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
10 * Authors: Ross Biro, <bir7@leland.Stanford.Edu>
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 sk->sk_debug debugging facility. */
65 #define SOCK_DEBUGGING
67 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && ((sk)->sk_debug)) \
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)); \
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
102 * This is the minimal network layer representation of sockets, the header
103 * for struct sock and struct tcp_tw_bucket.
106 unsigned short skc_family;
107 volatile unsigned char skc_state;
108 unsigned char skc_reuse;
109 int skc_bound_dev_if;
110 struct hlist_node skc_node;
111 struct hlist_node skc_bind_node;
116 * struct sock - network layer representation of sockets
117 * @__sk_common - shared layout with tcp_tw_bucket
118 * @sk_zapped - ax25 & ipx means !linked
119 * @sk_shutdown - mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
120 * @sk_use_write_queue - wheter to call sk->sk_write_space in sock_wfree
121 * @sk_userlocks - %SO_SNDBUF and %SO_RCVBUF settings
122 * @sk_lock - synchronizer
123 * @sk_rcvbuf - size of receive buffer in bytes
124 * @sk_sleep - sock wait queue
125 * @sk_dst_cache - destination cache
126 * @sk_dst_lock - destination cache lock
127 * @sk_policy - flow policy
128 * @sk_rmem_alloc - receive queue bytes committed
129 * @sk_receive_queue - incoming packets
130 * @sk_wmem_alloc - transmit queue bytes committed
131 * @sk_write_queue - Packet sending queue
132 * @sk_omem_alloc - "o" is "option" or "other"
133 * @sk_wmem_queued - persistent queue size
134 * @sk_forward_alloc - space allocated forward
135 * @sk_allocation - allocation mode
136 * @sk_sndbuf - size of send buffer in bytes
137 * @sk_flags - %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
138 * @sk_no_check - %SO_NO_CHECK setting, wether or not checkup packets
139 * @sk_debug - %SO_DEBUG setting
140 * @sk_rcvtstamp - %SO_TIMESTAMP setting
141 * @sk_no_largesend - whether to sent large segments or not
142 * @sk_route_caps - route capabilities (e.g. %NETIF_F_TSO)
143 * @sk_lingertime - %SO_LINGER l_linger setting
144 * @sk_hashent - hash entry in several tables (e.g. tcp_ehash)
145 * @sk_pair - socket pair (e.g. AF_UNIX/unix_peer)
146 * @sk_backlog - always used with the per-socket spinlock held
147 * @sk_callback_lock - used with the callbacks in the end of this struct
148 * @sk_error_queue - rarely used
149 * @sk_prot - protocol handlers inside a network family
150 * @sk_err - last error
151 * @sk_err_soft - errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
152 * @sk_ack_backlog - current listen backlog
153 * @sk_max_ack_backlog - listen backlog set in listen()
154 * @sk_priority - %SO_PRIORITY setting
155 * @sk_type - socket type (%SOCK_STREAM, etc)
156 * @sk_localroute - route locally only, %SO_DONTROUTE setting
157 * @sk_protocol - which protocol this socket belongs in this network family
158 * @sk_peercred - %SO_PEERCRED setting
159 * @sk_rcvlowat - %SO_RCVLOWAT setting
160 * @sk_rcvtimeo - %SO_RCVTIMEO setting
161 * @sk_sndtimeo - %SO_SNDTIMEO setting
162 * @sk_filter - socket filtering instructions
163 * @sk_protinfo - private area, net family specific, when not using slab
164 * @sk_slab - the slabcache this instance was allocated from
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 layer private data
169 * @sk_owner - module that owns this socket
170 * @sk_sndmsg_page - cached page for sendmsg
171 * @sk_sndmsg_off - cached offset for sendmsg
172 * @sk_send_head - front of stuff to transmit
173 * @sk_write_pending - a write to stream socket waits to start
174 * @sk_queue_shrunk - write queue has been shrunk recently
175 * @sk_state_change - callback to indicate change in the state of the sock
176 * @sk_data_ready - callback to indicate there is data to be processed
177 * @sk_write_space - callback to indicate there is bf sending space available
178 * @sk_error_report - callback to indicate errors (e.g. %MSG_ERRQUEUE)
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 tcp_tw_bucket 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 volatile unsigned char sk_zapped;
196 unsigned char sk_shutdown;
197 unsigned char sk_use_write_queue;
198 unsigned char sk_userlocks;
199 socket_lock_t sk_lock;
201 wait_queue_head_t *sk_sleep;
202 struct dst_entry *sk_dst_cache;
203 rwlock_t sk_dst_lock;
204 struct xfrm_policy *sk_policy[2];
205 atomic_t sk_rmem_alloc;
206 struct sk_buff_head sk_receive_queue;
207 atomic_t sk_wmem_alloc;
208 struct sk_buff_head sk_write_queue;
209 atomic_t sk_omem_alloc;
211 int sk_forward_alloc;
212 unsigned int sk_allocation;
214 unsigned long sk_flags;
216 unsigned char sk_debug;
217 unsigned char sk_rcvtstamp;
218 unsigned char sk_no_largesend;
220 unsigned long sk_lingertime;
222 struct sock *sk_pair;
224 * The backlog queue is special, it is always used with
225 * the per-socket spinlock held and requires low latency
226 * access. Therefore we special case it's implementation.
229 struct sk_buff *head;
230 struct sk_buff *tail;
232 rwlock_t sk_callback_lock;
233 struct sk_buff_head sk_error_queue;
234 struct proto *sk_prot;
237 unsigned short sk_ack_backlog;
238 unsigned short sk_max_ack_backlog;
240 unsigned short sk_type;
241 unsigned char sk_localroute;
242 unsigned char sk_protocol;
243 struct ucred sk_peercred;
247 struct sk_filter *sk_filter;
249 kmem_cache_t *sk_slab;
250 struct timer_list sk_timer;
251 struct timeval sk_stamp;
252 struct socket *sk_socket;
254 struct module *sk_owner;
255 struct page *sk_sndmsg_page;
257 struct sk_buff *sk_send_head;
258 int sk_write_pending;
260 __u8 sk_queue_shrunk;
261 /* three bytes hole, try to pack */
262 void (*sk_state_change)(struct sock *sk);
263 void (*sk_data_ready)(struct sock *sk, int bytes);
264 void (*sk_write_space)(struct sock *sk);
265 void (*sk_error_report)(struct sock *sk);
266 int (*sk_backlog_rcv)(struct sock *sk,
267 struct sk_buff *skb);
268 void (*sk_destruct)(struct sock *sk);
272 * Hashed lists helper routines
274 static inline struct sock *__sk_head(struct hlist_head *head)
276 return hlist_entry(head->first, struct sock, sk_node);
279 static inline struct sock *sk_head(struct hlist_head *head)
281 return hlist_empty(head) ? NULL : __sk_head(head);
284 static inline struct sock *sk_next(struct sock *sk)
286 return sk->sk_node.next ?
287 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
290 static inline int sk_unhashed(struct sock *sk)
292 return hlist_unhashed(&sk->sk_node);
295 static inline int sk_hashed(struct sock *sk)
297 return sk->sk_node.pprev != NULL;
300 static __inline__ void sk_node_init(struct hlist_node *node)
305 static __inline__ void __sk_del_node(struct sock *sk)
307 __hlist_del(&sk->sk_node);
310 static __inline__ int __sk_del_node_init(struct sock *sk)
314 sk_node_init(&sk->sk_node);
320 /* Grab socket reference count. This operation is valid only
321 when sk is ALREADY grabbed f.e. it is found in hash table
322 or a list and the lookup is made under lock preventing hash table
326 static inline void sock_hold(struct sock *sk)
328 atomic_inc(&sk->sk_refcnt);
331 /* Ungrab socket in the context, which assumes that socket refcnt
332 cannot hit zero, f.e. it is true in context of any socketcall.
334 static inline void __sock_put(struct sock *sk)
336 atomic_dec(&sk->sk_refcnt);
339 static __inline__ int sk_del_node_init(struct sock *sk)
341 int rc = __sk_del_node_init(sk);
344 /* paranoid for a while -acme */
345 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
351 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
353 hlist_add_head(&sk->sk_node, list);
356 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
359 __sk_add_node(sk, list);
362 static __inline__ void __sk_del_bind_node(struct sock *sk)
364 __hlist_del(&sk->sk_bind_node);
367 static __inline__ void sk_add_bind_node(struct sock *sk,
368 struct hlist_head *list)
370 hlist_add_head(&sk->sk_bind_node, list);
373 #define sk_for_each(__sk, node, list) \
374 hlist_for_each_entry(__sk, node, list, sk_node)
375 #define sk_for_each_from(__sk, node) \
376 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
377 hlist_for_each_entry_from(__sk, node, sk_node)
378 #define sk_for_each_continue(__sk, node) \
379 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
380 hlist_for_each_entry_continue(__sk, node, sk_node)
381 #define sk_for_each_safe(__sk, node, tmp, list) \
382 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
383 #define sk_for_each_bound(__sk, node, list) \
384 hlist_for_each_entry(__sk, node, list, sk_bind_node)
398 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
400 __set_bit(flag, &sk->sk_flags);
403 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
405 __clear_bit(flag, &sk->sk_flags);
408 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
410 return test_bit(flag, &sk->sk_flags);
413 static inline void sk_acceptq_removed(struct sock *sk)
415 sk->sk_ack_backlog--;
418 static inline void sk_acceptq_added(struct sock *sk)
420 sk->sk_ack_backlog++;
423 static inline int sk_acceptq_is_full(struct sock *sk)
425 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
429 * Compute minimal free write space needed to queue new packets.
431 static inline int sk_stream_min_wspace(struct sock *sk)
433 return sk->sk_wmem_queued / 2;
436 static inline int sk_stream_wspace(struct sock *sk)
438 return sk->sk_sndbuf - sk->sk_wmem_queued;
441 extern void sk_stream_write_space(struct sock *sk);
443 static inline int sk_stream_memory_free(struct sock *sk)
445 return sk->sk_wmem_queued < sk->sk_sndbuf;
448 extern void sk_stream_rfree(struct sk_buff *skb);
450 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
453 skb->destructor = sk_stream_rfree;
454 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
455 sk->sk_forward_alloc -= skb->truesize;
458 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
460 sk->sk_queue_shrunk = 1;
461 sk->sk_wmem_queued -= skb->truesize;
462 sk->sk_forward_alloc += skb->truesize;
466 /* The per-socket spinlock must be held here. */
467 #define sk_add_backlog(__sk, __skb) \
468 do { if (!(__sk)->sk_backlog.tail) { \
469 (__sk)->sk_backlog.head = \
470 (__sk)->sk_backlog.tail = (__skb); \
472 ((__sk)->sk_backlog.tail)->next = (__skb); \
473 (__sk)->sk_backlog.tail = (__skb); \
475 (__skb)->next = NULL; \
478 #define sk_wait_event(__sk, __timeo, __condition) \
480 release_sock(__sk); \
483 *(__timeo) = schedule_timeout(*(__timeo)); \
490 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
491 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
492 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
493 extern int sk_stream_error(struct sock *sk, int flags, int err);
494 extern void sk_stream_kill_queues(struct sock *sk);
496 extern int sk_wait_data(struct sock *sk, long *timeo);
498 /* Networking protocol blocks we attach to sockets.
499 * socket layer -> transport layer interface
500 * transport -> network interface is defined by struct inet_proto
503 void (*close)(struct sock *sk,
505 int (*connect)(struct sock *sk,
506 struct sockaddr *uaddr,
508 int (*disconnect)(struct sock *sk, int flags);
510 struct sock * (*accept) (struct sock *sk, int flags, int *err);
512 int (*ioctl)(struct sock *sk, int cmd,
514 int (*init)(struct sock *sk);
515 int (*destroy)(struct sock *sk);
516 void (*shutdown)(struct sock *sk, int how);
517 int (*setsockopt)(struct sock *sk, int level,
518 int optname, char __user *optval,
520 int (*getsockopt)(struct sock *sk, int level,
521 int optname, char __user *optval,
523 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
524 struct msghdr *msg, size_t len);
525 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
527 size_t len, int noblock, int flags,
529 int (*sendpage)(struct sock *sk, struct page *page,
530 int offset, size_t size, int flags);
531 int (*bind)(struct sock *sk,
532 struct sockaddr *uaddr, int addr_len);
534 int (*backlog_rcv) (struct sock *sk,
535 struct sk_buff *skb);
537 /* Keeping track of sk's, looking them up, and port selection methods. */
538 void (*hash)(struct sock *sk);
539 void (*unhash)(struct sock *sk);
540 int (*get_port)(struct sock *sk, unsigned short snum);
542 /* Memory pressure */
543 void (*enter_memory_pressure)(void);
544 atomic_t *memory_allocated; /* Current allocated memory. */
545 atomic_t *sockets_allocated; /* Current number of sockets. */
547 * Pressure flag: try to collapse.
548 * Technical note: it is used by multiple contexts non atomically.
549 * All the sk_stream_mem_schedule() is of this nature: accounting
550 * is strict, actions are advisory and have some latency.
552 int *memory_pressure;
562 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
566 static __inline__ void sk_set_owner(struct sock *sk, struct module *owner)
569 * One should use sk_set_owner just once, after struct sock creation,
570 * be it shortly after sk_alloc or after a function that returns a new
571 * struct sock (and that down the call chain called sk_alloc), e.g. the
572 * IPv4 and IPv6 modules share tcp_create_openreq_child, so if
573 * tcp_create_openreq_child called sk_set_owner IPv6 would have to
574 * change the ownership of this struct sock, with one not needed
575 * transient sk_set_owner call.
577 BUG_ON(sk->sk_owner != NULL);
579 sk->sk_owner = owner;
583 /* Called with local bh disabled */
584 static __inline__ void sock_prot_inc_use(struct proto *prot)
586 prot->stats[smp_processor_id()].inuse++;
589 static __inline__ void sock_prot_dec_use(struct proto *prot)
591 prot->stats[smp_processor_id()].inuse--;
594 /* About 10 seconds */
595 #define SOCK_DESTROY_TIME (10*HZ)
597 /* Sockets 0-1023 can't be bound to unless you are superuser */
598 #define PROT_SOCK 1024
600 #define SHUTDOWN_MASK 3
601 #define RCV_SHUTDOWN 1
602 #define SEND_SHUTDOWN 2
604 #define SOCK_SNDBUF_LOCK 1
605 #define SOCK_RCVBUF_LOCK 2
606 #define SOCK_BINDADDR_LOCK 4
607 #define SOCK_BINDPORT_LOCK 8
609 /* sock_iocb: used to kick off async processing of socket ios */
611 struct list_head list;
617 struct scm_cookie *scm;
618 struct msghdr *msg, async_msg;
619 struct iovec async_iov;
623 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
625 return (struct sock_iocb *)iocb->private;
628 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
633 struct socket_alloc {
634 struct socket socket;
635 struct inode vfs_inode;
638 static inline struct socket *SOCKET_I(struct inode *inode)
640 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
643 static inline struct inode *SOCK_INODE(struct socket *socket)
645 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
648 extern void __sk_stream_mem_reclaim(struct sock *sk);
649 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
651 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
653 static inline int sk_stream_pages(int amt)
655 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
658 static inline void sk_stream_mem_reclaim(struct sock *sk)
660 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
661 __sk_stream_mem_reclaim(sk);
664 static inline void sk_stream_writequeue_purge(struct sock *sk)
668 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
669 sk_stream_free_skb(sk, skb);
670 sk_stream_mem_reclaim(sk);
673 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
675 return (int)skb->truesize <= sk->sk_forward_alloc ||
676 sk_stream_mem_schedule(sk, skb->truesize, 1);
679 /* Used by processes to "lock" a socket state, so that
680 * interrupts and bottom half handlers won't change it
681 * from under us. It essentially blocks any incoming
682 * packets, so that we won't get any new data or any
683 * packets that change the state of the socket.
685 * While locked, BH processing will add new packets to
686 * the backlog queue. This queue is processed by the
687 * owner of the socket lock right before it is released.
689 * Since ~2.3.5 it is also exclusive sleep lock serializing
690 * accesses from user process context.
692 extern void __lock_sock(struct sock *sk);
693 extern void __release_sock(struct sock *sk);
694 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
696 extern void FASTCALL(lock_sock(struct sock *sk));
697 extern void FASTCALL(release_sock(struct sock *sk));
699 /* BH context may only use the following locking interface. */
700 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
701 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
703 extern struct sock * sk_alloc(int family, int priority, int zero_it,
705 extern void sk_free(struct sock *sk);
707 extern struct sk_buff *sock_wmalloc(struct sock *sk,
708 unsigned long size, int force,
710 extern struct sk_buff *sock_rmalloc(struct sock *sk,
711 unsigned long size, int force,
713 extern void sock_wfree(struct sk_buff *skb);
714 extern void sock_rfree(struct sk_buff *skb);
716 extern int sock_setsockopt(struct socket *sock, int level,
717 int op, char __user *optval,
720 extern int sock_getsockopt(struct socket *sock, int level,
721 int op, char __user *optval,
723 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
727 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
728 unsigned long header_len,
729 unsigned long data_len,
732 extern void *sock_kmalloc(struct sock *sk, int size, int priority);
733 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
734 extern void sk_send_sigurg(struct sock *sk);
737 * Functions to fill in entries in struct proto_ops when a protocol
738 * does not implement a particular function.
740 extern int sock_no_release(struct socket *);
741 extern int sock_no_bind(struct socket *,
742 struct sockaddr *, int);
743 extern int sock_no_connect(struct socket *,
744 struct sockaddr *, int, int);
745 extern int sock_no_socketpair(struct socket *,
747 extern int sock_no_accept(struct socket *,
748 struct socket *, int);
749 extern int sock_no_getname(struct socket *,
750 struct sockaddr *, int *, int);
751 extern unsigned int sock_no_poll(struct file *, struct socket *,
752 struct poll_table_struct *);
753 extern int sock_no_ioctl(struct socket *, unsigned int,
755 extern int sock_no_listen(struct socket *, int);
756 extern int sock_no_shutdown(struct socket *, int);
757 extern int sock_no_getsockopt(struct socket *, int , int,
758 char __user *, int __user *);
759 extern int sock_no_setsockopt(struct socket *, int, int,
761 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
762 struct msghdr *, size_t);
763 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
764 struct msghdr *, size_t, int);
765 extern int sock_no_mmap(struct file *file,
767 struct vm_area_struct *vma);
768 extern ssize_t sock_no_sendpage(struct socket *sock,
770 int offset, size_t size,
774 * Functions to fill in entries in struct proto_ops when a protocol
775 * uses the inet style.
777 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
778 char __user *optval, int __user *optlen);
779 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
780 struct msghdr *msg, size_t size, int flags);
781 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
782 char __user *optval, int optlen);
784 extern void sk_common_release(struct sock *sk);
787 * Default socket callbacks and setup code
790 extern void sock_def_destruct(struct sock *);
792 /* Initialise core socket variables */
793 extern void sock_init_data(struct socket *sock, struct sock *sk);
796 * sk_filter - run a packet through a socket filter
797 * @sk: sock associated with &sk_buff
798 * @skb: buffer to filter
799 * @needlock: set to 1 if the sock is not locked by caller.
801 * Run the filter code and then cut skb->data to correct size returned by
802 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
803 * than pkt_len we keep whole skb->data. This is the socket level
804 * wrapper to sk_run_filter. It returns 0 if the packet should
805 * be accepted or -EPERM if the packet should be tossed.
809 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
813 err = security_sock_rcv_skb(sk, skb);
818 struct sk_filter *filter;
823 filter = sk->sk_filter;
825 int pkt_len = sk_run_filter(skb, filter->insns,
830 skb_trim(skb, pkt_len);
840 * sk_filter_release: Release a socket filter
842 * @fp: filter to remove
844 * Remove a filter from a socket and release its resources.
847 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
849 unsigned int size = sk_filter_len(fp);
851 atomic_sub(size, &sk->sk_omem_alloc);
853 if (atomic_dec_and_test(&fp->refcnt))
857 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
859 atomic_inc(&fp->refcnt);
860 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
864 * Socket reference counting postulates.
866 * * Each user of socket SHOULD hold a reference count.
867 * * Each access point to socket (an hash table bucket, reference from a list,
868 * running timer, skb in flight MUST hold a reference count.
869 * * When reference count hits 0, it means it will never increase back.
870 * * When reference count hits 0, it means that no references from
871 * outside exist to this socket and current process on current CPU
872 * is last user and may/should destroy this socket.
873 * * sk_free is called from any context: process, BH, IRQ. When
874 * it is called, socket has no references from outside -> sk_free
875 * may release descendant resources allocated by the socket, but
876 * to the time when it is called, socket is NOT referenced by any
877 * hash tables, lists etc.
878 * * Packets, delivered from outside (from network or from another process)
879 * and enqueued on receive/error queues SHOULD NOT grab reference count,
880 * when they sit in queue. Otherwise, packets will leak to hole, when
881 * socket is looked up by one cpu and unhasing is made by another CPU.
882 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
883 * (leak to backlog). Packet socket does all the processing inside
884 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
885 * use separate SMP lock, so that they are prone too.
888 /* Ungrab socket and destroy it, if it was the last reference. */
889 static inline void sock_put(struct sock *sk)
891 if (atomic_dec_and_test(&sk->sk_refcnt))
895 /* Detach socket from process context.
896 * Announce socket dead, detach it from wait queue and inode.
897 * Note that parent inode held reference count on this struct sock,
898 * we do not release it in this function, because protocol
899 * probably wants some additional cleanups or even continuing
900 * to work with this socket (TCP).
902 static inline void sock_orphan(struct sock *sk)
904 write_lock_bh(&sk->sk_callback_lock);
905 sock_set_flag(sk, SOCK_DEAD);
906 sk->sk_socket = NULL;
908 write_unlock_bh(&sk->sk_callback_lock);
911 static inline void sock_graft(struct sock *sk, struct socket *parent)
913 write_lock_bh(&sk->sk_callback_lock);
914 sk->sk_sleep = &parent->wait;
916 sk->sk_socket = parent;
917 write_unlock_bh(&sk->sk_callback_lock);
920 extern int sock_i_uid(struct sock *sk);
921 extern unsigned long sock_i_ino(struct sock *sk);
923 static inline struct dst_entry *
924 __sk_dst_get(struct sock *sk)
926 return sk->sk_dst_cache;
929 static inline struct dst_entry *
930 sk_dst_get(struct sock *sk)
932 struct dst_entry *dst;
934 read_lock(&sk->sk_dst_lock);
935 dst = sk->sk_dst_cache;
938 read_unlock(&sk->sk_dst_lock);
943 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
945 struct dst_entry *old_dst;
947 old_dst = sk->sk_dst_cache;
948 sk->sk_dst_cache = dst;
949 dst_release(old_dst);
953 sk_dst_set(struct sock *sk, struct dst_entry *dst)
955 write_lock(&sk->sk_dst_lock);
956 __sk_dst_set(sk, dst);
957 write_unlock(&sk->sk_dst_lock);
961 __sk_dst_reset(struct sock *sk)
963 struct dst_entry *old_dst;
965 old_dst = sk->sk_dst_cache;
966 sk->sk_dst_cache = NULL;
967 dst_release(old_dst);
971 sk_dst_reset(struct sock *sk)
973 write_lock(&sk->sk_dst_lock);
975 write_unlock(&sk->sk_dst_lock);
978 static inline struct dst_entry *
979 __sk_dst_check(struct sock *sk, u32 cookie)
981 struct dst_entry *dst = sk->sk_dst_cache;
983 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
984 sk->sk_dst_cache = NULL;
991 static inline struct dst_entry *
992 sk_dst_check(struct sock *sk, u32 cookie)
994 struct dst_entry *dst = sk_dst_get(sk);
996 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1004 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1006 sk->sk_wmem_queued += skb->truesize;
1007 sk->sk_forward_alloc -= skb->truesize;
1010 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1011 struct sk_buff *skb, struct page *page,
1014 if (skb->ip_summed == CHECKSUM_NONE) {
1016 unsigned int csum = csum_and_copy_from_user(from,
1017 page_address(page) + off,
1021 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1022 } else if (copy_from_user(page_address(page) + off, from, copy))
1026 skb->data_len += copy;
1027 skb->truesize += copy;
1028 sk->sk_wmem_queued += copy;
1029 sk->sk_forward_alloc -= copy;
1034 * Queue a received datagram if it will fit. Stream and sequenced
1035 * protocols can't normally use this as they need to fit buffers in
1036 * and play with them.
1038 * Inlined as it's very short and called for pretty much every
1039 * packet ever received.
1042 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1046 skb->destructor = sock_wfree;
1047 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1050 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1053 skb->destructor = sock_rfree;
1054 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1057 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1058 unsigned long expires);
1060 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1062 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1067 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1068 number of warnings when compiling with -W --ANK
1070 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1071 (unsigned)sk->sk_rcvbuf) {
1076 /* It would be deadlock, if sock_queue_rcv_skb is used
1077 with socket lock! We assume that users of this
1078 function are lock free.
1080 err = sk_filter(sk, skb, 1);
1085 skb_set_owner_r(skb, sk);
1087 /* Cache the SKB length before we tack it onto the receive
1088 * queue. Once it is added it no longer belongs to us and
1089 * may be freed by other threads of control pulling packets
1094 skb_queue_tail(&sk->sk_receive_queue, skb);
1096 if (!sock_flag(sk, SOCK_DEAD))
1097 sk->sk_data_ready(sk, skb_len);
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)
1123 int err = xchg(&sk->sk_err, 0);
1127 static inline unsigned long sock_wspace(struct sock *sk)
1131 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1132 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1139 static inline void sk_wake_async(struct sock *sk, int how, int band)
1141 if (sk->sk_socket && sk->sk_socket->fasync_list)
1142 sock_wake_async(sk->sk_socket, how, band);
1145 #define SOCK_MIN_SNDBUF 2048
1146 #define SOCK_MIN_RCVBUF 256
1148 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1150 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1151 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1152 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1156 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1157 int size, int mem, int gfp)
1159 struct sk_buff *skb = alloc_skb(size + sk->sk_prot->max_header, gfp);
1162 skb->truesize += mem;
1163 if (sk->sk_forward_alloc >= (int)skb->truesize ||
1164 sk_stream_mem_schedule(sk, skb->truesize, 0)) {
1165 skb_reserve(skb, sk->sk_prot->max_header);
1170 sk->sk_prot->enter_memory_pressure();
1171 sk_stream_moderate_sndbuf(sk);
1176 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1179 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1182 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1184 struct page *page = NULL;
1186 if (sk->sk_forward_alloc >= (int)PAGE_SIZE ||
1187 sk_stream_mem_schedule(sk, PAGE_SIZE, 0))
1188 page = alloc_pages(sk->sk_allocation, 0);
1190 sk->sk_prot->enter_memory_pressure();
1191 sk_stream_moderate_sndbuf(sk);
1196 #define sk_stream_for_retrans_queue(skb, sk) \
1197 for (skb = (sk)->sk_write_queue.next; \
1198 (skb != (sk)->sk_send_head) && \
1199 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1203 * Default write policy as shown to user space via poll/select/SIGIO
1205 static inline int sock_writeable(const struct sock *sk)
1207 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1210 static inline int gfp_any(void)
1212 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1215 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1217 return noblock ? 0 : sk->sk_rcvtimeo;
1220 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1222 return noblock ? 0 : sk->sk_sndtimeo;
1225 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1227 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1230 /* Alas, with timeout socket operations are not restartable.
1231 * Compare this to poll().
1233 static inline int sock_intr_errno(long timeo)
1235 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1238 static __inline__ void
1239 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1241 struct timeval *stamp = &skb->stamp;
1242 if (sk->sk_rcvtstamp) {
1243 /* Race occurred between timestamp enabling and packet
1244 receiving. Fill in the current time for now. */
1245 if (stamp->tv_sec == 0)
1246 do_gettimeofday(stamp);
1247 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1250 sk->sk_stamp = *stamp;
1254 * sk_eat_skb - Release a skb if it is no longer needed
1255 * @sk - socket to eat this skb from
1256 * @skb - socket buffer to eat
1258 * This routine must be called with interrupts disabled or with the socket
1259 * locked so that the sk_buff queue operation is ok.
1261 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1263 __skb_unlink(skb, &sk->sk_receive_queue);
1267 extern atomic_t netstamp_needed;
1268 extern void sock_enable_timestamp(struct sock *sk);
1269 extern void sock_disable_timestamp(struct sock *sk);
1271 static inline void net_timestamp(struct timeval *stamp)
1273 if (atomic_read(&netstamp_needed))
1274 do_gettimeofday(stamp);
1281 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1284 * Enable debug/info messages
1288 #define NETDEBUG(x) do { } while (0)
1289 #define LIMIT_NETDEBUG(x) do {} while(0)
1291 #define NETDEBUG(x) do { x; } while (0)
1292 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1296 * Macros for sleeping on a socket. Use them like this:
1298 * SOCK_SLEEP_PRE(sk)
1301 * SOCK_SLEEP_POST(sk)
1303 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1304 * and when the last use of them in DECnet has gone, I'm intending to
1308 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1309 DECLARE_WAITQUEUE(wait, tsk); \
1310 tsk->state = TASK_INTERRUPTIBLE; \
1311 add_wait_queue((sk)->sk_sleep, &wait); \
1314 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1315 remove_wait_queue((sk)->sk_sleep, &wait); \
1319 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1322 sock_set_flag(sk, bit);
1324 sock_reset_flag(sk, bit);
1327 extern __u32 sysctl_wmem_max;
1328 extern __u32 sysctl_rmem_max;
1330 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1332 #endif /* _SOCK_H */