socket *, int);
812 extern int sock_no_getsockopt(struct socket *, int , int,
813 char __user *, int __user *);
814 extern int sock_no_setsockopt(struct socket *, int, int,
816 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
817 struct msghdr *, size_t);
818 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
819 struct msghdr *, size_t, int);
820 extern int sock_no_mmap(struct file *file,
822 struct vm_area_struct *vma);
823 extern ssize_t sock_no_sendpage(struct socket *sock,
825 int offset, size_t size,
829 * Functions to fill in entries in struct proto_ops when a protocol
830 * uses the inet style.
832 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
833 char __user *optval, int __user *optlen);
834 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
835 struct msghdr *msg, size_t size, int flags);
836 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
837 char __user *optval, int optlen);
838 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
839 int optname, char __user *optval, int __user *optlen);
840 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
841 int optname, char __user *optval, int optlen);
843 extern void sk_common_release(struct sock *sk);
846 * Default socket callbacks and setup code
849 /* Initialise core socket variables */
850 extern void sock_init_data(struct socket *sock, struct sock *sk);
853 * sk_filter - run a packet through a socket filter
854 * @sk: sock associated with &sk_buff
855 * @skb: buffer to filter
856 * @needlock: set to 1 if the sock is not locked by caller.
858 * Run the filter code and then cut skb->data to correct size returned by
859 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
860 * than pkt_len we keep whole skb->data. This is the socket level
861 * wrapper to sk_run_filter. It returns 0 if the packet should
862 * be accepted or -EPERM if the packet should be tossed.
866 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
870 err = security_sock_rcv_skb(sk, skb);
875 struct sk_filter *filter;
880 filter = sk->sk_filter;
882 unsigned int pkt_len = sk_run_filter(skb, filter->insns,
887 skb_trim(skb, pkt_len);
897 * sk_filter_release: Release a socket filter
899 * @fp: filter to remove
901 * Remove a filter from a socket and release its resources.
904 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
906 unsigned int size = sk_filter_len(fp);
908 atomic_sub(size, &sk->sk_omem_alloc);
910 if (atomic_dec_and_test(&fp->refcnt))
914 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
916 atomic_inc(&fp->refcnt);
917 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
921 * Socket reference counting postulates.
923 * * Each user of socket SHOULD hold a reference count.
924 * * Each access point to socket (an hash table bucket, reference from a list,
925 * running timer, skb in flight MUST hold a reference count.
926 * * When reference count hits 0, it means it will never increase back.
927 * * When reference count hits 0, it means that no references from
928 * outside exist to this socket and current process on current CPU
929 * is last user and may/should destroy this socket.
930 * * sk_free is called from any context: process, BH, IRQ. When
931 * it is called, socket has no references from outside -> sk_free
932 * may release descendant resources allocated by the socket, but
933 * to the time when it is called, socket is NOT referenced by any
934 * hash tables, lists etc.
935 * * Packets, delivered from outside (from network or from another process)
936 * and enqueued on receive/error queues SHOULD NOT grab reference count,
937 * when they sit in queue. Otherwise, packets will leak to hole, when
938 * socket is looked up by one cpu and unhasing is made by another CPU.
939 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
940 * (leak to backlog). Packet socket does all the processing inside
941 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
942 * use separate SMP lock, so that they are prone too.
945 /* Ungrab socket and destroy it, if it was the last reference. */
946 static inline void sock_put(struct sock *sk)
948 if (atomic_dec_and_test(&sk->sk_refcnt))
952 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb);
954 /* Detach socket from process context.
955 * Announce socket dead, detach it from wait queue and inode.
956 * Note that parent inode held reference count on this struct sock,
957 * we do not release it in this function, because protocol
958 * probably wants some additional cleanups or even continuing
959 * to work with this socket (TCP).
961 static inline void sock_orphan(struct sock *sk)
963 write_lock_bh(&sk->sk_callback_lock);
964 sock_set_flag(sk, SOCK_DEAD);
965 sk->sk_socket = NULL;
967 write_unlock_bh(&sk->sk_callback_lock);
970 static inline void sock_graft(struct sock *sk, struct socket *parent)
972 write_lock_bh(&sk->sk_callback_lock);
973 sk->sk_sleep = &parent->wait;
975 sk->sk_socket = parent;
976 write_unlock_bh(&sk->sk_callback_lock);
979 extern int sock_i_uid(struct sock *sk);
980 extern unsigned long sock_i_ino(struct sock *sk);
982 static inline struct dst_entry *
983 __sk_dst_get(struct sock *sk)
985 return sk->sk_dst_cache;
988 static inline struct dst_entry *
989 sk_dst_get(struct sock *sk)
991 struct dst_entry *dst;
993 read_lock(&sk->sk_dst_lock);
994 dst = sk->sk_dst_cache;
997 read_unlock(&sk->sk_dst_lock);
1002 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1004 struct dst_entry *old_dst;
1006 old_dst = sk->sk_dst_cache;
1007 sk->sk_dst_cache = dst;
1008 dst_release(old_dst);
1012 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1014 write_lock(&sk->sk_dst_lock);
1015 __sk_dst_set(sk, dst);
1016 write_unlock(&sk->sk_dst_lock);
1020 __sk_dst_reset(struct sock *sk)
1022 struct dst_entry *old_dst;
1024 old_dst = sk->sk_dst_cache;
1025 sk->sk_dst_cache = NULL;
1026 dst_release(old_dst);
1030 sk_dst_reset(struct sock *sk)
1032 write_lock(&sk->sk_dst_lock);
1034 write_unlock(&sk->sk_dst_lock);
1037 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1039 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1041 static inline void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1043 __sk_dst_set(sk, dst);
1044 sk->sk_route_caps = dst->dev->features;
1045 if (sk->sk_route_caps & NETIF_F_TSO) {
1046 if (sock_flag(sk, SOCK_NO_LARGESEND) || dst->header_len)
1047 sk->sk_route_caps &= ~NETIF_F_TSO;
1051 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1053 sk->sk_wmem_queued += skb->truesize;
1054 sk->sk_forward_alloc -= skb->truesize;
1057 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1058 struct sk_buff *skb, struct page *page,
1061 if (skb->ip_summed == CHECKSUM_NONE) {
1063 unsigned int csum = csum_and_copy_from_user(from,
1064 page_address(page) + off,
1068 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1069 } else if (copy_from_user(page_address(page) + off, from, copy))
1073 skb->data_len += copy;
1074 skb->truesize += copy;
1075 sk->sk_wmem_queued += copy;
1076 sk->sk_forward_alloc -= copy;
1081 * Queue a received datagram if it will fit. Stream and sequenced
1082 * protocols can't normally use this as they need to fit buffers in
1083 * and play with them.
1085 * Inlined as it's very short and called for pretty much every
1086 * packet ever received.
1089 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1093 skb->destructor = sock_wfree;
1094 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1097 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1100 skb->destructor = sock_rfree;
1101 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1104 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1105 unsigned long expires);
1107 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1109 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1111 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1113 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1114 number of warnings when compiling with -W --ANK
1116 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1117 (unsigned)sk->sk_rcvbuf)
1119 skb_set_owner_r(skb, sk);
1120 skb_queue_tail(&sk->sk_error_queue, skb);
1121 if (!sock_flag(sk, SOCK_DEAD))
1122 sk->sk_data_ready(sk, skb->len);
1127 * Recover an error report and clear atomically
1130 static inline int sock_error(struct sock *sk)
1133 if (likely(!sk->sk_err))
1135 err = xchg(&sk->sk_err, 0);
1139 static inline unsigned long sock_wspace(struct sock *sk)
1143 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1144 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1151 static inline void sk_wake_async(struct sock *sk, int how, int band)
1153 if (sk->sk_socket && sk->sk_socket->fasync_list)
1154 sock_wake_async(sk->sk_socket, how, band);
1157 #define SOCK_MIN_SNDBUF 2048
1158 #define SOCK_MIN_RCVBUF 256
1160 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1162 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1163 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1164 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1168 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1172 struct sk_buff *skb;
1175 hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header);
1176 skb = alloc_skb_fclone(size + hdr_len, gfp);
1178 skb->truesize += mem;
1179 if (sk_stream_wmem_schedule(sk, skb->truesize)) {
1180 skb_reserve(skb, hdr_len);
1185 sk->sk_prot->enter_memory_pressure();
1186 sk_stream_moderate_sndbuf(sk);
1191 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1195 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1198 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1200 struct page *page = NULL;
1202 page = alloc_pages(sk->sk_allocation, 0);
1204 sk->sk_prot->enter_memory_pressure();
1205 sk_stream_moderate_sndbuf(sk);
1210 #define sk_stream_for_retrans_queue(skb, sk) \
1211 for (skb = (sk)->sk_write_queue.next; \
1212 (skb != (sk)->sk_send_head) && \
1213 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1216 /*from STCP for fast SACK Process*/
1217 #define sk_stream_for_retrans_queue_from(skb, sk) \
1218 for (; (skb != (sk)->sk_send_head) && \
1219 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1223 * Default write policy as shown to user space via poll/select/SIGIO
1225 static inline int sock_writeable(const struct sock *sk)
1227 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1230 static inline gfp_t gfp_any(void)
1232 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1235 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1237 return noblock ? 0 : sk->sk_rcvtimeo;
1240 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1242 return noblock ? 0 : sk->sk_sndtimeo;
1245 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1247 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1250 /* Alas, with timeout socket operations are not restartable.
1251 * Compare this to poll().
1253 static inline int sock_intr_errno(long timeo)
1255 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1258 static __inline__ void
1259 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1261 struct timeval stamp;
1263 skb_get_timestamp(skb, &stamp);
1264 if (sock_flag(sk, SOCK_RCVTSTAMP)) {
1265 /* Race occurred between timestamp enabling and packet
1266 receiving. Fill in the current time for now. */
1267 if (stamp.tv_sec == 0)
1268 do_gettimeofday(&stamp);
1269 skb_set_timestamp(skb, &stamp);
1270 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1273 sk->sk_stamp = stamp;
1277 * sk_eat_skb - Release a skb if it is no longer needed
1278 * @sk: socket to eat this skb from
1279 * @skb: socket buffer to eat
1281 * This routine must be called with interrupts disabled or with the socket
1282 * locked so that the sk_buff queue operation is ok.
1284 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1286 __skb_unlink(skb, &sk->sk_receive_queue);
1290 extern void sock_enable_timestamp(struct sock *sk);
1291 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1294 * Enable debug/info messages
1297 #ifdef CONFIG_NETDEBUG
1298 #define NETDEBUG(fmt, args...) printk(fmt,##args)
1299 #define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0)
1301 #define NETDEBUG(fmt, args...) do { } while (0)
1302 #define LIMIT_NETDEBUG(fmt, args...) do { } while(0)
1306 * Macros for sleeping on a socket. Use them like this:
1308 * SOCK_SLEEP_PRE(sk)
1311 * SOCK_SLEEP_POST(sk)
1313 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1314 * and when the last use of them in DECnet has gone, I'm intending to
1318 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1319 DECLARE_WAITQUEUE(wait, tsk); \
1320 tsk->state = TASK_INTERRUPTIBLE; \
1321 add_wait_queue((sk)->sk_sleep, &wait); \
1324 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1325 remove_wait_queue((sk)->sk_sleep, &wait); \
1329 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1332 sock_set_flag(sk, bit);
1334 sock_reset_flag(sk, bit);
1337 extern __u32 sysctl_wmem_max;
1338 extern __u32 sysctl_rmem_max;
1341 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1343 static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
1349 extern void sk_init(void);
1351 #ifdef CONFIG_SYSCTL
1352 extern struct ctl_table core_table[];
1355 extern int sysctl_optmem_max;
1357 extern __u32 sysctl_wmem_default;
1358 extern __u32 sysctl_rmem_default;
1360 #endif /* _SOCK_H */