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
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/kernel.h>
44 #include <linux/list.h>
45 #include <linux/timer.h>
46 #include <linux/cache.h>
47 #include <linux/module.h>
48 #include <linux/lockdep.h>
49 #include <linux/netdevice.h>
50 #include <linux/skbuff.h> /* struct sk_buff */
52 #include <linux/security.h>
54 #include <linux/filter.h>
56 #include <asm/atomic.h>
58 #include <net/checksum.h>
59 #include <net/net_namespace.h>
62 * This structure really needs to be cleaned up.
63 * Most of it is for TCP, and not used by any of
64 * the other protocols.
67 /* Define this to get the SOCK_DBG debugging facility. */
68 #define SOCK_DEBUGGING
70 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
71 printk(KERN_DEBUG msg); } while (0)
73 #define SOCK_DEBUG(sk, msg...) do { } while (0)
76 /* This is the per-socket lock. The spinlock provides a synchronization
77 * between user contexts and software interrupt processing, whereas the
78 * mini-semaphore synchronizes multiple users amongst themselves.
85 * We express the mutex-alike socket_lock semantics
86 * to the lock validator by explicitly managing
87 * the slock as a lock variant (in addition to
90 #ifdef CONFIG_DEBUG_LOCK_ALLOC
91 struct lockdep_map dep_map;
99 * struct sock_common - minimal network layer representation of sockets
100 * @skc_family: network address family
101 * @skc_state: Connection state
102 * @skc_reuse: %SO_REUSEADDR setting
103 * @skc_bound_dev_if: bound device index if != 0
104 * @skc_node: main hash linkage for various protocol lookup tables
105 * @skc_bind_node: bind hash linkage for various protocol lookup tables
106 * @skc_refcnt: reference count
107 * @skc_hash: hash value used with various protocol lookup tables
108 * @skc_prot: protocol handlers inside a network family
109 * @skc_net: reference to the network namespace of this socket
111 * This is the minimal network layer representation of sockets, the header
112 * for struct sock and struct inet_timewait_sock.
115 unsigned short skc_family;
116 volatile unsigned char skc_state;
117 unsigned char skc_reuse;
118 int skc_bound_dev_if;
119 struct hlist_node skc_node;
120 struct hlist_node skc_bind_node;
122 unsigned int skc_hash;
123 struct proto *skc_prot;
128 * struct sock - network layer representation of sockets
129 * @__sk_common: shared layout with inet_timewait_sock
130 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
131 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
132 * @sk_lock: synchronizer
133 * @sk_rcvbuf: size of receive buffer in bytes
134 * @sk_sleep: sock wait queue
135 * @sk_dst_cache: destination cache
136 * @sk_dst_lock: destination cache lock
137 * @sk_policy: flow policy
138 * @sk_rmem_alloc: receive queue bytes committed
139 * @sk_receive_queue: incoming packets
140 * @sk_wmem_alloc: transmit queue bytes committed
141 * @sk_write_queue: Packet sending queue
142 * @sk_async_wait_queue: DMA copied packets
143 * @sk_omem_alloc: "o" is "option" or "other"
144 * @sk_wmem_queued: persistent queue size
145 * @sk_forward_alloc: space allocated forward
146 * @sk_allocation: allocation mode
147 * @sk_sndbuf: size of send buffer in bytes
148 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
149 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
150 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
151 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
152 * @sk_lingertime: %SO_LINGER l_linger setting
153 * @sk_backlog: always used with the per-socket spinlock held
154 * @sk_callback_lock: used with the callbacks in the end of this struct
155 * @sk_error_queue: rarely used
156 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
157 * @sk_err: last error
158 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
159 * @sk_ack_backlog: current listen backlog
160 * @sk_max_ack_backlog: listen backlog set in listen()
161 * @sk_priority: %SO_PRIORITY setting
162 * @sk_type: socket type (%SOCK_STREAM, etc)
163 * @sk_protocol: which protocol this socket belongs in this network family
164 * @sk_peercred: %SO_PEERCRED setting
165 * @sk_rcvlowat: %SO_RCVLOWAT setting
166 * @sk_rcvtimeo: %SO_RCVTIMEO setting
167 * @sk_sndtimeo: %SO_SNDTIMEO setting
168 * @sk_filter: socket filtering instructions
169 * @sk_protinfo: private area, net family specific, when not using slab
170 * @sk_timer: sock cleanup timer
171 * @sk_stamp: time stamp of last packet received
172 * @sk_socket: Identd and reporting IO signals
173 * @sk_user_data: RPC layer private data
174 * @sk_sndmsg_page: cached page for sendmsg
175 * @sk_sndmsg_off: cached offset for sendmsg
176 * @sk_send_head: front of stuff to transmit
177 * @sk_security: used by security modules
178 * @sk_write_pending: a write to stream socket waits to start
179 * @sk_state_change: callback to indicate change in the state of the sock
180 * @sk_data_ready: callback to indicate there is data to be processed
181 * @sk_write_space: callback to indicate there is bf sending space available
182 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
183 * @sk_backlog_rcv: callback to process the backlog
184 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
188 * Now struct inet_timewait_sock also uses sock_common, so please just
189 * don't add nothing before this first member (__sk_common) --acme
191 struct sock_common __sk_common;
192 #define sk_family __sk_common.skc_family
193 #define sk_state __sk_common.skc_state
194 #define sk_reuse __sk_common.skc_reuse
195 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
196 #define sk_node __sk_common.skc_node
197 #define sk_bind_node __sk_common.skc_bind_node
198 #define sk_refcnt __sk_common.skc_refcnt
199 #define sk_hash __sk_common.skc_hash
200 #define sk_prot __sk_common.skc_prot
201 #define sk_net __sk_common.skc_net
202 unsigned char sk_shutdown : 2,
205 unsigned char sk_protocol;
206 unsigned short sk_type;
208 socket_lock_t sk_lock;
210 * The backlog queue is special, it is always used with
211 * the per-socket spinlock held and requires low latency
212 * access. Therefore we special case it's implementation.
215 struct sk_buff *head;
216 struct sk_buff *tail;
218 wait_queue_head_t *sk_sleep;
219 struct dst_entry *sk_dst_cache;
220 struct xfrm_policy *sk_policy[2];
221 rwlock_t sk_dst_lock;
222 atomic_t sk_rmem_alloc;
223 atomic_t sk_wmem_alloc;
224 atomic_t sk_omem_alloc;
226 struct sk_buff_head sk_receive_queue;
227 struct sk_buff_head sk_write_queue;
228 struct sk_buff_head sk_async_wait_queue;
230 int sk_forward_alloc;
235 unsigned long sk_flags;
236 unsigned long sk_lingertime;
237 struct sk_buff_head sk_error_queue;
238 struct proto *sk_prot_creator;
239 rwlock_t sk_callback_lock;
242 unsigned short sk_ack_backlog;
243 unsigned short sk_max_ack_backlog;
245 struct ucred sk_peercred;
248 struct sk_filter *sk_filter;
250 struct timer_list sk_timer;
252 struct socket *sk_socket;
254 struct page *sk_sndmsg_page;
255 struct sk_buff *sk_send_head;
257 int sk_write_pending;
259 void (*sk_state_change)(struct sock *sk);
260 void (*sk_data_ready)(struct sock *sk, int bytes);
261 void (*sk_write_space)(struct sock *sk);
262 void (*sk_error_report)(struct sock *sk);
263 int (*sk_backlog_rcv)(struct sock *sk,
264 struct sk_buff *skb);
265 void (*sk_destruct)(struct sock *sk);
269 * Hashed lists helper routines
271 static inline struct sock *__sk_head(const struct hlist_head *head)
273 return hlist_entry(head->first, struct sock, sk_node);
276 static inline struct sock *sk_head(const struct hlist_head *head)
278 return hlist_empty(head) ? NULL : __sk_head(head);
281 static inline struct sock *sk_next(const struct sock *sk)
283 return sk->sk_node.next ?
284 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
287 static inline int sk_unhashed(const struct sock *sk)
289 return hlist_unhashed(&sk->sk_node);
292 static inline int sk_hashed(const struct sock *sk)
294 return !sk_unhashed(sk);
297 static __inline__ void sk_node_init(struct hlist_node *node)
302 static __inline__ void __sk_del_node(struct sock *sk)
304 __hlist_del(&sk->sk_node);
307 static __inline__ int __sk_del_node_init(struct sock *sk)
311 sk_node_init(&sk->sk_node);
317 /* Grab socket reference count. This operation is valid only
318 when sk is ALREADY grabbed f.e. it is found in hash table
319 or a list and the lookup is made under lock preventing hash table
323 static inline void sock_hold(struct sock *sk)
325 atomic_inc(&sk->sk_refcnt);
328 /* Ungrab socket in the context, which assumes that socket refcnt
329 cannot hit zero, f.e. it is true in context of any socketcall.
331 static inline void __sock_put(struct sock *sk)
333 atomic_dec(&sk->sk_refcnt);
336 static __inline__ int sk_del_node_init(struct sock *sk)
338 int rc = __sk_del_node_init(sk);
341 /* paranoid for a while -acme */
342 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
348 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
350 hlist_add_head(&sk->sk_node, list);
353 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
356 __sk_add_node(sk, list);
359 static __inline__ void __sk_del_bind_node(struct sock *sk)
361 __hlist_del(&sk->sk_bind_node);
364 static __inline__ void sk_add_bind_node(struct sock *sk,
365 struct hlist_head *list)
367 hlist_add_head(&sk->sk_bind_node, list);
370 #define sk_for_each(__sk, node, list) \
371 hlist_for_each_entry(__sk, node, list, sk_node)
372 #define sk_for_each_from(__sk, node) \
373 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
374 hlist_for_each_entry_from(__sk, node, sk_node)
375 #define sk_for_each_continue(__sk, node) \
376 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
377 hlist_for_each_entry_continue(__sk, node, sk_node)
378 #define sk_for_each_safe(__sk, node, tmp, list) \
379 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
380 #define sk_for_each_bound(__sk, node, list) \
381 hlist_for_each_entry(__sk, node, list, sk_bind_node)
394 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
395 SOCK_DBG, /* %SO_DEBUG setting */
396 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
397 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
398 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
399 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
402 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
404 nsk->sk_flags = osk->sk_flags;
407 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
409 __set_bit(flag, &sk->sk_flags);
412 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
414 __clear_bit(flag, &sk->sk_flags);
417 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
419 return test_bit(flag, &sk->sk_flags);
422 static inline void sk_acceptq_removed(struct sock *sk)
424 sk->sk_ack_backlog--;
427 static inline void sk_acceptq_added(struct sock *sk)
429 sk->sk_ack_backlog++;
432 static inline int sk_acceptq_is_full(struct sock *sk)
434 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
438 * Compute minimal free write space needed to queue new packets.
440 static inline int sk_stream_min_wspace(struct sock *sk)
442 return sk->sk_wmem_queued / 2;
445 static inline int sk_stream_wspace(struct sock *sk)
447 return sk->sk_sndbuf - sk->sk_wmem_queued;
450 extern void sk_stream_write_space(struct sock *sk);
452 static inline int sk_stream_memory_free(struct sock *sk)
454 return sk->sk_wmem_queued < sk->sk_sndbuf;
457 extern void sk_stream_rfree(struct sk_buff *skb);
459 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
462 skb->destructor = sk_stream_rfree;
463 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
464 sk->sk_forward_alloc -= skb->truesize;
467 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
469 skb_truesize_check(skb);
470 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
471 sk->sk_wmem_queued -= skb->truesize;
472 sk->sk_forward_alloc += skb->truesize;
476 /* The per-socket spinlock must be held here. */
477 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
479 if (!sk->sk_backlog.tail) {
480 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
482 sk->sk_backlog.tail->next = skb;
483 sk->sk_backlog.tail = skb;
488 #define sk_wait_event(__sk, __timeo, __condition) \
490 release_sock(__sk); \
491 __rc = __condition; \
493 *(__timeo) = schedule_timeout(*(__timeo)); \
496 __rc = __condition; \
500 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
501 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
502 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
503 extern int sk_stream_error(struct sock *sk, int flags, int err);
504 extern void sk_stream_kill_queues(struct sock *sk);
506 extern int sk_wait_data(struct sock *sk, long *timeo);
508 struct request_sock_ops;
509 struct timewait_sock_ops;
511 /* Networking protocol blocks we attach to sockets.
512 * socket layer -> transport layer interface
513 * transport -> network interface is defined by struct inet_proto
516 void (*close)(struct sock *sk,
518 int (*connect)(struct sock *sk,
519 struct sockaddr *uaddr,
521 int (*disconnect)(struct sock *sk, int flags);
523 struct sock * (*accept) (struct sock *sk, int flags, int *err);
525 int (*ioctl)(struct sock *sk, int cmd,
527 int (*init)(struct sock *sk);
528 int (*destroy)(struct sock *sk);
529 void (*shutdown)(struct sock *sk, int how);
530 int (*setsockopt)(struct sock *sk, int level,
531 int optname, char __user *optval,
533 int (*getsockopt)(struct sock *sk, int level,
534 int optname, char __user *optval,
536 int (*compat_setsockopt)(struct sock *sk,
538 int optname, char __user *optval,
540 int (*compat_getsockopt)(struct sock *sk,
542 int optname, char __user *optval,
544 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
545 struct msghdr *msg, size_t len);
546 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
548 size_t len, int noblock, int flags,
550 int (*sendpage)(struct sock *sk, struct page *page,
551 int offset, size_t size, int flags);
552 int (*bind)(struct sock *sk,
553 struct sockaddr *uaddr, int addr_len);
555 int (*backlog_rcv) (struct sock *sk,
556 struct sk_buff *skb);
558 /* Keeping track of sk's, looking them up, and port selection methods. */
559 void (*hash)(struct sock *sk);
560 void (*unhash)(struct sock *sk);
561 int (*get_port)(struct sock *sk, unsigned short snum);
563 /* Memory pressure */
564 void (*enter_memory_pressure)(void);
565 atomic_t *memory_allocated; /* Current allocated memory. */
566 atomic_t *sockets_allocated; /* Current number of sockets. */
568 * Pressure flag: try to collapse.
569 * Technical note: it is used by multiple contexts non atomically.
570 * All the sk_stream_mem_schedule() is of this nature: accounting
571 * is strict, actions are advisory and have some latency.
573 int *memory_pressure;
579 struct kmem_cache *slab;
580 unsigned int obj_size;
582 atomic_t *orphan_count;
584 struct request_sock_ops *rsk_prot;
585 struct timewait_sock_ops *twsk_prot;
587 struct module *owner;
591 struct list_head node;
592 #ifdef SOCK_REFCNT_DEBUG
597 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
601 extern int proto_register(struct proto *prot, int alloc_slab);
602 extern void proto_unregister(struct proto *prot);
604 #ifdef SOCK_REFCNT_DEBUG
605 static inline void sk_refcnt_debug_inc(struct sock *sk)
607 atomic_inc(&sk->sk_prot->socks);
610 static inline void sk_refcnt_debug_dec(struct sock *sk)
612 atomic_dec(&sk->sk_prot->socks);
613 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
614 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
617 static inline void sk_refcnt_debug_release(const struct sock *sk)
619 if (atomic_read(&sk->sk_refcnt) != 1)
620 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
621 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
623 #else /* SOCK_REFCNT_DEBUG */
624 #define sk_refcnt_debug_inc(sk) do { } while (0)
625 #define sk_refcnt_debug_dec(sk) do { } while (0)
626 #define sk_refcnt_debug_release(sk) do { } while (0)
627 #endif /* SOCK_REFCNT_DEBUG */
629 /* Called with local bh disabled */
630 static __inline__ void sock_prot_inc_use(struct proto *prot)
632 prot->stats[smp_processor_id()].inuse++;
635 static __inline__ void sock_prot_dec_use(struct proto *prot)
637 prot->stats[smp_processor_id()].inuse--;
640 /* With per-bucket locks this operation is not-atomic, so that
641 * this version is not worse.
643 static inline void __sk_prot_rehash(struct sock *sk)
645 sk->sk_prot->unhash(sk);
646 sk->sk_prot->hash(sk);
649 /* About 10 seconds */
650 #define SOCK_DESTROY_TIME (10*HZ)
652 /* Sockets 0-1023 can't be bound to unless you are superuser */
653 #define PROT_SOCK 1024
655 #define SHUTDOWN_MASK 3
656 #define RCV_SHUTDOWN 1
657 #define SEND_SHUTDOWN 2
659 #define SOCK_SNDBUF_LOCK 1
660 #define SOCK_RCVBUF_LOCK 2
661 #define SOCK_BINDADDR_LOCK 4
662 #define SOCK_BINDPORT_LOCK 8
664 /* sock_iocb: used to kick off async processing of socket ios */
666 struct list_head list;
672 struct scm_cookie *scm;
673 struct msghdr *msg, async_msg;
677 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
679 return (struct sock_iocb *)iocb->private;
682 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
687 struct socket_alloc {
688 struct socket socket;
689 struct inode vfs_inode;
692 static inline struct socket *SOCKET_I(struct inode *inode)
694 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
697 static inline struct inode *SOCK_INODE(struct socket *socket)
699 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
702 extern void __sk_stream_mem_reclaim(struct sock *sk);
703 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
705 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
707 static inline int sk_stream_pages(int amt)
709 return DIV_ROUND_UP(amt, SK_STREAM_MEM_QUANTUM);
712 static inline void sk_stream_mem_reclaim(struct sock *sk)
714 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
715 __sk_stream_mem_reclaim(sk);
718 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
720 return (int)skb->truesize <= sk->sk_forward_alloc ||
721 sk_stream_mem_schedule(sk, skb->truesize, 1);
724 static inline int sk_stream_wmem_schedule(struct sock *sk, int size)
726 return size <= sk->sk_forward_alloc ||
727 sk_stream_mem_schedule(sk, size, 0);
730 /* Used by processes to "lock" a socket state, so that
731 * interrupts and bottom half handlers won't change it
732 * from under us. It essentially blocks any incoming
733 * packets, so that we won't get any new data or any
734 * packets that change the state of the socket.
736 * While locked, BH processing will add new packets to
737 * the backlog queue. This queue is processed by the
738 * owner of the socket lock right before it is released.
740 * Since ~2.3.5 it is also exclusive sleep lock serializing
741 * accesses from user process context.
743 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
746 * Macro so as to not evaluate some arguments when
747 * lockdep is not enabled.
749 * Mark both the sk_lock and the sk_lock.slock as a
750 * per-address-family lock class.
752 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
754 sk->sk_lock.owned = 0; \
755 init_waitqueue_head(&sk->sk_lock.wq); \
756 spin_lock_init(&(sk)->sk_lock.slock); \
757 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
758 sizeof((sk)->sk_lock)); \
759 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
761 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
764 extern void FASTCALL(lock_sock_nested(struct sock *sk, int subclass));
766 static inline void lock_sock(struct sock *sk)
768 lock_sock_nested(sk, 0);
771 extern void FASTCALL(release_sock(struct sock *sk));
773 /* BH context may only use the following locking interface. */
774 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
775 #define bh_lock_sock_nested(__sk) \
776 spin_lock_nested(&((__sk)->sk_lock.slock), \
777 SINGLE_DEPTH_NESTING)
778 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
780 extern struct sock *sk_alloc(struct net *net, int family,
782 struct proto *prot, int zero_it);
783 extern void sk_free(struct sock *sk);
784 extern struct sock *sk_clone(const struct sock *sk,
785 const gfp_t priority);
787 extern struct sk_buff *sock_wmalloc(struct sock *sk,
788 unsigned long size, int force,
790 extern struct sk_buff *sock_rmalloc(struct sock *sk,
791 unsigned long size, int force,
793 extern void sock_wfree(struct sk_buff *skb);
794 extern void sock_rfree(struct sk_buff *skb);
796 extern int sock_setsockopt(struct socket *sock, int level,
797 int op, char __user *optval,
800 extern int sock_getsockopt(struct socket *sock, int level,
801 int op, char __user *optval,
803 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
807 extern void *sock_kmalloc(struct sock *sk, int size,
809 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
810 extern void sk_send_sigurg(struct sock *sk);
813 * Functions to fill in entries in struct proto_ops when a protocol
814 * does not implement a particular function.
816 extern int sock_no_bind(struct socket *,
817 struct sockaddr *, int);
818 extern int sock_no_connect(struct socket *,
819 struct sockaddr *, int, int);
820 extern int sock_no_socketpair(struct socket *,
822 extern int sock_no_accept(struct socket *,
823 struct socket *, int);
824 extern int sock_no_getname(struct socket *,
825 struct sockaddr *, int *, int);
826 extern unsigned int sock_no_poll(struct file *, struct socket *,
827 struct poll_table_struct *);
828 extern int sock_no_ioctl(struct socket *, unsigned int,
830 extern int sock_no_listen(struct socket *, int);
831 extern int sock_no_shutdown(struct socket *, int);
832 extern int sock_no_getsockopt(struct socket *, int , int,
833 char __user *, int __user *);
834 extern int sock_no_setsockopt(struct socket *, int, int,
836 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
837 struct msghdr *, size_t);
838 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
839 struct msghdr *, size_t, int);
840 extern int sock_no_mmap(struct file *file,
842 struct vm_area_struct *vma);
843 extern ssize_t sock_no_sendpage(struct socket *sock,
845 int offset, size_t size,
849 * Functions to fill in entries in struct proto_ops when a protocol
850 * uses the inet style.
852 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
853 char __user *optval, int __user *optlen);
854 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
855 struct msghdr *msg, size_t size, int flags);
856 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
857 char __user *optval, int optlen);
858 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
859 int optname, char __user *optval, int __user *optlen);
860 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
861 int optname, char __user *optval, int optlen);
863 extern void sk_common_release(struct sock *sk);
866 * Default socket callbacks and setup code
869 /* Initialise core socket variables */
870 extern void sock_init_data(struct socket *sock, struct sock *sk);
873 * sk_filter - run a packet through a socket filter
874 * @sk: sock associated with &sk_buff
875 * @skb: buffer to filter
876 * @needlock: set to 1 if the sock is not locked by caller.
878 * Run the filter code and then cut skb->data to correct size returned by
879 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
880 * than pkt_len we keep whole skb->data. This is the socket level
881 * wrapper to sk_run_filter. It returns 0 if the packet should
882 * be accepted or -EPERM if the packet should be tossed.
886 static inline int sk_filter(struct sock *sk, struct sk_buff *skb)
889 struct sk_filter *filter;
891 err = security_sock_rcv_skb(sk, skb);
896 filter = sk->sk_filter;
898 unsigned int pkt_len = sk_run_filter(skb, filter->insns,
900 err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM;
902 rcu_read_unlock_bh();
908 * sk_filter_rcu_free: Free a socket filter
909 * @rcu: rcu_head that contains the sk_filter to free
911 static inline void sk_filter_rcu_free(struct rcu_head *rcu)
913 struct sk_filter *fp = container_of(rcu, struct sk_filter, rcu);
918 * sk_filter_release: Release a socket filter
920 * @fp: filter to remove
922 * Remove a filter from a socket and release its resources.
925 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
927 unsigned int size = sk_filter_len(fp);
929 atomic_sub(size, &sk->sk_omem_alloc);
931 if (atomic_dec_and_test(&fp->refcnt))
932 call_rcu_bh(&fp->rcu, sk_filter_rcu_free);
935 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
937 atomic_inc(&fp->refcnt);
938 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
942 * Socket reference counting postulates.
944 * * Each user of socket SHOULD hold a reference count.
945 * * Each access point to socket (an hash table bucket, reference from a list,
946 * running timer, skb in flight MUST hold a reference count.
947 * * When reference count hits 0, it means it will never increase back.
948 * * When reference count hits 0, it means that no references from
949 * outside exist to this socket and current process on current CPU
950 * is last user and may/should destroy this socket.
951 * * sk_free is called from any context: process, BH, IRQ. When
952 * it is called, socket has no references from outside -> sk_free
953 * may release descendant resources allocated by the socket, but
954 * to the time when it is called, socket is NOT referenced by any
955 * hash tables, lists etc.
956 * * Packets, delivered from outside (from network or from another process)
957 * and enqueued on receive/error queues SHOULD NOT grab reference count,
958 * when they sit in queue. Otherwise, packets will leak to hole, when
959 * socket is looked up by one cpu and unhasing is made by another CPU.
960 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
961 * (leak to backlog). Packet socket does all the processing inside
962 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
963 * use separate SMP lock, so that they are prone too.
966 /* Ungrab socket and destroy it, if it was the last reference. */
967 static inline void sock_put(struct sock *sk)
969 if (atomic_dec_and_test(&sk->sk_refcnt))
973 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
976 /* Detach socket from process context.
977 * Announce socket dead, detach it from wait queue and inode.
978 * Note that parent inode held reference count on this struct sock,
979 * we do not release it in this function, because protocol
980 * probably wants some additional cleanups or even continuing
981 * to work with this socket (TCP).
983 static inline void sock_orphan(struct sock *sk)
985 write_lock_bh(&sk->sk_callback_lock);
986 sock_set_flag(sk, SOCK_DEAD);
987 sk->sk_socket = NULL;
989 write_unlock_bh(&sk->sk_callback_lock);
992 static inline void sock_graft(struct sock *sk, struct socket *parent)
994 write_lock_bh(&sk->sk_callback_lock);
995 sk->sk_sleep = &parent->wait;
997 sk->sk_socket = parent;
998 security_sock_graft(sk, parent);
999 write_unlock_bh(&sk->sk_callback_lock);
1002 static inline void sock_copy(struct sock *nsk, const struct sock *osk)
1004 #ifdef CONFIG_SECURITY_NETWORK
1005 void *sptr = nsk->sk_security;
1008 memcpy(nsk, osk, osk->sk_prot->obj_size);
1009 get_net(nsk->sk_net);
1010 #ifdef CONFIG_SECURITY_NETWORK
1011 nsk->sk_security = sptr;
1012 security_sk_clone(osk, nsk);
1016 extern int sock_i_uid(struct sock *sk);
1017 extern unsigned long sock_i_ino(struct sock *sk);
1019 static inline struct dst_entry *
1020 __sk_dst_get(struct sock *sk)
1022 return sk->sk_dst_cache;
1025 static inline struct dst_entry *
1026 sk_dst_get(struct sock *sk)
1028 struct dst_entry *dst;
1030 read_lock(&sk->sk_dst_lock);
1031 dst = sk->sk_dst_cache;
1034 read_unlock(&sk->sk_dst_lock);
1039 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1041 struct dst_entry *old_dst;
1043 old_dst = sk->sk_dst_cache;
1044 sk->sk_dst_cache = dst;
1045 dst_release(old_dst);
1049 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1051 write_lock(&sk->sk_dst_lock);
1052 __sk_dst_set(sk, dst);
1053 write_unlock(&sk->sk_dst_lock);
1057 __sk_dst_reset(struct sock *sk)
1059 struct dst_entry *old_dst;
1061 old_dst = sk->sk_dst_cache;
1062 sk->sk_dst_cache = NULL;
1063 dst_release(old_dst);
1067 sk_dst_reset(struct sock *sk)
1069 write_lock(&sk->sk_dst_lock);
1071 write_unlock(&sk->sk_dst_lock);
1074 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1076 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1078 static inline int sk_can_gso(const struct sock *sk)
1080 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1083 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1085 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1087 sk->sk_wmem_queued += skb->truesize;
1088 sk->sk_forward_alloc -= skb->truesize;
1091 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1092 struct sk_buff *skb, struct page *page,
1095 if (skb->ip_summed == CHECKSUM_NONE) {
1097 __wsum csum = csum_and_copy_from_user(from,
1098 page_address(page) + off,
1102 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1103 } else if (copy_from_user(page_address(page) + off, from, copy))
1107 skb->data_len += copy;
1108 skb->truesize += copy;
1109 sk->sk_wmem_queued += copy;
1110 sk->sk_forward_alloc -= copy;
1115 * Queue a received datagram if it will fit. Stream and sequenced
1116 * protocols can't normally use this as they need to fit buffers in
1117 * and play with them.
1119 * Inlined as it's very short and called for pretty much every
1120 * packet ever received.
1123 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1127 skb->destructor = sock_wfree;
1128 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1131 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1134 skb->destructor = sock_rfree;
1135 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1138 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1139 unsigned long expires);
1141 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1143 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1145 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1147 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1148 number of warnings when compiling with -W --ANK
1150 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1151 (unsigned)sk->sk_rcvbuf)
1153 skb_set_owner_r(skb, sk);
1154 skb_queue_tail(&sk->sk_error_queue, skb);
1155 if (!sock_flag(sk, SOCK_DEAD))
1156 sk->sk_data_ready(sk, skb->len);
1161 * Recover an error report and clear atomically
1164 static inline int sock_error(struct sock *sk)
1167 if (likely(!sk->sk_err))
1169 err = xchg(&sk->sk_err, 0);
1173 static inline unsigned long sock_wspace(struct sock *sk)
1177 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1178 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1185 static inline void sk_wake_async(struct sock *sk, int how, int band)
1187 if (sk->sk_socket && sk->sk_socket->fasync_list)
1188 sock_wake_async(sk->sk_socket, how, band);
1191 #define SOCK_MIN_SNDBUF 2048
1192 #define SOCK_MIN_RCVBUF 256
1194 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1196 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1197 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1198 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1202 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1206 struct sk_buff *skb;
1209 hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header);
1210 skb = alloc_skb_fclone(size + hdr_len, gfp);
1212 skb->truesize += mem;
1213 if (sk_stream_wmem_schedule(sk, skb->truesize)) {
1214 skb_reserve(skb, hdr_len);
1219 sk->sk_prot->enter_memory_pressure();
1220 sk_stream_moderate_sndbuf(sk);
1225 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1229 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1232 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1234 struct page *page = NULL;
1236 page = alloc_pages(sk->sk_allocation, 0);
1238 sk->sk_prot->enter_memory_pressure();
1239 sk_stream_moderate_sndbuf(sk);
1245 * Default write policy as shown to user space via poll/select/SIGIO
1247 static inline int sock_writeable(const struct sock *sk)
1249 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1252 static inline gfp_t gfp_any(void)
1254 return in_atomic() ? GFP_ATOMIC : GFP_KERNEL;
1257 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1259 return noblock ? 0 : sk->sk_rcvtimeo;
1262 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1264 return noblock ? 0 : sk->sk_sndtimeo;
1267 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1269 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1272 /* Alas, with timeout socket operations are not restartable.
1273 * Compare this to poll().
1275 static inline int sock_intr_errno(long timeo)
1277 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1280 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1281 struct sk_buff *skb);
1283 static __inline__ void
1284 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1286 ktime_t kt = skb->tstamp;
1288 if (sock_flag(sk, SOCK_RCVTSTAMP))
1289 __sock_recv_timestamp(msg, sk, skb);
1295 * sk_eat_skb - Release a skb if it is no longer needed
1296 * @sk: socket to eat this skb from
1297 * @skb: socket buffer to eat
1298 * @copied_early: flag indicating whether DMA operations copied this data early
1300 * This routine must be called with interrupts disabled or with the socket
1301 * locked so that the sk_buff queue operation is ok.
1303 #ifdef CONFIG_NET_DMA
1304 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1306 __skb_unlink(skb, &sk->sk_receive_queue);
1310 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1313 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1315 __skb_unlink(skb, &sk->sk_receive_queue);
1320 extern void sock_enable_timestamp(struct sock *sk);
1321 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1322 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1325 * Enable debug/info messages
1327 extern int net_msg_warn;
1328 #define NETDEBUG(fmt, args...) \
1329 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1331 #define LIMIT_NETDEBUG(fmt, args...) \
1332 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1335 * Macros for sleeping on a socket. Use them like this:
1337 * SOCK_SLEEP_PRE(sk)
1340 * SOCK_SLEEP_POST(sk)
1342 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1343 * and when the last use of them in DECnet has gone, I'm intending to
1347 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1348 DECLARE_WAITQUEUE(wait, tsk); \
1349 tsk->state = TASK_INTERRUPTIBLE; \
1350 add_wait_queue((sk)->sk_sleep, &wait); \
1353 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1354 remove_wait_queue((sk)->sk_sleep, &wait); \
1358 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1361 sock_set_flag(sk, bit);
1363 sock_reset_flag(sk, bit);
1366 extern __u32 sysctl_wmem_max;
1367 extern __u32 sysctl_rmem_max;
1369 extern void sk_init(void);
1371 #ifdef CONFIG_SYSCTL
1372 extern struct ctl_table core_table[];
1375 extern int sysctl_optmem_max;
1377 extern __u32 sysctl_wmem_default;
1378 extern __u32 sysctl_rmem_default;
1380 #endif /* _SOCK_H */