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;
114 struct vx_info *skc_vx_info;
116 struct nx_info *skc_nx_info;
120 * struct sock - network layer representation of sockets
121 * @__sk_common - shared layout with tcp_tw_bucket
122 * @sk_zapped - ax25 & ipx means !linked
123 * @sk_shutdown - mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
124 * @sk_use_write_queue - wheter to call sk->sk_write_space in sock_wfree
125 * @sk_userlocks - %SO_SNDBUF and %SO_RCVBUF settings
126 * @sk_lock - synchronizer
127 * @sk_rcvbuf - size of receive buffer in bytes
128 * @sk_sleep - sock wait queue
129 * @sk_dst_cache - destination cache
130 * @sk_dst_lock - destination cache lock
131 * @sk_policy - flow policy
132 * @sk_rmem_alloc - receive queue bytes committed
133 * @sk_receive_queue - incoming packets
134 * @sk_wmem_alloc - transmit queue bytes committed
135 * @sk_write_queue - Packet sending queue
136 * @sk_omem_alloc - "o" is "option" or "other"
137 * @sk_wmem_queued - persistent queue size
138 * @sk_forward_alloc - space allocated forward
139 * @sk_allocation - allocation mode
140 * @sk_sndbuf - size of send buffer in bytes
141 * @sk_flags - %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
142 * @sk_no_check - %SO_NO_CHECK setting, wether or not checkup packets
143 * @sk_debug - %SO_DEBUG setting
144 * @sk_rcvtstamp - %SO_TIMESTAMP setting
145 * @sk_no_largesend - whether to sent large segments or not
146 * @sk_route_caps - route capabilities (e.g. %NETIF_F_TSO)
147 * @sk_lingertime - %SO_LINGER l_linger setting
148 * @sk_hashent - hash entry in several tables (e.g. tcp_ehash)
149 * @sk_pair - socket pair (e.g. AF_UNIX/unix_peer)
150 * @sk_backlog - always used with the per-socket spinlock held
151 * @sk_callback_lock - used with the callbacks in the end of this struct
152 * @sk_error_queue - rarely used
153 * @sk_prot - protocol handlers inside a network family
154 * @sk_err - last error
155 * @sk_err_soft - errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
156 * @sk_ack_backlog - current listen backlog
157 * @sk_max_ack_backlog - listen backlog set in listen()
158 * @sk_priority - %SO_PRIORITY setting
159 * @sk_type - socket type (%SOCK_STREAM, etc)
160 * @sk_localroute - route locally only, %SO_DONTROUTE setting
161 * @sk_protocol - which protocol this socket belongs in this network family
162 * @sk_peercred - %SO_PEERCRED setting
163 * @sk_rcvlowat - %SO_RCVLOWAT setting
164 * @sk_rcvtimeo - %SO_RCVTIMEO setting
165 * @sk_sndtimeo - %SO_SNDTIMEO setting
166 * @sk_filter - socket filtering instructions
167 * @sk_protinfo - private area, net family specific, when not using slab
168 * @sk_slab - the slabcache this instance was allocated from
169 * @sk_timer - sock cleanup timer
170 * @sk_stamp - time stamp of last packet received
171 * @sk_socket - Identd and reporting IO signals
172 * @sk_user_data - RPC and Tux layer private data
173 * @sk_owner - module that owns this socket
174 * @sk_sndmsg_page - cached page for sendmsg
175 * @sk_sndmsg_off - cached offset for sendmsg
176 * @sk_send_head - front of stuff to transmit
177 * @sk_write_pending - a write to stream socket waits to start
178 * @sk_queue_shrunk - write queue has been shrunk recently
179 * @sk_state_change - callback to indicate change in the state of the sock
180 * @sk_data_ready - callback to indicate there is data to be processed
181 * @sk_write_space - callback to indicate there is bf sending space available
182 * @sk_error_report - callback to indicate errors (e.g. %MSG_ERRQUEUE)
183 * @sk_create_child - callback to get new socket events
184 * @sk_backlog_rcv - callback to process the backlog
185 * @sk_destruct - called at sock freeing time, i.e. when all refcnt == 0
189 * Now struct tcp_tw_bucket also uses sock_common, so please just
190 * don't add nothing before this first member (__sk_common) --acme
192 struct sock_common __sk_common;
193 #define sk_family __sk_common.skc_family
194 #define sk_state __sk_common.skc_state
195 #define sk_reuse __sk_common.skc_reuse
196 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
197 #define sk_node __sk_common.skc_node
198 #define sk_bind_node __sk_common.skc_bind_node
199 #define sk_refcnt __sk_common.skc_refcnt
200 #define sk_xid __sk_common.skc_xid
201 #define sk_vx_info __sk_common.skc_vx_info
202 #define sk_nid __sk_common.skc_nid
203 #define sk_nx_info __sk_common.skc_nx_info
204 volatile unsigned char sk_zapped;
205 unsigned char sk_shutdown;
206 unsigned char sk_use_write_queue;
207 unsigned char sk_userlocks;
208 socket_lock_t sk_lock;
210 wait_queue_head_t *sk_sleep;
211 struct dst_entry *sk_dst_cache;
212 rwlock_t sk_dst_lock;
213 struct xfrm_policy *sk_policy[2];
214 atomic_t sk_rmem_alloc;
215 struct sk_buff_head sk_receive_queue;
216 atomic_t sk_wmem_alloc;
217 struct sk_buff_head sk_write_queue;
218 atomic_t sk_omem_alloc;
220 int sk_forward_alloc;
221 unsigned int sk_allocation;
223 unsigned long sk_flags;
225 unsigned char sk_debug;
226 unsigned char sk_rcvtstamp;
227 unsigned char sk_no_largesend;
229 unsigned long sk_lingertime;
231 struct sock *sk_pair;
233 * The backlog queue is special, it is always used with
234 * the per-socket spinlock held and requires low latency
235 * access. Therefore we special case it's implementation.
238 struct sk_buff *head;
239 struct sk_buff *tail;
241 rwlock_t sk_callback_lock;
242 struct sk_buff_head sk_error_queue;
243 struct proto *sk_prot;
246 unsigned short sk_ack_backlog;
247 unsigned short sk_max_ack_backlog;
249 unsigned short sk_type;
250 unsigned char sk_localroute;
251 unsigned char sk_protocol;
252 struct ucred sk_peercred;
256 struct sk_filter *sk_filter;
258 kmem_cache_t *sk_slab;
259 struct timer_list sk_timer;
260 struct timeval sk_stamp;
261 struct socket *sk_socket;
263 void *sk_ns; // For use by CKRM
264 struct module *sk_owner;
265 struct page *sk_sndmsg_page;
267 struct sk_buff *sk_send_head;
268 int sk_write_pending;
270 __u8 sk_queue_shrunk;
271 /* three bytes hole, try to pack */
272 void (*sk_state_change)(struct sock *sk);
273 void (*sk_data_ready)(struct sock *sk, int bytes);
274 void (*sk_write_space)(struct sock *sk);
275 void (*sk_error_report)(struct sock *sk);
276 int (*sk_backlog_rcv)(struct sock *sk,
277 struct sk_buff *skb);
278 void (*sk_create_child)(struct sock *sk, struct sock *newsk);
279 void (*sk_destruct)(struct sock *sk);
283 * Hashed lists helper routines
285 static inline struct sock *__sk_head(struct hlist_head *head)
287 return hlist_entry(head->first, struct sock, sk_node);
290 static inline struct sock *sk_head(struct hlist_head *head)
292 return hlist_empty(head) ? NULL : __sk_head(head);
295 static inline struct sock *sk_next(struct sock *sk)
297 return sk->sk_node.next ?
298 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
301 static inline int sk_unhashed(struct sock *sk)
303 return hlist_unhashed(&sk->sk_node);
306 static inline int sk_hashed(struct sock *sk)
308 return sk->sk_node.pprev != NULL;
311 static __inline__ void sk_node_init(struct hlist_node *node)
316 static __inline__ void __sk_del_node(struct sock *sk)
318 __hlist_del(&sk->sk_node);
321 static __inline__ int __sk_del_node_init(struct sock *sk)
325 sk_node_init(&sk->sk_node);
331 /* Grab socket reference count. This operation is valid only
332 when sk is ALREADY grabbed f.e. it is found in hash table
333 or a list and the lookup is made under lock preventing hash table
337 static inline void sock_hold(struct sock *sk)
339 atomic_inc(&sk->sk_refcnt);
342 /* Ungrab socket in the context, which assumes that socket refcnt
343 cannot hit zero, f.e. it is true in context of any socketcall.
345 static inline void __sock_put(struct sock *sk)
347 atomic_dec(&sk->sk_refcnt);
350 static __inline__ int sk_del_node_init(struct sock *sk)
352 int rc = __sk_del_node_init(sk);
355 /* paranoid for a while -acme */
356 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
362 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
364 hlist_add_head(&sk->sk_node, list);
367 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
370 __sk_add_node(sk, list);
373 static __inline__ void __sk_del_bind_node(struct sock *sk)
375 __hlist_del(&sk->sk_bind_node);
378 static __inline__ void sk_add_bind_node(struct sock *sk,
379 struct hlist_head *list)
381 hlist_add_head(&sk->sk_bind_node, list);
384 #define sk_for_each(__sk, node, list) \
385 hlist_for_each_entry(__sk, node, list, sk_node)
386 #define sk_for_each_from(__sk, node) \
387 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
388 hlist_for_each_entry_from(__sk, node, sk_node)
389 #define sk_for_each_continue(__sk, node) \
390 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
391 hlist_for_each_entry_continue(__sk, node, sk_node)
392 #define sk_for_each_safe(__sk, node, tmp, list) \
393 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
394 #define sk_for_each_bound(__sk, node, list) \
395 hlist_for_each_entry(__sk, node, list, sk_bind_node)
409 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
411 __set_bit(flag, &sk->sk_flags);
414 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
416 __clear_bit(flag, &sk->sk_flags);
419 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
421 return test_bit(flag, &sk->sk_flags);
424 #ifndef CONFIG_ACCEPT_QUEUES
425 static inline void sk_acceptq_removed(struct sock *sk)
427 sk->sk_ack_backlog--;
430 static inline void sk_acceptq_added(struct sock *sk)
432 sk->sk_ack_backlog++;
435 static inline int sk_acceptq_is_full(struct sock *sk)
437 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
442 * Compute minimal free write space needed to queue new packets.
444 static inline int sk_stream_min_wspace(struct sock *sk)
446 return sk->sk_wmem_queued / 2;
449 static inline int sk_stream_wspace(struct sock *sk)
451 return sk->sk_sndbuf - sk->sk_wmem_queued;
454 extern void sk_stream_write_space(struct sock *sk);
456 static inline int sk_stream_memory_free(struct sock *sk)
458 return sk->sk_wmem_queued < sk->sk_sndbuf;
461 extern void sk_stream_rfree(struct sk_buff *skb);
463 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
466 skb->destructor = sk_stream_rfree;
467 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
468 sk->sk_forward_alloc -= skb->truesize;
471 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
473 sk->sk_queue_shrunk = 1;
474 sk->sk_wmem_queued -= skb->truesize;
475 sk->sk_forward_alloc += skb->truesize;
479 /* The per-socket spinlock must be held here. */
480 #define sk_add_backlog(__sk, __skb) \
481 do { if (!(__sk)->sk_backlog.tail) { \
482 (__sk)->sk_backlog.head = \
483 (__sk)->sk_backlog.tail = (__skb); \
485 ((__sk)->sk_backlog.tail)->next = (__skb); \
486 (__sk)->sk_backlog.tail = (__skb); \
488 (__skb)->next = NULL; \
491 #define sk_wait_event(__sk, __timeo, __condition) \
493 release_sock(__sk); \
496 *(__timeo) = schedule_timeout(*(__timeo)); \
503 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
504 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
505 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
506 extern int sk_stream_error(struct sock *sk, int flags, int err);
507 extern void sk_stream_kill_queues(struct sock *sk);
509 extern int sk_wait_data(struct sock *sk, long *timeo);
511 /* Networking protocol blocks we attach to sockets.
512 * socket layer -> transport layer interface
513 * transport -> network interface is defined by struct inet_proto
516 void (*close)(struct sock *sk,
518 int (*connect)(struct sock *sk,
519 struct sockaddr *uaddr,
521 int (*disconnect)(struct sock *sk, int flags);
523 struct sock * (*accept) (struct sock *sk, int flags, int *err);
525 int (*ioctl)(struct sock *sk, int cmd,
527 int (*init)(struct sock *sk);
528 int (*destroy)(struct sock *sk);
529 void (*shutdown)(struct sock *sk, int how);
530 int (*setsockopt)(struct sock *sk, int level,
531 int optname, char __user *optval,
533 int (*getsockopt)(struct sock *sk, int level,
534 int optname, char __user *optval,
536 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
537 struct msghdr *msg, size_t len);
538 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
540 size_t len, int noblock, int flags,
542 int (*sendpage)(struct sock *sk, struct page *page,
543 int offset, size_t size, int flags);
544 int (*bind)(struct sock *sk,
545 struct sockaddr *uaddr, int addr_len);
547 int (*backlog_rcv) (struct sock *sk,
548 struct sk_buff *skb);
550 /* Keeping track of sk's, looking them up, and port selection methods. */
551 void (*hash)(struct sock *sk);
552 void (*unhash)(struct sock *sk);
553 int (*get_port)(struct sock *sk, unsigned short snum);
555 /* Memory pressure */
556 void (*enter_memory_pressure)(void);
557 atomic_t *memory_allocated; /* Current allocated memory. */
558 atomic_t *sockets_allocated; /* Current number of sockets. */
560 * Pressure flag: try to collapse.
561 * Technical note: it is used by multiple contexts non atomically.
562 * All the sk_stream_mem_schedule() is of this nature: accounting
563 * is strict, actions are advisory and have some latency.
565 int *memory_pressure;
575 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
579 static __inline__ void sk_set_owner(struct sock *sk, struct module *owner)
582 * One should use sk_set_owner just once, after struct sock creation,
583 * be it shortly after sk_alloc or after a function that returns a new
584 * struct sock (and that down the call chain called sk_alloc), e.g. the
585 * IPv4 and IPv6 modules share tcp_create_openreq_child, so if
586 * tcp_create_openreq_child called sk_set_owner IPv6 would have to
587 * change the ownership of this struct sock, with one not needed
588 * transient sk_set_owner call.
590 BUG_ON(sk->sk_owner != NULL);
592 sk->sk_owner = owner;
596 /* Called with local bh disabled */
597 static __inline__ void sock_prot_inc_use(struct proto *prot)
599 prot->stats[smp_processor_id()].inuse++;
602 static __inline__ void sock_prot_dec_use(struct proto *prot)
604 prot->stats[smp_processor_id()].inuse--;
607 /* About 10 seconds */
608 #define SOCK_DESTROY_TIME (10*HZ)
610 /* Sockets 0-1023 can't be bound to unless you are superuser */
611 #define PROT_SOCK 1024
613 #define SHUTDOWN_MASK 3
614 #define RCV_SHUTDOWN 1
615 #define SEND_SHUTDOWN 2
617 #define SOCK_SNDBUF_LOCK 1
618 #define SOCK_RCVBUF_LOCK 2
619 #define SOCK_BINDADDR_LOCK 4
620 #define SOCK_BINDPORT_LOCK 8
622 /* sock_iocb: used to kick off async processing of socket ios */
624 struct list_head list;
630 struct scm_cookie *scm;
631 struct msghdr *msg, async_msg;
632 struct iovec async_iov;
636 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
638 return (struct sock_iocb *)iocb->private;
641 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
646 struct socket_alloc {
647 struct socket socket;
648 struct inode vfs_inode;
651 static inline struct socket *SOCKET_I(struct inode *inode)
653 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
656 static inline struct inode *SOCK_INODE(struct socket *socket)
658 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
661 extern void __sk_stream_mem_reclaim(struct sock *sk);
662 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
664 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
666 static inline int sk_stream_pages(int amt)
668 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
671 static inline void sk_stream_mem_reclaim(struct sock *sk)
673 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
674 __sk_stream_mem_reclaim(sk);
677 static inline void sk_stream_writequeue_purge(struct sock *sk)
681 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
682 sk_stream_free_skb(sk, skb);
683 sk_stream_mem_reclaim(sk);
686 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
688 return (int)skb->truesize <= sk->sk_forward_alloc ||
689 sk_stream_mem_schedule(sk, skb->truesize, 1);
692 /* Used by processes to "lock" a socket state, so that
693 * interrupts and bottom half handlers won't change it
694 * from under us. It essentially blocks any incoming
695 * packets, so that we won't get any new data or any
696 * packets that change the state of the socket.
698 * While locked, BH processing will add new packets to
699 * the backlog queue. This queue is processed by the
700 * owner of the socket lock right before it is released.
702 * Since ~2.3.5 it is also exclusive sleep lock serializing
703 * accesses from user process context.
705 extern void __lock_sock(struct sock *sk);
706 extern void __release_sock(struct sock *sk);
707 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
709 extern void FASTCALL(lock_sock(struct sock *sk));
710 extern void FASTCALL(release_sock(struct sock *sk));
712 /* BH context may only use the following locking interface. */
713 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
714 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
716 extern struct sock * sk_alloc(int family, int priority, int zero_it,
718 extern void sk_free(struct sock *sk);
720 extern struct sk_buff *sock_wmalloc(struct sock *sk,
721 unsigned long size, int force,
723 extern struct sk_buff *sock_rmalloc(struct sock *sk,
724 unsigned long size, int force,
726 extern void sock_wfree(struct sk_buff *skb);
727 extern void sock_rfree(struct sk_buff *skb);
729 extern int sock_setsockopt(struct socket *sock, int level,
730 int op, char __user *optval,
733 extern int sock_getsockopt(struct socket *sock, int level,
734 int op, char __user *optval,
736 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
740 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
741 unsigned long header_len,
742 unsigned long data_len,
745 extern void *sock_kmalloc(struct sock *sk, int size, int priority);
746 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
747 extern void sk_send_sigurg(struct sock *sk);
750 * Functions to fill in entries in struct proto_ops when a protocol
751 * does not implement a particular function.
753 extern int sock_no_release(struct socket *);
754 extern int sock_no_bind(struct socket *,
755 struct sockaddr *, int);
756 extern int sock_no_connect(struct socket *,
757 struct sockaddr *, int, int);
758 extern int sock_no_socketpair(struct socket *,
760 extern int sock_no_accept(struct socket *,
761 struct socket *, int);
762 extern int sock_no_getname(struct socket *,
763 struct sockaddr *, int *, int);
764 extern unsigned int sock_no_poll(struct file *, struct socket *,
765 struct poll_table_struct *);
766 extern int sock_no_ioctl(struct socket *, unsigned int,
768 extern int sock_no_listen(struct socket *, int);
769 extern int sock_no_shutdown(struct socket *, int);
770 extern int sock_no_getsockopt(struct socket *, int , int,
771 char __user *, int __user *);
772 extern int sock_no_setsockopt(struct socket *, int, int,
774 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
775 struct msghdr *, size_t);
776 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
777 struct msghdr *, size_t, int);
778 extern int sock_no_mmap(struct file *file,
780 struct vm_area_struct *vma);
781 extern ssize_t sock_no_sendpage(struct socket *sock,
783 int offset, size_t size,
787 * Functions to fill in entries in struct proto_ops when a protocol
788 * uses the inet style.
790 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
791 char __user *optval, int __user *optlen);
792 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
793 struct msghdr *msg, size_t size, int flags);
794 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
795 char __user *optval, int optlen);
797 extern void sk_common_release(struct sock *sk);
800 * Default socket callbacks and setup code
803 extern void sock_def_destruct(struct sock *);
805 /* Initialise core socket variables */
806 extern void sock_init_data(struct socket *sock, struct sock *sk);
809 * sk_filter - run a packet through a socket filter
810 * @sk: sock associated with &sk_buff
811 * @skb: buffer to filter
812 * @needlock: set to 1 if the sock is not locked by caller.
814 * Run the filter code and then cut skb->data to correct size returned by
815 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
816 * than pkt_len we keep whole skb->data. This is the socket level
817 * wrapper to sk_run_filter. It returns 0 if the packet should
818 * be accepted or -EPERM if the packet should be tossed.
822 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
826 err = security_sock_rcv_skb(sk, skb);
831 struct sk_filter *filter;
836 filter = sk->sk_filter;
838 int pkt_len = sk_run_filter(skb, filter->insns,
843 skb_trim(skb, pkt_len);
853 * sk_filter_release: Release a socket filter
855 * @fp: filter to remove
857 * Remove a filter from a socket and release its resources.
860 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
862 unsigned int size = sk_filter_len(fp);
864 atomic_sub(size, &sk->sk_omem_alloc);
866 if (atomic_dec_and_test(&fp->refcnt))
870 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
872 atomic_inc(&fp->refcnt);
873 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
877 * Socket reference counting postulates.
879 * * Each user of socket SHOULD hold a reference count.
880 * * Each access point to socket (an hash table bucket, reference from a list,
881 * running timer, skb in flight MUST hold a reference count.
882 * * When reference count hits 0, it means it will never increase back.
883 * * When reference count hits 0, it means that no references from
884 * outside exist to this socket and current process on current CPU
885 * is last user and may/should destroy this socket.
886 * * sk_free is called from any context: process, BH, IRQ. When
887 * it is called, socket has no references from outside -> sk_free
888 * may release descendant resources allocated by the socket, but
889 * to the time when it is called, socket is NOT referenced by any
890 * hash tables, lists etc.
891 * * Packets, delivered from outside (from network or from another process)
892 * and enqueued on receive/error queues SHOULD NOT grab reference count,
893 * when they sit in queue. Otherwise, packets will leak to hole, when
894 * socket is looked up by one cpu and unhasing is made by another CPU.
895 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
896 * (leak to backlog). Packet socket does all the processing inside
897 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
898 * use separate SMP lock, so that they are prone too.
901 /* Ungrab socket and destroy it, if it was the last reference. */
902 static inline void sock_put(struct sock *sk)
904 if (atomic_dec_and_test(&sk->sk_refcnt))
908 /* Detach socket from process context.
909 * Announce socket dead, detach it from wait queue and inode.
910 * Note that parent inode held reference count on this struct sock,
911 * we do not release it in this function, because protocol
912 * probably wants some additional cleanups or even continuing
913 * to work with this socket (TCP).
915 static inline void sock_orphan(struct sock *sk)
917 write_lock_bh(&sk->sk_callback_lock);
918 sock_set_flag(sk, SOCK_DEAD);
919 sk->sk_socket = NULL;
921 write_unlock_bh(&sk->sk_callback_lock);
924 static inline void sock_graft(struct sock *sk, struct socket *parent)
926 write_lock_bh(&sk->sk_callback_lock);
927 sk->sk_sleep = &parent->wait;
929 sk->sk_socket = parent;
930 write_unlock_bh(&sk->sk_callback_lock);
933 extern int sock_i_uid(struct sock *sk);
934 extern unsigned long sock_i_ino(struct sock *sk);
936 static inline struct dst_entry *
937 __sk_dst_get(struct sock *sk)
939 return sk->sk_dst_cache;
942 static inline struct dst_entry *
943 sk_dst_get(struct sock *sk)
945 struct dst_entry *dst;
947 read_lock(&sk->sk_dst_lock);
948 dst = sk->sk_dst_cache;
951 read_unlock(&sk->sk_dst_lock);
956 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
958 struct dst_entry *old_dst;
960 old_dst = sk->sk_dst_cache;
961 sk->sk_dst_cache = dst;
962 dst_release(old_dst);
966 sk_dst_set(struct sock *sk, struct dst_entry *dst)
968 write_lock(&sk->sk_dst_lock);
969 __sk_dst_set(sk, dst);
970 write_unlock(&sk->sk_dst_lock);
974 __sk_dst_reset(struct sock *sk)
976 struct dst_entry *old_dst;
978 old_dst = sk->sk_dst_cache;
979 sk->sk_dst_cache = NULL;
980 dst_release(old_dst);
984 sk_dst_reset(struct sock *sk)
986 write_lock(&sk->sk_dst_lock);
988 write_unlock(&sk->sk_dst_lock);
991 static inline struct dst_entry *
992 __sk_dst_check(struct sock *sk, u32 cookie)
994 struct dst_entry *dst = sk->sk_dst_cache;
996 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
997 sk->sk_dst_cache = NULL;
1004 static inline struct dst_entry *
1005 sk_dst_check(struct sock *sk, u32 cookie)
1007 struct dst_entry *dst = sk_dst_get(sk);
1009 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1017 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1019 sk->sk_wmem_queued += skb->truesize;
1020 sk->sk_forward_alloc -= skb->truesize;
1023 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1024 struct sk_buff *skb, struct page *page,
1027 if (skb->ip_summed == CHECKSUM_NONE) {
1029 unsigned int csum = csum_and_copy_from_user(from,
1030 page_address(page) + off,
1034 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1035 } else if (copy_from_user(page_address(page) + off, from, copy))
1039 skb->data_len += copy;
1040 skb->truesize += copy;
1041 sk->sk_wmem_queued += copy;
1042 sk->sk_forward_alloc -= copy;
1047 * Queue a received datagram if it will fit. Stream and sequenced
1048 * protocols can't normally use this as they need to fit buffers in
1049 * and play with them.
1051 * Inlined as it's very short and called for pretty much every
1052 * packet ever received.
1055 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1059 skb->destructor = sock_wfree;
1060 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1063 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1066 skb->destructor = sock_rfree;
1067 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1070 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1071 unsigned long expires);
1073 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1075 extern struct proto_ops inet_stream_ops;
1077 extern int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len);
1079 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1084 /* Silently drop if VNET is active (if INET bind() has been
1085 * overridden) and the context is not entitled to read the
1088 if (inet_stream_ops.bind != inet_bind &&
1089 (int) sk->sk_xid > 0 && sk->sk_xid != skb->xid) {
1090 err = 0; /* ensures that it is silent */
1091 kfree_skb(skb); /* discard frame */
1095 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1096 number of warnings when compiling with -W --ANK
1098 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1099 (unsigned)sk->sk_rcvbuf) {
1104 /* It would be deadlock, if sock_queue_rcv_skb is used
1105 with socket lock! We assume that users of this
1106 function are lock free.
1108 err = sk_filter(sk, skb, 1);
1113 skb_set_owner_r(skb, sk);
1115 /* Cache the SKB length before we tack it onto the receive
1116 * queue. Once it is added it no longer belongs to us and
1117 * may be freed by other threads of control pulling packets
1122 skb_queue_tail(&sk->sk_receive_queue, skb);
1124 if (!sock_flag(sk, SOCK_DEAD))
1125 sk->sk_data_ready(sk, skb_len);
1130 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1132 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1133 number of warnings when compiling with -W --ANK
1135 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1136 (unsigned)sk->sk_rcvbuf)
1138 skb_set_owner_r(skb, sk);
1139 skb_queue_tail(&sk->sk_error_queue, skb);
1140 if (!sock_flag(sk, SOCK_DEAD))
1141 sk->sk_data_ready(sk, skb->len);
1146 * Recover an error report and clear atomically
1149 static inline int sock_error(struct sock *sk)
1151 int err = xchg(&sk->sk_err, 0);
1155 static inline unsigned long sock_wspace(struct sock *sk)
1159 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1160 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1167 static inline void sk_wake_async(struct sock *sk, int how, int band)
1169 if (sk->sk_socket && sk->sk_socket->fasync_list)
1170 sock_wake_async(sk->sk_socket, how, band);
1173 #define SOCK_MIN_SNDBUF 2048
1174 #define SOCK_MIN_RCVBUF 256
1176 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1178 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1179 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1180 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1184 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1185 int size, int mem, int gfp)
1187 struct sk_buff *skb = alloc_skb(size + sk->sk_prot->max_header, gfp);
1190 skb->truesize += mem;
1191 if (sk->sk_forward_alloc >= (int)skb->truesize ||
1192 sk_stream_mem_schedule(sk, skb->truesize, 0)) {
1193 skb_reserve(skb, sk->sk_prot->max_header);
1198 sk->sk_prot->enter_memory_pressure();
1199 sk_stream_moderate_sndbuf(sk);
1204 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1207 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1210 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1212 struct page *page = NULL;
1214 if (sk->sk_forward_alloc >= (int)PAGE_SIZE ||
1215 sk_stream_mem_schedule(sk, PAGE_SIZE, 0))
1216 page = alloc_pages(sk->sk_allocation, 0);
1218 sk->sk_prot->enter_memory_pressure();
1219 sk_stream_moderate_sndbuf(sk);
1224 #define sk_stream_for_retrans_queue(skb, sk) \
1225 for (skb = (sk)->sk_write_queue.next; \
1226 (skb != (sk)->sk_send_head) && \
1227 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1231 * Default write policy as shown to user space via poll/select/SIGIO
1233 static inline int sock_writeable(const struct sock *sk)
1235 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1238 static inline int gfp_any(void)
1240 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1243 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1245 return noblock ? 0 : sk->sk_rcvtimeo;
1248 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1250 return noblock ? 0 : sk->sk_sndtimeo;
1253 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1255 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1258 /* Alas, with timeout socket operations are not restartable.
1259 * Compare this to poll().
1261 static inline int sock_intr_errno(long timeo)
1263 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1266 static __inline__ void
1267 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1269 struct timeval *stamp = &skb->stamp;
1270 if (sk->sk_rcvtstamp) {
1271 /* Race occurred between timestamp enabling and packet
1272 receiving. Fill in the current time for now. */
1273 if (stamp->tv_sec == 0)
1274 do_gettimeofday(stamp);
1275 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1278 sk->sk_stamp = *stamp;
1282 * sk_eat_skb - Release a skb if it is no longer needed
1283 * @sk - socket to eat this skb from
1284 * @skb - socket buffer to eat
1286 * This routine must be called with interrupts disabled or with the socket
1287 * locked so that the sk_buff queue operation is ok.
1289 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1291 __skb_unlink(skb, &sk->sk_receive_queue);
1295 extern atomic_t netstamp_needed;
1296 extern void sock_enable_timestamp(struct sock *sk);
1297 extern void sock_disable_timestamp(struct sock *sk);
1299 static inline void net_timestamp(struct timeval *stamp)
1301 if (atomic_read(&netstamp_needed))
1302 do_gettimeofday(stamp);
1309 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1312 * Enable debug/info messages
1316 #define NETDEBUG(x) do { } while (0)
1317 #define LIMIT_NETDEBUG(x) do {} while(0)
1319 #define NETDEBUG(x) do { x; } while (0)
1320 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1324 * Macros for sleeping on a socket. Use them like this:
1326 * SOCK_SLEEP_PRE(sk)
1329 * SOCK_SLEEP_POST(sk)
1331 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1332 * and when the last use of them in DECnet has gone, I'm intending to
1336 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1337 DECLARE_WAITQUEUE(wait, tsk); \
1338 tsk->state = TASK_INTERRUPTIBLE; \
1339 add_wait_queue((sk)->sk_sleep, &wait); \
1342 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1343 remove_wait_queue((sk)->sk_sleep, &wait); \
1347 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1350 sock_set_flag(sk, bit);
1352 sock_reset_flag(sk, bit);
1355 extern __u32 sysctl_wmem_max;
1356 extern __u32 sysctl_rmem_max;
1358 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1360 #endif /* _SOCK_H */