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
10 * Authors: Ross Biro, <bir7@leland.Stanford.Edu>
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;
116 * struct sock - network layer representation of sockets
117 * @__sk_common: shared layout with tcp_tw_bucket
118 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
119 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
120 * @sk_lock: synchronizer
121 * @sk_rcvbuf: size of receive buffer in bytes
122 * @sk_sleep: sock wait queue
123 * @sk_dst_cache: destination cache
124 * @sk_dst_lock: destination cache lock
125 * @sk_policy: flow policy
126 * @sk_rmem_alloc: receive queue bytes committed
127 * @sk_receive_queue: incoming packets
128 * @sk_wmem_alloc: transmit queue bytes committed
129 * @sk_write_queue: Packet sending queue
130 * @sk_omem_alloc: "o" is "option" or "other"
131 * @sk_wmem_queued: persistent queue size
132 * @sk_forward_alloc: space allocated forward
133 * @sk_allocation: allocation mode
134 * @sk_sndbuf: size of send buffer in bytes
135 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
136 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
137 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
138 * @sk_lingertime: %SO_LINGER l_linger setting
139 * @sk_hashent: hash entry in several tables (e.g. tcp_ehash)
140 * @sk_backlog: always used with the per-socket spinlock held
141 * @sk_callback_lock: used with the callbacks in the end of this struct
142 * @sk_error_queue: rarely used
143 * @sk_prot: protocol handlers inside a network family
144 * @sk_err: last error
145 * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
146 * @sk_ack_backlog: current listen backlog
147 * @sk_max_ack_backlog: listen backlog set in listen()
148 * @sk_priority: %SO_PRIORITY setting
149 * @sk_type: socket type (%SOCK_STREAM, etc)
150 * @sk_protocol: which protocol this socket belongs in this network family
151 * @sk_peercred: %SO_PEERCRED setting
152 * @sk_rcvlowat: %SO_RCVLOWAT setting
153 * @sk_rcvtimeo: %SO_RCVTIMEO setting
154 * @sk_sndtimeo: %SO_SNDTIMEO setting
155 * @sk_filter: socket filtering instructions
156 * @sk_protinfo: private area, net family specific, when not using slab
157 * @sk_timer: sock cleanup timer
158 * @sk_stamp: time stamp of last packet received
159 * @sk_socket: Identd and reporting IO signals
160 * @sk_user_data: RPC layer private data
161 * @sk_sndmsg_page: cached page for sendmsg
162 * @sk_sndmsg_off: cached offset for sendmsg
163 * @sk_send_head: front of stuff to transmit
164 * @sk_security: used by security modules
165 * @sk_write_pending: a write to stream socket waits to start
166 * @sk_state_change: callback to indicate change in the state of the sock
167 * @sk_data_ready: callback to indicate there is data to be processed
168 * @sk_write_space: callback to indicate there is bf sending space available
169 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
170 * @sk_backlog_rcv: callback to process the backlog
171 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
175 * Now struct tcp_tw_bucket also uses sock_common, so please just
176 * don't add nothing before this first member (__sk_common) --acme
178 struct sock_common __sk_common;
179 #define sk_family __sk_common.skc_family
180 #define sk_state __sk_common.skc_state
181 #define sk_reuse __sk_common.skc_reuse
182 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
183 #define sk_node __sk_common.skc_node
184 #define sk_bind_node __sk_common.skc_bind_node
185 #define sk_refcnt __sk_common.skc_refcnt
186 unsigned char sk_shutdown : 2,
189 unsigned char sk_protocol;
190 unsigned short sk_type;
192 socket_lock_t sk_lock;
193 wait_queue_head_t *sk_sleep;
194 struct dst_entry *sk_dst_cache;
195 struct xfrm_policy *sk_policy[2];
196 rwlock_t sk_dst_lock;
197 atomic_t sk_rmem_alloc;
198 atomic_t sk_wmem_alloc;
199 atomic_t sk_omem_alloc;
200 struct sk_buff_head sk_receive_queue;
201 struct sk_buff_head sk_write_queue;
203 int sk_forward_alloc;
204 unsigned int sk_allocation;
208 unsigned long sk_flags;
209 unsigned long sk_lingertime;
211 * The backlog queue is special, it is always used with
212 * the per-socket spinlock held and requires low latency
213 * access. Therefore we special case it's implementation.
216 struct sk_buff *head;
217 struct sk_buff *tail;
219 struct sk_buff_head sk_error_queue;
220 struct proto *sk_prot;
221 rwlock_t sk_callback_lock;
224 unsigned short sk_ack_backlog;
225 unsigned short sk_max_ack_backlog;
227 struct ucred sk_peercred;
231 struct sk_filter *sk_filter;
233 struct timer_list sk_timer;
234 struct timeval sk_stamp;
235 struct socket *sk_socket;
237 struct page *sk_sndmsg_page;
238 struct sk_buff *sk_send_head;
240 int sk_write_pending;
242 void (*sk_state_change)(struct sock *sk);
243 void (*sk_data_ready)(struct sock *sk, int bytes);
244 void (*sk_write_space)(struct sock *sk);
245 void (*sk_error_report)(struct sock *sk);
246 int (*sk_backlog_rcv)(struct sock *sk,
247 struct sk_buff *skb);
248 void (*sk_destruct)(struct sock *sk);
252 * Hashed lists helper routines
254 static inline struct sock *__sk_head(struct hlist_head *head)
256 return hlist_entry(head->first, struct sock, sk_node);
259 static inline struct sock *sk_head(struct hlist_head *head)
261 return hlist_empty(head) ? NULL : __sk_head(head);
264 static inline struct sock *sk_next(struct sock *sk)
266 return sk->sk_node.next ?
267 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
270 static inline int sk_unhashed(struct sock *sk)
272 return hlist_unhashed(&sk->sk_node);
275 static inline int sk_hashed(struct sock *sk)
277 return sk->sk_node.pprev != NULL;
280 static __inline__ void sk_node_init(struct hlist_node *node)
285 static __inline__ void __sk_del_node(struct sock *sk)
287 __hlist_del(&sk->sk_node);
290 static __inline__ int __sk_del_node_init(struct sock *sk)
294 sk_node_init(&sk->sk_node);
300 /* Grab socket reference count. This operation is valid only
301 when sk is ALREADY grabbed f.e. it is found in hash table
302 or a list and the lookup is made under lock preventing hash table
306 static inline void sock_hold(struct sock *sk)
308 atomic_inc(&sk->sk_refcnt);
311 /* Ungrab socket in the context, which assumes that socket refcnt
312 cannot hit zero, f.e. it is true in context of any socketcall.
314 static inline void __sock_put(struct sock *sk)
316 atomic_dec(&sk->sk_refcnt);
319 static __inline__ int sk_del_node_init(struct sock *sk)
321 int rc = __sk_del_node_init(sk);
324 /* paranoid for a while -acme */
325 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
331 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
333 hlist_add_head(&sk->sk_node, list);
336 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
339 __sk_add_node(sk, list);
342 static __inline__ void __sk_del_bind_node(struct sock *sk)
344 __hlist_del(&sk->sk_bind_node);
347 static __inline__ void sk_add_bind_node(struct sock *sk,
348 struct hlist_head *list)
350 hlist_add_head(&sk->sk_bind_node, list);
353 #define sk_for_each(__sk, node, list) \
354 hlist_for_each_entry(__sk, node, list, sk_node)
355 #define sk_for_each_from(__sk, node) \
356 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
357 hlist_for_each_entry_from(__sk, node, sk_node)
358 #define sk_for_each_continue(__sk, node) \
359 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
360 hlist_for_each_entry_continue(__sk, node, sk_node)
361 #define sk_for_each_safe(__sk, node, tmp, list) \
362 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
363 #define sk_for_each_bound(__sk, node, list) \
364 hlist_for_each_entry(__sk, node, list, sk_bind_node)
377 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
378 SOCK_DBG, /* %SO_DEBUG setting */
379 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
380 SOCK_NO_LARGESEND, /* whether to sent large segments or not */
381 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
382 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
385 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
387 __set_bit(flag, &sk->sk_flags);
390 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
392 __clear_bit(flag, &sk->sk_flags);
395 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
397 return test_bit(flag, &sk->sk_flags);
400 static inline void sk_acceptq_removed(struct sock *sk)
402 sk->sk_ack_backlog--;
405 static inline void sk_acceptq_added(struct sock *sk)
407 sk->sk_ack_backlog++;
410 static inline int sk_acceptq_is_full(struct sock *sk)
412 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
416 * Compute minimal free write space needed to queue new packets.
418 static inline int sk_stream_min_wspace(struct sock *sk)
420 return sk->sk_wmem_queued / 2;
423 static inline int sk_stream_wspace(struct sock *sk)
425 return sk->sk_sndbuf - sk->sk_wmem_queued;
428 extern void sk_stream_write_space(struct sock *sk);
430 static inline int sk_stream_memory_free(struct sock *sk)
432 return sk->sk_wmem_queued < sk->sk_sndbuf;
435 extern void sk_stream_rfree(struct sk_buff *skb);
437 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
440 skb->destructor = sk_stream_rfree;
441 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
442 sk->sk_forward_alloc -= skb->truesize;
445 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
447 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
448 sk->sk_wmem_queued -= skb->truesize;
449 sk->sk_forward_alloc += skb->truesize;
453 /* The per-socket spinlock must be held here. */
454 #define sk_add_backlog(__sk, __skb) \
455 do { if (!(__sk)->sk_backlog.tail) { \
456 (__sk)->sk_backlog.head = \
457 (__sk)->sk_backlog.tail = (__skb); \
459 ((__sk)->sk_backlog.tail)->next = (__skb); \
460 (__sk)->sk_backlog.tail = (__skb); \
462 (__skb)->next = NULL; \
465 #define sk_wait_event(__sk, __timeo, __condition) \
467 release_sock(__sk); \
470 *(__timeo) = schedule_timeout(*(__timeo)); \
477 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
478 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
479 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
480 extern int sk_stream_error(struct sock *sk, int flags, int err);
481 extern void sk_stream_kill_queues(struct sock *sk);
483 extern int sk_wait_data(struct sock *sk, long *timeo);
485 /* Networking protocol blocks we attach to sockets.
486 * socket layer -> transport layer interface
487 * transport -> network interface is defined by struct inet_proto
490 void (*close)(struct sock *sk,
492 int (*connect)(struct sock *sk,
493 struct sockaddr *uaddr,
495 int (*disconnect)(struct sock *sk, int flags);
497 struct sock * (*accept) (struct sock *sk, int flags, int *err);
499 int (*ioctl)(struct sock *sk, int cmd,
501 int (*init)(struct sock *sk);
502 int (*destroy)(struct sock *sk);
503 void (*shutdown)(struct sock *sk, int how);
504 int (*setsockopt)(struct sock *sk, int level,
505 int optname, char __user *optval,
507 int (*getsockopt)(struct sock *sk, int level,
508 int optname, char __user *optval,
510 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
511 struct msghdr *msg, size_t len);
512 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
514 size_t len, int noblock, int flags,
516 int (*sendpage)(struct sock *sk, struct page *page,
517 int offset, size_t size, int flags);
518 int (*bind)(struct sock *sk,
519 struct sockaddr *uaddr, int addr_len);
521 int (*backlog_rcv) (struct sock *sk,
522 struct sk_buff *skb);
524 /* Keeping track of sk's, looking them up, and port selection methods. */
525 void (*hash)(struct sock *sk);
526 void (*unhash)(struct sock *sk);
527 int (*get_port)(struct sock *sk, unsigned short snum);
529 /* Memory pressure */
530 void (*enter_memory_pressure)(void);
531 atomic_t *memory_allocated; /* Current allocated memory. */
532 atomic_t *sockets_allocated; /* Current number of sockets. */
534 * Pressure flag: try to collapse.
535 * Technical note: it is used by multiple contexts non atomically.
536 * All the sk_stream_mem_schedule() is of this nature: accounting
537 * is strict, actions are advisory and have some latency.
539 int *memory_pressure;
546 unsigned int obj_size;
548 struct module *owner;
552 struct list_head node;
556 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
560 extern int proto_register(struct proto *prot, int alloc_slab);
561 extern void proto_unregister(struct proto *prot);
563 /* Called with local bh disabled */
564 static __inline__ void sock_prot_inc_use(struct proto *prot)
566 prot->stats[smp_processor_id()].inuse++;
569 static __inline__ void sock_prot_dec_use(struct proto *prot)
571 prot->stats[smp_processor_id()].inuse--;
574 /* About 10 seconds */
575 #define SOCK_DESTROY_TIME (10*HZ)
577 /* Sockets 0-1023 can't be bound to unless you are superuser */
578 #define PROT_SOCK 1024
580 #define SHUTDOWN_MASK 3
581 #define RCV_SHUTDOWN 1
582 #define SEND_SHUTDOWN 2
584 #define SOCK_SNDBUF_LOCK 1
585 #define SOCK_RCVBUF_LOCK 2
586 #define SOCK_BINDADDR_LOCK 4
587 #define SOCK_BINDPORT_LOCK 8
589 /* sock_iocb: used to kick off async processing of socket ios */
591 struct list_head list;
597 struct scm_cookie *scm;
598 struct msghdr *msg, async_msg;
599 struct iovec async_iov;
603 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
605 return (struct sock_iocb *)iocb->private;
608 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
613 struct socket_alloc {
614 struct socket socket;
615 struct inode vfs_inode;
618 static inline struct socket *SOCKET_I(struct inode *inode)
620 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
623 static inline struct inode *SOCK_INODE(struct socket *socket)
625 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
628 extern void __sk_stream_mem_reclaim(struct sock *sk);
629 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
631 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
633 static inline int sk_stream_pages(int amt)
635 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
638 static inline void sk_stream_mem_reclaim(struct sock *sk)
640 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
641 __sk_stream_mem_reclaim(sk);
644 static inline void sk_stream_writequeue_purge(struct sock *sk)
648 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
649 sk_stream_free_skb(sk, skb);
650 sk_stream_mem_reclaim(sk);
653 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
655 return (int)skb->truesize <= sk->sk_forward_alloc ||
656 sk_stream_mem_schedule(sk, skb->truesize, 1);
659 /* Used by processes to "lock" a socket state, so that
660 * interrupts and bottom half handlers won't change it
661 * from under us. It essentially blocks any incoming
662 * packets, so that we won't get any new data or any
663 * packets that change the state of the socket.
665 * While locked, BH processing will add new packets to
666 * the backlog queue. This queue is processed by the
667 * owner of the socket lock right before it is released.
669 * Since ~2.3.5 it is also exclusive sleep lock serializing
670 * accesses from user process context.
672 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
674 extern void FASTCALL(lock_sock(struct sock *sk));
675 extern void FASTCALL(release_sock(struct sock *sk));
677 /* BH context may only use the following locking interface. */
678 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
679 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
681 extern struct sock *sk_alloc(int family, int priority,
682 struct proto *prot, int zero_it);
683 extern void sk_free(struct sock *sk);
685 extern struct sk_buff *sock_wmalloc(struct sock *sk,
686 unsigned long size, int force,
688 extern struct sk_buff *sock_rmalloc(struct sock *sk,
689 unsigned long size, int force,
691 extern void sock_wfree(struct sk_buff *skb);
692 extern void sock_rfree(struct sk_buff *skb);
694 extern int sock_setsockopt(struct socket *sock, int level,
695 int op, char __user *optval,
698 extern int sock_getsockopt(struct socket *sock, int level,
699 int op, char __user *optval,
701 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
705 extern void *sock_kmalloc(struct sock *sk, int size, int priority);
706 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
707 extern void sk_send_sigurg(struct sock *sk);
710 * Functions to fill in entries in struct proto_ops when a protocol
711 * does not implement a particular function.
713 extern int sock_no_bind(struct socket *,
714 struct sockaddr *, int);
715 extern int sock_no_connect(struct socket *,
716 struct sockaddr *, int, int);
717 extern int sock_no_socketpair(struct socket *,
719 extern int sock_no_accept(struct socket *,
720 struct socket *, int);
721 extern int sock_no_getname(struct socket *,
722 struct sockaddr *, int *, int);
723 extern unsigned int sock_no_poll(struct file *, struct socket *,
724 struct poll_table_struct *);
725 extern int sock_no_ioctl(struct socket *, unsigned int,
727 extern int sock_no_listen(struct socket *, int);
728 extern int sock_no_shutdown(struct socket *, int);
729 extern int sock_no_getsockopt(struct socket *, int , int,
730 char __user *, int __user *);
731 extern int sock_no_setsockopt(struct socket *, int, int,
733 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
734 struct msghdr *, size_t);
735 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
736 struct msghdr *, size_t, int);
737 extern int sock_no_mmap(struct file *file,
739 struct vm_area_struct *vma);
740 extern ssize_t sock_no_sendpage(struct socket *sock,
742 int offset, size_t size,
746 * Functions to fill in entries in struct proto_ops when a protocol
747 * uses the inet style.
749 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
750 char __user *optval, int __user *optlen);
751 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
752 struct msghdr *msg, size_t size, int flags);
753 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
754 char __user *optval, int optlen);
756 extern void sk_common_release(struct sock *sk);
759 * Default socket callbacks and setup code
762 /* Initialise core socket variables */
763 extern void sock_init_data(struct socket *sock, struct sock *sk);
766 * sk_filter - run a packet through a socket filter
767 * @sk: sock associated with &sk_buff
768 * @skb: buffer to filter
769 * @needlock: set to 1 if the sock is not locked by caller.
771 * Run the filter code and then cut skb->data to correct size returned by
772 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
773 * than pkt_len we keep whole skb->data. This is the socket level
774 * wrapper to sk_run_filter. It returns 0 if the packet should
775 * be accepted or -EPERM if the packet should be tossed.
779 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
783 err = security_sock_rcv_skb(sk, skb);
788 struct sk_filter *filter;
793 filter = sk->sk_filter;
795 int pkt_len = sk_run_filter(skb, filter->insns,
800 skb_trim(skb, pkt_len);
810 * sk_filter_release: Release a socket filter
812 * @fp: filter to remove
814 * Remove a filter from a socket and release its resources.
817 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
819 unsigned int size = sk_filter_len(fp);
821 atomic_sub(size, &sk->sk_omem_alloc);
823 if (atomic_dec_and_test(&fp->refcnt))
827 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
829 atomic_inc(&fp->refcnt);
830 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
834 * Socket reference counting postulates.
836 * * Each user of socket SHOULD hold a reference count.
837 * * Each access point to socket (an hash table bucket, reference from a list,
838 * running timer, skb in flight MUST hold a reference count.
839 * * When reference count hits 0, it means it will never increase back.
840 * * When reference count hits 0, it means that no references from
841 * outside exist to this socket and current process on current CPU
842 * is last user and may/should destroy this socket.
843 * * sk_free is called from any context: process, BH, IRQ. When
844 * it is called, socket has no references from outside -> sk_free
845 * may release descendant resources allocated by the socket, but
846 * to the time when it is called, socket is NOT referenced by any
847 * hash tables, lists etc.
848 * * Packets, delivered from outside (from network or from another process)
849 * and enqueued on receive/error queues SHOULD NOT grab reference count,
850 * when they sit in queue. Otherwise, packets will leak to hole, when
851 * socket is looked up by one cpu and unhasing is made by another CPU.
852 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
853 * (leak to backlog). Packet socket does all the processing inside
854 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
855 * use separate SMP lock, so that they are prone too.
858 /* Ungrab socket and destroy it, if it was the last reference. */
859 static inline void sock_put(struct sock *sk)
861 if (atomic_dec_and_test(&sk->sk_refcnt))
865 /* Detach socket from process context.
866 * Announce socket dead, detach it from wait queue and inode.
867 * Note that parent inode held reference count on this struct sock,
868 * we do not release it in this function, because protocol
869 * probably wants some additional cleanups or even continuing
870 * to work with this socket (TCP).
872 static inline void sock_orphan(struct sock *sk)
874 write_lock_bh(&sk->sk_callback_lock);
875 sock_set_flag(sk, SOCK_DEAD);
876 sk->sk_socket = NULL;
878 write_unlock_bh(&sk->sk_callback_lock);
881 static inline void sock_graft(struct sock *sk, struct socket *parent)
883 write_lock_bh(&sk->sk_callback_lock);
884 sk->sk_sleep = &parent->wait;
886 sk->sk_socket = parent;
887 write_unlock_bh(&sk->sk_callback_lock);
890 extern int sock_i_uid(struct sock *sk);
891 extern unsigned long sock_i_ino(struct sock *sk);
893 static inline struct dst_entry *
894 __sk_dst_get(struct sock *sk)
896 return sk->sk_dst_cache;
899 static inline struct dst_entry *
900 sk_dst_get(struct sock *sk)
902 struct dst_entry *dst;
904 read_lock(&sk->sk_dst_lock);
905 dst = sk->sk_dst_cache;
908 read_unlock(&sk->sk_dst_lock);
913 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
915 struct dst_entry *old_dst;
917 old_dst = sk->sk_dst_cache;
918 sk->sk_dst_cache = dst;
919 dst_release(old_dst);
923 sk_dst_set(struct sock *sk, struct dst_entry *dst)
925 write_lock(&sk->sk_dst_lock);
926 __sk_dst_set(sk, dst);
927 write_unlock(&sk->sk_dst_lock);
931 __sk_dst_reset(struct sock *sk)
933 struct dst_entry *old_dst;
935 old_dst = sk->sk_dst_cache;
936 sk->sk_dst_cache = NULL;
937 dst_release(old_dst);
941 sk_dst_reset(struct sock *sk)
943 write_lock(&sk->sk_dst_lock);
945 write_unlock(&sk->sk_dst_lock);
948 static inline struct dst_entry *
949 __sk_dst_check(struct sock *sk, u32 cookie)
951 struct dst_entry *dst = sk->sk_dst_cache;
953 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
954 sk->sk_dst_cache = NULL;
962 static inline struct dst_entry *
963 sk_dst_check(struct sock *sk, u32 cookie)
965 struct dst_entry *dst = sk_dst_get(sk);
967 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
976 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
978 sk->sk_wmem_queued += skb->truesize;
979 sk->sk_forward_alloc -= skb->truesize;
982 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
983 struct sk_buff *skb, struct page *page,
986 if (skb->ip_summed == CHECKSUM_NONE) {
988 unsigned int csum = csum_and_copy_from_user(from,
989 page_address(page) + off,
993 skb->csum = csum_block_add(skb->csum, csum, skb->len);
994 } else if (copy_from_user(page_address(page) + off, from, copy))
998 skb->data_len += copy;
999 skb->truesize += copy;
1000 sk->sk_wmem_queued += copy;
1001 sk->sk_forward_alloc -= copy;
1006 * Queue a received datagram if it will fit. Stream and sequenced
1007 * protocols can't normally use this as they need to fit buffers in
1008 * and play with them.
1010 * Inlined as it's very short and called for pretty much every
1011 * packet ever received.
1014 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1018 skb->destructor = sock_wfree;
1019 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1022 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1025 skb->destructor = sock_rfree;
1026 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1029 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1030 unsigned long expires);
1032 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1034 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1039 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1040 number of warnings when compiling with -W --ANK
1042 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1043 (unsigned)sk->sk_rcvbuf) {
1048 /* It would be deadlock, if sock_queue_rcv_skb is used
1049 with socket lock! We assume that users of this
1050 function are lock free.
1052 err = sk_filter(sk, skb, 1);
1057 skb_set_owner_r(skb, sk);
1059 /* Cache the SKB length before we tack it onto the receive
1060 * queue. Once it is added it no longer belongs to us and
1061 * may be freed by other threads of control pulling packets
1066 skb_queue_tail(&sk->sk_receive_queue, skb);
1068 if (!sock_flag(sk, SOCK_DEAD))
1069 sk->sk_data_ready(sk, skb_len);
1074 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1076 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1077 number of warnings when compiling with -W --ANK
1079 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1080 (unsigned)sk->sk_rcvbuf)
1082 skb_set_owner_r(skb, sk);
1083 skb_queue_tail(&sk->sk_error_queue, skb);
1084 if (!sock_flag(sk, SOCK_DEAD))
1085 sk->sk_data_ready(sk, skb->len);
1090 * Recover an error report and clear atomically
1093 static inline int sock_error(struct sock *sk)
1095 int err = xchg(&sk->sk_err, 0);
1099 static inline unsigned long sock_wspace(struct sock *sk)
1103 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1104 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1111 static inline void sk_wake_async(struct sock *sk, int how, int band)
1113 if (sk->sk_socket && sk->sk_socket->fasync_list)
1114 sock_wake_async(sk->sk_socket, how, band);
1117 #define SOCK_MIN_SNDBUF 2048
1118 #define SOCK_MIN_RCVBUF 256
1120 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1122 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1123 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1124 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1128 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1129 int size, int mem, int gfp)
1131 struct sk_buff *skb = alloc_skb(size + sk->sk_prot->max_header, gfp);
1134 skb->truesize += mem;
1135 if (sk->sk_forward_alloc >= (int)skb->truesize ||
1136 sk_stream_mem_schedule(sk, skb->truesize, 0)) {
1137 skb_reserve(skb, sk->sk_prot->max_header);
1142 sk->sk_prot->enter_memory_pressure();
1143 sk_stream_moderate_sndbuf(sk);
1148 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1151 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1154 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1156 struct page *page = NULL;
1158 if (sk->sk_forward_alloc >= (int)PAGE_SIZE ||
1159 sk_stream_mem_schedule(sk, PAGE_SIZE, 0))
1160 page = alloc_pages(sk->sk_allocation, 0);
1162 sk->sk_prot->enter_memory_pressure();
1163 sk_stream_moderate_sndbuf(sk);
1168 #define sk_stream_for_retrans_queue(skb, sk) \
1169 for (skb = (sk)->sk_write_queue.next; \
1170 (skb != (sk)->sk_send_head) && \
1171 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1175 * Default write policy as shown to user space via poll/select/SIGIO
1177 static inline int sock_writeable(const struct sock *sk)
1179 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1182 static inline int gfp_any(void)
1184 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1187 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1189 return noblock ? 0 : sk->sk_rcvtimeo;
1192 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1194 return noblock ? 0 : sk->sk_sndtimeo;
1197 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1199 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1202 /* Alas, with timeout socket operations are not restartable.
1203 * Compare this to poll().
1205 static inline int sock_intr_errno(long timeo)
1207 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1210 static __inline__ void
1211 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1213 struct timeval *stamp = &skb->stamp;
1214 if (sock_flag(sk, SOCK_RCVTSTAMP)) {
1215 /* Race occurred between timestamp enabling and packet
1216 receiving. Fill in the current time for now. */
1217 if (stamp->tv_sec == 0)
1218 do_gettimeofday(stamp);
1219 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1222 sk->sk_stamp = *stamp;
1226 * sk_eat_skb - Release a skb if it is no longer needed
1227 * @sk: socket to eat this skb from
1228 * @skb: socket buffer to eat
1230 * This routine must be called with interrupts disabled or with the socket
1231 * locked so that the sk_buff queue operation is ok.
1233 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1235 __skb_unlink(skb, &sk->sk_receive_queue);
1239 extern void sock_enable_timestamp(struct sock *sk);
1240 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1243 * Enable debug/info messages
1247 #define NETDEBUG(x) do { } while (0)
1248 #define LIMIT_NETDEBUG(x) do {} while(0)
1250 #define NETDEBUG(x) do { x; } while (0)
1251 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1255 * Macros for sleeping on a socket. Use them like this:
1257 * SOCK_SLEEP_PRE(sk)
1260 * SOCK_SLEEP_POST(sk)
1262 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1263 * and when the last use of them in DECnet has gone, I'm intending to
1267 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1268 DECLARE_WAITQUEUE(wait, tsk); \
1269 tsk->state = TASK_INTERRUPTIBLE; \
1270 add_wait_queue((sk)->sk_sleep, &wait); \
1273 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1274 remove_wait_queue((sk)->sk_sleep, &wait); \
1278 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1281 sock_set_flag(sk, bit);
1283 sock_reset_flag(sk, bit);
1286 extern __u32 sysctl_wmem_max;
1287 extern __u32 sysctl_rmem_max;
1290 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1292 static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
1298 #endif /* _SOCK_H */