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 /* Validate arguments and do nothing */
74 static void inline int __attribute__ ((format (printf, 2, 3)))
75 SOCK_DEBUG(struct sock *sk, const char *msg, ...)
80 /* This is the per-socket lock. The spinlock provides a synchronization
81 * between user contexts and software interrupt processing, whereas the
82 * mini-semaphore synchronizes multiple users amongst themselves.
89 * We express the mutex-alike socket_lock semantics
90 * to the lock validator by explicitly managing
91 * the slock as a lock variant (in addition to
94 #ifdef CONFIG_DEBUG_LOCK_ALLOC
95 struct lockdep_map dep_map;
104 * struct sock_common - minimal network layer representation of sockets
105 * @skc_family: network address family
106 * @skc_state: Connection state
107 * @skc_reuse: %SO_REUSEADDR setting
108 * @skc_bound_dev_if: bound device index if != 0
109 * @skc_node: main hash linkage for various protocol lookup tables
110 * @skc_bind_node: bind hash linkage for various protocol lookup tables
111 * @skc_refcnt: reference count
112 * @skc_hash: hash value used with various protocol lookup tables
113 * @skc_prot: protocol handlers inside a network family
114 * @skc_net: reference to the network namespace of this socket
116 * This is the minimal network layer representation of sockets, the header
117 * for struct sock and struct inet_timewait_sock.
120 unsigned short skc_family;
121 volatile unsigned char skc_state;
122 unsigned char skc_reuse;
123 int skc_bound_dev_if;
124 struct hlist_node skc_node;
125 struct hlist_node skc_bind_node;
127 unsigned int skc_hash;
128 struct proto *skc_prot;
133 * struct sock - network layer representation of sockets
134 * @__sk_common: shared layout with inet_timewait_sock
135 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
136 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
137 * @sk_lock: synchronizer
138 * @sk_rcvbuf: size of receive buffer in bytes
139 * @sk_sleep: sock wait queue
140 * @sk_dst_cache: destination cache
141 * @sk_dst_lock: destination cache lock
142 * @sk_policy: flow policy
143 * @sk_rmem_alloc: receive queue bytes committed
144 * @sk_receive_queue: incoming packets
145 * @sk_wmem_alloc: transmit queue bytes committed
146 * @sk_write_queue: Packet sending queue
147 * @sk_async_wait_queue: DMA copied packets
148 * @sk_omem_alloc: "o" is "option" or "other"
149 * @sk_wmem_queued: persistent queue size
150 * @sk_forward_alloc: space allocated forward
151 * @sk_allocation: allocation mode
152 * @sk_sndbuf: size of send buffer in bytes
153 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
154 * %SO_OOBINLINE settings
155 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
156 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
157 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
158 * @sk_gso_max_size: Maximum GSO segment size to build
159 * @sk_lingertime: %SO_LINGER l_linger setting
160 * @sk_backlog: always used with the per-socket spinlock held
161 * @sk_callback_lock: used with the callbacks in the end of this struct
162 * @sk_error_queue: rarely used
163 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
164 * IPV6_ADDRFORM for instance)
165 * @sk_err: last error
166 * @sk_err_soft: errors that don't cause failure but are the cause of a
167 * persistent failure not just 'timed out'
168 * @sk_drops: raw drops counter
169 * @sk_ack_backlog: current listen backlog
170 * @sk_max_ack_backlog: listen backlog set in listen()
171 * @sk_priority: %SO_PRIORITY setting
172 * @sk_type: socket type (%SOCK_STREAM, etc)
173 * @sk_protocol: which protocol this socket belongs in this network family
174 * @sk_peercred: %SO_PEERCRED setting
175 * @sk_rcvlowat: %SO_RCVLOWAT setting
176 * @sk_rcvtimeo: %SO_RCVTIMEO setting
177 * @sk_sndtimeo: %SO_SNDTIMEO setting
178 * @sk_filter: socket filtering instructions
179 * @sk_protinfo: private area, net family specific, when not using slab
180 * @sk_timer: sock cleanup timer
181 * @sk_stamp: time stamp of last packet received
182 * @sk_socket: Identd and reporting IO signals
183 * @sk_user_data: RPC layer private data
184 * @sk_sndmsg_page: cached page for sendmsg
185 * @sk_sndmsg_off: cached offset for sendmsg
186 * @sk_send_head: front of stuff to transmit
187 * @sk_security: used by security modules
188 * @sk_mark: generic packet mark
189 * @sk_write_pending: a write to stream socket waits to start
190 * @sk_state_change: callback to indicate change in the state of the sock
191 * @sk_data_ready: callback to indicate there is data to be processed
192 * @sk_write_space: callback to indicate there is bf sending space available
193 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
194 * @sk_backlog_rcv: callback to process the backlog
195 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
199 * Now struct inet_timewait_sock also uses sock_common, so please just
200 * don't add nothing before this first member (__sk_common) --acme
202 struct sock_common __sk_common;
203 #define sk_family __sk_common.skc_family
204 #define sk_state __sk_common.skc_state
205 #define sk_reuse __sk_common.skc_reuse
206 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
207 #define sk_node __sk_common.skc_node
208 #define sk_bind_node __sk_common.skc_bind_node
209 #define sk_refcnt __sk_common.skc_refcnt
210 #define sk_hash __sk_common.skc_hash
211 #define sk_prot __sk_common.skc_prot
212 #define sk_net __sk_common.skc_net
213 unsigned char sk_shutdown : 2,
216 unsigned char sk_protocol;
217 unsigned short sk_type;
219 socket_lock_t sk_lock;
221 * The backlog queue is special, it is always used with
222 * the per-socket spinlock held and requires low latency
223 * access. Therefore we special case it's implementation.
226 struct sk_buff *head;
227 struct sk_buff *tail;
229 wait_queue_head_t *sk_sleep;
230 struct dst_entry *sk_dst_cache;
231 struct xfrm_policy *sk_policy[2];
232 rwlock_t sk_dst_lock;
233 atomic_t sk_rmem_alloc;
234 atomic_t sk_wmem_alloc;
235 atomic_t sk_omem_alloc;
237 struct sk_buff_head sk_receive_queue;
238 struct sk_buff_head sk_write_queue;
239 struct sk_buff_head sk_async_wait_queue;
241 int sk_forward_alloc;
245 unsigned int sk_gso_max_size;
247 unsigned long sk_flags;
248 unsigned long sk_lingertime;
249 struct sk_buff_head sk_error_queue;
250 struct proto *sk_prot_creator;
251 rwlock_t sk_callback_lock;
255 unsigned short sk_ack_backlog;
256 unsigned short sk_max_ack_backlog;
258 struct ucred sk_peercred;
261 struct sk_filter *sk_filter;
263 struct timer_list sk_timer;
265 struct socket *sk_socket;
267 struct page *sk_sndmsg_page;
268 struct sk_buff *sk_send_head;
270 int sk_write_pending;
273 /* XXX 4 bytes hole on 64 bit */
274 void (*sk_state_change)(struct sock *sk);
275 void (*sk_data_ready)(struct sock *sk, int bytes);
276 void (*sk_write_space)(struct sock *sk);
277 void (*sk_error_report)(struct sock *sk);
278 int (*sk_backlog_rcv)(struct sock *sk,
279 struct sk_buff *skb);
280 void (*sk_destruct)(struct sock *sk);
284 * Hashed lists helper routines
286 static inline struct sock *__sk_head(const struct hlist_head *head)
288 return hlist_entry(head->first, struct sock, sk_node);
291 static inline struct sock *sk_head(const struct hlist_head *head)
293 return hlist_empty(head) ? NULL : __sk_head(head);
296 static inline struct sock *sk_next(const struct sock *sk)
298 return sk->sk_node.next ?
299 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
302 static inline int sk_unhashed(const struct sock *sk)
304 return hlist_unhashed(&sk->sk_node);
307 static inline int sk_hashed(const struct sock *sk)
309 return !sk_unhashed(sk);
312 static __inline__ void sk_node_init(struct hlist_node *node)
317 static __inline__ void __sk_del_node(struct sock *sk)
319 __hlist_del(&sk->sk_node);
322 static __inline__ int __sk_del_node_init(struct sock *sk)
326 sk_node_init(&sk->sk_node);
332 /* Grab socket reference count. This operation is valid only
333 when sk is ALREADY grabbed f.e. it is found in hash table
334 or a list and the lookup is made under lock preventing hash table
338 static inline void sock_hold(struct sock *sk)
340 atomic_inc(&sk->sk_refcnt);
343 /* Ungrab socket in the context, which assumes that socket refcnt
344 cannot hit zero, f.e. it is true in context of any socketcall.
346 static inline void __sock_put(struct sock *sk)
348 atomic_dec(&sk->sk_refcnt);
351 static __inline__ int sk_del_node_init(struct sock *sk)
353 int rc = __sk_del_node_init(sk);
356 /* paranoid for a while -acme */
357 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
363 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
365 hlist_add_head(&sk->sk_node, list);
368 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
371 __sk_add_node(sk, list);
374 static __inline__ void __sk_del_bind_node(struct sock *sk)
376 __hlist_del(&sk->sk_bind_node);
379 static __inline__ void sk_add_bind_node(struct sock *sk,
380 struct hlist_head *list)
382 hlist_add_head(&sk->sk_bind_node, list);
385 #define sk_for_each(__sk, node, list) \
386 hlist_for_each_entry(__sk, node, list, sk_node)
387 #define sk_for_each_from(__sk, node) \
388 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
389 hlist_for_each_entry_from(__sk, node, sk_node)
390 #define sk_for_each_continue(__sk, node) \
391 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
392 hlist_for_each_entry_continue(__sk, node, sk_node)
393 #define sk_for_each_safe(__sk, node, tmp, list) \
394 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
395 #define sk_for_each_bound(__sk, node, list) \
396 hlist_for_each_entry(__sk, node, list, sk_bind_node)
409 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
410 SOCK_DBG, /* %SO_DEBUG setting */
411 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
412 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
413 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
414 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
417 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
419 nsk->sk_flags = osk->sk_flags;
422 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
424 __set_bit(flag, &sk->sk_flags);
427 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
429 __clear_bit(flag, &sk->sk_flags);
432 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
434 return test_bit(flag, &sk->sk_flags);
437 static inline void sk_acceptq_removed(struct sock *sk)
439 sk->sk_ack_backlog--;
442 static inline void sk_acceptq_added(struct sock *sk)
444 sk->sk_ack_backlog++;
447 static inline int sk_acceptq_is_full(struct sock *sk)
449 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
453 * Compute minimal free write space needed to queue new packets.
455 static inline int sk_stream_min_wspace(struct sock *sk)
457 return sk->sk_wmem_queued >> 1;
460 static inline int sk_stream_wspace(struct sock *sk)
462 return sk->sk_sndbuf - sk->sk_wmem_queued;
465 extern void sk_stream_write_space(struct sock *sk);
467 static inline int sk_stream_memory_free(struct sock *sk)
469 return sk->sk_wmem_queued < sk->sk_sndbuf;
472 /* The per-socket spinlock must be held here. */
473 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
475 if (!sk->sk_backlog.tail) {
476 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
478 sk->sk_backlog.tail->next = skb;
479 sk->sk_backlog.tail = skb;
484 #define sk_wait_event(__sk, __timeo, __condition) \
486 release_sock(__sk); \
487 __rc = __condition; \
489 *(__timeo) = schedule_timeout(*(__timeo)); \
492 __rc = __condition; \
496 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
497 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
498 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
499 extern int sk_stream_error(struct sock *sk, int flags, int err);
500 extern void sk_stream_kill_queues(struct sock *sk);
502 extern int sk_wait_data(struct sock *sk, long *timeo);
504 struct request_sock_ops;
505 struct timewait_sock_ops;
506 struct inet_hashinfo;
508 /* Networking protocol blocks we attach to sockets.
509 * socket layer -> transport layer interface
510 * transport -> network interface is defined by struct inet_proto
513 void (*close)(struct sock *sk,
515 int (*connect)(struct sock *sk,
516 struct sockaddr *uaddr,
518 int (*disconnect)(struct sock *sk, int flags);
520 struct sock * (*accept) (struct sock *sk, int flags, int *err);
522 int (*ioctl)(struct sock *sk, int cmd,
524 int (*init)(struct sock *sk);
525 int (*destroy)(struct sock *sk);
526 void (*shutdown)(struct sock *sk, int how);
527 int (*setsockopt)(struct sock *sk, int level,
528 int optname, char __user *optval,
530 int (*getsockopt)(struct sock *sk, int level,
531 int optname, char __user *optval,
533 int (*compat_setsockopt)(struct sock *sk,
535 int optname, char __user *optval,
537 int (*compat_getsockopt)(struct sock *sk,
539 int optname, char __user *optval,
541 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
542 struct msghdr *msg, size_t len);
543 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
545 size_t len, int noblock, int flags,
547 int (*sendpage)(struct sock *sk, struct page *page,
548 int offset, size_t size, int flags);
549 int (*bind)(struct sock *sk,
550 struct sockaddr *uaddr, int addr_len);
552 int (*backlog_rcv) (struct sock *sk,
553 struct sk_buff *skb);
555 /* Keeping track of sk's, looking them up, and port selection methods. */
556 void (*hash)(struct sock *sk);
557 void (*unhash)(struct sock *sk);
558 int (*get_port)(struct sock *sk, unsigned short snum);
560 /* Keeping track of sockets in use */
561 #ifdef CONFIG_PROC_FS
562 struct pcounter inuse;
565 /* Memory pressure */
566 void (*enter_memory_pressure)(void);
567 atomic_t *memory_allocated; /* Current allocated memory. */
568 atomic_t *sockets_allocated; /* Current number of sockets. */
570 * Pressure flag: try to collapse.
571 * Technical note: it is used by multiple contexts non atomically.
572 * All the __sk_mem_schedule() is of this nature: accounting
573 * is strict, actions are advisory and have some latency.
575 int *memory_pressure;
581 struct kmem_cache *slab;
582 unsigned int obj_size;
584 atomic_t *orphan_count;
586 struct request_sock_ops *rsk_prot;
587 struct timewait_sock_ops *twsk_prot;
589 struct inet_hashinfo *hashinfo;
591 struct module *owner;
595 struct list_head node;
596 #ifdef SOCK_REFCNT_DEBUG
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 */
630 #ifdef CONFIG_PROC_FS
631 # define DEFINE_PROTO_INUSE(NAME) DEFINE_PCOUNTER(NAME)
632 # define REF_PROTO_INUSE(NAME) PCOUNTER_MEMBER_INITIALIZER(NAME, .inuse)
633 /* Called with local bh disabled */
634 static inline void sock_prot_inuse_add(struct proto *prot, int inc)
636 pcounter_add(&prot->inuse, inc);
638 static inline int sock_prot_inuse_init(struct proto *proto)
640 return pcounter_alloc(&proto->inuse);
642 static inline int sock_prot_inuse_get(struct proto *proto)
644 return pcounter_getval(&proto->inuse);
646 static inline void sock_prot_inuse_free(struct proto *proto)
648 pcounter_free(&proto->inuse);
651 # define DEFINE_PROTO_INUSE(NAME)
652 # define REF_PROTO_INUSE(NAME)
653 static void inline sock_prot_inuse_add(struct proto *prot, int inc)
656 static int inline sock_prot_inuse_init(struct proto *proto)
660 static void inline sock_prot_inuse_free(struct proto *proto)
666 /* With per-bucket locks this operation is not-atomic, so that
667 * this version is not worse.
669 static inline void __sk_prot_rehash(struct sock *sk)
671 sk->sk_prot->unhash(sk);
672 sk->sk_prot->hash(sk);
675 /* About 10 seconds */
676 #define SOCK_DESTROY_TIME (10*HZ)
678 /* Sockets 0-1023 can't be bound to unless you are superuser */
679 #define PROT_SOCK 1024
681 #define SHUTDOWN_MASK 3
682 #define RCV_SHUTDOWN 1
683 #define SEND_SHUTDOWN 2
685 #define SOCK_SNDBUF_LOCK 1
686 #define SOCK_RCVBUF_LOCK 2
687 #define SOCK_BINDADDR_LOCK 4
688 #define SOCK_BINDPORT_LOCK 8
690 /* sock_iocb: used to kick off async processing of socket ios */
692 struct list_head list;
698 struct scm_cookie *scm;
699 struct msghdr *msg, async_msg;
703 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
705 return (struct sock_iocb *)iocb->private;
708 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
713 struct socket_alloc {
714 struct socket socket;
715 struct inode vfs_inode;
718 static inline struct socket *SOCKET_I(struct inode *inode)
720 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
723 static inline struct inode *SOCK_INODE(struct socket *socket)
725 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
729 * Functions for memory accounting
731 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
732 extern void __sk_mem_reclaim(struct sock *sk);
734 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
735 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
736 #define SK_MEM_SEND 0
737 #define SK_MEM_RECV 1
739 static inline int sk_mem_pages(int amt)
741 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
744 static inline int sk_has_account(struct sock *sk)
746 /* return true if protocol supports memory accounting */
747 return !!sk->sk_prot->memory_allocated;
750 static inline int sk_wmem_schedule(struct sock *sk, int size)
752 if (!sk_has_account(sk))
754 return size <= sk->sk_forward_alloc ||
755 __sk_mem_schedule(sk, size, SK_MEM_SEND);
758 static inline int sk_rmem_schedule(struct sock *sk, int size)
760 if (!sk_has_account(sk))
762 return size <= sk->sk_forward_alloc ||
763 __sk_mem_schedule(sk, size, SK_MEM_RECV);
766 static inline void sk_mem_reclaim(struct sock *sk)
768 if (!sk_has_account(sk))
770 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
771 __sk_mem_reclaim(sk);
774 static inline void sk_mem_reclaim_partial(struct sock *sk)
776 if (!sk_has_account(sk))
778 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
779 __sk_mem_reclaim(sk);
782 static inline void sk_mem_charge(struct sock *sk, int size)
784 if (!sk_has_account(sk))
786 sk->sk_forward_alloc -= size;
789 static inline void sk_mem_uncharge(struct sock *sk, int size)
791 if (!sk_has_account(sk))
793 sk->sk_forward_alloc += size;
796 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
798 skb_truesize_check(skb);
799 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
800 sk->sk_wmem_queued -= skb->truesize;
801 sk_mem_uncharge(sk, skb->truesize);
805 /* Used by processes to "lock" a socket state, so that
806 * interrupts and bottom half handlers won't change it
807 * from under us. It essentially blocks any incoming
808 * packets, so that we won't get any new data or any
809 * packets that change the state of the socket.
811 * While locked, BH processing will add new packets to
812 * the backlog queue. This queue is processed by the
813 * owner of the socket lock right before it is released.
815 * Since ~2.3.5 it is also exclusive sleep lock serializing
816 * accesses from user process context.
818 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
821 * Macro so as to not evaluate some arguments when
822 * lockdep is not enabled.
824 * Mark both the sk_lock and the sk_lock.slock as a
825 * per-address-family lock class.
827 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
829 sk->sk_lock.owned = 0; \
830 init_waitqueue_head(&sk->sk_lock.wq); \
831 spin_lock_init(&(sk)->sk_lock.slock); \
832 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
833 sizeof((sk)->sk_lock)); \
834 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
836 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
839 extern void lock_sock_nested(struct sock *sk, int subclass);
841 static inline void lock_sock(struct sock *sk)
843 lock_sock_nested(sk, 0);
846 extern void release_sock(struct sock *sk);
848 /* BH context may only use the following locking interface. */
849 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
850 #define bh_lock_sock_nested(__sk) \
851 spin_lock_nested(&((__sk)->sk_lock.slock), \
852 SINGLE_DEPTH_NESTING)
853 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
855 extern struct sock *sk_alloc(struct net *net, int family,
858 extern void sk_free(struct sock *sk);
859 extern void sk_release_kernel(struct sock *sk);
860 extern struct sock *sk_clone(const struct sock *sk,
861 const gfp_t priority);
863 extern struct sk_buff *sock_wmalloc(struct sock *sk,
864 unsigned long size, int force,
866 extern struct sk_buff *sock_rmalloc(struct sock *sk,
867 unsigned long size, int force,
869 extern void sock_wfree(struct sk_buff *skb);
870 extern void sock_rfree(struct sk_buff *skb);
872 extern int sock_setsockopt(struct socket *sock, int level,
873 int op, char __user *optval,
876 extern int sock_getsockopt(struct socket *sock, int level,
877 int op, char __user *optval,
879 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
883 extern void *sock_kmalloc(struct sock *sk, int size,
885 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
886 extern void sk_send_sigurg(struct sock *sk);
889 * Functions to fill in entries in struct proto_ops when a protocol
890 * does not implement a particular function.
892 extern int sock_no_bind(struct socket *,
893 struct sockaddr *, int);
894 extern int sock_no_connect(struct socket *,
895 struct sockaddr *, int, int);
896 extern int sock_no_socketpair(struct socket *,
898 extern int sock_no_accept(struct socket *,
899 struct socket *, int);
900 extern int sock_no_getname(struct socket *,
901 struct sockaddr *, int *, int);
902 extern unsigned int sock_no_poll(struct file *, struct socket *,
903 struct poll_table_struct *);
904 extern int sock_no_ioctl(struct socket *, unsigned int,
906 extern int sock_no_listen(struct socket *, int);
907 extern int sock_no_shutdown(struct socket *, int);
908 extern int sock_no_getsockopt(struct socket *, int , int,
909 char __user *, int __user *);
910 extern int sock_no_setsockopt(struct socket *, int, int,
912 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
913 struct msghdr *, size_t);
914 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
915 struct msghdr *, size_t, int);
916 extern int sock_no_mmap(struct file *file,
918 struct vm_area_struct *vma);
919 extern ssize_t sock_no_sendpage(struct socket *sock,
921 int offset, size_t size,
925 * Functions to fill in entries in struct proto_ops when a protocol
926 * uses the inet style.
928 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
929 char __user *optval, int __user *optlen);
930 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
931 struct msghdr *msg, size_t size, int flags);
932 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
933 char __user *optval, int optlen);
934 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
935 int optname, char __user *optval, int __user *optlen);
936 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
937 int optname, char __user *optval, int optlen);
939 extern void sk_common_release(struct sock *sk);
942 * Default socket callbacks and setup code
945 /* Initialise core socket variables */
946 extern void sock_init_data(struct socket *sock, struct sock *sk);
949 * sk_filter - run a packet through a socket filter
950 * @sk: sock associated with &sk_buff
951 * @skb: buffer to filter
952 * @needlock: set to 1 if the sock is not locked by caller.
954 * Run the filter code and then cut skb->data to correct size returned by
955 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
956 * than pkt_len we keep whole skb->data. This is the socket level
957 * wrapper to sk_run_filter. It returns 0 if the packet should
958 * be accepted or -EPERM if the packet should be tossed.
962 static inline int sk_filter(struct sock *sk, struct sk_buff *skb)
965 struct sk_filter *filter;
967 err = security_sock_rcv_skb(sk, skb);
972 filter = rcu_dereference(sk->sk_filter);
974 unsigned int pkt_len = sk_run_filter(skb, filter->insns,
976 err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM;
978 rcu_read_unlock_bh();
984 * sk_filter_release: Release a socket filter
986 * @fp: filter to remove
988 * Remove a filter from a socket and release its resources.
991 static inline void sk_filter_release(struct sk_filter *fp)
993 if (atomic_dec_and_test(&fp->refcnt))
997 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
999 unsigned int size = sk_filter_len(fp);
1001 atomic_sub(size, &sk->sk_omem_alloc);
1002 sk_filter_release(fp);
1005 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1007 atomic_inc(&fp->refcnt);
1008 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1012 * Socket reference counting postulates.
1014 * * Each user of socket SHOULD hold a reference count.
1015 * * Each access point to socket (an hash table bucket, reference from a list,
1016 * running timer, skb in flight MUST hold a reference count.
1017 * * When reference count hits 0, it means it will never increase back.
1018 * * When reference count hits 0, it means that no references from
1019 * outside exist to this socket and current process on current CPU
1020 * is last user and may/should destroy this socket.
1021 * * sk_free is called from any context: process, BH, IRQ. When
1022 * it is called, socket has no references from outside -> sk_free
1023 * may release descendant resources allocated by the socket, but
1024 * to the time when it is called, socket is NOT referenced by any
1025 * hash tables, lists etc.
1026 * * Packets, delivered from outside (from network or from another process)
1027 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1028 * when they sit in queue. Otherwise, packets will leak to hole, when
1029 * socket is looked up by one cpu and unhasing is made by another CPU.
1030 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1031 * (leak to backlog). Packet socket does all the processing inside
1032 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1033 * use separate SMP lock, so that they are prone too.
1036 /* Ungrab socket and destroy it, if it was the last reference. */
1037 static inline void sock_put(struct sock *sk)
1039 if (atomic_dec_and_test(&sk->sk_refcnt))
1043 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1046 /* Detach socket from process context.
1047 * Announce socket dead, detach it from wait queue and inode.
1048 * Note that parent inode held reference count on this struct sock,
1049 * we do not release it in this function, because protocol
1050 * probably wants some additional cleanups or even continuing
1051 * to work with this socket (TCP).
1053 static inline void sock_orphan(struct sock *sk)
1055 write_lock_bh(&sk->sk_callback_lock);
1056 sock_set_flag(sk, SOCK_DEAD);
1057 sk->sk_socket = NULL;
1058 sk->sk_sleep = NULL;
1059 write_unlock_bh(&sk->sk_callback_lock);
1062 static inline void sock_graft(struct sock *sk, struct socket *parent)
1064 write_lock_bh(&sk->sk_callback_lock);
1065 sk->sk_sleep = &parent->wait;
1067 sk->sk_socket = parent;
1068 security_sock_graft(sk, parent);
1069 write_unlock_bh(&sk->sk_callback_lock);
1072 extern int sock_i_uid(struct sock *sk);
1073 extern unsigned long sock_i_ino(struct sock *sk);
1075 static inline struct dst_entry *
1076 __sk_dst_get(struct sock *sk)
1078 return sk->sk_dst_cache;
1081 static inline struct dst_entry *
1082 sk_dst_get(struct sock *sk)
1084 struct dst_entry *dst;
1086 read_lock(&sk->sk_dst_lock);
1087 dst = sk->sk_dst_cache;
1090 read_unlock(&sk->sk_dst_lock);
1095 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1097 struct dst_entry *old_dst;
1099 old_dst = sk->sk_dst_cache;
1100 sk->sk_dst_cache = dst;
1101 dst_release(old_dst);
1105 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1107 write_lock(&sk->sk_dst_lock);
1108 __sk_dst_set(sk, dst);
1109 write_unlock(&sk->sk_dst_lock);
1113 __sk_dst_reset(struct sock *sk)
1115 struct dst_entry *old_dst;
1117 old_dst = sk->sk_dst_cache;
1118 sk->sk_dst_cache = NULL;
1119 dst_release(old_dst);
1123 sk_dst_reset(struct sock *sk)
1125 write_lock(&sk->sk_dst_lock);
1127 write_unlock(&sk->sk_dst_lock);
1130 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1132 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1134 static inline int sk_can_gso(const struct sock *sk)
1136 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1139 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1141 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1142 struct sk_buff *skb, struct page *page,
1145 if (skb->ip_summed == CHECKSUM_NONE) {
1147 __wsum csum = csum_and_copy_from_user(from,
1148 page_address(page) + off,
1152 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1153 } else if (copy_from_user(page_address(page) + off, from, copy))
1157 skb->data_len += copy;
1158 skb->truesize += copy;
1159 sk->sk_wmem_queued += copy;
1160 sk_mem_charge(sk, copy);
1165 * Queue a received datagram if it will fit. Stream and sequenced
1166 * protocols can't normally use this as they need to fit buffers in
1167 * and play with them.
1169 * Inlined as it's very short and called for pretty much every
1170 * packet ever received.
1173 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1177 skb->destructor = sock_wfree;
1178 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1181 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1184 skb->destructor = sock_rfree;
1185 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1186 sk_mem_charge(sk, skb->truesize);
1189 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1190 unsigned long expires);
1192 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1194 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1196 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1198 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1199 number of warnings when compiling with -W --ANK
1201 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1202 (unsigned)sk->sk_rcvbuf)
1204 skb_set_owner_r(skb, sk);
1205 skb_queue_tail(&sk->sk_error_queue, skb);
1206 if (!sock_flag(sk, SOCK_DEAD))
1207 sk->sk_data_ready(sk, skb->len);
1212 * Recover an error report and clear atomically
1215 static inline int sock_error(struct sock *sk)
1218 if (likely(!sk->sk_err))
1220 err = xchg(&sk->sk_err, 0);
1224 static inline unsigned long sock_wspace(struct sock *sk)
1228 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1229 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1236 static inline void sk_wake_async(struct sock *sk, int how, int band)
1238 if (sk->sk_socket && sk->sk_socket->fasync_list)
1239 sock_wake_async(sk->sk_socket, how, band);
1242 #define SOCK_MIN_SNDBUF 2048
1243 #define SOCK_MIN_RCVBUF 256
1245 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1247 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1248 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1249 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1253 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1255 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1257 struct page *page = NULL;
1259 page = alloc_pages(sk->sk_allocation, 0);
1261 sk->sk_prot->enter_memory_pressure();
1262 sk_stream_moderate_sndbuf(sk);
1268 * Default write policy as shown to user space via poll/select/SIGIO
1270 static inline int sock_writeable(const struct sock *sk)
1272 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1275 static inline gfp_t gfp_any(void)
1277 return in_atomic() ? GFP_ATOMIC : GFP_KERNEL;
1280 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1282 return noblock ? 0 : sk->sk_rcvtimeo;
1285 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1287 return noblock ? 0 : sk->sk_sndtimeo;
1290 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1292 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1295 /* Alas, with timeout socket operations are not restartable.
1296 * Compare this to poll().
1298 static inline int sock_intr_errno(long timeo)
1300 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1303 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1304 struct sk_buff *skb);
1306 static __inline__ void
1307 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1309 ktime_t kt = skb->tstamp;
1311 if (sock_flag(sk, SOCK_RCVTSTAMP))
1312 __sock_recv_timestamp(msg, sk, skb);
1318 * sk_eat_skb - Release a skb if it is no longer needed
1319 * @sk: socket to eat this skb from
1320 * @skb: socket buffer to eat
1321 * @copied_early: flag indicating whether DMA operations copied this data early
1323 * This routine must be called with interrupts disabled or with the socket
1324 * locked so that the sk_buff queue operation is ok.
1326 #ifdef CONFIG_NET_DMA
1327 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1329 __skb_unlink(skb, &sk->sk_receive_queue);
1333 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1336 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1338 __skb_unlink(skb, &sk->sk_receive_queue);
1344 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1345 * They should not hold a referrence to a namespace in order to allow
1347 * Sockets after sk_change_net should be released using sk_release_kernel
1349 static inline void sk_change_net(struct sock *sk, struct net *net)
1351 put_net(sk->sk_net);
1355 extern void sock_enable_timestamp(struct sock *sk);
1356 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1357 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1360 * Enable debug/info messages
1362 extern int net_msg_warn;
1363 #define NETDEBUG(fmt, args...) \
1364 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1366 #define LIMIT_NETDEBUG(fmt, args...) \
1367 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1370 * Macros for sleeping on a socket. Use them like this:
1372 * SOCK_SLEEP_PRE(sk)
1375 * SOCK_SLEEP_POST(sk)
1377 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1378 * and when the last use of them in DECnet has gone, I'm intending to
1382 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1383 DECLARE_WAITQUEUE(wait, tsk); \
1384 tsk->state = TASK_INTERRUPTIBLE; \
1385 add_wait_queue((sk)->sk_sleep, &wait); \
1388 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1389 remove_wait_queue((sk)->sk_sleep, &wait); \
1393 extern __u32 sysctl_wmem_max;
1394 extern __u32 sysctl_rmem_max;
1396 extern void sk_init(void);
1398 extern int sysctl_optmem_max;
1400 extern __u32 sysctl_wmem_default;
1401 extern __u32 sysctl_rmem_default;
1403 #endif /* _SOCK_H */