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/config.h>
44 #include <linux/list.h>
45 #include <linux/timer.h>
46 #include <linux/cache.h>
47 #include <linux/module.h>
48 #include <linux/netdevice.h>
49 #include <linux/skbuff.h> /* struct sk_buff */
50 #include <linux/security.h>
52 #include <linux/filter.h>
54 #include <asm/atomic.h>
56 #include <net/checksum.h>
59 * This structure really needs to be cleaned up.
60 * Most of it is for TCP, and not used by any of
61 * the other protocols.
64 /* Define this to get the SOCK_DBG debugging facility. */
65 #define SOCK_DEBUGGING
67 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
68 printk(KERN_DEBUG msg); } while (0)
70 #define SOCK_DEBUG(sk, msg...) do { } while (0)
73 /* This is the per-socket lock. The spinlock provides a synchronization
74 * between user contexts and software interrupt processing, whereas the
75 * mini-semaphore synchronizes multiple users amongst themselves.
80 struct sock_iocb *owner;
84 #define sock_lock_init(__sk) \
85 do { spin_lock_init(&((__sk)->sk_lock.slock)); \
86 (__sk)->sk_lock.owner = NULL; \
87 init_waitqueue_head(&((__sk)->sk_lock.wq)); \
93 * struct sock_common - minimal network layer representation of sockets
94 * @skc_family: network address family
95 * @skc_state: Connection state
96 * @skc_reuse: %SO_REUSEADDR setting
97 * @skc_bound_dev_if: bound device index if != 0
98 * @skc_node: main hash linkage for various protocol lookup tables
99 * @skc_bind_node: bind hash linkage for various protocol lookup tables
100 * @skc_refcnt: reference count
102 * This is the minimal network layer representation of sockets, the header
103 * for struct sock and struct tcp_tw_bucket.
106 unsigned short skc_family;
107 volatile unsigned char skc_state;
108 unsigned char skc_reuse;
109 int skc_bound_dev_if;
110 struct hlist_node skc_node;
111 struct hlist_node skc_bind_node;
114 struct vx_info *skc_vx_info;
116 struct nx_info *skc_nx_info;
120 * struct sock - network layer representation of sockets
121 * @__sk_common: shared layout with tcp_tw_bucket
122 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
123 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
124 * @sk_lock: synchronizer
125 * @sk_rcvbuf: size of receive buffer in bytes
126 * @sk_sleep: sock wait queue
127 * @sk_dst_cache: destination cache
128 * @sk_dst_lock: destination cache lock
129 * @sk_policy: flow policy
130 * @sk_rmem_alloc: receive queue bytes committed
131 * @sk_receive_queue: incoming packets
132 * @sk_wmem_alloc: transmit queue bytes committed
133 * @sk_write_queue: Packet sending queue
134 * @sk_omem_alloc: "o" is "option" or "other"
135 * @sk_wmem_queued: persistent queue size
136 * @sk_forward_alloc: space allocated forward
137 * @sk_allocation: allocation mode
138 * @sk_sndbuf: size of send buffer in bytes
139 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
140 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
141 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
142 * @sk_lingertime: %SO_LINGER l_linger setting
143 * @sk_hashent: hash entry in several tables (e.g. tcp_ehash)
144 * @sk_backlog: always used with the per-socket spinlock held
145 * @sk_callback_lock: used with the callbacks in the end of this struct
146 * @sk_error_queue: rarely used
147 * @sk_prot: protocol handlers inside a network family
148 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
149 * @sk_err: last error
150 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
151 * @sk_ack_backlog: current listen backlog
152 * @sk_max_ack_backlog: listen backlog set in listen()
153 * @sk_priority: %SO_PRIORITY setting
154 * @sk_type: socket type (%SOCK_STREAM, etc)
155 * @sk_protocol: which protocol this socket belongs in this network family
156 * @sk_peercred: %SO_PEERCRED setting
157 * @sk_rcvlowat: %SO_RCVLOWAT setting
158 * @sk_rcvtimeo: %SO_RCVTIMEO setting
159 * @sk_sndtimeo: %SO_SNDTIMEO setting
160 * @sk_filter: socket filtering instructions
161 * @sk_protinfo: private area, net family specific, when not using slab
162 * @sk_timer: sock cleanup timer
163 * @sk_stamp: time stamp of last packet received
164 * @sk_socket: Identd and reporting IO signals
165 * @sk_user_data: RPC layer private data
166 * @sk_sndmsg_page: cached page for sendmsg
167 * @sk_sndmsg_off: cached offset for sendmsg
168 * @sk_send_head: front of stuff to transmit
169 * @sk_security: used by security modules
170 * @sk_write_pending: a write to stream socket waits to start
171 * @sk_state_change: callback to indicate change in the state of the sock
172 * @sk_data_ready: callback to indicate there is data to be processed
173 * @sk_write_space: callback to indicate there is bf sending space available
174 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
175 * @sk_backlog_rcv: callback to process the backlog
176 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
180 * Now struct tcp_tw_bucket also uses sock_common, so please just
181 * don't add nothing before this first member (__sk_common) --acme
183 struct sock_common __sk_common;
184 #define sk_family __sk_common.skc_family
185 #define sk_state __sk_common.skc_state
186 #define sk_reuse __sk_common.skc_reuse
187 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
188 #define sk_node __sk_common.skc_node
189 #define sk_bind_node __sk_common.skc_bind_node
190 #define sk_refcnt __sk_common.skc_refcnt
191 #define sk_xid __sk_common.skc_xid
192 #define sk_vx_info __sk_common.skc_vx_info
193 #define sk_nid __sk_common.skc_nid
194 #define sk_nx_info __sk_common.skc_nx_info
195 unsigned char sk_shutdown : 2,
198 unsigned char sk_protocol;
199 unsigned short sk_type;
201 socket_lock_t sk_lock;
202 wait_queue_head_t *sk_sleep;
203 struct dst_entry *sk_dst_cache;
204 struct xfrm_policy *sk_policy[2];
205 rwlock_t sk_dst_lock;
206 atomic_t sk_rmem_alloc;
207 atomic_t sk_wmem_alloc;
208 atomic_t sk_omem_alloc;
209 struct sk_buff_head sk_receive_queue;
210 struct sk_buff_head sk_write_queue;
212 int sk_forward_alloc;
213 unsigned int sk_allocation;
217 unsigned long sk_flags;
218 unsigned long sk_lingertime;
220 * The backlog queue is special, it is always used with
221 * the per-socket spinlock held and requires low latency
222 * access. Therefore we special case it's implementation.
225 struct sk_buff *head;
226 struct sk_buff *tail;
228 struct sk_buff_head sk_error_queue;
229 struct proto *sk_prot;
230 struct proto *sk_prot_creator;
231 rwlock_t sk_callback_lock;
234 unsigned short sk_ack_backlog;
235 unsigned short sk_max_ack_backlog;
237 struct ucred sk_peercred;
241 struct sk_filter *sk_filter;
243 struct timer_list sk_timer;
244 struct timeval sk_stamp;
245 struct socket *sk_socket;
247 struct page *sk_sndmsg_page;
248 struct sk_buff *sk_send_head;
250 int sk_write_pending;
252 void (*sk_state_change)(struct sock *sk);
253 void (*sk_data_ready)(struct sock *sk, int bytes);
254 void (*sk_write_space)(struct sock *sk);
255 void (*sk_error_report)(struct sock *sk);
256 int (*sk_backlog_rcv)(struct sock *sk,
257 struct sk_buff *skb);
258 void (*sk_destruct)(struct sock *sk);
262 * Hashed lists helper routines
264 static inline struct sock *__sk_head(struct hlist_head *head)
266 return hlist_entry(head->first, struct sock, sk_node);
269 static inline struct sock *sk_head(struct hlist_head *head)
271 return hlist_empty(head) ? NULL : __sk_head(head);
274 static inline struct sock *sk_next(struct sock *sk)
276 return sk->sk_node.next ?
277 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
280 static inline int sk_unhashed(struct sock *sk)
282 return hlist_unhashed(&sk->sk_node);
285 static inline int sk_hashed(struct sock *sk)
287 return sk->sk_node.pprev != NULL;
290 static __inline__ void sk_node_init(struct hlist_node *node)
295 static __inline__ void __sk_del_node(struct sock *sk)
297 __hlist_del(&sk->sk_node);
300 static __inline__ int __sk_del_node_init(struct sock *sk)
304 sk_node_init(&sk->sk_node);
310 /* Grab socket reference count. This operation is valid only
311 when sk is ALREADY grabbed f.e. it is found in hash table
312 or a list and the lookup is made under lock preventing hash table
316 static inline void sock_hold(struct sock *sk)
318 atomic_inc(&sk->sk_refcnt);
321 /* Ungrab socket in the context, which assumes that socket refcnt
322 cannot hit zero, f.e. it is true in context of any socketcall.
324 static inline void __sock_put(struct sock *sk)
326 atomic_dec(&sk->sk_refcnt);
329 static __inline__ int sk_del_node_init(struct sock *sk)
331 int rc = __sk_del_node_init(sk);
334 /* paranoid for a while -acme */
335 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
341 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
343 hlist_add_head(&sk->sk_node, list);
346 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
349 __sk_add_node(sk, list);
352 static __inline__ void __sk_del_bind_node(struct sock *sk)
354 __hlist_del(&sk->sk_bind_node);
357 static __inline__ void sk_add_bind_node(struct sock *sk,
358 struct hlist_head *list)
360 hlist_add_head(&sk->sk_bind_node, list);
363 #define sk_for_each(__sk, node, list) \
364 hlist_for_each_entry(__sk, node, list, sk_node)
365 #define sk_for_each_from(__sk, node) \
366 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
367 hlist_for_each_entry_from(__sk, node, sk_node)
368 #define sk_for_each_continue(__sk, node) \
369 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
370 hlist_for_each_entry_continue(__sk, node, sk_node)
371 #define sk_for_each_safe(__sk, node, tmp, list) \
372 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
373 #define sk_for_each_bound(__sk, node, list) \
374 hlist_for_each_entry(__sk, node, list, sk_bind_node)
387 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
388 SOCK_DBG, /* %SO_DEBUG setting */
389 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
390 SOCK_NO_LARGESEND, /* whether to sent large segments or not */
391 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
392 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
395 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
397 __set_bit(flag, &sk->sk_flags);
400 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
402 __clear_bit(flag, &sk->sk_flags);
405 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
407 return test_bit(flag, &sk->sk_flags);
410 static inline void sk_acceptq_removed(struct sock *sk)
412 sk->sk_ack_backlog--;
415 static inline void sk_acceptq_added(struct sock *sk)
417 sk->sk_ack_backlog++;
420 static inline int sk_acceptq_is_full(struct sock *sk)
422 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
426 * Compute minimal free write space needed to queue new packets.
428 static inline int sk_stream_min_wspace(struct sock *sk)
430 return sk->sk_wmem_queued / 2;
433 static inline int sk_stream_wspace(struct sock *sk)
435 return sk->sk_sndbuf - sk->sk_wmem_queued;
438 extern void sk_stream_write_space(struct sock *sk);
440 static inline int sk_stream_memory_free(struct sock *sk)
442 return sk->sk_wmem_queued < sk->sk_sndbuf;
445 extern void sk_stream_rfree(struct sk_buff *skb);
447 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
450 skb->destructor = sk_stream_rfree;
451 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
452 sk->sk_forward_alloc -= skb->truesize;
455 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
457 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
458 sk->sk_wmem_queued -= skb->truesize;
459 sk->sk_forward_alloc += skb->truesize;
463 /* The per-socket spinlock must be held here. */
464 #define sk_add_backlog(__sk, __skb) \
465 do { if (!(__sk)->sk_backlog.tail) { \
466 (__sk)->sk_backlog.head = \
467 (__sk)->sk_backlog.tail = (__skb); \
469 ((__sk)->sk_backlog.tail)->next = (__skb); \
470 (__sk)->sk_backlog.tail = (__skb); \
472 (__skb)->next = NULL; \
475 #define sk_wait_event(__sk, __timeo, __condition) \
477 release_sock(__sk); \
480 *(__timeo) = schedule_timeout(*(__timeo)); \
487 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
488 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
489 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
490 extern int sk_stream_error(struct sock *sk, int flags, int err);
491 extern void sk_stream_kill_queues(struct sock *sk);
493 extern int sk_wait_data(struct sock *sk, long *timeo);
495 /* Networking protocol blocks we attach to sockets.
496 * socket layer -> transport layer interface
497 * transport -> network interface is defined by struct inet_proto
500 void (*close)(struct sock *sk,
502 int (*connect)(struct sock *sk,
503 struct sockaddr *uaddr,
505 int (*disconnect)(struct sock *sk, int flags);
507 struct sock * (*accept) (struct sock *sk, int flags, int *err);
509 int (*ioctl)(struct sock *sk, int cmd,
511 int (*init)(struct sock *sk);
512 int (*destroy)(struct sock *sk);
513 void (*shutdown)(struct sock *sk, int how);
514 int (*setsockopt)(struct sock *sk, int level,
515 int optname, char __user *optval,
517 int (*getsockopt)(struct sock *sk, int level,
518 int optname, char __user *optval,
520 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
521 struct msghdr *msg, size_t len);
522 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
524 size_t len, int noblock, int flags,
526 int (*sendpage)(struct sock *sk, struct page *page,
527 int offset, size_t size, int flags);
528 int (*bind)(struct sock *sk,
529 struct sockaddr *uaddr, int addr_len);
531 int (*backlog_rcv) (struct sock *sk,
532 struct sk_buff *skb);
534 /* Keeping track of sk's, looking them up, and port selection methods. */
535 void (*hash)(struct sock *sk);
536 void (*unhash)(struct sock *sk);
537 int (*get_port)(struct sock *sk, unsigned short snum);
539 /* Memory pressure */
540 void (*enter_memory_pressure)(void);
541 atomic_t *memory_allocated; /* Current allocated memory. */
542 atomic_t *sockets_allocated; /* Current number of sockets. */
544 * Pressure flag: try to collapse.
545 * Technical note: it is used by multiple contexts non atomically.
546 * All the sk_stream_mem_schedule() is of this nature: accounting
547 * is strict, actions are advisory and have some latency.
549 int *memory_pressure;
556 unsigned int obj_size;
558 struct module *owner;
562 struct list_head node;
566 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
570 extern int proto_register(struct proto *prot, int alloc_slab);
571 extern void proto_unregister(struct proto *prot);
573 /* Called with local bh disabled */
574 static __inline__ void sock_prot_inc_use(struct proto *prot)
576 prot->stats[smp_processor_id()].inuse++;
579 static __inline__ void sock_prot_dec_use(struct proto *prot)
581 prot->stats[smp_processor_id()].inuse--;
584 /* About 10 seconds */
585 #define SOCK_DESTROY_TIME (10*HZ)
587 /* Sockets 0-1023 can't be bound to unless you are superuser */
588 #define PROT_SOCK 1024
590 #define SHUTDOWN_MASK 3
591 #define RCV_SHUTDOWN 1
592 #define SEND_SHUTDOWN 2
594 #define SOCK_SNDBUF_LOCK 1
595 #define SOCK_RCVBUF_LOCK 2
596 #define SOCK_BINDADDR_LOCK 4
597 #define SOCK_BINDPORT_LOCK 8
599 /* sock_iocb: used to kick off async processing of socket ios */
601 struct list_head list;
607 struct scm_cookie *scm;
608 struct msghdr *msg, async_msg;
609 struct iovec async_iov;
613 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
615 return (struct sock_iocb *)iocb->private;
618 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
623 struct socket_alloc {
624 struct socket socket;
625 struct inode vfs_inode;
628 static inline struct socket *SOCKET_I(struct inode *inode)
630 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
633 static inline struct inode *SOCK_INODE(struct socket *socket)
635 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
638 extern void __sk_stream_mem_reclaim(struct sock *sk);
639 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
641 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
643 static inline int sk_stream_pages(int amt)
645 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
648 static inline void sk_stream_mem_reclaim(struct sock *sk)
650 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
651 __sk_stream_mem_reclaim(sk);
654 static inline void sk_stream_writequeue_purge(struct sock *sk)
658 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
659 sk_stream_free_skb(sk, skb);
660 sk_stream_mem_reclaim(sk);
663 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
665 return (int)skb->truesize <= sk->sk_forward_alloc ||
666 sk_stream_mem_schedule(sk, skb->truesize, 1);
669 /* Used by processes to "lock" a socket state, so that
670 * interrupts and bottom half handlers won't change it
671 * from under us. It essentially blocks any incoming
672 * packets, so that we won't get any new data or any
673 * packets that change the state of the socket.
675 * While locked, BH processing will add new packets to
676 * the backlog queue. This queue is processed by the
677 * owner of the socket lock right before it is released.
679 * Since ~2.3.5 it is also exclusive sleep lock serializing
680 * accesses from user process context.
682 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
684 extern void FASTCALL(lock_sock(struct sock *sk));
685 extern void FASTCALL(release_sock(struct sock *sk));
687 /* BH context may only use the following locking interface. */
688 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
689 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
691 extern struct sock *sk_alloc(int family, int priority,
692 struct proto *prot, int zero_it);
693 extern void sk_free(struct sock *sk);
695 extern struct sk_buff *sock_wmalloc(struct sock *sk,
696 unsigned long size, int force,
698 extern struct sk_buff *sock_rmalloc(struct sock *sk,
699 unsigned long size, int force,
701 extern void sock_wfree(struct sk_buff *skb);
702 extern void sock_rfree(struct sk_buff *skb);
704 extern int sock_setsockopt(struct socket *sock, int level,
705 int op, char __user *optval,
708 extern int sock_getsockopt(struct socket *sock, int level,
709 int op, char __user *optval,
711 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
715 extern void *sock_kmalloc(struct sock *sk, int size, int priority);
716 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
717 extern void sk_send_sigurg(struct sock *sk);
720 * Functions to fill in entries in struct proto_ops when a protocol
721 * does not implement a particular function.
723 extern int sock_no_bind(struct socket *,
724 struct sockaddr *, int);
725 extern int sock_no_connect(struct socket *,
726 struct sockaddr *, int, int);
727 extern int sock_no_socketpair(struct socket *,
729 extern int sock_no_accept(struct socket *,
730 struct socket *, int);
731 extern int sock_no_getname(struct socket *,
732 struct sockaddr *, int *, int);
733 extern unsigned int sock_no_poll(struct file *, struct socket *,
734 struct poll_table_struct *);
735 extern int sock_no_ioctl(struct socket *, unsigned int,
737 extern int sock_no_listen(struct socket *, int);
738 extern int sock_no_shutdown(struct socket *, int);
739 extern int sock_no_getsockopt(struct socket *, int , int,
740 char __user *, int __user *);
741 extern int sock_no_setsockopt(struct socket *, int, int,
743 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
744 struct msghdr *, size_t);
745 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
746 struct msghdr *, size_t, int);
747 extern int sock_no_mmap(struct file *file,
749 struct vm_area_struct *vma);
750 extern ssize_t sock_no_sendpage(struct socket *sock,
752 int offset, size_t size,
756 * Functions to fill in entries in struct proto_ops when a protocol
757 * uses the inet style.
759 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
760 char __user *optval, int __user *optlen);
761 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
762 struct msghdr *msg, size_t size, int flags);
763 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
764 char __user *optval, int optlen);
766 extern void sk_common_release(struct sock *sk);
769 * Default socket callbacks and setup code
772 /* Initialise core socket variables */
773 extern void sock_init_data(struct socket *sock, struct sock *sk);
776 * sk_filter - run a packet through a socket filter
777 * @sk: sock associated with &sk_buff
778 * @skb: buffer to filter
779 * @needlock: set to 1 if the sock is not locked by caller.
781 * Run the filter code and then cut skb->data to correct size returned by
782 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
783 * than pkt_len we keep whole skb->data. This is the socket level
784 * wrapper to sk_run_filter. It returns 0 if the packet should
785 * be accepted or -EPERM if the packet should be tossed.
789 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
793 err = security_sock_rcv_skb(sk, skb);
798 struct sk_filter *filter;
803 filter = sk->sk_filter;
805 int pkt_len = sk_run_filter(skb, filter->insns,
810 skb_trim(skb, pkt_len);
820 * sk_filter_release: Release a socket filter
822 * @fp: filter to remove
824 * Remove a filter from a socket and release its resources.
827 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
829 unsigned int size = sk_filter_len(fp);
831 atomic_sub(size, &sk->sk_omem_alloc);
833 if (atomic_dec_and_test(&fp->refcnt))
837 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
839 atomic_inc(&fp->refcnt);
840 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
844 * Socket reference counting postulates.
846 * * Each user of socket SHOULD hold a reference count.
847 * * Each access point to socket (an hash table bucket, reference from a list,
848 * running timer, skb in flight MUST hold a reference count.
849 * * When reference count hits 0, it means it will never increase back.
850 * * When reference count hits 0, it means that no references from
851 * outside exist to this socket and current process on current CPU
852 * is last user and may/should destroy this socket.
853 * * sk_free is called from any context: process, BH, IRQ. When
854 * it is called, socket has no references from outside -> sk_free
855 * may release descendant resources allocated by the socket, but
856 * to the time when it is called, socket is NOT referenced by any
857 * hash tables, lists etc.
858 * * Packets, delivered from outside (from network or from another process)
859 * and enqueued on receive/error queues SHOULD NOT grab reference count,
860 * when they sit in queue. Otherwise, packets will leak to hole, when
861 * socket is looked up by one cpu and unhasing is made by another CPU.
862 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
863 * (leak to backlog). Packet socket does all the processing inside
864 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
865 * use separate SMP lock, so that they are prone too.
868 /* Ungrab socket and destroy it, if it was the last reference. */
869 static inline void sock_put(struct sock *sk)
871 if (atomic_dec_and_test(&sk->sk_refcnt))
875 /* Detach socket from process context.
876 * Announce socket dead, detach it from wait queue and inode.
877 * Note that parent inode held reference count on this struct sock,
878 * we do not release it in this function, because protocol
879 * probably wants some additional cleanups or even continuing
880 * to work with this socket (TCP).
882 static inline void sock_orphan(struct sock *sk)
884 write_lock_bh(&sk->sk_callback_lock);
885 sock_set_flag(sk, SOCK_DEAD);
886 sk->sk_socket = NULL;
888 write_unlock_bh(&sk->sk_callback_lock);
891 static inline void sock_graft(struct sock *sk, struct socket *parent)
893 write_lock_bh(&sk->sk_callback_lock);
894 sk->sk_sleep = &parent->wait;
896 sk->sk_socket = parent;
897 write_unlock_bh(&sk->sk_callback_lock);
900 extern int sock_i_uid(struct sock *sk);
901 extern unsigned long sock_i_ino(struct sock *sk);
903 static inline struct dst_entry *
904 __sk_dst_get(struct sock *sk)
906 return sk->sk_dst_cache;
909 static inline struct dst_entry *
910 sk_dst_get(struct sock *sk)
912 struct dst_entry *dst;
914 read_lock(&sk->sk_dst_lock);
915 dst = sk->sk_dst_cache;
918 read_unlock(&sk->sk_dst_lock);
923 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
925 struct dst_entry *old_dst;
927 old_dst = sk->sk_dst_cache;
928 sk->sk_dst_cache = dst;
929 dst_release(old_dst);
933 sk_dst_set(struct sock *sk, struct dst_entry *dst)
935 write_lock(&sk->sk_dst_lock);
936 __sk_dst_set(sk, dst);
937 write_unlock(&sk->sk_dst_lock);
941 __sk_dst_reset(struct sock *sk)
943 struct dst_entry *old_dst;
945 old_dst = sk->sk_dst_cache;
946 sk->sk_dst_cache = NULL;
947 dst_release(old_dst);
951 sk_dst_reset(struct sock *sk)
953 write_lock(&sk->sk_dst_lock);
955 write_unlock(&sk->sk_dst_lock);
958 static inline struct dst_entry *
959 __sk_dst_check(struct sock *sk, u32 cookie)
961 struct dst_entry *dst = sk->sk_dst_cache;
963 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
964 sk->sk_dst_cache = NULL;
972 static inline struct dst_entry *
973 sk_dst_check(struct sock *sk, u32 cookie)
975 struct dst_entry *dst = sk_dst_get(sk);
977 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
986 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
988 sk->sk_wmem_queued += skb->truesize;
989 sk->sk_forward_alloc -= skb->truesize;
992 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
993 struct sk_buff *skb, struct page *page,
996 if (skb->ip_summed == CHECKSUM_NONE) {
998 unsigned int csum = csum_and_copy_from_user(from,
999 page_address(page) + off,
1003 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1004 } else if (copy_from_user(page_address(page) + off, from, copy))
1008 skb->data_len += copy;
1009 skb->truesize += copy;
1010 sk->sk_wmem_queued += copy;
1011 sk->sk_forward_alloc -= copy;
1016 * Queue a received datagram if it will fit. Stream and sequenced
1017 * protocols can't normally use this as they need to fit buffers in
1018 * and play with them.
1020 * Inlined as it's very short and called for pretty much every
1021 * packet ever received.
1024 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1028 skb->destructor = sock_wfree;
1029 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1032 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1035 skb->destructor = sock_rfree;
1036 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1039 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1040 unsigned long expires);
1042 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1044 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1049 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1050 number of warnings when compiling with -W --ANK
1052 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1053 (unsigned)sk->sk_rcvbuf) {
1058 /* It would be deadlock, if sock_queue_rcv_skb is used
1059 with socket lock! We assume that users of this
1060 function are lock free.
1062 err = sk_filter(sk, skb, 1);
1067 skb_set_owner_r(skb, sk);
1069 /* Cache the SKB length before we tack it onto the receive
1070 * queue. Once it is added it no longer belongs to us and
1071 * may be freed by other threads of control pulling packets
1076 skb_queue_tail(&sk->sk_receive_queue, skb);
1078 if (!sock_flag(sk, SOCK_DEAD))
1079 sk->sk_data_ready(sk, skb_len);
1084 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1086 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1087 number of warnings when compiling with -W --ANK
1089 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1090 (unsigned)sk->sk_rcvbuf)
1092 skb_set_owner_r(skb, sk);
1093 skb_queue_tail(&sk->sk_error_queue, skb);
1094 if (!sock_flag(sk, SOCK_DEAD))
1095 sk->sk_data_ready(sk, skb->len);
1100 * Recover an error report and clear atomically
1103 static inline int sock_error(struct sock *sk)
1105 int err = xchg(&sk->sk_err, 0);
1109 static inline unsigned long sock_wspace(struct sock *sk)
1113 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1114 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1121 static inline void sk_wake_async(struct sock *sk, int how, int band)
1123 if (sk->sk_socket && sk->sk_socket->fasync_list)
1124 sock_wake_async(sk->sk_socket, how, band);
1127 #define SOCK_MIN_SNDBUF 2048
1128 #define SOCK_MIN_RCVBUF 256
1130 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1132 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1133 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1134 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1138 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1139 int size, int mem, int gfp)
1141 struct sk_buff *skb = alloc_skb(size + sk->sk_prot->max_header, gfp);
1144 skb->truesize += mem;
1145 if (sk->sk_forward_alloc >= (int)skb->truesize ||
1146 sk_stream_mem_schedule(sk, skb->truesize, 0)) {
1147 skb_reserve(skb, sk->sk_prot->max_header);
1152 sk->sk_prot->enter_memory_pressure();
1153 sk_stream_moderate_sndbuf(sk);
1158 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1161 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1164 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1166 struct page *page = NULL;
1168 if (sk->sk_forward_alloc >= (int)PAGE_SIZE ||
1169 sk_stream_mem_schedule(sk, PAGE_SIZE, 0))
1170 page = alloc_pages(sk->sk_allocation, 0);
1172 sk->sk_prot->enter_memory_pressure();
1173 sk_stream_moderate_sndbuf(sk);
1178 #define sk_stream_for_retrans_queue(skb, sk) \
1179 for (skb = (sk)->sk_write_queue.next; \
1180 (skb != (sk)->sk_send_head) && \
1181 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1185 * Default write policy as shown to user space via poll/select/SIGIO
1187 static inline int sock_writeable(const struct sock *sk)
1189 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1192 static inline int gfp_any(void)
1194 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1197 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1199 return noblock ? 0 : sk->sk_rcvtimeo;
1202 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1204 return noblock ? 0 : sk->sk_sndtimeo;
1207 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1209 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1212 /* Alas, with timeout socket operations are not restartable.
1213 * Compare this to poll().
1215 static inline int sock_intr_errno(long timeo)
1217 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1220 static __inline__ void
1221 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1223 struct timeval *stamp = &skb->stamp;
1224 if (sock_flag(sk, SOCK_RCVTSTAMP)) {
1225 /* Race occurred between timestamp enabling and packet
1226 receiving. Fill in the current time for now. */
1227 if (stamp->tv_sec == 0)
1228 do_gettimeofday(stamp);
1229 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1232 sk->sk_stamp = *stamp;
1236 * sk_eat_skb - Release a skb if it is no longer needed
1237 * @sk: socket to eat this skb from
1238 * @skb: socket buffer to eat
1240 * This routine must be called with interrupts disabled or with the socket
1241 * locked so that the sk_buff queue operation is ok.
1243 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1245 __skb_unlink(skb, &sk->sk_receive_queue);
1249 extern void sock_enable_timestamp(struct sock *sk);
1250 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1253 * Enable debug/info messages
1257 #define NETDEBUG(x) do { } while (0)
1258 #define LIMIT_NETDEBUG(x) do {} while(0)
1260 #define NETDEBUG(x) do { x; } while (0)
1261 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1265 * Macros for sleeping on a socket. Use them like this:
1267 * SOCK_SLEEP_PRE(sk)
1270 * SOCK_SLEEP_POST(sk)
1272 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1273 * and when the last use of them in DECnet has gone, I'm intending to
1277 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1278 DECLARE_WAITQUEUE(wait, tsk); \
1279 tsk->state = TASK_INTERRUPTIBLE; \
1280 add_wait_queue((sk)->sk_sleep, &wait); \
1283 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1284 remove_wait_queue((sk)->sk_sleep, &wait); \
1288 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1291 sock_set_flag(sk, bit);
1293 sock_reset_flag(sk, bit);
1296 extern __u32 sysctl_wmem_max;
1297 extern __u32 sysctl_rmem_max;
1300 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1302 static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
1308 #endif /* _SOCK_H */