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/pcounter.h>
51 #include <linux/skbuff.h> /* struct sk_buff */
53 #include <linux/security.h>
55 #include <linux/filter.h>
57 #include <asm/atomic.h>
59 #include <net/checksum.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;
100 * struct sock_common - minimal network layer representation of sockets
101 * @skc_family: network address family
102 * @skc_state: Connection state
103 * @skc_reuse: %SO_REUSEADDR setting
104 * @skc_bound_dev_if: bound device index if != 0
105 * @skc_node: main hash linkage for various protocol lookup tables
106 * @skc_bind_node: bind hash linkage for various protocol lookup tables
107 * @skc_refcnt: reference count
108 * @skc_hash: hash value used with various protocol lookup tables
109 * @skc_prot: protocol handlers inside a network family
110 * @skc_net: reference to the network namespace of this socket
112 * This is the minimal network layer representation of sockets, the header
113 * for struct sock and struct inet_timewait_sock.
116 unsigned short skc_family;
117 volatile unsigned char skc_state;
118 unsigned char skc_reuse;
119 int skc_bound_dev_if;
120 struct hlist_node skc_node;
121 struct hlist_node skc_bind_node;
123 unsigned int skc_hash;
124 struct proto *skc_prot;
129 * struct sock - network layer representation of sockets
130 * @__sk_common: shared layout with inet_timewait_sock
131 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
132 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
133 * @sk_lock: synchronizer
134 * @sk_rcvbuf: size of receive buffer in bytes
135 * @sk_sleep: sock wait queue
136 * @sk_dst_cache: destination cache
137 * @sk_dst_lock: destination cache lock
138 * @sk_policy: flow policy
139 * @sk_rmem_alloc: receive queue bytes committed
140 * @sk_receive_queue: incoming packets
141 * @sk_wmem_alloc: transmit queue bytes committed
142 * @sk_write_queue: Packet sending queue
143 * @sk_async_wait_queue: DMA copied packets
144 * @sk_omem_alloc: "o" is "option" or "other"
145 * @sk_wmem_queued: persistent queue size
146 * @sk_forward_alloc: space allocated forward
147 * @sk_allocation: allocation mode
148 * @sk_sndbuf: size of send buffer in bytes
149 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
150 * %SO_OOBINLINE settings
151 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
152 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
153 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
154 * @sk_lingertime: %SO_LINGER l_linger setting
155 * @sk_backlog: always used with the per-socket spinlock held
156 * @sk_callback_lock: used with the callbacks in the end of this struct
157 * @sk_error_queue: rarely used
158 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
159 * IPV6_ADDRFORM for instance)
160 * @sk_err: last error
161 * @sk_err_soft: errors that don't cause failure but are the cause of a
162 * persistent failure not just 'timed out'
163 * @sk_drops: raw drops counter
164 * @sk_ack_backlog: current listen backlog
165 * @sk_max_ack_backlog: listen backlog set in listen()
166 * @sk_priority: %SO_PRIORITY setting
167 * @sk_type: socket type (%SOCK_STREAM, etc)
168 * @sk_protocol: which protocol this socket belongs in this network family
169 * @sk_peercred: %SO_PEERCRED setting
170 * @sk_rcvlowat: %SO_RCVLOWAT setting
171 * @sk_rcvtimeo: %SO_RCVTIMEO setting
172 * @sk_sndtimeo: %SO_SNDTIMEO setting
173 * @sk_filter: socket filtering instructions
174 * @sk_protinfo: private area, net family specific, when not using slab
175 * @sk_timer: sock cleanup timer
176 * @sk_stamp: time stamp of last packet received
177 * @sk_socket: Identd and reporting IO signals
178 * @sk_user_data: RPC layer private data
179 * @sk_sndmsg_page: cached page for sendmsg
180 * @sk_sndmsg_off: cached offset for sendmsg
181 * @sk_send_head: front of stuff to transmit
182 * @sk_security: used by security modules
183 * @sk_mark: generic packet mark
184 * @sk_write_pending: a write to stream socket waits to start
185 * @sk_state_change: callback to indicate change in the state of the sock
186 * @sk_data_ready: callback to indicate there is data to be processed
187 * @sk_write_space: callback to indicate there is bf sending space available
188 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
189 * @sk_backlog_rcv: callback to process the backlog
190 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
194 * Now struct inet_timewait_sock also uses sock_common, so please just
195 * don't add nothing before this first member (__sk_common) --acme
197 struct sock_common __sk_common;
198 #define sk_family __sk_common.skc_family
199 #define sk_state __sk_common.skc_state
200 #define sk_reuse __sk_common.skc_reuse
201 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
202 #define sk_node __sk_common.skc_node
203 #define sk_bind_node __sk_common.skc_bind_node
204 #define sk_refcnt __sk_common.skc_refcnt
205 #define sk_hash __sk_common.skc_hash
206 #define sk_prot __sk_common.skc_prot
207 #define sk_net __sk_common.skc_net
208 unsigned char sk_shutdown : 2,
211 unsigned char sk_protocol;
212 unsigned short sk_type;
214 socket_lock_t sk_lock;
216 * The backlog queue is special, it is always used with
217 * the per-socket spinlock held and requires low latency
218 * access. Therefore we special case it's implementation.
221 struct sk_buff *head;
222 struct sk_buff *tail;
224 wait_queue_head_t *sk_sleep;
225 struct dst_entry *sk_dst_cache;
226 struct xfrm_policy *sk_policy[2];
227 rwlock_t sk_dst_lock;
228 atomic_t sk_rmem_alloc;
229 atomic_t sk_wmem_alloc;
230 atomic_t sk_omem_alloc;
232 struct sk_buff_head sk_receive_queue;
233 struct sk_buff_head sk_write_queue;
234 struct sk_buff_head sk_async_wait_queue;
236 int sk_forward_alloc;
241 unsigned long sk_flags;
242 unsigned long sk_lingertime;
243 struct sk_buff_head sk_error_queue;
244 struct proto *sk_prot_creator;
245 rwlock_t sk_callback_lock;
249 unsigned short sk_ack_backlog;
250 unsigned short sk_max_ack_backlog;
252 struct ucred sk_peercred;
255 struct sk_filter *sk_filter;
257 struct timer_list sk_timer;
259 struct socket *sk_socket;
261 struct page *sk_sndmsg_page;
262 struct sk_buff *sk_send_head;
264 int sk_write_pending;
267 /* XXX 4 bytes hole on 64 bit */
268 void (*sk_state_change)(struct sock *sk);
269 void (*sk_data_ready)(struct sock *sk, int bytes);
270 void (*sk_write_space)(struct sock *sk);
271 void (*sk_error_report)(struct sock *sk);
272 int (*sk_backlog_rcv)(struct sock *sk,
273 struct sk_buff *skb);
274 void (*sk_destruct)(struct sock *sk);
278 * Hashed lists helper routines
280 static inline struct sock *__sk_head(const struct hlist_head *head)
282 return hlist_entry(head->first, struct sock, sk_node);
285 static inline struct sock *sk_head(const struct hlist_head *head)
287 return hlist_empty(head) ? NULL : __sk_head(head);
290 static inline struct sock *sk_next(const struct sock *sk)
292 return sk->sk_node.next ?
293 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
296 static inline int sk_unhashed(const struct sock *sk)
298 return hlist_unhashed(&sk->sk_node);
301 static inline int sk_hashed(const struct sock *sk)
303 return !sk_unhashed(sk);
306 static __inline__ void sk_node_init(struct hlist_node *node)
311 static __inline__ void __sk_del_node(struct sock *sk)
313 __hlist_del(&sk->sk_node);
316 static __inline__ int __sk_del_node_init(struct sock *sk)
320 sk_node_init(&sk->sk_node);
326 /* Grab socket reference count. This operation is valid only
327 when sk is ALREADY grabbed f.e. it is found in hash table
328 or a list and the lookup is made under lock preventing hash table
332 static inline void sock_hold(struct sock *sk)
334 atomic_inc(&sk->sk_refcnt);
337 /* Ungrab socket in the context, which assumes that socket refcnt
338 cannot hit zero, f.e. it is true in context of any socketcall.
340 static inline void __sock_put(struct sock *sk)
342 atomic_dec(&sk->sk_refcnt);
345 static __inline__ int sk_del_node_init(struct sock *sk)
347 int rc = __sk_del_node_init(sk);
350 /* paranoid for a while -acme */
351 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
357 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
359 hlist_add_head(&sk->sk_node, list);
362 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
365 __sk_add_node(sk, list);
368 static __inline__ void __sk_del_bind_node(struct sock *sk)
370 __hlist_del(&sk->sk_bind_node);
373 static __inline__ void sk_add_bind_node(struct sock *sk,
374 struct hlist_head *list)
376 hlist_add_head(&sk->sk_bind_node, list);
379 #define sk_for_each(__sk, node, list) \
380 hlist_for_each_entry(__sk, node, list, sk_node)
381 #define sk_for_each_from(__sk, node) \
382 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
383 hlist_for_each_entry_from(__sk, node, sk_node)
384 #define sk_for_each_continue(__sk, node) \
385 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
386 hlist_for_each_entry_continue(__sk, node, sk_node)
387 #define sk_for_each_safe(__sk, node, tmp, list) \
388 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
389 #define sk_for_each_bound(__sk, node, list) \
390 hlist_for_each_entry(__sk, node, list, sk_bind_node)
403 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
404 SOCK_DBG, /* %SO_DEBUG setting */
405 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
406 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
407 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
408 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
411 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
413 nsk->sk_flags = osk->sk_flags;
416 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
418 __set_bit(flag, &sk->sk_flags);
421 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
423 __clear_bit(flag, &sk->sk_flags);
426 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
428 return test_bit(flag, &sk->sk_flags);
431 static inline void sk_acceptq_removed(struct sock *sk)
433 sk->sk_ack_backlog--;
436 static inline void sk_acceptq_added(struct sock *sk)
438 sk->sk_ack_backlog++;
441 static inline int sk_acceptq_is_full(struct sock *sk)
443 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
447 * Compute minimal free write space needed to queue new packets.
449 static inline int sk_stream_min_wspace(struct sock *sk)
451 return sk->sk_wmem_queued >> 1;
454 static inline int sk_stream_wspace(struct sock *sk)
456 return sk->sk_sndbuf - sk->sk_wmem_queued;
459 extern void sk_stream_write_space(struct sock *sk);
461 static inline int sk_stream_memory_free(struct sock *sk)
463 return sk->sk_wmem_queued < sk->sk_sndbuf;
466 /* The per-socket spinlock must be held here. */
467 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
469 if (!sk->sk_backlog.tail) {
470 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
472 sk->sk_backlog.tail->next = skb;
473 sk->sk_backlog.tail = skb;
478 #define sk_wait_event(__sk, __timeo, __condition) \
480 release_sock(__sk); \
481 __rc = __condition; \
483 *(__timeo) = schedule_timeout(*(__timeo)); \
486 __rc = __condition; \
490 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
491 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
492 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
493 extern int sk_stream_error(struct sock *sk, int flags, int err);
494 extern void sk_stream_kill_queues(struct sock *sk);
496 extern int sk_wait_data(struct sock *sk, long *timeo);
498 struct request_sock_ops;
499 struct timewait_sock_ops;
500 struct inet_hashinfo;
502 /* Networking protocol blocks we attach to sockets.
503 * socket layer -> transport layer interface
504 * transport -> network interface is defined by struct inet_proto
507 void (*close)(struct sock *sk,
509 int (*connect)(struct sock *sk,
510 struct sockaddr *uaddr,
512 int (*disconnect)(struct sock *sk, int flags);
514 struct sock * (*accept) (struct sock *sk, int flags, int *err);
516 int (*ioctl)(struct sock *sk, int cmd,
518 int (*init)(struct sock *sk);
519 int (*destroy)(struct sock *sk);
520 void (*shutdown)(struct sock *sk, int how);
521 int (*setsockopt)(struct sock *sk, int level,
522 int optname, char __user *optval,
524 int (*getsockopt)(struct sock *sk, int level,
525 int optname, char __user *optval,
527 int (*compat_setsockopt)(struct sock *sk,
529 int optname, char __user *optval,
531 int (*compat_getsockopt)(struct sock *sk,
533 int optname, char __user *optval,
535 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
536 struct msghdr *msg, size_t len);
537 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
539 size_t len, int noblock, int flags,
541 int (*sendpage)(struct sock *sk, struct page *page,
542 int offset, size_t size, int flags);
543 int (*bind)(struct sock *sk,
544 struct sockaddr *uaddr, int addr_len);
546 int (*backlog_rcv) (struct sock *sk,
547 struct sk_buff *skb);
549 /* Keeping track of sk's, looking them up, and port selection methods. */
550 void (*hash)(struct sock *sk);
551 void (*unhash)(struct sock *sk);
552 int (*get_port)(struct sock *sk, unsigned short snum);
554 /* Keeping track of sockets in use */
555 #ifdef CONFIG_PROC_FS
556 struct pcounter inuse;
559 /* Memory pressure */
560 void (*enter_memory_pressure)(void);
561 atomic_t *memory_allocated; /* Current allocated memory. */
562 atomic_t *sockets_allocated; /* Current number of sockets. */
564 * Pressure flag: try to collapse.
565 * Technical note: it is used by multiple contexts non atomically.
566 * All the __sk_mem_schedule() is of this nature: accounting
567 * is strict, actions are advisory and have some latency.
569 int *memory_pressure;
575 struct kmem_cache *slab;
576 unsigned int obj_size;
578 atomic_t *orphan_count;
580 struct request_sock_ops *rsk_prot;
581 struct timewait_sock_ops *twsk_prot;
583 struct inet_hashinfo *hashinfo;
585 struct module *owner;
589 struct list_head node;
590 #ifdef SOCK_REFCNT_DEBUG
595 extern int proto_register(struct proto *prot, int alloc_slab);
596 extern void proto_unregister(struct proto *prot);
598 #ifdef SOCK_REFCNT_DEBUG
599 static inline void sk_refcnt_debug_inc(struct sock *sk)
601 atomic_inc(&sk->sk_prot->socks);
604 static inline void sk_refcnt_debug_dec(struct sock *sk)
606 atomic_dec(&sk->sk_prot->socks);
607 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
608 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
611 static inline void sk_refcnt_debug_release(const struct sock *sk)
613 if (atomic_read(&sk->sk_refcnt) != 1)
614 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
615 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
617 #else /* SOCK_REFCNT_DEBUG */
618 #define sk_refcnt_debug_inc(sk) do { } while (0)
619 #define sk_refcnt_debug_dec(sk) do { } while (0)
620 #define sk_refcnt_debug_release(sk) do { } while (0)
621 #endif /* SOCK_REFCNT_DEBUG */
624 #ifdef CONFIG_PROC_FS
625 # define DEFINE_PROTO_INUSE(NAME) DEFINE_PCOUNTER(NAME)
626 # define REF_PROTO_INUSE(NAME) PCOUNTER_MEMBER_INITIALIZER(NAME, .inuse)
627 /* Called with local bh disabled */
628 static inline void sock_prot_inuse_add(struct proto *prot, int inc)
630 pcounter_add(&prot->inuse, inc);
632 static inline int sock_prot_inuse_init(struct proto *proto)
634 return pcounter_alloc(&proto->inuse);
636 static inline int sock_prot_inuse_get(struct proto *proto)
638 return pcounter_getval(&proto->inuse);
640 static inline void sock_prot_inuse_free(struct proto *proto)
642 pcounter_free(&proto->inuse);
645 # define DEFINE_PROTO_INUSE(NAME)
646 # define REF_PROTO_INUSE(NAME)
647 static void inline sock_prot_inuse_add(struct proto *prot, int inc)
650 static int inline sock_prot_inuse_init(struct proto *proto)
654 static void inline sock_prot_inuse_free(struct proto *proto)
660 /* With per-bucket locks this operation is not-atomic, so that
661 * this version is not worse.
663 static inline void __sk_prot_rehash(struct sock *sk)
665 sk->sk_prot->unhash(sk);
666 sk->sk_prot->hash(sk);
669 /* About 10 seconds */
670 #define SOCK_DESTROY_TIME (10*HZ)
672 /* Sockets 0-1023 can't be bound to unless you are superuser */
673 #define PROT_SOCK 1024
675 #define SHUTDOWN_MASK 3
676 #define RCV_SHUTDOWN 1
677 #define SEND_SHUTDOWN 2
679 #define SOCK_SNDBUF_LOCK 1
680 #define SOCK_RCVBUF_LOCK 2
681 #define SOCK_BINDADDR_LOCK 4
682 #define SOCK_BINDPORT_LOCK 8
684 /* sock_iocb: used to kick off async processing of socket ios */
686 struct list_head list;
692 struct scm_cookie *scm;
693 struct msghdr *msg, async_msg;
697 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
699 return (struct sock_iocb *)iocb->private;
702 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
707 struct socket_alloc {
708 struct socket socket;
709 struct inode vfs_inode;
712 static inline struct socket *SOCKET_I(struct inode *inode)
714 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
717 static inline struct inode *SOCK_INODE(struct socket *socket)
719 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
723 * Functions for memory accounting
725 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
726 extern void __sk_mem_reclaim(struct sock *sk);
728 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
729 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
730 #define SK_MEM_SEND 0
731 #define SK_MEM_RECV 1
733 static inline int sk_mem_pages(int amt)
735 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
738 static inline int sk_has_account(struct sock *sk)
740 /* return true if protocol supports memory accounting */
741 return !!sk->sk_prot->memory_allocated;
744 static inline int sk_wmem_schedule(struct sock *sk, int size)
746 if (!sk_has_account(sk))
748 return size <= sk->sk_forward_alloc ||
749 __sk_mem_schedule(sk, size, SK_MEM_SEND);
752 static inline int sk_rmem_schedule(struct sock *sk, int size)
754 if (!sk_has_account(sk))
756 return size <= sk->sk_forward_alloc ||
757 __sk_mem_schedule(sk, size, SK_MEM_RECV);
760 static inline void sk_mem_reclaim(struct sock *sk)
762 if (!sk_has_account(sk))
764 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
765 __sk_mem_reclaim(sk);
768 static inline void sk_mem_reclaim_partial(struct sock *sk)
770 if (!sk_has_account(sk))
772 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
773 __sk_mem_reclaim(sk);
776 static inline void sk_mem_charge(struct sock *sk, int size)
778 if (!sk_has_account(sk))
780 sk->sk_forward_alloc -= size;
783 static inline void sk_mem_uncharge(struct sock *sk, int size)
785 if (!sk_has_account(sk))
787 sk->sk_forward_alloc += size;
790 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
792 skb_truesize_check(skb);
793 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
794 sk->sk_wmem_queued -= skb->truesize;
795 sk_mem_uncharge(sk, skb->truesize);
799 /* Used by processes to "lock" a socket state, so that
800 * interrupts and bottom half handlers won't change it
801 * from under us. It essentially blocks any incoming
802 * packets, so that we won't get any new data or any
803 * packets that change the state of the socket.
805 * While locked, BH processing will add new packets to
806 * the backlog queue. This queue is processed by the
807 * owner of the socket lock right before it is released.
809 * Since ~2.3.5 it is also exclusive sleep lock serializing
810 * accesses from user process context.
812 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
815 * Macro so as to not evaluate some arguments when
816 * lockdep is not enabled.
818 * Mark both the sk_lock and the sk_lock.slock as a
819 * per-address-family lock class.
821 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
823 sk->sk_lock.owned = 0; \
824 init_waitqueue_head(&sk->sk_lock.wq); \
825 spin_lock_init(&(sk)->sk_lock.slock); \
826 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
827 sizeof((sk)->sk_lock)); \
828 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
830 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
833 extern void lock_sock_nested(struct sock *sk, int subclass);
835 static inline void lock_sock(struct sock *sk)
837 lock_sock_nested(sk, 0);
840 extern void release_sock(struct sock *sk);
842 /* BH context may only use the following locking interface. */
843 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
844 #define bh_lock_sock_nested(__sk) \
845 spin_lock_nested(&((__sk)->sk_lock.slock), \
846 SINGLE_DEPTH_NESTING)
847 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
849 extern struct sock *sk_alloc(struct net *net, int family,
852 extern void sk_free(struct sock *sk);
853 extern struct sock *sk_clone(const struct sock *sk,
854 const gfp_t priority);
856 extern struct sk_buff *sock_wmalloc(struct sock *sk,
857 unsigned long size, int force,
859 extern struct sk_buff *sock_rmalloc(struct sock *sk,
860 unsigned long size, int force,
862 extern void sock_wfree(struct sk_buff *skb);
863 extern void sock_rfree(struct sk_buff *skb);
865 extern int sock_setsockopt(struct socket *sock, int level,
866 int op, char __user *optval,
869 extern int sock_getsockopt(struct socket *sock, int level,
870 int op, char __user *optval,
872 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
876 extern void *sock_kmalloc(struct sock *sk, int size,
878 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
879 extern void sk_send_sigurg(struct sock *sk);
882 * Functions to fill in entries in struct proto_ops when a protocol
883 * does not implement a particular function.
885 extern int sock_no_bind(struct socket *,
886 struct sockaddr *, int);
887 extern int sock_no_connect(struct socket *,
888 struct sockaddr *, int, int);
889 extern int sock_no_socketpair(struct socket *,
891 extern int sock_no_accept(struct socket *,
892 struct socket *, int);
893 extern int sock_no_getname(struct socket *,
894 struct sockaddr *, int *, int);
895 extern unsigned int sock_no_poll(struct file *, struct socket *,
896 struct poll_table_struct *);
897 extern int sock_no_ioctl(struct socket *, unsigned int,
899 extern int sock_no_listen(struct socket *, int);
900 extern int sock_no_shutdown(struct socket *, int);
901 extern int sock_no_getsockopt(struct socket *, int , int,
902 char __user *, int __user *);
903 extern int sock_no_setsockopt(struct socket *, int, int,
905 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
906 struct msghdr *, size_t);
907 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
908 struct msghdr *, size_t, int);
909 extern int sock_no_mmap(struct file *file,
911 struct vm_area_struct *vma);
912 extern ssize_t sock_no_sendpage(struct socket *sock,
914 int offset, size_t size,
918 * Functions to fill in entries in struct proto_ops when a protocol
919 * uses the inet style.
921 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
922 char __user *optval, int __user *optlen);
923 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
924 struct msghdr *msg, size_t size, int flags);
925 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
926 char __user *optval, int optlen);
927 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
928 int optname, char __user *optval, int __user *optlen);
929 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
930 int optname, char __user *optval, int optlen);
932 extern void sk_common_release(struct sock *sk);
935 * Default socket callbacks and setup code
938 /* Initialise core socket variables */
939 extern void sock_init_data(struct socket *sock, struct sock *sk);
942 * sk_filter - run a packet through a socket filter
943 * @sk: sock associated with &sk_buff
944 * @skb: buffer to filter
945 * @needlock: set to 1 if the sock is not locked by caller.
947 * Run the filter code and then cut skb->data to correct size returned by
948 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
949 * than pkt_len we keep whole skb->data. This is the socket level
950 * wrapper to sk_run_filter. It returns 0 if the packet should
951 * be accepted or -EPERM if the packet should be tossed.
955 static inline int sk_filter(struct sock *sk, struct sk_buff *skb)
958 struct sk_filter *filter;
960 err = security_sock_rcv_skb(sk, skb);
965 filter = rcu_dereference(sk->sk_filter);
967 unsigned int pkt_len = sk_run_filter(skb, filter->insns,
969 err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM;
971 rcu_read_unlock_bh();
977 * sk_filter_release: Release a socket filter
979 * @fp: filter to remove
981 * Remove a filter from a socket and release its resources.
984 static inline void sk_filter_release(struct sk_filter *fp)
986 if (atomic_dec_and_test(&fp->refcnt))
990 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
992 unsigned int size = sk_filter_len(fp);
994 atomic_sub(size, &sk->sk_omem_alloc);
995 sk_filter_release(fp);
998 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1000 atomic_inc(&fp->refcnt);
1001 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1005 * Socket reference counting postulates.
1007 * * Each user of socket SHOULD hold a reference count.
1008 * * Each access point to socket (an hash table bucket, reference from a list,
1009 * running timer, skb in flight MUST hold a reference count.
1010 * * When reference count hits 0, it means it will never increase back.
1011 * * When reference count hits 0, it means that no references from
1012 * outside exist to this socket and current process on current CPU
1013 * is last user and may/should destroy this socket.
1014 * * sk_free is called from any context: process, BH, IRQ. When
1015 * it is called, socket has no references from outside -> sk_free
1016 * may release descendant resources allocated by the socket, but
1017 * to the time when it is called, socket is NOT referenced by any
1018 * hash tables, lists etc.
1019 * * Packets, delivered from outside (from network or from another process)
1020 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1021 * when they sit in queue. Otherwise, packets will leak to hole, when
1022 * socket is looked up by one cpu and unhasing is made by another CPU.
1023 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1024 * (leak to backlog). Packet socket does all the processing inside
1025 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1026 * use separate SMP lock, so that they are prone too.
1029 /* Ungrab socket and destroy it, if it was the last reference. */
1030 static inline void sock_put(struct sock *sk)
1032 if (atomic_dec_and_test(&sk->sk_refcnt))
1036 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1039 /* Detach socket from process context.
1040 * Announce socket dead, detach it from wait queue and inode.
1041 * Note that parent inode held reference count on this struct sock,
1042 * we do not release it in this function, because protocol
1043 * probably wants some additional cleanups or even continuing
1044 * to work with this socket (TCP).
1046 static inline void sock_orphan(struct sock *sk)
1048 write_lock_bh(&sk->sk_callback_lock);
1049 sock_set_flag(sk, SOCK_DEAD);
1050 sk->sk_socket = NULL;
1051 sk->sk_sleep = NULL;
1052 write_unlock_bh(&sk->sk_callback_lock);
1055 static inline void sock_graft(struct sock *sk, struct socket *parent)
1057 write_lock_bh(&sk->sk_callback_lock);
1058 sk->sk_sleep = &parent->wait;
1060 sk->sk_socket = parent;
1061 security_sock_graft(sk, parent);
1062 write_unlock_bh(&sk->sk_callback_lock);
1065 extern int sock_i_uid(struct sock *sk);
1066 extern unsigned long sock_i_ino(struct sock *sk);
1068 static inline struct dst_entry *
1069 __sk_dst_get(struct sock *sk)
1071 return sk->sk_dst_cache;
1074 static inline struct dst_entry *
1075 sk_dst_get(struct sock *sk)
1077 struct dst_entry *dst;
1079 read_lock(&sk->sk_dst_lock);
1080 dst = sk->sk_dst_cache;
1083 read_unlock(&sk->sk_dst_lock);
1088 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1090 struct dst_entry *old_dst;
1092 old_dst = sk->sk_dst_cache;
1093 sk->sk_dst_cache = dst;
1094 dst_release(old_dst);
1098 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1100 write_lock(&sk->sk_dst_lock);
1101 __sk_dst_set(sk, dst);
1102 write_unlock(&sk->sk_dst_lock);
1106 __sk_dst_reset(struct sock *sk)
1108 struct dst_entry *old_dst;
1110 old_dst = sk->sk_dst_cache;
1111 sk->sk_dst_cache = NULL;
1112 dst_release(old_dst);
1116 sk_dst_reset(struct sock *sk)
1118 write_lock(&sk->sk_dst_lock);
1120 write_unlock(&sk->sk_dst_lock);
1123 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1125 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1127 static inline int sk_can_gso(const struct sock *sk)
1129 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1132 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1134 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1135 struct sk_buff *skb, struct page *page,
1138 if (skb->ip_summed == CHECKSUM_NONE) {
1140 __wsum csum = csum_and_copy_from_user(from,
1141 page_address(page) + off,
1145 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1146 } else if (copy_from_user(page_address(page) + off, from, copy))
1150 skb->data_len += copy;
1151 skb->truesize += copy;
1152 sk->sk_wmem_queued += copy;
1153 sk_mem_charge(sk, copy);
1158 * Queue a received datagram if it will fit. Stream and sequenced
1159 * protocols can't normally use this as they need to fit buffers in
1160 * and play with them.
1162 * Inlined as it's very short and called for pretty much every
1163 * packet ever received.
1166 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1170 skb->destructor = sock_wfree;
1171 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1174 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1177 skb->destructor = sock_rfree;
1178 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1179 sk_mem_charge(sk, skb->truesize);
1182 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1183 unsigned long expires);
1185 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1187 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1189 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1191 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1192 number of warnings when compiling with -W --ANK
1194 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1195 (unsigned)sk->sk_rcvbuf)
1197 skb_set_owner_r(skb, sk);
1198 skb_queue_tail(&sk->sk_error_queue, skb);
1199 if (!sock_flag(sk, SOCK_DEAD))
1200 sk->sk_data_ready(sk, skb->len);
1205 * Recover an error report and clear atomically
1208 static inline int sock_error(struct sock *sk)
1211 if (likely(!sk->sk_err))
1213 err = xchg(&sk->sk_err, 0);
1217 static inline unsigned long sock_wspace(struct sock *sk)
1221 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1222 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1229 static inline void sk_wake_async(struct sock *sk, int how, int band)
1231 if (sk->sk_socket && sk->sk_socket->fasync_list)
1232 sock_wake_async(sk->sk_socket, how, band);
1235 #define SOCK_MIN_SNDBUF 2048
1236 #define SOCK_MIN_RCVBUF 256
1238 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1240 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1241 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1242 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1246 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1248 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1250 struct page *page = NULL;
1252 page = alloc_pages(sk->sk_allocation, 0);
1254 sk->sk_prot->enter_memory_pressure();
1255 sk_stream_moderate_sndbuf(sk);
1261 * Default write policy as shown to user space via poll/select/SIGIO
1263 static inline int sock_writeable(const struct sock *sk)
1265 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1268 static inline gfp_t gfp_any(void)
1270 return in_atomic() ? GFP_ATOMIC : GFP_KERNEL;
1273 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1275 return noblock ? 0 : sk->sk_rcvtimeo;
1278 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1280 return noblock ? 0 : sk->sk_sndtimeo;
1283 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1285 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1288 /* Alas, with timeout socket operations are not restartable.
1289 * Compare this to poll().
1291 static inline int sock_intr_errno(long timeo)
1293 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1296 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1297 struct sk_buff *skb);
1299 static __inline__ void
1300 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1302 ktime_t kt = skb->tstamp;
1304 if (sock_flag(sk, SOCK_RCVTSTAMP))
1305 __sock_recv_timestamp(msg, sk, skb);
1311 * sk_eat_skb - Release a skb if it is no longer needed
1312 * @sk: socket to eat this skb from
1313 * @skb: socket buffer to eat
1314 * @copied_early: flag indicating whether DMA operations copied this data early
1316 * This routine must be called with interrupts disabled or with the socket
1317 * locked so that the sk_buff queue operation is ok.
1319 #ifdef CONFIG_NET_DMA
1320 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1322 __skb_unlink(skb, &sk->sk_receive_queue);
1326 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1329 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1331 __skb_unlink(skb, &sk->sk_receive_queue);
1336 extern void sock_enable_timestamp(struct sock *sk);
1337 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1338 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1341 * Enable debug/info messages
1343 extern int net_msg_warn;
1344 #define NETDEBUG(fmt, args...) \
1345 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1347 #define LIMIT_NETDEBUG(fmt, args...) \
1348 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1351 * Macros for sleeping on a socket. Use them like this:
1353 * SOCK_SLEEP_PRE(sk)
1356 * SOCK_SLEEP_POST(sk)
1358 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1359 * and when the last use of them in DECnet has gone, I'm intending to
1363 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1364 DECLARE_WAITQUEUE(wait, tsk); \
1365 tsk->state = TASK_INTERRUPTIBLE; \
1366 add_wait_queue((sk)->sk_sleep, &wait); \
1369 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1370 remove_wait_queue((sk)->sk_sleep, &wait); \
1374 extern __u32 sysctl_wmem_max;
1375 extern __u32 sysctl_rmem_max;
1377 extern void sk_init(void);
1379 extern int sysctl_optmem_max;
1381 extern __u32 sysctl_wmem_default;
1382 extern __u32 sysctl_rmem_default;
1384 #endif /* _SOCK_H */