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/list_nulls.h>
46 #include <linux/timer.h>
47 #include <linux/cache.h>
48 #include <linux/module.h>
49 #include <linux/lockdep.h>
50 #include <linux/netdevice.h>
51 #include <linux/skbuff.h> /* struct sk_buff */
53 #include <linux/security.h>
55 #include <linux/filter.h>
56 #include <linux/rculist_nulls.h>
58 #include <asm/atomic.h>
60 #include <net/checksum.h>
63 * This structure really needs to be cleaned up.
64 * Most of it is for TCP, and not used by any of
65 * the other protocols.
68 /* Define this to get the SOCK_DBG debugging facility. */
69 #define SOCK_DEBUGGING
71 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
72 printk(KERN_DEBUG msg); } while (0)
74 /* Validate arguments and do nothing */
75 static void inline int __attribute__ ((format (printf, 2, 3)))
76 SOCK_DEBUG(struct sock *sk, const char *msg, ...)
81 /* This is the per-socket lock. The spinlock provides a synchronization
82 * between user contexts and software interrupt processing, whereas the
83 * mini-semaphore synchronizes multiple users amongst themselves.
90 * We express the mutex-alike socket_lock semantics
91 * to the lock validator by explicitly managing
92 * the slock as a lock variant (in addition to
95 #ifdef CONFIG_DEBUG_LOCK_ALLOC
96 struct lockdep_map dep_map;
105 * struct sock_common - minimal network layer representation of sockets
106 * @skc_family: network address family
107 * @skc_state: Connection state
108 * @skc_reuse: %SO_REUSEADDR setting
109 * @skc_bound_dev_if: bound device index if != 0
110 * @skc_node: main hash linkage for various protocol lookup tables
111 * @skc_nulls_node: main hash linkage for UDP/UDP-Lite protocol
112 * @skc_bind_node: bind hash linkage for various protocol lookup tables
113 * @skc_refcnt: reference count
114 * @skc_hash: hash value used with various protocol lookup tables
115 * @skc_prot: protocol handlers inside a network family
116 * @skc_net: reference to the network namespace of this socket
118 * This is the minimal network layer representation of sockets, the header
119 * for struct sock and struct inet_timewait_sock.
122 unsigned short skc_family;
123 volatile unsigned char skc_state;
124 unsigned char skc_reuse;
125 int skc_bound_dev_if;
127 struct hlist_node skc_node;
128 struct hlist_nulls_node skc_nulls_node;
130 struct hlist_node skc_bind_node;
132 unsigned int skc_hash;
133 struct proto *skc_prot;
140 * struct sock - network layer representation of sockets
141 * @__sk_common: shared layout with inet_timewait_sock
142 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
143 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
144 * @sk_lock: synchronizer
145 * @sk_rcvbuf: size of receive buffer in bytes
146 * @sk_sleep: sock wait queue
147 * @sk_dst_cache: destination cache
148 * @sk_dst_lock: destination cache lock
149 * @sk_policy: flow policy
150 * @sk_rmem_alloc: receive queue bytes committed
151 * @sk_receive_queue: incoming packets
152 * @sk_wmem_alloc: transmit queue bytes committed
153 * @sk_write_queue: Packet sending queue
154 * @sk_async_wait_queue: DMA copied packets
155 * @sk_omem_alloc: "o" is "option" or "other"
156 * @sk_wmem_queued: persistent queue size
157 * @sk_forward_alloc: space allocated forward
158 * @sk_allocation: allocation mode
159 * @sk_sndbuf: size of send buffer in bytes
160 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
161 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
162 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
163 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
164 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
165 * @sk_gso_max_size: Maximum GSO segment size to build
166 * @sk_lingertime: %SO_LINGER l_linger setting
167 * @sk_backlog: always used with the per-socket spinlock held
168 * @sk_callback_lock: used with the callbacks in the end of this struct
169 * @sk_error_queue: rarely used
170 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
171 * IPV6_ADDRFORM for instance)
172 * @sk_err: last error
173 * @sk_err_soft: errors that don't cause failure but are the cause of a
174 * persistent failure not just 'timed out'
175 * @sk_drops: raw/udp drops counter
176 * @sk_ack_backlog: current listen backlog
177 * @sk_max_ack_backlog: listen backlog set in listen()
178 * @sk_priority: %SO_PRIORITY setting
179 * @sk_type: socket type (%SOCK_STREAM, etc)
180 * @sk_protocol: which protocol this socket belongs in this network family
181 * @sk_peercred: %SO_PEERCRED setting
182 * @sk_rcvlowat: %SO_RCVLOWAT setting
183 * @sk_rcvtimeo: %SO_RCVTIMEO setting
184 * @sk_sndtimeo: %SO_SNDTIMEO setting
185 * @sk_filter: socket filtering instructions
186 * @sk_protinfo: private area, net family specific, when not using slab
187 * @sk_timer: sock cleanup timer
188 * @sk_stamp: time stamp of last packet received
189 * @sk_socket: Identd and reporting IO signals
190 * @sk_user_data: RPC layer private data
191 * @sk_sndmsg_page: cached page for sendmsg
192 * @sk_sndmsg_off: cached offset for sendmsg
193 * @sk_send_head: front of stuff to transmit
194 * @sk_security: used by security modules
195 * @sk_mark: generic packet mark
196 * @sk_write_pending: a write to stream socket waits to start
197 * @sk_state_change: callback to indicate change in the state of the sock
198 * @sk_data_ready: callback to indicate there is data to be processed
199 * @sk_write_space: callback to indicate there is bf sending space available
200 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
201 * @sk_backlog_rcv: callback to process the backlog
202 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
206 * Now struct inet_timewait_sock also uses sock_common, so please just
207 * don't add nothing before this first member (__sk_common) --acme
209 struct sock_common __sk_common;
210 #define sk_family __sk_common.skc_family
211 #define sk_state __sk_common.skc_state
212 #define sk_reuse __sk_common.skc_reuse
213 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
214 #define sk_node __sk_common.skc_node
215 #define sk_nulls_node __sk_common.skc_nulls_node
216 #define sk_bind_node __sk_common.skc_bind_node
217 #define sk_refcnt __sk_common.skc_refcnt
218 #define sk_hash __sk_common.skc_hash
219 #define sk_prot __sk_common.skc_prot
220 #define sk_net __sk_common.skc_net
221 unsigned char sk_shutdown : 2,
224 unsigned char sk_protocol;
225 unsigned short sk_type;
227 socket_lock_t sk_lock;
229 * The backlog queue is special, it is always used with
230 * the per-socket spinlock held and requires low latency
231 * access. Therefore we special case it's implementation.
234 struct sk_buff *head;
235 struct sk_buff *tail;
237 wait_queue_head_t *sk_sleep;
238 struct dst_entry *sk_dst_cache;
240 struct xfrm_policy *sk_policy[2];
242 rwlock_t sk_dst_lock;
243 atomic_t sk_rmem_alloc;
244 atomic_t sk_wmem_alloc;
245 atomic_t sk_omem_alloc;
247 struct sk_buff_head sk_receive_queue;
248 struct sk_buff_head sk_write_queue;
249 #ifdef CONFIG_NET_DMA
250 struct sk_buff_head sk_async_wait_queue;
253 int sk_forward_alloc;
257 unsigned int sk_gso_max_size;
259 unsigned long sk_flags;
260 unsigned long sk_lingertime;
261 struct sk_buff_head sk_error_queue;
262 struct proto *sk_prot_creator;
263 rwlock_t sk_callback_lock;
267 unsigned short sk_ack_backlog;
268 unsigned short sk_max_ack_backlog;
270 struct ucred sk_peercred;
273 struct sk_filter *sk_filter;
275 struct timer_list sk_timer;
277 struct socket *sk_socket;
279 struct page *sk_sndmsg_page;
280 struct sk_buff *sk_send_head;
282 int sk_write_pending;
283 #ifdef CONFIG_SECURITY
287 /* XXX 4 bytes hole on 64 bit */
288 void (*sk_state_change)(struct sock *sk);
289 void (*sk_data_ready)(struct sock *sk, int bytes);
290 void (*sk_write_space)(struct sock *sk);
291 void (*sk_error_report)(struct sock *sk);
292 int (*sk_backlog_rcv)(struct sock *sk,
293 struct sk_buff *skb);
294 void (*sk_destruct)(struct sock *sk);
298 * Hashed lists helper routines
300 static inline struct sock *__sk_head(const struct hlist_head *head)
302 return hlist_entry(head->first, struct sock, sk_node);
305 static inline struct sock *sk_head(const struct hlist_head *head)
307 return hlist_empty(head) ? NULL : __sk_head(head);
310 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
312 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
315 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
317 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
320 static inline struct sock *sk_next(const struct sock *sk)
322 return sk->sk_node.next ?
323 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
326 static inline struct sock *sk_nulls_next(const struct sock *sk)
328 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
329 hlist_nulls_entry(sk->sk_nulls_node.next,
330 struct sock, sk_nulls_node) :
334 static inline int sk_unhashed(const struct sock *sk)
336 return hlist_unhashed(&sk->sk_node);
339 static inline int sk_hashed(const struct sock *sk)
341 return !sk_unhashed(sk);
344 static __inline__ void sk_node_init(struct hlist_node *node)
349 static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
354 static __inline__ void __sk_del_node(struct sock *sk)
356 __hlist_del(&sk->sk_node);
359 static __inline__ int __sk_del_node_init(struct sock *sk)
363 sk_node_init(&sk->sk_node);
369 /* Grab socket reference count. This operation is valid only
370 when sk is ALREADY grabbed f.e. it is found in hash table
371 or a list and the lookup is made under lock preventing hash table
375 static inline void sock_hold(struct sock *sk)
377 atomic_inc(&sk->sk_refcnt);
380 /* Ungrab socket in the context, which assumes that socket refcnt
381 cannot hit zero, f.e. it is true in context of any socketcall.
383 static inline void __sock_put(struct sock *sk)
385 atomic_dec(&sk->sk_refcnt);
388 static __inline__ int sk_del_node_init(struct sock *sk)
390 int rc = __sk_del_node_init(sk);
393 /* paranoid for a while -acme */
394 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
400 static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
403 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
409 static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
411 int rc = __sk_nulls_del_node_init_rcu(sk);
414 /* paranoid for a while -acme */
415 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
421 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
423 hlist_add_head(&sk->sk_node, list);
426 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
429 __sk_add_node(sk, list);
432 static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
434 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
437 static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
440 __sk_nulls_add_node_rcu(sk, list);
443 static __inline__ void __sk_del_bind_node(struct sock *sk)
445 __hlist_del(&sk->sk_bind_node);
448 static __inline__ void sk_add_bind_node(struct sock *sk,
449 struct hlist_head *list)
451 hlist_add_head(&sk->sk_bind_node, list);
454 #define sk_for_each(__sk, node, list) \
455 hlist_for_each_entry(__sk, node, list, sk_node)
456 #define sk_nulls_for_each(__sk, node, list) \
457 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
458 #define sk_nulls_for_each_rcu(__sk, node, list) \
459 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
460 #define sk_for_each_from(__sk, node) \
461 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
462 hlist_for_each_entry_from(__sk, node, sk_node)
463 #define sk_nulls_for_each_from(__sk, node) \
464 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
465 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
466 #define sk_for_each_continue(__sk, node) \
467 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
468 hlist_for_each_entry_continue(__sk, node, sk_node)
469 #define sk_for_each_safe(__sk, node, tmp, list) \
470 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
471 #define sk_for_each_bound(__sk, node, list) \
472 hlist_for_each_entry(__sk, node, list, sk_bind_node)
485 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
486 SOCK_DBG, /* %SO_DEBUG setting */
487 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
488 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
489 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
490 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
491 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
492 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
493 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
494 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
495 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
496 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
497 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
500 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
502 nsk->sk_flags = osk->sk_flags;
505 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
507 __set_bit(flag, &sk->sk_flags);
510 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
512 __clear_bit(flag, &sk->sk_flags);
515 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
517 return test_bit(flag, &sk->sk_flags);
520 static inline void sk_acceptq_removed(struct sock *sk)
522 sk->sk_ack_backlog--;
525 static inline void sk_acceptq_added(struct sock *sk)
527 sk->sk_ack_backlog++;
530 static inline int sk_acceptq_is_full(struct sock *sk)
532 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
536 * Compute minimal free write space needed to queue new packets.
538 static inline int sk_stream_min_wspace(struct sock *sk)
540 return sk->sk_wmem_queued >> 1;
543 static inline int sk_stream_wspace(struct sock *sk)
545 return sk->sk_sndbuf - sk->sk_wmem_queued;
548 extern void sk_stream_write_space(struct sock *sk);
550 static inline int sk_stream_memory_free(struct sock *sk)
552 return sk->sk_wmem_queued < sk->sk_sndbuf;
555 /* The per-socket spinlock must be held here. */
556 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
558 if (!sk->sk_backlog.tail) {
559 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
561 sk->sk_backlog.tail->next = skb;
562 sk->sk_backlog.tail = skb;
567 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
569 return sk->sk_backlog_rcv(sk, skb);
572 #define sk_wait_event(__sk, __timeo, __condition) \
574 release_sock(__sk); \
575 __rc = __condition; \
577 *(__timeo) = schedule_timeout(*(__timeo)); \
580 __rc = __condition; \
584 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
585 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
586 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
587 extern int sk_stream_error(struct sock *sk, int flags, int err);
588 extern void sk_stream_kill_queues(struct sock *sk);
590 extern int sk_wait_data(struct sock *sk, long *timeo);
592 struct request_sock_ops;
593 struct timewait_sock_ops;
594 struct inet_hashinfo;
597 /* Networking protocol blocks we attach to sockets.
598 * socket layer -> transport layer interface
599 * transport -> network interface is defined by struct inet_proto
602 void (*close)(struct sock *sk,
604 int (*connect)(struct sock *sk,
605 struct sockaddr *uaddr,
607 int (*disconnect)(struct sock *sk, int flags);
609 struct sock * (*accept) (struct sock *sk, int flags, int *err);
611 int (*ioctl)(struct sock *sk, int cmd,
613 int (*init)(struct sock *sk);
614 void (*destroy)(struct sock *sk);
615 void (*shutdown)(struct sock *sk, int how);
616 int (*setsockopt)(struct sock *sk, int level,
617 int optname, char __user *optval,
619 int (*getsockopt)(struct sock *sk, int level,
620 int optname, char __user *optval,
623 int (*compat_setsockopt)(struct sock *sk,
625 int optname, char __user *optval,
627 int (*compat_getsockopt)(struct sock *sk,
629 int optname, char __user *optval,
632 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
633 struct msghdr *msg, size_t len);
634 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
636 size_t len, int noblock, int flags,
638 int (*sendpage)(struct sock *sk, struct page *page,
639 int offset, size_t size, int flags);
640 int (*bind)(struct sock *sk,
641 struct sockaddr *uaddr, int addr_len);
643 int (*backlog_rcv) (struct sock *sk,
644 struct sk_buff *skb);
646 /* Keeping track of sk's, looking them up, and port selection methods. */
647 void (*hash)(struct sock *sk);
648 void (*unhash)(struct sock *sk);
649 int (*get_port)(struct sock *sk, unsigned short snum);
651 /* Keeping track of sockets in use */
652 #ifdef CONFIG_PROC_FS
653 unsigned int inuse_idx;
656 /* Memory pressure */
657 void (*enter_memory_pressure)(struct sock *sk);
658 atomic_t *memory_allocated; /* Current allocated memory. */
659 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
661 * Pressure flag: try to collapse.
662 * Technical note: it is used by multiple contexts non atomically.
663 * All the __sk_mem_schedule() is of this nature: accounting
664 * is strict, actions are advisory and have some latency.
666 int *memory_pressure;
672 struct kmem_cache *slab;
673 unsigned int obj_size;
676 struct percpu_counter *orphan_count;
678 struct request_sock_ops *rsk_prot;
679 struct timewait_sock_ops *twsk_prot;
682 struct inet_hashinfo *hashinfo;
683 struct udp_table *udp_table;
684 struct raw_hashinfo *raw_hash;
687 struct module *owner;
691 struct list_head node;
692 #ifdef SOCK_REFCNT_DEBUG
697 extern int proto_register(struct proto *prot, int alloc_slab);
698 extern void proto_unregister(struct proto *prot);
700 #ifdef SOCK_REFCNT_DEBUG
701 static inline void sk_refcnt_debug_inc(struct sock *sk)
703 atomic_inc(&sk->sk_prot->socks);
706 static inline void sk_refcnt_debug_dec(struct sock *sk)
708 atomic_dec(&sk->sk_prot->socks);
709 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
710 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
713 static inline void sk_refcnt_debug_release(const struct sock *sk)
715 if (atomic_read(&sk->sk_refcnt) != 1)
716 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
717 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
719 #else /* SOCK_REFCNT_DEBUG */
720 #define sk_refcnt_debug_inc(sk) do { } while (0)
721 #define sk_refcnt_debug_dec(sk) do { } while (0)
722 #define sk_refcnt_debug_release(sk) do { } while (0)
723 #endif /* SOCK_REFCNT_DEBUG */
726 #ifdef CONFIG_PROC_FS
727 /* Called with local bh disabled */
728 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
729 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
731 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
738 /* With per-bucket locks this operation is not-atomic, so that
739 * this version is not worse.
741 static inline void __sk_prot_rehash(struct sock *sk)
743 sk->sk_prot->unhash(sk);
744 sk->sk_prot->hash(sk);
747 /* About 10 seconds */
748 #define SOCK_DESTROY_TIME (10*HZ)
750 /* Sockets 0-1023 can't be bound to unless you are superuser */
751 #define PROT_SOCK 1024
753 #define SHUTDOWN_MASK 3
754 #define RCV_SHUTDOWN 1
755 #define SEND_SHUTDOWN 2
757 #define SOCK_SNDBUF_LOCK 1
758 #define SOCK_RCVBUF_LOCK 2
759 #define SOCK_BINDADDR_LOCK 4
760 #define SOCK_BINDPORT_LOCK 8
762 /* sock_iocb: used to kick off async processing of socket ios */
764 struct list_head list;
770 struct scm_cookie *scm;
771 struct msghdr *msg, async_msg;
775 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
777 return (struct sock_iocb *)iocb->private;
780 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
785 struct socket_alloc {
786 struct socket socket;
787 struct inode vfs_inode;
790 static inline struct socket *SOCKET_I(struct inode *inode)
792 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
795 static inline struct inode *SOCK_INODE(struct socket *socket)
797 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
801 * Functions for memory accounting
803 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
804 extern void __sk_mem_reclaim(struct sock *sk);
806 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
807 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
808 #define SK_MEM_SEND 0
809 #define SK_MEM_RECV 1
811 static inline int sk_mem_pages(int amt)
813 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
816 static inline int sk_has_account(struct sock *sk)
818 /* return true if protocol supports memory accounting */
819 return !!sk->sk_prot->memory_allocated;
822 static inline int sk_wmem_schedule(struct sock *sk, int size)
824 if (!sk_has_account(sk))
826 return size <= sk->sk_forward_alloc ||
827 __sk_mem_schedule(sk, size, SK_MEM_SEND);
830 static inline int sk_rmem_schedule(struct sock *sk, int size)
832 if (!sk_has_account(sk))
834 return size <= sk->sk_forward_alloc ||
835 __sk_mem_schedule(sk, size, SK_MEM_RECV);
838 static inline void sk_mem_reclaim(struct sock *sk)
840 if (!sk_has_account(sk))
842 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
843 __sk_mem_reclaim(sk);
846 static inline void sk_mem_reclaim_partial(struct sock *sk)
848 if (!sk_has_account(sk))
850 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
851 __sk_mem_reclaim(sk);
854 static inline void sk_mem_charge(struct sock *sk, int size)
856 if (!sk_has_account(sk))
858 sk->sk_forward_alloc -= size;
861 static inline void sk_mem_uncharge(struct sock *sk, int size)
863 if (!sk_has_account(sk))
865 sk->sk_forward_alloc += size;
868 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
870 skb_truesize_check(skb);
871 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
872 sk->sk_wmem_queued -= skb->truesize;
873 sk_mem_uncharge(sk, skb->truesize);
877 /* Used by processes to "lock" a socket state, so that
878 * interrupts and bottom half handlers won't change it
879 * from under us. It essentially blocks any incoming
880 * packets, so that we won't get any new data or any
881 * packets that change the state of the socket.
883 * While locked, BH processing will add new packets to
884 * the backlog queue. This queue is processed by the
885 * owner of the socket lock right before it is released.
887 * Since ~2.3.5 it is also exclusive sleep lock serializing
888 * accesses from user process context.
890 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
893 * Macro so as to not evaluate some arguments when
894 * lockdep is not enabled.
896 * Mark both the sk_lock and the sk_lock.slock as a
897 * per-address-family lock class.
899 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
901 sk->sk_lock.owned = 0; \
902 init_waitqueue_head(&sk->sk_lock.wq); \
903 spin_lock_init(&(sk)->sk_lock.slock); \
904 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
905 sizeof((sk)->sk_lock)); \
906 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
908 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
911 extern void lock_sock_nested(struct sock *sk, int subclass);
913 static inline void lock_sock(struct sock *sk)
915 lock_sock_nested(sk, 0);
918 extern void release_sock(struct sock *sk);
920 /* BH context may only use the following locking interface. */
921 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
922 #define bh_lock_sock_nested(__sk) \
923 spin_lock_nested(&((__sk)->sk_lock.slock), \
924 SINGLE_DEPTH_NESTING)
925 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
927 extern struct sock *sk_alloc(struct net *net, int family,
930 extern void sk_free(struct sock *sk);
931 extern void sk_release_kernel(struct sock *sk);
932 extern struct sock *sk_clone(const struct sock *sk,
933 const gfp_t priority);
935 extern struct sk_buff *sock_wmalloc(struct sock *sk,
936 unsigned long size, int force,
938 extern struct sk_buff *sock_rmalloc(struct sock *sk,
939 unsigned long size, int force,
941 extern void sock_wfree(struct sk_buff *skb);
942 extern void sock_rfree(struct sk_buff *skb);
944 extern int sock_setsockopt(struct socket *sock, int level,
945 int op, char __user *optval,
948 extern int sock_getsockopt(struct socket *sock, int level,
949 int op, char __user *optval,
951 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
955 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
956 unsigned long header_len,
957 unsigned long data_len,
960 extern void *sock_kmalloc(struct sock *sk, int size,
962 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
963 extern void sk_send_sigurg(struct sock *sk);
966 * Functions to fill in entries in struct proto_ops when a protocol
967 * does not implement a particular function.
969 extern int sock_no_bind(struct socket *,
970 struct sockaddr *, int);
971 extern int sock_no_connect(struct socket *,
972 struct sockaddr *, int, int);
973 extern int sock_no_socketpair(struct socket *,
975 extern int sock_no_accept(struct socket *,
976 struct socket *, int);
977 extern int sock_no_getname(struct socket *,
978 struct sockaddr *, int *, int);
979 extern unsigned int sock_no_poll(struct file *, struct socket *,
980 struct poll_table_struct *);
981 extern int sock_no_ioctl(struct socket *, unsigned int,
983 extern int sock_no_listen(struct socket *, int);
984 extern int sock_no_shutdown(struct socket *, int);
985 extern int sock_no_getsockopt(struct socket *, int , int,
986 char __user *, int __user *);
987 extern int sock_no_setsockopt(struct socket *, int, int,
989 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
990 struct msghdr *, size_t);
991 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
992 struct msghdr *, size_t, int);
993 extern int sock_no_mmap(struct file *file,
995 struct vm_area_struct *vma);
996 extern ssize_t sock_no_sendpage(struct socket *sock,
998 int offset, size_t size,
1002 * Functions to fill in entries in struct proto_ops when a protocol
1003 * uses the inet style.
1005 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1006 char __user *optval, int __user *optlen);
1007 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1008 struct msghdr *msg, size_t size, int flags);
1009 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1010 char __user *optval, int optlen);
1011 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1012 int optname, char __user *optval, int __user *optlen);
1013 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1014 int optname, char __user *optval, int optlen);
1016 extern void sk_common_release(struct sock *sk);
1019 * Default socket callbacks and setup code
1022 /* Initialise core socket variables */
1023 extern void sock_init_data(struct socket *sock, struct sock *sk);
1026 * sk_filter_release: Release a socket filter
1027 * @fp: filter to remove
1029 * Remove a filter from a socket and release its resources.
1032 static inline void sk_filter_release(struct sk_filter *fp)
1034 if (atomic_dec_and_test(&fp->refcnt))
1038 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1040 unsigned int size = sk_filter_len(fp);
1042 atomic_sub(size, &sk->sk_omem_alloc);
1043 sk_filter_release(fp);
1046 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1048 atomic_inc(&fp->refcnt);
1049 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1053 * Socket reference counting postulates.
1055 * * Each user of socket SHOULD hold a reference count.
1056 * * Each access point to socket (an hash table bucket, reference from a list,
1057 * running timer, skb in flight MUST hold a reference count.
1058 * * When reference count hits 0, it means it will never increase back.
1059 * * When reference count hits 0, it means that no references from
1060 * outside exist to this socket and current process on current CPU
1061 * is last user and may/should destroy this socket.
1062 * * sk_free is called from any context: process, BH, IRQ. When
1063 * it is called, socket has no references from outside -> sk_free
1064 * may release descendant resources allocated by the socket, but
1065 * to the time when it is called, socket is NOT referenced by any
1066 * hash tables, lists etc.
1067 * * Packets, delivered from outside (from network or from another process)
1068 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1069 * when they sit in queue. Otherwise, packets will leak to hole, when
1070 * socket is looked up by one cpu and unhasing is made by another CPU.
1071 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1072 * (leak to backlog). Packet socket does all the processing inside
1073 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1074 * use separate SMP lock, so that they are prone too.
1077 /* Ungrab socket and destroy it, if it was the last reference. */
1078 static inline void sock_put(struct sock *sk)
1080 if (atomic_dec_and_test(&sk->sk_refcnt))
1084 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1087 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1089 sk->sk_socket = sock;
1092 /* Detach socket from process context.
1093 * Announce socket dead, detach it from wait queue and inode.
1094 * Note that parent inode held reference count on this struct sock,
1095 * we do not release it in this function, because protocol
1096 * probably wants some additional cleanups or even continuing
1097 * to work with this socket (TCP).
1099 static inline void sock_orphan(struct sock *sk)
1101 write_lock_bh(&sk->sk_callback_lock);
1102 sock_set_flag(sk, SOCK_DEAD);
1103 sk_set_socket(sk, NULL);
1104 sk->sk_sleep = NULL;
1105 write_unlock_bh(&sk->sk_callback_lock);
1108 static inline void sock_graft(struct sock *sk, struct socket *parent)
1110 write_lock_bh(&sk->sk_callback_lock);
1111 sk->sk_sleep = &parent->wait;
1113 sk_set_socket(sk, parent);
1114 security_sock_graft(sk, parent);
1115 write_unlock_bh(&sk->sk_callback_lock);
1118 extern int sock_i_uid(struct sock *sk);
1119 extern unsigned long sock_i_ino(struct sock *sk);
1121 static inline struct dst_entry *
1122 __sk_dst_get(struct sock *sk)
1124 return sk->sk_dst_cache;
1127 static inline struct dst_entry *
1128 sk_dst_get(struct sock *sk)
1130 struct dst_entry *dst;
1132 read_lock(&sk->sk_dst_lock);
1133 dst = sk->sk_dst_cache;
1136 read_unlock(&sk->sk_dst_lock);
1141 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1143 struct dst_entry *old_dst;
1145 old_dst = sk->sk_dst_cache;
1146 sk->sk_dst_cache = dst;
1147 dst_release(old_dst);
1151 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1153 write_lock(&sk->sk_dst_lock);
1154 __sk_dst_set(sk, dst);
1155 write_unlock(&sk->sk_dst_lock);
1159 __sk_dst_reset(struct sock *sk)
1161 struct dst_entry *old_dst;
1163 old_dst = sk->sk_dst_cache;
1164 sk->sk_dst_cache = NULL;
1165 dst_release(old_dst);
1169 sk_dst_reset(struct sock *sk)
1171 write_lock(&sk->sk_dst_lock);
1173 write_unlock(&sk->sk_dst_lock);
1176 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1178 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1180 static inline int sk_can_gso(const struct sock *sk)
1182 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1185 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1187 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1188 struct sk_buff *skb, struct page *page,
1191 if (skb->ip_summed == CHECKSUM_NONE) {
1193 __wsum csum = csum_and_copy_from_user(from,
1194 page_address(page) + off,
1198 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1199 } else if (copy_from_user(page_address(page) + off, from, copy))
1203 skb->data_len += copy;
1204 skb->truesize += copy;
1205 sk->sk_wmem_queued += copy;
1206 sk_mem_charge(sk, copy);
1211 * Queue a received datagram if it will fit. Stream and sequenced
1212 * protocols can't normally use this as they need to fit buffers in
1213 * and play with them.
1215 * Inlined as it's very short and called for pretty much every
1216 * packet ever received.
1219 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1223 skb->destructor = sock_wfree;
1224 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1227 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1230 skb->destructor = sock_rfree;
1231 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1232 sk_mem_charge(sk, skb->truesize);
1235 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1236 unsigned long expires);
1238 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1240 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1242 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1244 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1245 number of warnings when compiling with -W --ANK
1247 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1248 (unsigned)sk->sk_rcvbuf)
1250 skb_set_owner_r(skb, sk);
1251 skb_queue_tail(&sk->sk_error_queue, skb);
1252 if (!sock_flag(sk, SOCK_DEAD))
1253 sk->sk_data_ready(sk, skb->len);
1258 * Recover an error report and clear atomically
1261 static inline int sock_error(struct sock *sk)
1264 if (likely(!sk->sk_err))
1266 err = xchg(&sk->sk_err, 0);
1270 static inline unsigned long sock_wspace(struct sock *sk)
1274 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1275 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1282 static inline void sk_wake_async(struct sock *sk, int how, int band)
1284 if (sk->sk_socket && sk->sk_socket->fasync_list)
1285 sock_wake_async(sk->sk_socket, how, band);
1288 #define SOCK_MIN_SNDBUF 2048
1289 #define SOCK_MIN_RCVBUF 256
1291 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1293 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1294 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1295 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1299 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1301 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1303 struct page *page = NULL;
1305 page = alloc_pages(sk->sk_allocation, 0);
1307 sk->sk_prot->enter_memory_pressure(sk);
1308 sk_stream_moderate_sndbuf(sk);
1314 * Default write policy as shown to user space via poll/select/SIGIO
1316 static inline int sock_writeable(const struct sock *sk)
1318 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1321 static inline gfp_t gfp_any(void)
1323 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1326 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1328 return noblock ? 0 : sk->sk_rcvtimeo;
1331 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1333 return noblock ? 0 : sk->sk_sndtimeo;
1336 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1338 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1341 /* Alas, with timeout socket operations are not restartable.
1342 * Compare this to poll().
1344 static inline int sock_intr_errno(long timeo)
1346 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1349 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1350 struct sk_buff *skb);
1352 static __inline__ void
1353 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1355 ktime_t kt = skb->tstamp;
1356 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
1359 * generate control messages if
1360 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1361 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1362 * - software time stamp available and wanted
1363 * (SOCK_TIMESTAMPING_SOFTWARE)
1364 * - hardware time stamps available and wanted
1365 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1366 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1368 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1369 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1370 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1371 (hwtstamps->hwtstamp.tv64 &&
1372 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1373 (hwtstamps->syststamp.tv64 &&
1374 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
1375 __sock_recv_timestamp(msg, sk, skb);
1381 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1382 * @msg: outgoing packet
1383 * @sk: socket sending this packet
1384 * @shtx: filled with instructions for time stamping
1386 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1387 * parameters are invalid.
1389 extern int sock_tx_timestamp(struct msghdr *msg,
1391 union skb_shared_tx *shtx);
1395 * sk_eat_skb - Release a skb if it is no longer needed
1396 * @sk: socket to eat this skb from
1397 * @skb: socket buffer to eat
1398 * @copied_early: flag indicating whether DMA operations copied this data early
1400 * This routine must be called with interrupts disabled or with the socket
1401 * locked so that the sk_buff queue operation is ok.
1403 #ifdef CONFIG_NET_DMA
1404 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1406 __skb_unlink(skb, &sk->sk_receive_queue);
1410 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1413 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1415 __skb_unlink(skb, &sk->sk_receive_queue);
1421 struct net *sock_net(const struct sock *sk)
1423 #ifdef CONFIG_NET_NS
1431 void sock_net_set(struct sock *sk, struct net *net)
1433 #ifdef CONFIG_NET_NS
1439 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1440 * They should not hold a referrence to a namespace in order to allow
1442 * Sockets after sk_change_net should be released using sk_release_kernel
1444 static inline void sk_change_net(struct sock *sk, struct net *net)
1446 put_net(sock_net(sk));
1447 sock_net_set(sk, hold_net(net));
1450 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
1452 if (unlikely(skb->sk)) {
1453 struct sock *sk = skb->sk;
1455 skb->destructor = NULL;
1462 extern void sock_enable_timestamp(struct sock *sk, int flag);
1463 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1464 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1467 * Enable debug/info messages
1469 extern int net_msg_warn;
1470 #define NETDEBUG(fmt, args...) \
1471 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1473 #define LIMIT_NETDEBUG(fmt, args...) \
1474 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1476 extern __u32 sysctl_wmem_max;
1477 extern __u32 sysctl_rmem_max;
1479 extern void sk_init(void);
1481 extern int sysctl_optmem_max;
1483 extern __u32 sysctl_wmem_default;
1484 extern __u32 sysctl_rmem_default;
1486 #endif /* _SOCK_H */