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 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_backlog_rcv - callback to process the backlog
178 * @sk_destruct - called at sock freeing time, i.e. when all refcnt == 0
182 * Now struct tcp_tw_bucket also uses sock_common, so please just
183 * don't add nothing before this first member (__sk_common) --acme
185 struct sock_common __sk_common;
186 #define sk_family __sk_common.skc_family
187 #define sk_state __sk_common.skc_state
188 #define sk_reuse __sk_common.skc_reuse
189 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
190 #define sk_node __sk_common.skc_node
191 #define sk_bind_node __sk_common.skc_bind_node
192 #define sk_refcnt __sk_common.skc_refcnt
193 #define sk_xid __sk_common.skc_xid
194 #define sk_vx_info __sk_common.skc_vx_info
195 #define sk_nid __sk_common.skc_nid
196 #define sk_nx_info __sk_common.skc_nx_info
197 volatile unsigned char sk_zapped;
198 unsigned char sk_shutdown;
199 unsigned char sk_use_write_queue;
200 unsigned char sk_userlocks;
201 socket_lock_t sk_lock;
203 wait_queue_head_t *sk_sleep;
204 struct dst_entry *sk_dst_cache;
205 rwlock_t sk_dst_lock;
206 struct xfrm_policy *sk_policy[2];
207 atomic_t sk_rmem_alloc;
208 struct sk_buff_head sk_receive_queue;
209 atomic_t sk_wmem_alloc;
210 struct sk_buff_head sk_write_queue;
211 atomic_t sk_omem_alloc;
213 int sk_forward_alloc;
214 unsigned int sk_allocation;
216 unsigned long sk_flags;
218 unsigned char sk_debug;
219 unsigned char sk_rcvtstamp;
220 unsigned char sk_no_largesend;
222 unsigned long sk_lingertime;
224 struct sock *sk_pair;
226 * The backlog queue is special, it is always used with
227 * the per-socket spinlock held and requires low latency
228 * access. Therefore we special case it's implementation.
231 struct sk_buff *head;
232 struct sk_buff *tail;
234 rwlock_t sk_callback_lock;
235 struct sk_buff_head sk_error_queue;
236 struct proto *sk_prot;
239 unsigned short sk_ack_backlog;
240 unsigned short sk_max_ack_backlog;
242 unsigned short sk_type;
243 unsigned char sk_localroute;
244 unsigned char sk_protocol;
245 struct ucred sk_peercred;
249 struct sk_filter *sk_filter;
251 kmem_cache_t *sk_slab;
252 struct timer_list sk_timer;
253 struct timeval sk_stamp;
254 struct socket *sk_socket;
256 struct module *sk_owner;
258 void (*sk_state_change)(struct sock *sk);
259 void (*sk_data_ready)(struct sock *sk, int bytes);
260 void (*sk_write_space)(struct sock *sk);
261 void (*sk_error_report)(struct sock *sk);
262 int (*sk_backlog_rcv)(struct sock *sk,
263 struct sk_buff *skb);
264 void (*sk_destruct)(struct sock *sk);
268 * Hashed lists helper routines
270 static inline struct sock *__sk_head(struct hlist_head *head)
272 return hlist_entry(head->first, struct sock, sk_node);
275 static inline struct sock *sk_head(struct hlist_head *head)
277 return hlist_empty(head) ? NULL : __sk_head(head);
280 static inline struct sock *sk_next(struct sock *sk)
282 return sk->sk_node.next ?
283 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
286 static inline int sk_unhashed(struct sock *sk)
288 return hlist_unhashed(&sk->sk_node);
291 static inline int sk_hashed(struct sock *sk)
293 return sk->sk_node.pprev != NULL;
296 static __inline__ void sk_node_init(struct hlist_node *node)
301 static __inline__ void __sk_del_node(struct sock *sk)
303 __hlist_del(&sk->sk_node);
306 static __inline__ int __sk_del_node_init(struct sock *sk)
310 sk_node_init(&sk->sk_node);
316 /* Grab socket reference count. This operation is valid only
317 when sk is ALREADY grabbed f.e. it is found in hash table
318 or a list and the lookup is made under lock preventing hash table
322 static inline void sock_hold(struct sock *sk)
324 atomic_inc(&sk->sk_refcnt);
327 /* Ungrab socket in the context, which assumes that socket refcnt
328 cannot hit zero, f.e. it is true in context of any socketcall.
330 static inline void __sock_put(struct sock *sk)
332 atomic_dec(&sk->sk_refcnt);
335 static __inline__ int sk_del_node_init(struct sock *sk)
337 int rc = __sk_del_node_init(sk);
340 /* paranoid for a while -acme */
341 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
347 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
349 hlist_add_head(&sk->sk_node, list);
352 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
355 __sk_add_node(sk, list);
358 static __inline__ void __sk_del_bind_node(struct sock *sk)
360 __hlist_del(&sk->sk_bind_node);
363 static __inline__ void sk_add_bind_node(struct sock *sk,
364 struct hlist_head *list)
366 hlist_add_head(&sk->sk_bind_node, list);
369 #define sk_for_each(__sk, node, list) \
370 hlist_for_each_entry(__sk, node, list, sk_node)
371 #define sk_for_each_from(__sk, node) \
372 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
373 hlist_for_each_entry_from(__sk, node, sk_node)
374 #define sk_for_each_continue(__sk, node) \
375 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
376 hlist_for_each_entry_continue(__sk, node, sk_node)
377 #define sk_for_each_safe(__sk, node, tmp, list) \
378 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
379 #define sk_for_each_bound(__sk, node, list) \
380 hlist_for_each_entry(__sk, node, list, sk_bind_node)
394 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
396 __set_bit(flag, &sk->sk_flags);
399 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
401 __clear_bit(flag, &sk->sk_flags);
404 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
406 return test_bit(flag, &sk->sk_flags);
409 static inline void sk_acceptq_removed(struct sock *sk)
411 sk->sk_ack_backlog--;
414 static inline void sk_acceptq_added(struct sock *sk)
416 sk->sk_ack_backlog++;
419 static inline int sk_acceptq_is_full(struct sock *sk)
421 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
424 /* The per-socket spinlock must be held here. */
425 #define sk_add_backlog(__sk, __skb) \
426 do { if (!(__sk)->sk_backlog.tail) { \
427 (__sk)->sk_backlog.head = \
428 (__sk)->sk_backlog.tail = (__skb); \
430 ((__sk)->sk_backlog.tail)->next = (__skb); \
431 (__sk)->sk_backlog.tail = (__skb); \
433 (__skb)->next = NULL; \
436 #define sk_wait_event(__sk, __timeo, __condition) \
438 release_sock(__sk); \
441 *(__timeo) = schedule_timeout(*(__timeo)); \
448 extern int sk_wait_data(struct sock *sk, long *timeo);
450 /* IP protocol blocks we attach to sockets.
451 * socket layer -> transport layer interface
452 * transport -> network interface is defined by struct inet_proto
455 void (*close)(struct sock *sk,
457 int (*connect)(struct sock *sk,
458 struct sockaddr *uaddr,
460 int (*disconnect)(struct sock *sk, int flags);
462 struct sock * (*accept) (struct sock *sk, int flags, int *err);
464 int (*ioctl)(struct sock *sk, int cmd,
466 int (*init)(struct sock *sk);
467 int (*destroy)(struct sock *sk);
468 void (*shutdown)(struct sock *sk, int how);
469 int (*setsockopt)(struct sock *sk, int level,
470 int optname, char __user *optval,
472 int (*getsockopt)(struct sock *sk, int level,
473 int optname, char __user *optval,
475 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
476 struct msghdr *msg, size_t len);
477 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
479 size_t len, int noblock, int flags,
481 int (*sendpage)(struct sock *sk, struct page *page,
482 int offset, size_t size, int flags);
483 int (*bind)(struct sock *sk,
484 struct sockaddr *uaddr, int addr_len);
486 int (*backlog_rcv) (struct sock *sk,
487 struct sk_buff *skb);
489 /* Keeping track of sk's, looking them up, and port selection methods. */
490 void (*hash)(struct sock *sk);
491 void (*unhash)(struct sock *sk);
492 int (*get_port)(struct sock *sk, unsigned short snum);
498 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
502 static __inline__ void sk_set_owner(struct sock *sk, struct module *owner)
505 * One should use sk_set_owner just once, after struct sock creation,
506 * be it shortly after sk_alloc or after a function that returns a new
507 * struct sock (and that down the call chain called sk_alloc), e.g. the
508 * IPv4 and IPv6 modules share tcp_create_openreq_child, so if
509 * tcp_create_openreq_child called sk_set_owner IPv6 would have to
510 * change the ownership of this struct sock, with one not needed
511 * transient sk_set_owner call.
513 BUG_ON(sk->sk_owner != NULL);
515 sk->sk_owner = owner;
519 /* Called with local bh disabled */
520 static __inline__ void sock_prot_inc_use(struct proto *prot)
522 prot->stats[smp_processor_id()].inuse++;
525 static __inline__ void sock_prot_dec_use(struct proto *prot)
527 prot->stats[smp_processor_id()].inuse--;
530 /* About 10 seconds */
531 #define SOCK_DESTROY_TIME (10*HZ)
533 /* Sockets 0-1023 can't be bound to unless you are superuser */
534 #define PROT_SOCK 1024
536 #define SHUTDOWN_MASK 3
537 #define RCV_SHUTDOWN 1
538 #define SEND_SHUTDOWN 2
540 #define SOCK_SNDBUF_LOCK 1
541 #define SOCK_RCVBUF_LOCK 2
542 #define SOCK_BINDADDR_LOCK 4
543 #define SOCK_BINDPORT_LOCK 8
545 /* sock_iocb: used to kick off async processing of socket ios */
547 struct list_head list;
553 struct scm_cookie *scm;
554 struct msghdr *msg, async_msg;
555 struct iovec async_iov;
558 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
560 BUG_ON(sizeof(struct sock_iocb) > KIOCB_PRIVATE_SIZE);
561 return (struct sock_iocb *)iocb->private;
564 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
566 return container_of((void *)si, struct kiocb, private);
569 struct socket_alloc {
570 struct socket socket;
571 struct inode vfs_inode;
574 static inline struct socket *SOCKET_I(struct inode *inode)
576 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
579 static inline struct inode *SOCK_INODE(struct socket *socket)
581 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
584 /* Used by processes to "lock" a socket state, so that
585 * interrupts and bottom half handlers won't change it
586 * from under us. It essentially blocks any incoming
587 * packets, so that we won't get any new data or any
588 * packets that change the state of the socket.
590 * While locked, BH processing will add new packets to
591 * the backlog queue. This queue is processed by the
592 * owner of the socket lock right before it is released.
594 * Since ~2.3.5 it is also exclusive sleep lock serializing
595 * accesses from user process context.
597 extern void __lock_sock(struct sock *sk);
598 extern void __release_sock(struct sock *sk);
599 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
601 extern void FASTCALL(lock_sock(struct sock *sk));
602 extern void FASTCALL(release_sock(struct sock *sk));
604 /* BH context may only use the following locking interface. */
605 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
606 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
608 extern struct sock * sk_alloc(int family, int priority, int zero_it,
610 extern void sk_free(struct sock *sk);
612 extern struct sk_buff *sock_wmalloc(struct sock *sk,
613 unsigned long size, int force,
615 extern struct sk_buff *sock_rmalloc(struct sock *sk,
616 unsigned long size, int force,
618 extern void sock_wfree(struct sk_buff *skb);
619 extern void sock_rfree(struct sk_buff *skb);
621 extern int sock_setsockopt(struct socket *sock, int level,
622 int op, char __user *optval,
625 extern int sock_getsockopt(struct socket *sock, int level,
626 int op, char __user *optval,
628 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
632 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
633 unsigned long header_len,
634 unsigned long data_len,
637 extern void *sock_kmalloc(struct sock *sk, int size, int priority);
638 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
639 extern void sk_send_sigurg(struct sock *sk);
642 * Functions to fill in entries in struct proto_ops when a protocol
643 * does not implement a particular function.
645 extern int sock_no_release(struct socket *);
646 extern int sock_no_bind(struct socket *,
647 struct sockaddr *, int);
648 extern int sock_no_connect(struct socket *,
649 struct sockaddr *, int, int);
650 extern int sock_no_socketpair(struct socket *,
652 extern int sock_no_accept(struct socket *,
653 struct socket *, int);
654 extern int sock_no_getname(struct socket *,
655 struct sockaddr *, int *, int);
656 extern unsigned int sock_no_poll(struct file *, struct socket *,
657 struct poll_table_struct *);
658 extern int sock_no_ioctl(struct socket *, unsigned int,
660 extern int sock_no_listen(struct socket *, int);
661 extern int sock_no_shutdown(struct socket *, int);
662 extern int sock_no_getsockopt(struct socket *, int , int,
663 char __user *, int __user *);
664 extern int sock_no_setsockopt(struct socket *, int, int,
666 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
667 struct msghdr *, size_t);
668 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
669 struct msghdr *, size_t, int);
670 extern int sock_no_mmap(struct file *file,
672 struct vm_area_struct *vma);
673 extern ssize_t sock_no_sendpage(struct socket *sock,
675 int offset, size_t size,
679 * Default socket callbacks and setup code
682 extern void sock_def_destruct(struct sock *);
684 /* Initialise core socket variables */
685 extern void sock_init_data(struct socket *sock, struct sock *sk);
688 * sk_filter - run a packet through a socket filter
689 * @sk: sock associated with &sk_buff
690 * @skb: buffer to filter
691 * @needlock: set to 1 if the sock is not locked by caller.
693 * Run the filter code and then cut skb->data to correct size returned by
694 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
695 * than pkt_len we keep whole skb->data. This is the socket level
696 * wrapper to sk_run_filter. It returns 0 if the packet should
697 * be accepted or -EPERM if the packet should be tossed.
701 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
705 err = security_sock_rcv_skb(sk, skb);
710 struct sk_filter *filter;
715 filter = sk->sk_filter;
717 int pkt_len = sk_run_filter(skb, filter->insns,
722 skb_trim(skb, pkt_len);
732 * sk_filter_release: Release a socket filter
734 * @fp: filter to remove
736 * Remove a filter from a socket and release its resources.
739 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
741 unsigned int size = sk_filter_len(fp);
743 atomic_sub(size, &sk->sk_omem_alloc);
745 if (atomic_dec_and_test(&fp->refcnt))
749 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
751 atomic_inc(&fp->refcnt);
752 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
756 * Socket reference counting postulates.
758 * * Each user of socket SHOULD hold a reference count.
759 * * Each access point to socket (an hash table bucket, reference from a list,
760 * running timer, skb in flight MUST hold a reference count.
761 * * When reference count hits 0, it means it will never increase back.
762 * * When reference count hits 0, it means that no references from
763 * outside exist to this socket and current process on current CPU
764 * is last user and may/should destroy this socket.
765 * * sk_free is called from any context: process, BH, IRQ. When
766 * it is called, socket has no references from outside -> sk_free
767 * may release descendant resources allocated by the socket, but
768 * to the time when it is called, socket is NOT referenced by any
769 * hash tables, lists etc.
770 * * Packets, delivered from outside (from network or from another process)
771 * and enqueued on receive/error queues SHOULD NOT grab reference count,
772 * when they sit in queue. Otherwise, packets will leak to hole, when
773 * socket is looked up by one cpu and unhasing is made by another CPU.
774 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
775 * (leak to backlog). Packet socket does all the processing inside
776 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
777 * use separate SMP lock, so that they are prone too.
780 /* Ungrab socket and destroy it, if it was the last reference. */
781 static inline void sock_put(struct sock *sk)
783 if (atomic_dec_and_test(&sk->sk_refcnt))
787 /* Detach socket from process context.
788 * Announce socket dead, detach it from wait queue and inode.
789 * Note that parent inode held reference count on this struct sock,
790 * we do not release it in this function, because protocol
791 * probably wants some additional cleanups or even continuing
792 * to work with this socket (TCP).
794 static inline void sock_orphan(struct sock *sk)
796 write_lock_bh(&sk->sk_callback_lock);
797 sock_set_flag(sk, SOCK_DEAD);
798 sk->sk_socket = NULL;
800 write_unlock_bh(&sk->sk_callback_lock);
803 static inline void sock_graft(struct sock *sk, struct socket *parent)
805 write_lock_bh(&sk->sk_callback_lock);
806 sk->sk_sleep = &parent->wait;
808 sk->sk_socket = parent;
809 write_unlock_bh(&sk->sk_callback_lock);
812 static inline int sock_i_uid(struct sock *sk)
816 read_lock(&sk->sk_callback_lock);
817 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
818 read_unlock(&sk->sk_callback_lock);
822 static inline unsigned long sock_i_ino(struct sock *sk)
826 read_lock(&sk->sk_callback_lock);
827 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
828 read_unlock(&sk->sk_callback_lock);
832 static inline struct dst_entry *
833 __sk_dst_get(struct sock *sk)
835 return sk->sk_dst_cache;
838 static inline struct dst_entry *
839 sk_dst_get(struct sock *sk)
841 struct dst_entry *dst;
843 read_lock(&sk->sk_dst_lock);
844 dst = sk->sk_dst_cache;
847 read_unlock(&sk->sk_dst_lock);
852 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
854 struct dst_entry *old_dst;
856 old_dst = sk->sk_dst_cache;
857 sk->sk_dst_cache = dst;
858 dst_release(old_dst);
862 sk_dst_set(struct sock *sk, struct dst_entry *dst)
864 write_lock(&sk->sk_dst_lock);
865 __sk_dst_set(sk, dst);
866 write_unlock(&sk->sk_dst_lock);
870 __sk_dst_reset(struct sock *sk)
872 struct dst_entry *old_dst;
874 old_dst = sk->sk_dst_cache;
875 sk->sk_dst_cache = NULL;
876 dst_release(old_dst);
880 sk_dst_reset(struct sock *sk)
882 write_lock(&sk->sk_dst_lock);
884 write_unlock(&sk->sk_dst_lock);
887 static inline struct dst_entry *
888 __sk_dst_check(struct sock *sk, u32 cookie)
890 struct dst_entry *dst = sk->sk_dst_cache;
892 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
893 sk->sk_dst_cache = NULL;
900 static inline struct dst_entry *
901 sk_dst_check(struct sock *sk, u32 cookie)
903 struct dst_entry *dst = sk_dst_get(sk);
905 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
915 * Queue a received datagram if it will fit. Stream and sequenced
916 * protocols can't normally use this as they need to fit buffers in
917 * and play with them.
919 * Inlined as it's very short and called for pretty much every
920 * packet ever received.
923 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
927 skb->destructor = sock_wfree;
928 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
931 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
934 skb->destructor = sock_rfree;
935 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
938 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
939 unsigned long expires);
941 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
943 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
948 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
949 number of warnings when compiling with -W --ANK
951 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
952 (unsigned)sk->sk_rcvbuf) {
957 /* It would be deadlock, if sock_queue_rcv_skb is used
958 with socket lock! We assume that users of this
959 function are lock free.
961 err = sk_filter(sk, skb, 1);
966 skb_set_owner_r(skb, sk);
968 /* Cache the SKB length before we tack it onto the receive
969 * queue. Once it is added it no longer belongs to us and
970 * may be freed by other threads of control pulling packets
975 skb_queue_tail(&sk->sk_receive_queue, skb);
977 if (!sock_flag(sk, SOCK_DEAD))
978 sk->sk_data_ready(sk, skb_len);
983 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
985 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
986 number of warnings when compiling with -W --ANK
988 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
989 (unsigned)sk->sk_rcvbuf)
991 skb_set_owner_r(skb, sk);
992 skb_queue_tail(&sk->sk_error_queue, skb);
993 if (!sock_flag(sk, SOCK_DEAD))
994 sk->sk_data_ready(sk, skb->len);
999 * Recover an error report and clear atomically
1002 static inline int sock_error(struct sock *sk)
1004 int err = xchg(&sk->sk_err, 0);
1008 static inline unsigned long sock_wspace(struct sock *sk)
1012 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1013 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1020 static inline void sk_wake_async(struct sock *sk, int how, int band)
1022 if (sk->sk_socket && sk->sk_socket->fasync_list)
1023 sock_wake_async(sk->sk_socket, how, band);
1026 #define SOCK_MIN_SNDBUF 2048
1027 #define SOCK_MIN_RCVBUF 256
1030 * Default write policy as shown to user space via poll/select/SIGIO
1032 static inline int sock_writeable(struct sock *sk)
1034 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1037 static inline int gfp_any(void)
1039 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1042 static inline long sock_rcvtimeo(struct sock *sk, int noblock)
1044 return noblock ? 0 : sk->sk_rcvtimeo;
1047 static inline long sock_sndtimeo(struct sock *sk, int noblock)
1049 return noblock ? 0 : sk->sk_sndtimeo;
1052 static inline int sock_rcvlowat(struct sock *sk, int waitall, int len)
1054 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1057 /* Alas, with timeout socket operations are not restartable.
1058 * Compare this to poll().
1060 static inline int sock_intr_errno(long timeo)
1062 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1065 static __inline__ void
1066 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1068 struct timeval *stamp = &skb->stamp;
1069 if (sk->sk_rcvtstamp) {
1070 /* Race occurred between timestamp enabling and packet
1071 receiving. Fill in the current time for now. */
1072 if (stamp->tv_sec == 0)
1073 do_gettimeofday(stamp);
1074 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1077 sk->sk_stamp = *stamp;
1081 * sk_eat_skb - Release a skb if it is no longer needed
1082 * @sk - socket to eat this skb from
1083 * @skb - socket buffer to eat
1085 * This routine must be called with interrupts disabled or with the socket
1086 * locked so that the sk_buff queue operation is ok.
1088 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1090 __skb_unlink(skb, &sk->sk_receive_queue);
1094 extern atomic_t netstamp_needed;
1095 extern void sock_enable_timestamp(struct sock *sk);
1096 extern void sock_disable_timestamp(struct sock *sk);
1098 static inline void net_timestamp(struct timeval *stamp)
1100 if (atomic_read(&netstamp_needed))
1101 do_gettimeofday(stamp);
1108 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1111 * Enable debug/info messages
1115 #define NETDEBUG(x) do { } while (0)
1116 #define LIMIT_NETDEBUG(x) do {} while(0)
1118 #define NETDEBUG(x) do { x; } while (0)
1119 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1123 * Macros for sleeping on a socket. Use them like this:
1125 * SOCK_SLEEP_PRE(sk)
1128 * SOCK_SLEEP_POST(sk)
1130 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1131 * and when the last use of them in DECnet has gone, I'm intending to
1135 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1136 DECLARE_WAITQUEUE(wait, tsk); \
1137 tsk->state = TASK_INTERRUPTIBLE; \
1138 add_wait_queue((sk)->sk_sleep, &wait); \
1141 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1142 remove_wait_queue((sk)->sk_sleep, &wait); \
1146 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1149 sock_set_flag(sk, bit);
1151 sock_reset_flag(sk, bit);
1154 extern __u32 sysctl_wmem_max;
1155 extern __u32 sysctl_rmem_max;
1157 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1159 #endif /* _SOCK_H */