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;
113 struct vx_info *skc_vx_info;
115 struct nx_info *skc_nx_info;
119 * struct sock - network layer representation of sockets
120 * @__sk_common - shared layout with tcp_tw_bucket
121 * @sk_zapped - ax25 & ipx means !linked
122 * @sk_shutdown - mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
123 * @sk_use_write_queue - wheter to call sk->sk_write_space in sock_wfree
124 * @sk_userlocks - %SO_SNDBUF and %SO_RCVBUF settings
125 * @sk_lock - synchronizer
126 * @sk_rcvbuf - size of receive buffer in bytes
127 * @sk_sleep - sock wait queue
128 * @sk_dst_cache - destination cache
129 * @sk_dst_lock - destination cache lock
130 * @sk_policy - flow policy
131 * @sk_rmem_alloc - receive queue bytes committed
132 * @sk_receive_queue - incoming packets
133 * @sk_wmem_alloc - transmit queue bytes committed
134 * @sk_write_queue - Packet sending queue
135 * @sk_omem_alloc - "o" is "option" or "other"
136 * @sk_wmem_queued - persistent queue size
137 * @sk_forward_alloc - space allocated forward
138 * @sk_allocation - allocation mode
139 * @sk_sndbuf - size of send buffer in bytes
140 * @sk_flags - %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
141 * @sk_no_check - %SO_NO_CHECK setting, wether or not checkup packets
142 * @sk_debug - %SO_DEBUG setting
143 * @sk_rcvtstamp - %SO_TIMESTAMP setting
144 * @sk_no_largesend - whether to sent large segments or not
145 * @sk_route_caps - route capabilities (e.g. %NETIF_F_TSO)
146 * @sk_lingertime - %SO_LINGER l_linger setting
147 * @sk_hashent - hash entry in several tables (e.g. tcp_ehash)
148 * @sk_pair - socket pair (e.g. AF_UNIX/unix_peer)
149 * @sk_backlog - always used with the per-socket spinlock held
150 * @sk_callback_lock - used with the callbacks in the end of this struct
151 * @sk_error_queue - rarely used
152 * @sk_prot - protocol handlers inside a network family
153 * @sk_err - last error
154 * @sk_err_soft - errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
155 * @sk_ack_backlog - current listen backlog
156 * @sk_max_ack_backlog - listen backlog set in listen()
157 * @sk_priority - %SO_PRIORITY setting
158 * @sk_type - socket type (%SOCK_STREAM, etc)
159 * @sk_localroute - route locally only, %SO_DONTROUTE setting
160 * @sk_protocol - which protocol this socket belongs in this network family
161 * @sk_peercred - %SO_PEERCRED setting
162 * @sk_rcvlowat - %SO_RCVLOWAT setting
163 * @sk_rcvtimeo - %SO_RCVTIMEO setting
164 * @sk_sndtimeo - %SO_SNDTIMEO setting
165 * @sk_filter - socket filtering instructions
166 * @sk_protinfo - private area, net family specific, when not using slab
167 * @sk_slab - the slabcache this instance was allocated from
168 * @sk_timer - sock cleanup timer
169 * @sk_stamp - time stamp of last packet received
170 * @sk_socket - Identd and reporting IO signals
171 * @sk_user_data - RPC and Tux layer private data
172 * @sk_owner - module that owns this socket
173 * @sk_state_change - callback to indicate change in the state of the sock
174 * @sk_data_ready - callback to indicate there is data to be processed
175 * @sk_write_space - callback to indicate there is bf sending space available
176 * @sk_error_report - callback to indicate errors (e.g. %MSG_ERRQUEUE)
177 * @sk_create_child - callback to get new socket events
178 * @sk_backlog_rcv - callback to process the backlog
179 * @sk_destruct - called at sock freeing time, i.e. when all refcnt == 0
183 * Now struct tcp_tw_bucket also uses sock_common, so please just
184 * don't add nothing before this first member (__sk_common) --acme
186 struct sock_common __sk_common;
187 #define sk_family __sk_common.skc_family
188 #define sk_state __sk_common.skc_state
189 #define sk_reuse __sk_common.skc_reuse
190 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
191 #define sk_node __sk_common.skc_node
192 #define sk_bind_node __sk_common.skc_bind_node
193 #define sk_refcnt __sk_common.skc_refcnt
194 #define sk_xid __sk_common.skc_xid
195 #define sk_vx_info __sk_common.skc_vx_info
196 #define sk_nid __sk_common.skc_nid
197 #define sk_nx_info __sk_common.skc_nx_info
198 volatile unsigned char sk_zapped;
199 unsigned char sk_shutdown;
200 unsigned char sk_use_write_queue;
201 unsigned char sk_userlocks;
202 socket_lock_t sk_lock;
204 wait_queue_head_t *sk_sleep;
205 struct dst_entry *sk_dst_cache;
206 rwlock_t sk_dst_lock;
207 struct xfrm_policy *sk_policy[2];
208 atomic_t sk_rmem_alloc;
209 struct sk_buff_head sk_receive_queue;
210 atomic_t sk_wmem_alloc;
211 struct sk_buff_head sk_write_queue;
212 atomic_t sk_omem_alloc;
214 int sk_forward_alloc;
215 unsigned int sk_allocation;
217 unsigned long sk_flags;
219 unsigned char sk_debug;
220 unsigned char sk_rcvtstamp;
221 unsigned char sk_no_largesend;
223 unsigned long sk_lingertime;
225 struct sock *sk_pair;
227 * The backlog queue is special, it is always used with
228 * the per-socket spinlock held and requires low latency
229 * access. Therefore we special case it's implementation.
232 struct sk_buff *head;
233 struct sk_buff *tail;
235 rwlock_t sk_callback_lock;
236 struct sk_buff_head sk_error_queue;
237 struct proto *sk_prot;
240 unsigned short sk_ack_backlog;
241 unsigned short sk_max_ack_backlog;
243 unsigned short sk_type;
244 unsigned char sk_localroute;
245 unsigned char sk_protocol;
246 struct ucred sk_peercred;
250 struct sk_filter *sk_filter;
252 kmem_cache_t *sk_slab;
253 struct timer_list sk_timer;
254 struct timeval sk_stamp;
255 struct socket *sk_socket;
257 void *sk_ns; // For use by CKRM
258 struct module *sk_owner;
260 void (*sk_state_change)(struct sock *sk);
261 void (*sk_data_ready)(struct sock *sk, int bytes);
262 void (*sk_write_space)(struct sock *sk);
263 void (*sk_error_report)(struct sock *sk);
264 int (*sk_backlog_rcv)(struct sock *sk,
265 struct sk_buff *skb);
266 void (*sk_create_child)(struct sock *sk, struct sock *newsk);
267 void (*sk_destruct)(struct sock *sk);
271 * Hashed lists helper routines
273 static inline struct sock *__sk_head(struct hlist_head *head)
275 return hlist_entry(head->first, struct sock, sk_node);
278 static inline struct sock *sk_head(struct hlist_head *head)
280 return hlist_empty(head) ? NULL : __sk_head(head);
283 static inline struct sock *sk_next(struct sock *sk)
285 return sk->sk_node.next ?
286 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
289 static inline int sk_unhashed(struct sock *sk)
291 return hlist_unhashed(&sk->sk_node);
294 static inline int sk_hashed(struct sock *sk)
296 return sk->sk_node.pprev != NULL;
299 static __inline__ void sk_node_init(struct hlist_node *node)
304 static __inline__ void __sk_del_node(struct sock *sk)
306 __hlist_del(&sk->sk_node);
309 static __inline__ int __sk_del_node_init(struct sock *sk)
313 sk_node_init(&sk->sk_node);
319 /* Grab socket reference count. This operation is valid only
320 when sk is ALREADY grabbed f.e. it is found in hash table
321 or a list and the lookup is made under lock preventing hash table
325 static inline void sock_hold(struct sock *sk)
327 atomic_inc(&sk->sk_refcnt);
330 /* Ungrab socket in the context, which assumes that socket refcnt
331 cannot hit zero, f.e. it is true in context of any socketcall.
333 static inline void __sock_put(struct sock *sk)
335 atomic_dec(&sk->sk_refcnt);
338 static __inline__ int sk_del_node_init(struct sock *sk)
340 int rc = __sk_del_node_init(sk);
343 /* paranoid for a while -acme */
344 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
350 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
352 hlist_add_head(&sk->sk_node, list);
355 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
358 __sk_add_node(sk, list);
361 static __inline__ void __sk_del_bind_node(struct sock *sk)
363 __hlist_del(&sk->sk_bind_node);
366 static __inline__ void sk_add_bind_node(struct sock *sk,
367 struct hlist_head *list)
369 hlist_add_head(&sk->sk_bind_node, list);
372 #define sk_for_each(__sk, node, list) \
373 hlist_for_each_entry(__sk, node, list, sk_node)
374 #define sk_for_each_from(__sk, node) \
375 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
376 hlist_for_each_entry_from(__sk, node, sk_node)
377 #define sk_for_each_continue(__sk, node) \
378 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
379 hlist_for_each_entry_continue(__sk, node, sk_node)
380 #define sk_for_each_safe(__sk, node, tmp, list) \
381 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
382 #define sk_for_each_bound(__sk, node, list) \
383 hlist_for_each_entry(__sk, node, list, sk_bind_node)
397 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
399 __set_bit(flag, &sk->sk_flags);
402 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
404 __clear_bit(flag, &sk->sk_flags);
407 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
409 return test_bit(flag, &sk->sk_flags);
412 static inline void sk_acceptq_removed(struct sock *sk)
414 sk->sk_ack_backlog--;
417 static inline void sk_acceptq_added(struct sock *sk)
419 sk->sk_ack_backlog++;
422 static inline int sk_acceptq_is_full(struct sock *sk)
424 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
428 * Compute minimal free write space needed to queue new packets.
430 static inline int sk_stream_min_wspace(struct sock *sk)
432 return sk->sk_wmem_queued / 2;
435 static inline int sk_stream_wspace(struct sock *sk)
437 return sk->sk_sndbuf - sk->sk_wmem_queued;
440 extern void sk_stream_write_space(struct sock *sk);
442 /* The per-socket spinlock must be held here. */
443 #define sk_add_backlog(__sk, __skb) \
444 do { if (!(__sk)->sk_backlog.tail) { \
445 (__sk)->sk_backlog.head = \
446 (__sk)->sk_backlog.tail = (__skb); \
448 ((__sk)->sk_backlog.tail)->next = (__skb); \
449 (__sk)->sk_backlog.tail = (__skb); \
451 (__skb)->next = NULL; \
454 #define sk_wait_event(__sk, __timeo, __condition) \
456 release_sock(__sk); \
459 *(__timeo) = schedule_timeout(*(__timeo)); \
466 extern int sk_wait_data(struct sock *sk, long *timeo);
468 /* IP protocol blocks we attach to sockets.
469 * socket layer -> transport layer interface
470 * transport -> network interface is defined by struct inet_proto
473 void (*close)(struct sock *sk,
475 int (*connect)(struct sock *sk,
476 struct sockaddr *uaddr,
478 int (*disconnect)(struct sock *sk, int flags);
480 struct sock * (*accept) (struct sock *sk, int flags, int *err);
482 int (*ioctl)(struct sock *sk, int cmd,
484 int (*init)(struct sock *sk);
485 int (*destroy)(struct sock *sk);
486 void (*shutdown)(struct sock *sk, int how);
487 int (*setsockopt)(struct sock *sk, int level,
488 int optname, char __user *optval,
490 int (*getsockopt)(struct sock *sk, int level,
491 int optname, char __user *optval,
493 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
494 struct msghdr *msg, size_t len);
495 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
497 size_t len, int noblock, int flags,
499 int (*sendpage)(struct sock *sk, struct page *page,
500 int offset, size_t size, int flags);
501 int (*bind)(struct sock *sk,
502 struct sockaddr *uaddr, int addr_len);
504 int (*backlog_rcv) (struct sock *sk,
505 struct sk_buff *skb);
507 /* Keeping track of sk's, looking them up, and port selection methods. */
508 void (*hash)(struct sock *sk);
509 void (*unhash)(struct sock *sk);
510 int (*get_port)(struct sock *sk, unsigned short snum);
516 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
520 static __inline__ void sk_set_owner(struct sock *sk, struct module *owner)
523 * One should use sk_set_owner just once, after struct sock creation,
524 * be it shortly after sk_alloc or after a function that returns a new
525 * struct sock (and that down the call chain called sk_alloc), e.g. the
526 * IPv4 and IPv6 modules share tcp_create_openreq_child, so if
527 * tcp_create_openreq_child called sk_set_owner IPv6 would have to
528 * change the ownership of this struct sock, with one not needed
529 * transient sk_set_owner call.
531 BUG_ON(sk->sk_owner != NULL);
533 sk->sk_owner = owner;
537 /* Called with local bh disabled */
538 static __inline__ void sock_prot_inc_use(struct proto *prot)
540 prot->stats[smp_processor_id()].inuse++;
543 static __inline__ void sock_prot_dec_use(struct proto *prot)
545 prot->stats[smp_processor_id()].inuse--;
548 /* About 10 seconds */
549 #define SOCK_DESTROY_TIME (10*HZ)
551 /* Sockets 0-1023 can't be bound to unless you are superuser */
552 #define PROT_SOCK 1024
554 #define SHUTDOWN_MASK 3
555 #define RCV_SHUTDOWN 1
556 #define SEND_SHUTDOWN 2
558 #define SOCK_SNDBUF_LOCK 1
559 #define SOCK_RCVBUF_LOCK 2
560 #define SOCK_BINDADDR_LOCK 4
561 #define SOCK_BINDPORT_LOCK 8
563 /* sock_iocb: used to kick off async processing of socket ios */
565 struct list_head list;
571 struct scm_cookie *scm;
572 struct msghdr *msg, async_msg;
573 struct iovec async_iov;
576 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
578 BUG_ON(sizeof(struct sock_iocb) > KIOCB_PRIVATE_SIZE);
579 return (struct sock_iocb *)iocb->private;
582 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
584 return container_of((void *)si, struct kiocb, private);
587 struct socket_alloc {
588 struct socket socket;
589 struct inode vfs_inode;
592 static inline struct socket *SOCKET_I(struct inode *inode)
594 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
597 static inline struct inode *SOCK_INODE(struct socket *socket)
599 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
602 /* Used by processes to "lock" a socket state, so that
603 * interrupts and bottom half handlers won't change it
604 * from under us. It essentially blocks any incoming
605 * packets, so that we won't get any new data or any
606 * packets that change the state of the socket.
608 * While locked, BH processing will add new packets to
609 * the backlog queue. This queue is processed by the
610 * owner of the socket lock right before it is released.
612 * Since ~2.3.5 it is also exclusive sleep lock serializing
613 * accesses from user process context.
615 extern void __lock_sock(struct sock *sk);
616 extern void __release_sock(struct sock *sk);
617 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
619 extern void FASTCALL(lock_sock(struct sock *sk));
620 extern void FASTCALL(release_sock(struct sock *sk));
622 /* BH context may only use the following locking interface. */
623 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
624 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
626 extern struct sock * sk_alloc(int family, int priority, int zero_it,
628 extern void sk_free(struct sock *sk);
630 extern struct sk_buff *sock_wmalloc(struct sock *sk,
631 unsigned long size, int force,
633 extern struct sk_buff *sock_rmalloc(struct sock *sk,
634 unsigned long size, int force,
636 extern void sock_wfree(struct sk_buff *skb);
637 extern void sock_rfree(struct sk_buff *skb);
639 extern int sock_setsockopt(struct socket *sock, int level,
640 int op, char __user *optval,
643 extern int sock_getsockopt(struct socket *sock, int level,
644 int op, char __user *optval,
646 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
650 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
651 unsigned long header_len,
652 unsigned long data_len,
655 extern void *sock_kmalloc(struct sock *sk, int size, int priority);
656 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
657 extern void sk_send_sigurg(struct sock *sk);
660 * Functions to fill in entries in struct proto_ops when a protocol
661 * does not implement a particular function.
663 extern int sock_no_release(struct socket *);
664 extern int sock_no_bind(struct socket *,
665 struct sockaddr *, int);
666 extern int sock_no_connect(struct socket *,
667 struct sockaddr *, int, int);
668 extern int sock_no_socketpair(struct socket *,
670 extern int sock_no_accept(struct socket *,
671 struct socket *, int);
672 extern int sock_no_getname(struct socket *,
673 struct sockaddr *, int *, int);
674 extern unsigned int sock_no_poll(struct file *, struct socket *,
675 struct poll_table_struct *);
676 extern int sock_no_ioctl(struct socket *, unsigned int,
678 extern int sock_no_listen(struct socket *, int);
679 extern int sock_no_shutdown(struct socket *, int);
680 extern int sock_no_getsockopt(struct socket *, int , int,
681 char __user *, int __user *);
682 extern int sock_no_setsockopt(struct socket *, int, int,
684 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
685 struct msghdr *, size_t);
686 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
687 struct msghdr *, size_t, int);
688 extern int sock_no_mmap(struct file *file,
690 struct vm_area_struct *vma);
691 extern ssize_t sock_no_sendpage(struct socket *sock,
693 int offset, size_t size,
697 * Default socket callbacks and setup code
700 extern void sock_def_destruct(struct sock *);
702 /* Initialise core socket variables */
703 extern void sock_init_data(struct socket *sock, struct sock *sk);
706 * sk_filter - run a packet through a socket filter
707 * @sk: sock associated with &sk_buff
708 * @skb: buffer to filter
709 * @needlock: set to 1 if the sock is not locked by caller.
711 * Run the filter code and then cut skb->data to correct size returned by
712 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
713 * than pkt_len we keep whole skb->data. This is the socket level
714 * wrapper to sk_run_filter. It returns 0 if the packet should
715 * be accepted or -EPERM if the packet should be tossed.
719 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
723 err = security_sock_rcv_skb(sk, skb);
728 struct sk_filter *filter;
733 filter = sk->sk_filter;
735 int pkt_len = sk_run_filter(skb, filter->insns,
740 skb_trim(skb, pkt_len);
750 * sk_filter_release: Release a socket filter
752 * @fp: filter to remove
754 * Remove a filter from a socket and release its resources.
757 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
759 unsigned int size = sk_filter_len(fp);
761 atomic_sub(size, &sk->sk_omem_alloc);
763 if (atomic_dec_and_test(&fp->refcnt))
767 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
769 atomic_inc(&fp->refcnt);
770 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
774 * Socket reference counting postulates.
776 * * Each user of socket SHOULD hold a reference count.
777 * * Each access point to socket (an hash table bucket, reference from a list,
778 * running timer, skb in flight MUST hold a reference count.
779 * * When reference count hits 0, it means it will never increase back.
780 * * When reference count hits 0, it means that no references from
781 * outside exist to this socket and current process on current CPU
782 * is last user and may/should destroy this socket.
783 * * sk_free is called from any context: process, BH, IRQ. When
784 * it is called, socket has no references from outside -> sk_free
785 * may release descendant resources allocated by the socket, but
786 * to the time when it is called, socket is NOT referenced by any
787 * hash tables, lists etc.
788 * * Packets, delivered from outside (from network or from another process)
789 * and enqueued on receive/error queues SHOULD NOT grab reference count,
790 * when they sit in queue. Otherwise, packets will leak to hole, when
791 * socket is looked up by one cpu and unhasing is made by another CPU.
792 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
793 * (leak to backlog). Packet socket does all the processing inside
794 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
795 * use separate SMP lock, so that they are prone too.
798 /* Ungrab socket and destroy it, if it was the last reference. */
799 static inline void sock_put(struct sock *sk)
801 if (atomic_dec_and_test(&sk->sk_refcnt))
805 /* Detach socket from process context.
806 * Announce socket dead, detach it from wait queue and inode.
807 * Note that parent inode held reference count on this struct sock,
808 * we do not release it in this function, because protocol
809 * probably wants some additional cleanups or even continuing
810 * to work with this socket (TCP).
812 static inline void sock_orphan(struct sock *sk)
814 write_lock_bh(&sk->sk_callback_lock);
815 sock_set_flag(sk, SOCK_DEAD);
816 sk->sk_socket = NULL;
818 write_unlock_bh(&sk->sk_callback_lock);
821 static inline void sock_graft(struct sock *sk, struct socket *parent)
823 write_lock_bh(&sk->sk_callback_lock);
824 sk->sk_sleep = &parent->wait;
826 sk->sk_socket = parent;
827 write_unlock_bh(&sk->sk_callback_lock);
830 static inline int sock_i_uid(struct sock *sk)
834 read_lock(&sk->sk_callback_lock);
835 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
836 read_unlock(&sk->sk_callback_lock);
840 static inline unsigned long sock_i_ino(struct sock *sk)
844 read_lock(&sk->sk_callback_lock);
845 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
846 read_unlock(&sk->sk_callback_lock);
850 static inline struct dst_entry *
851 __sk_dst_get(struct sock *sk)
853 return sk->sk_dst_cache;
856 static inline struct dst_entry *
857 sk_dst_get(struct sock *sk)
859 struct dst_entry *dst;
861 read_lock(&sk->sk_dst_lock);
862 dst = sk->sk_dst_cache;
865 read_unlock(&sk->sk_dst_lock);
870 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
872 struct dst_entry *old_dst;
874 old_dst = sk->sk_dst_cache;
875 sk->sk_dst_cache = dst;
876 dst_release(old_dst);
880 sk_dst_set(struct sock *sk, struct dst_entry *dst)
882 write_lock(&sk->sk_dst_lock);
883 __sk_dst_set(sk, dst);
884 write_unlock(&sk->sk_dst_lock);
888 __sk_dst_reset(struct sock *sk)
890 struct dst_entry *old_dst;
892 old_dst = sk->sk_dst_cache;
893 sk->sk_dst_cache = NULL;
894 dst_release(old_dst);
898 sk_dst_reset(struct sock *sk)
900 write_lock(&sk->sk_dst_lock);
902 write_unlock(&sk->sk_dst_lock);
905 static inline struct dst_entry *
906 __sk_dst_check(struct sock *sk, u32 cookie)
908 struct dst_entry *dst = sk->sk_dst_cache;
910 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
911 sk->sk_dst_cache = NULL;
918 static inline struct dst_entry *
919 sk_dst_check(struct sock *sk, u32 cookie)
921 struct dst_entry *dst = sk_dst_get(sk);
923 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
931 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
933 sk->sk_wmem_queued += skb->truesize;
934 sk->sk_forward_alloc -= skb->truesize;
938 * Queue a received datagram if it will fit. Stream and sequenced
939 * protocols can't normally use this as they need to fit buffers in
940 * and play with them.
942 * Inlined as it's very short and called for pretty much every
943 * packet ever received.
946 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
950 skb->destructor = sock_wfree;
951 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
954 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
957 skb->destructor = sock_rfree;
958 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
961 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
962 unsigned long expires);
964 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
966 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
971 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
972 number of warnings when compiling with -W --ANK
974 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
975 (unsigned)sk->sk_rcvbuf) {
980 /* It would be deadlock, if sock_queue_rcv_skb is used
981 with socket lock! We assume that users of this
982 function are lock free.
984 err = sk_filter(sk, skb, 1);
989 skb_set_owner_r(skb, sk);
991 /* Cache the SKB length before we tack it onto the receive
992 * queue. Once it is added it no longer belongs to us and
993 * may be freed by other threads of control pulling packets
998 skb_queue_tail(&sk->sk_receive_queue, skb);
1000 if (!sock_flag(sk, SOCK_DEAD))
1001 sk->sk_data_ready(sk, skb_len);
1006 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1008 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1009 number of warnings when compiling with -W --ANK
1011 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1012 (unsigned)sk->sk_rcvbuf)
1014 skb_set_owner_r(skb, sk);
1015 skb_queue_tail(&sk->sk_error_queue, skb);
1016 if (!sock_flag(sk, SOCK_DEAD))
1017 sk->sk_data_ready(sk, skb->len);
1022 * Recover an error report and clear atomically
1025 static inline int sock_error(struct sock *sk)
1027 int err = xchg(&sk->sk_err, 0);
1031 static inline unsigned long sock_wspace(struct sock *sk)
1035 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1036 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1043 static inline void sk_wake_async(struct sock *sk, int how, int band)
1045 if (sk->sk_socket && sk->sk_socket->fasync_list)
1046 sock_wake_async(sk->sk_socket, how, band);
1049 #define SOCK_MIN_SNDBUF 2048
1050 #define SOCK_MIN_RCVBUF 256
1053 * Default write policy as shown to user space via poll/select/SIGIO
1055 static inline int sock_writeable(struct sock *sk)
1057 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1060 static inline int gfp_any(void)
1062 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1065 static inline long sock_rcvtimeo(struct sock *sk, int noblock)
1067 return noblock ? 0 : sk->sk_rcvtimeo;
1070 static inline long sock_sndtimeo(struct sock *sk, int noblock)
1072 return noblock ? 0 : sk->sk_sndtimeo;
1075 static inline int sock_rcvlowat(struct sock *sk, int waitall, int len)
1077 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1080 /* Alas, with timeout socket operations are not restartable.
1081 * Compare this to poll().
1083 static inline int sock_intr_errno(long timeo)
1085 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1088 static __inline__ void
1089 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1091 struct timeval *stamp = &skb->stamp;
1092 if (sk->sk_rcvtstamp) {
1093 /* Race occurred between timestamp enabling and packet
1094 receiving. Fill in the current time for now. */
1095 if (stamp->tv_sec == 0)
1096 do_gettimeofday(stamp);
1097 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1100 sk->sk_stamp = *stamp;
1104 * sk_eat_skb - Release a skb if it is no longer needed
1105 * @sk - socket to eat this skb from
1106 * @skb - socket buffer to eat
1108 * This routine must be called with interrupts disabled or with the socket
1109 * locked so that the sk_buff queue operation is ok.
1111 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1113 __skb_unlink(skb, &sk->sk_receive_queue);
1117 extern atomic_t netstamp_needed;
1118 extern void sock_enable_timestamp(struct sock *sk);
1119 extern void sock_disable_timestamp(struct sock *sk);
1121 static inline void net_timestamp(struct timeval *stamp)
1123 if (atomic_read(&netstamp_needed))
1124 do_gettimeofday(stamp);
1131 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1134 * Enable debug/info messages
1138 #define NETDEBUG(x) do { } while (0)
1139 #define LIMIT_NETDEBUG(x) do {} while(0)
1141 #define NETDEBUG(x) do { x; } while (0)
1142 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1146 * Macros for sleeping on a socket. Use them like this:
1148 * SOCK_SLEEP_PRE(sk)
1151 * SOCK_SLEEP_POST(sk)
1153 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1154 * and when the last use of them in DECnet has gone, I'm intending to
1158 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1159 DECLARE_WAITQUEUE(wait, tsk); \
1160 tsk->state = TASK_INTERRUPTIBLE; \
1161 add_wait_queue((sk)->sk_sleep, &wait); \
1164 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1165 remove_wait_queue((sk)->sk_sleep, &wait); \
1169 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1172 sock_set_flag(sk, bit);
1174 sock_reset_flag(sk, bit);
1177 extern __u32 sysctl_wmem_max;
1178 extern __u32 sysctl_rmem_max;
1180 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1182 #endif /* _SOCK_H */