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>
58 * This structure really needs to be cleaned up.
59 * Most of it is for TCP, and not used by any of
60 * the other protocols.
63 /* Define this to get the sk->sk_debug debugging facility. */
64 //#define SOCK_DEBUGGING
66 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && ((sk)->sk_debug)) \
67 printk(KERN_DEBUG msg); } while (0)
69 #define SOCK_DEBUG(sk, msg...) do { } while (0)
72 /* This is the per-socket lock. The spinlock provides a synchronization
73 * between user contexts and software interrupt processing, whereas the
74 * mini-semaphore synchronizes multiple users amongst themselves.
79 struct sock_iocb *owner;
83 #define sock_lock_init(__sk) \
84 do { spin_lock_init(&((__sk)->sk_lock.slock)); \
85 (__sk)->sk_lock.owner = NULL; \
86 init_waitqueue_head(&((__sk)->sk_lock.wq)); \
92 * struct sock_common - minimal network layer representation of sockets
93 * @skc_family - network address family
94 * @skc_state - Connection state
95 * @skc_reuse - %SO_REUSEADDR setting
96 * @skc_bound_dev_if - bound device index if != 0
97 * @skc_node - main hash linkage for various protocol lookup tables
98 * @skc_bind_node - bind hash linkage for various protocol lookup tables
99 * @skc_refcnt - reference count
101 * This is the minimal network layer representation of sockets, the header
102 * for struct sock and struct tcp_tw_bucket.
105 unsigned short skc_family;
106 volatile unsigned char skc_state;
107 unsigned char skc_reuse;
108 int skc_bound_dev_if;
109 struct hlist_node skc_node;
110 struct hlist_node skc_bind_node;
115 * struct sock - network layer representation of sockets
116 * @__sk_common - shared layout with tcp_tw_bucket
117 * @sk_zapped - ax25 & ipx means !linked
118 * @sk_shutdown - mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
119 * @sk_use_write_queue - wheter to call sk->sk_write_space in sock_wfree
120 * @sk_userlocks - %SO_SNDBUF and %SO_RCVBUF settings
121 * @sk_lock - synchronizer
122 * @sk_rcvbuf - size of receive buffer in bytes
123 * @sk_sleep - sock wait queue
124 * @sk_dst_cache - destination cache
125 * @sk_dst_lock - destination cache lock
126 * @sk_policy - flow policy
127 * @sk_rmem_alloc - receive queue bytes committed
128 * @sk_receive_queue - incoming packets
129 * @sk_wmem_alloc - transmit queue bytes committed
130 * @sk_write_queue - Packet sending queue
131 * @sk_omem_alloc - "o" is "option" or "other"
132 * @sk_wmem_queued - persistent queue size
133 * @sk_forward_alloc - space allocated forward
134 * @sk_allocation - allocation mode
135 * @sk_sndbuf - size of send buffer in bytes
136 * @sk_flags - %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
137 * @sk_no_check - %SO_NO_CHECK setting, wether or not checkup packets
138 * @sk_debug - %SO_DEBUG setting
139 * @sk_rcvtstamp - %SO_TIMESTAMP setting
140 * @sk_no_largesend - whether to sent large segments or not
141 * @sk_route_caps - route capabilities (e.g. %NETIF_F_TSO)
142 * @sk_lingertime - %SO_LINGER l_linger setting
143 * @sk_hashent - hash entry in several tables (e.g. tcp_ehash)
144 * @sk_pair - socket pair (e.g. AF_UNIX/unix_peer)
145 * @sk_backlog - always used with the per-socket spinlock held
146 * @sk_callback_lock - used with the callbacks in the end of this struct
147 * @sk_error_queue - rarely used
148 * @sk_prot - protocol handlers inside a network family
149 * @sk_err - last error
150 * @sk_err_soft - errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
151 * @sk_ack_backlog - current listen backlog
152 * @sk_max_ack_backlog - listen backlog set in listen()
153 * @sk_priority - %SO_PRIORITY setting
154 * @sk_type - socket type (%SOCK_STREAM, etc)
155 * @sk_localroute - route locally only, %SO_DONTROUTE setting
156 * @sk_protocol - which protocol this socket belongs in this network family
157 * @sk_peercred - %SO_PEERCRED setting
158 * @sk_rcvlowat - %SO_RCVLOWAT setting
159 * @sk_rcvtimeo - %SO_RCVTIMEO setting
160 * @sk_sndtimeo - %SO_SNDTIMEO setting
161 * @sk_filter - socket filtering instructions
162 * @sk_protinfo - private area, net family specific, when not using slab
163 * @sk_slab - the slabcache this instance was allocated from
164 * @sk_timer - sock cleanup timer
165 * @sk_stamp - time stamp of last packet received
166 * @sk_socket - Identd and reporting IO signals
167 * @sk_user_data - RPC and Tux layer private data
168 * @sk_owner - module that owns this socket
169 * @sk_state_change - callback to indicate change in the state of the sock
170 * @sk_data_ready - callback to indicate there is data to be processed
171 * @sk_write_space - callback to indicate there is bf sending space available
172 * @sk_error_report - callback to indicate errors (e.g. %MSG_ERRQUEUE)
173 * @sk_create_child - callback to get new socket events
174 * @sk_backlog_rcv - callback to process the backlog
175 * @sk_destruct - called at sock freeing time, i.e. when all refcnt == 0
179 * Now struct tcp_tw_bucket also uses sock_common, so please just
180 * don't add nothing before this first member (__sk_common) --acme
182 struct sock_common __sk_common;
183 #define sk_family __sk_common.skc_family
184 #define sk_state __sk_common.skc_state
185 #define sk_reuse __sk_common.skc_reuse
186 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
187 #define sk_node __sk_common.skc_node
188 #define sk_bind_node __sk_common.skc_bind_node
189 #define sk_refcnt __sk_common.skc_refcnt
190 volatile unsigned char sk_zapped;
191 unsigned char sk_shutdown;
192 unsigned char sk_use_write_queue;
193 unsigned char sk_userlocks;
194 socket_lock_t sk_lock;
196 wait_queue_head_t *sk_sleep;
197 struct dst_entry *sk_dst_cache;
198 rwlock_t sk_dst_lock;
199 struct xfrm_policy *sk_policy[2];
200 atomic_t sk_rmem_alloc;
201 struct sk_buff_head sk_receive_queue;
202 atomic_t sk_wmem_alloc;
203 struct sk_buff_head sk_write_queue;
204 atomic_t sk_omem_alloc;
206 int sk_forward_alloc;
207 unsigned int sk_allocation;
209 unsigned long sk_flags;
211 unsigned char sk_debug;
212 unsigned char sk_rcvtstamp;
213 unsigned char sk_no_largesend;
215 unsigned long sk_lingertime;
217 struct sock *sk_pair;
219 * The backlog queue is special, it is always used with
220 * the per-socket spinlock held and requires low latency
221 * access. Therefore we special case it's implementation.
224 struct sk_buff *head;
225 struct sk_buff *tail;
227 rwlock_t sk_callback_lock;
228 struct sk_buff_head sk_error_queue;
229 struct proto *sk_prot;
232 unsigned short sk_ack_backlog;
233 unsigned short sk_max_ack_backlog;
235 unsigned short sk_type;
236 unsigned char sk_localroute;
237 unsigned char sk_protocol;
238 struct ucred sk_peercred;
242 struct sk_filter *sk_filter;
244 kmem_cache_t *sk_slab;
245 struct timer_list sk_timer;
246 struct timeval sk_stamp;
247 struct socket *sk_socket;
249 struct module *sk_owner;
251 void (*sk_state_change)(struct sock *sk);
252 void (*sk_data_ready)(struct sock *sk, int bytes);
253 void (*sk_write_space)(struct sock *sk);
254 void (*sk_error_report)(struct sock *sk);
255 int (*sk_backlog_rcv)(struct sock *sk,
256 struct sk_buff *skb);
257 void (*sk_create_child)(struct sock *sk, struct sock *newsk);
258 void (*sk_destruct)(struct sock *sk);
262 * Hashed lists helper routines
264 static inline struct sock *__sk_head(struct hlist_head *head)
266 return hlist_entry(head->first, struct sock, sk_node);
269 static inline struct sock *sk_head(struct hlist_head *head)
271 return hlist_empty(head) ? NULL : __sk_head(head);
274 static inline struct sock *sk_next(struct sock *sk)
276 return sk->sk_node.next ?
277 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
280 static inline int sk_unhashed(struct sock *sk)
282 return hlist_unhashed(&sk->sk_node);
285 static inline int sk_hashed(struct sock *sk)
287 return sk->sk_node.pprev != NULL;
290 static __inline__ void sk_node_init(struct hlist_node *node)
295 static __inline__ void __sk_del_node(struct sock *sk)
297 __hlist_del(&sk->sk_node);
300 static __inline__ int __sk_del_node_init(struct sock *sk)
304 sk_node_init(&sk->sk_node);
310 /* Grab socket reference count. This operation is valid only
311 when sk is ALREADY grabbed f.e. it is found in hash table
312 or a list and the lookup is made under lock preventing hash table
316 static inline void sock_hold(struct sock *sk)
318 atomic_inc(&sk->sk_refcnt);
321 /* Ungrab socket in the context, which assumes that socket refcnt
322 cannot hit zero, f.e. it is true in context of any socketcall.
324 static inline void __sock_put(struct sock *sk)
326 atomic_dec(&sk->sk_refcnt);
329 static __inline__ int sk_del_node_init(struct sock *sk)
331 int rc = __sk_del_node_init(sk);
334 /* paranoid for a while -acme */
335 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
341 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
343 hlist_add_head(&sk->sk_node, list);
346 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
349 __sk_add_node(sk, list);
352 static __inline__ void __sk_del_bind_node(struct sock *sk)
354 __hlist_del(&sk->sk_bind_node);
357 static __inline__ void sk_add_bind_node(struct sock *sk,
358 struct hlist_head *list)
360 hlist_add_head(&sk->sk_bind_node, list);
363 #define sk_for_each(__sk, node, list) \
364 hlist_for_each_entry(__sk, node, list, sk_node)
365 #define sk_for_each_from(__sk, node) \
366 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
367 hlist_for_each_entry_from(__sk, node, sk_node)
368 #define sk_for_each_continue(__sk, node) \
369 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
370 hlist_for_each_entry_continue(__sk, node, sk_node)
371 #define sk_for_each_safe(__sk, node, tmp, list) \
372 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
373 #define sk_for_each_bound(__sk, node, list) \
374 hlist_for_each_entry(__sk, node, list, sk_bind_node)
388 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
390 __set_bit(flag, &sk->sk_flags);
393 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
395 __clear_bit(flag, &sk->sk_flags);
398 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
400 return test_bit(flag, &sk->sk_flags);
403 static inline void sk_acceptq_removed(struct sock *sk)
405 sk->sk_ack_backlog--;
408 static inline void sk_acceptq_added(struct sock *sk)
410 sk->sk_ack_backlog++;
413 static inline int sk_acceptq_is_full(struct sock *sk)
415 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
419 * Compute minimal free write space needed to queue new packets.
421 static inline int sk_stream_min_wspace(struct sock *sk)
423 return sk->sk_wmem_queued / 2;
426 static inline int sk_stream_wspace(struct sock *sk)
428 return sk->sk_sndbuf - sk->sk_wmem_queued;
431 extern void sk_stream_write_space(struct sock *sk);
433 /* The per-socket spinlock must be held here. */
434 #define sk_add_backlog(__sk, __skb) \
435 do { if (!(__sk)->sk_backlog.tail) { \
436 (__sk)->sk_backlog.head = \
437 (__sk)->sk_backlog.tail = (__skb); \
439 ((__sk)->sk_backlog.tail)->next = (__skb); \
440 (__sk)->sk_backlog.tail = (__skb); \
442 (__skb)->next = NULL; \
445 #define sk_wait_event(__sk, __timeo, __condition) \
447 release_sock(__sk); \
450 *(__timeo) = schedule_timeout(*(__timeo)); \
457 extern int sk_wait_data(struct sock *sk, long *timeo);
459 /* IP protocol blocks we attach to sockets.
460 * socket layer -> transport layer interface
461 * transport -> network interface is defined by struct inet_proto
464 void (*close)(struct sock *sk,
466 int (*connect)(struct sock *sk,
467 struct sockaddr *uaddr,
469 int (*disconnect)(struct sock *sk, int flags);
471 struct sock * (*accept) (struct sock *sk, int flags, int *err);
473 int (*ioctl)(struct sock *sk, int cmd,
475 int (*init)(struct sock *sk);
476 int (*destroy)(struct sock *sk);
477 void (*shutdown)(struct sock *sk, int how);
478 int (*setsockopt)(struct sock *sk, int level,
479 int optname, char __user *optval,
481 int (*getsockopt)(struct sock *sk, int level,
482 int optname, char __user *optval,
484 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
485 struct msghdr *msg, size_t len);
486 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
488 size_t len, int noblock, int flags,
490 int (*sendpage)(struct sock *sk, struct page *page,
491 int offset, size_t size, int flags);
492 int (*bind)(struct sock *sk,
493 struct sockaddr *uaddr, int addr_len);
495 int (*backlog_rcv) (struct sock *sk,
496 struct sk_buff *skb);
498 /* Keeping track of sk's, looking them up, and port selection methods. */
499 void (*hash)(struct sock *sk);
500 void (*unhash)(struct sock *sk);
501 int (*get_port)(struct sock *sk, unsigned short snum);
507 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
511 static __inline__ void sk_set_owner(struct sock *sk, struct module *owner)
514 * One should use sk_set_owner just once, after struct sock creation,
515 * be it shortly after sk_alloc or after a function that returns a new
516 * struct sock (and that down the call chain called sk_alloc), e.g. the
517 * IPv4 and IPv6 modules share tcp_create_openreq_child, so if
518 * tcp_create_openreq_child called sk_set_owner IPv6 would have to
519 * change the ownership of this struct sock, with one not needed
520 * transient sk_set_owner call.
522 BUG_ON(sk->sk_owner != NULL);
524 sk->sk_owner = owner;
528 /* Called with local bh disabled */
529 static __inline__ void sock_prot_inc_use(struct proto *prot)
531 prot->stats[smp_processor_id()].inuse++;
534 static __inline__ void sock_prot_dec_use(struct proto *prot)
536 prot->stats[smp_processor_id()].inuse--;
539 /* About 10 seconds */
540 #define SOCK_DESTROY_TIME (10*HZ)
542 /* Sockets 0-1023 can't be bound to unless you are superuser */
543 #define PROT_SOCK 1024
545 #define SHUTDOWN_MASK 3
546 #define RCV_SHUTDOWN 1
547 #define SEND_SHUTDOWN 2
549 #define SOCK_SNDBUF_LOCK 1
550 #define SOCK_RCVBUF_LOCK 2
551 #define SOCK_BINDADDR_LOCK 4
552 #define SOCK_BINDPORT_LOCK 8
554 /* sock_iocb: used to kick off async processing of socket ios */
556 struct list_head list;
562 struct scm_cookie *scm;
563 struct msghdr *msg, async_msg;
564 struct iovec async_iov;
567 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
569 BUG_ON(sizeof(struct sock_iocb) > KIOCB_PRIVATE_SIZE);
570 return (struct sock_iocb *)iocb->private;
573 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
575 return container_of((void *)si, struct kiocb, private);
578 struct socket_alloc {
579 struct socket socket;
580 struct inode vfs_inode;
583 static inline struct socket *SOCKET_I(struct inode *inode)
585 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
588 static inline struct inode *SOCK_INODE(struct socket *socket)
590 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
593 /* Used by processes to "lock" a socket state, so that
594 * interrupts and bottom half handlers won't change it
595 * from under us. It essentially blocks any incoming
596 * packets, so that we won't get any new data or any
597 * packets that change the state of the socket.
599 * While locked, BH processing will add new packets to
600 * the backlog queue. This queue is processed by the
601 * owner of the socket lock right before it is released.
603 * Since ~2.3.5 it is also exclusive sleep lock serializing
604 * accesses from user process context.
606 extern void __lock_sock(struct sock *sk);
607 extern void __release_sock(struct sock *sk);
608 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
610 extern void FASTCALL(lock_sock(struct sock *sk));
611 extern void FASTCALL(release_sock(struct sock *sk));
613 /* BH context may only use the following locking interface. */
614 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
615 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
617 extern struct sock * sk_alloc(int family, int priority, int zero_it,
619 extern void sk_free(struct sock *sk);
621 extern struct sk_buff *sock_wmalloc(struct sock *sk,
622 unsigned long size, int force,
624 extern struct sk_buff *sock_rmalloc(struct sock *sk,
625 unsigned long size, int force,
627 extern void sock_wfree(struct sk_buff *skb);
628 extern void sock_rfree(struct sk_buff *skb);
630 extern int sock_setsockopt(struct socket *sock, int level,
631 int op, char __user *optval,
634 extern int sock_getsockopt(struct socket *sock, int level,
635 int op, char __user *optval,
637 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
641 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
642 unsigned long header_len,
643 unsigned long data_len,
646 extern void *sock_kmalloc(struct sock *sk, int size, int priority);
647 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
648 extern void sk_send_sigurg(struct sock *sk);
651 * Functions to fill in entries in struct proto_ops when a protocol
652 * does not implement a particular function.
654 extern int sock_no_release(struct socket *);
655 extern int sock_no_bind(struct socket *,
656 struct sockaddr *, int);
657 extern int sock_no_connect(struct socket *,
658 struct sockaddr *, int, int);
659 extern int sock_no_socketpair(struct socket *,
661 extern int sock_no_accept(struct socket *,
662 struct socket *, int);
663 extern int sock_no_getname(struct socket *,
664 struct sockaddr *, int *, int);
665 extern unsigned int sock_no_poll(struct file *, struct socket *,
666 struct poll_table_struct *);
667 extern int sock_no_ioctl(struct socket *, unsigned int,
669 extern int sock_no_listen(struct socket *, int);
670 extern int sock_no_shutdown(struct socket *, int);
671 extern int sock_no_getsockopt(struct socket *, int , int,
672 char __user *, int __user *);
673 extern int sock_no_setsockopt(struct socket *, int, int,
675 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
676 struct msghdr *, size_t);
677 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
678 struct msghdr *, size_t, int);
679 extern int sock_no_mmap(struct file *file,
681 struct vm_area_struct *vma);
682 extern ssize_t sock_no_sendpage(struct socket *sock,
684 int offset, size_t size,
688 * Default socket callbacks and setup code
691 extern void sock_def_destruct(struct sock *);
693 /* Initialise core socket variables */
694 extern void sock_init_data(struct socket *sock, struct sock *sk);
697 * sk_filter - run a packet through a socket filter
698 * @sk: sock associated with &sk_buff
699 * @skb: buffer to filter
700 * @needlock: set to 1 if the sock is not locked by caller.
702 * Run the filter code and then cut skb->data to correct size returned by
703 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
704 * than pkt_len we keep whole skb->data. This is the socket level
705 * wrapper to sk_run_filter. It returns 0 if the packet should
706 * be accepted or -EPERM if the packet should be tossed.
710 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
714 err = security_sock_rcv_skb(sk, skb);
719 struct sk_filter *filter;
724 filter = sk->sk_filter;
726 int pkt_len = sk_run_filter(skb, filter->insns,
731 skb_trim(skb, pkt_len);
741 * sk_filter_release: Release a socket filter
743 * @fp: filter to remove
745 * Remove a filter from a socket and release its resources.
748 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
750 unsigned int size = sk_filter_len(fp);
752 atomic_sub(size, &sk->sk_omem_alloc);
754 if (atomic_dec_and_test(&fp->refcnt))
758 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
760 atomic_inc(&fp->refcnt);
761 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
765 * Socket reference counting postulates.
767 * * Each user of socket SHOULD hold a reference count.
768 * * Each access point to socket (an hash table bucket, reference from a list,
769 * running timer, skb in flight MUST hold a reference count.
770 * * When reference count hits 0, it means it will never increase back.
771 * * When reference count hits 0, it means that no references from
772 * outside exist to this socket and current process on current CPU
773 * is last user and may/should destroy this socket.
774 * * sk_free is called from any context: process, BH, IRQ. When
775 * it is called, socket has no references from outside -> sk_free
776 * may release descendant resources allocated by the socket, but
777 * to the time when it is called, socket is NOT referenced by any
778 * hash tables, lists etc.
779 * * Packets, delivered from outside (from network or from another process)
780 * and enqueued on receive/error queues SHOULD NOT grab reference count,
781 * when they sit in queue. Otherwise, packets will leak to hole, when
782 * socket is looked up by one cpu and unhasing is made by another CPU.
783 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
784 * (leak to backlog). Packet socket does all the processing inside
785 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
786 * use separate SMP lock, so that they are prone too.
789 /* Ungrab socket and destroy it, if it was the last reference. */
790 static inline void sock_put(struct sock *sk)
792 if (atomic_dec_and_test(&sk->sk_refcnt))
796 /* Detach socket from process context.
797 * Announce socket dead, detach it from wait queue and inode.
798 * Note that parent inode held reference count on this struct sock,
799 * we do not release it in this function, because protocol
800 * probably wants some additional cleanups or even continuing
801 * to work with this socket (TCP).
803 static inline void sock_orphan(struct sock *sk)
805 write_lock_bh(&sk->sk_callback_lock);
806 sock_set_flag(sk, SOCK_DEAD);
807 sk->sk_socket = NULL;
809 write_unlock_bh(&sk->sk_callback_lock);
812 static inline void sock_graft(struct sock *sk, struct socket *parent)
814 write_lock_bh(&sk->sk_callback_lock);
815 sk->sk_sleep = &parent->wait;
817 sk->sk_socket = parent;
818 write_unlock_bh(&sk->sk_callback_lock);
821 static inline int sock_i_uid(struct sock *sk)
825 read_lock(&sk->sk_callback_lock);
826 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
827 read_unlock(&sk->sk_callback_lock);
831 static inline unsigned long sock_i_ino(struct sock *sk)
835 read_lock(&sk->sk_callback_lock);
836 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
837 read_unlock(&sk->sk_callback_lock);
841 static inline struct dst_entry *
842 __sk_dst_get(struct sock *sk)
844 return sk->sk_dst_cache;
847 static inline struct dst_entry *
848 sk_dst_get(struct sock *sk)
850 struct dst_entry *dst;
852 read_lock(&sk->sk_dst_lock);
853 dst = sk->sk_dst_cache;
856 read_unlock(&sk->sk_dst_lock);
861 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
863 struct dst_entry *old_dst;
865 old_dst = sk->sk_dst_cache;
866 sk->sk_dst_cache = dst;
867 dst_release(old_dst);
871 sk_dst_set(struct sock *sk, struct dst_entry *dst)
873 write_lock(&sk->sk_dst_lock);
874 __sk_dst_set(sk, dst);
875 write_unlock(&sk->sk_dst_lock);
879 __sk_dst_reset(struct sock *sk)
881 struct dst_entry *old_dst;
883 old_dst = sk->sk_dst_cache;
884 sk->sk_dst_cache = NULL;
885 dst_release(old_dst);
889 sk_dst_reset(struct sock *sk)
891 write_lock(&sk->sk_dst_lock);
893 write_unlock(&sk->sk_dst_lock);
896 static inline struct dst_entry *
897 __sk_dst_check(struct sock *sk, u32 cookie)
899 struct dst_entry *dst = sk->sk_dst_cache;
901 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
902 sk->sk_dst_cache = NULL;
909 static inline struct dst_entry *
910 sk_dst_check(struct sock *sk, u32 cookie)
912 struct dst_entry *dst = sk_dst_get(sk);
914 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
922 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
924 sk->sk_wmem_queued += skb->truesize;
925 sk->sk_forward_alloc -= skb->truesize;
929 * Queue a received datagram if it will fit. Stream and sequenced
930 * protocols can't normally use this as they need to fit buffers in
931 * and play with them.
933 * Inlined as it's very short and called for pretty much every
934 * packet ever received.
937 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
941 skb->destructor = sock_wfree;
942 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
945 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
948 skb->destructor = sock_rfree;
949 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
952 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
953 unsigned long expires);
955 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
957 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
962 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
963 number of warnings when compiling with -W --ANK
965 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
966 (unsigned)sk->sk_rcvbuf) {
971 /* It would be deadlock, if sock_queue_rcv_skb is used
972 with socket lock! We assume that users of this
973 function are lock free.
975 err = sk_filter(sk, skb, 1);
980 skb_set_owner_r(skb, sk);
982 /* Cache the SKB length before we tack it onto the receive
983 * queue. Once it is added it no longer belongs to us and
984 * may be freed by other threads of control pulling packets
989 skb_queue_tail(&sk->sk_receive_queue, skb);
991 if (!sock_flag(sk, SOCK_DEAD))
992 sk->sk_data_ready(sk, skb_len);
997 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
999 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1000 number of warnings when compiling with -W --ANK
1002 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1003 (unsigned)sk->sk_rcvbuf)
1005 skb_set_owner_r(skb, sk);
1006 skb_queue_tail(&sk->sk_error_queue, skb);
1007 if (!sock_flag(sk, SOCK_DEAD))
1008 sk->sk_data_ready(sk, skb->len);
1013 * Recover an error report and clear atomically
1016 static inline int sock_error(struct sock *sk)
1018 int err = xchg(&sk->sk_err, 0);
1022 static inline unsigned long sock_wspace(struct sock *sk)
1026 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1027 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1034 static inline void sk_wake_async(struct sock *sk, int how, int band)
1036 if (sk->sk_socket && sk->sk_socket->fasync_list)
1037 sock_wake_async(sk->sk_socket, how, band);
1040 #define SOCK_MIN_SNDBUF 2048
1041 #define SOCK_MIN_RCVBUF 256
1044 * Default write policy as shown to user space via poll/select/SIGIO
1046 static inline int sock_writeable(struct sock *sk)
1048 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1051 static inline int gfp_any(void)
1053 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1056 static inline long sock_rcvtimeo(struct sock *sk, int noblock)
1058 return noblock ? 0 : sk->sk_rcvtimeo;
1061 static inline long sock_sndtimeo(struct sock *sk, int noblock)
1063 return noblock ? 0 : sk->sk_sndtimeo;
1066 static inline int sock_rcvlowat(struct sock *sk, int waitall, int len)
1068 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1071 /* Alas, with timeout socket operations are not restartable.
1072 * Compare this to poll().
1074 static inline int sock_intr_errno(long timeo)
1076 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1079 static __inline__ void
1080 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1082 struct timeval *stamp = &skb->stamp;
1083 if (sk->sk_rcvtstamp) {
1084 /* Race occurred between timestamp enabling and packet
1085 receiving. Fill in the current time for now. */
1086 if (stamp->tv_sec == 0)
1087 do_gettimeofday(stamp);
1088 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1091 sk->sk_stamp = *stamp;
1095 * sk_eat_skb - Release a skb if it is no longer needed
1096 * @sk - socket to eat this skb from
1097 * @skb - socket buffer to eat
1099 * This routine must be called with interrupts disabled or with the socket
1100 * locked so that the sk_buff queue operation is ok.
1102 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1104 __skb_unlink(skb, &sk->sk_receive_queue);
1108 extern atomic_t netstamp_needed;
1109 extern void sock_enable_timestamp(struct sock *sk);
1110 extern void sock_disable_timestamp(struct sock *sk);
1112 static inline void net_timestamp(struct timeval *stamp)
1114 if (atomic_read(&netstamp_needed))
1115 do_gettimeofday(stamp);
1122 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1125 * Enable debug/info messages
1129 #define NETDEBUG(x) do { } while (0)
1130 #define LIMIT_NETDEBUG(x) do {} while(0)
1132 #define NETDEBUG(x) do { x; } while (0)
1133 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1137 * Macros for sleeping on a socket. Use them like this:
1139 * SOCK_SLEEP_PRE(sk)
1142 * SOCK_SLEEP_POST(sk)
1144 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1145 * and when the last use of them in DECnet has gone, I'm intending to
1149 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1150 DECLARE_WAITQUEUE(wait, tsk); \
1151 tsk->state = TASK_INTERRUPTIBLE; \
1152 add_wait_queue((sk)->sk_sleep, &wait); \
1155 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1156 remove_wait_queue((sk)->sk_sleep, &wait); \
1160 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1163 sock_set_flag(sk, bit);
1165 sock_reset_flag(sk, bit);
1168 extern __u32 sysctl_wmem_max;
1169 extern __u32 sysctl_rmem_max;
1171 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1173 #endif /* _SOCK_H */