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>
53 #include <linux/vinline.h>
55 #include <asm/atomic.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 sk->sk_debug debugging facility. */
65 #define SOCK_DEBUGGING
67 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && ((sk)->sk_debug)) \
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_zapped - ax25 & ipx means !linked
123 * @sk_shutdown - mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
124 * @sk_use_write_queue - wheter to call sk->sk_write_space in sock_wfree
125 * @sk_userlocks - %SO_SNDBUF and %SO_RCVBUF settings
126 * @sk_lock - synchronizer
127 * @sk_rcvbuf - size of receive buffer in bytes
128 * @sk_sleep - sock wait queue
129 * @sk_dst_cache - destination cache
130 * @sk_dst_lock - destination cache lock
131 * @sk_policy - flow policy
132 * @sk_rmem_alloc - receive queue bytes committed
133 * @sk_receive_queue - incoming packets
134 * @sk_wmem_alloc - transmit queue bytes committed
135 * @sk_write_queue - Packet sending queue
136 * @sk_omem_alloc - "o" is "option" or "other"
137 * @sk_wmem_queued - persistent queue size
138 * @sk_forward_alloc - space allocated forward
139 * @sk_allocation - allocation mode
140 * @sk_sndbuf - size of send buffer in bytes
141 * @sk_flags - %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
142 * @sk_no_check - %SO_NO_CHECK setting, wether or not checkup packets
143 * @sk_debug - %SO_DEBUG setting
144 * @sk_rcvtstamp - %SO_TIMESTAMP setting
145 * @sk_no_largesend - whether to sent large segments or not
146 * @sk_route_caps - route capabilities (e.g. %NETIF_F_TSO)
147 * @sk_lingertime - %SO_LINGER l_linger setting
148 * @sk_hashent - hash entry in several tables (e.g. tcp_ehash)
149 * @sk_pair - socket pair (e.g. AF_UNIX/unix_peer)
150 * @sk_backlog - always used with the per-socket spinlock held
151 * @sk_callback_lock - used with the callbacks in the end of this struct
152 * @sk_error_queue - rarely used
153 * @sk_prot - protocol handlers inside a network family
154 * @sk_err - last error
155 * @sk_err_soft - errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
156 * @sk_ack_backlog - current listen backlog
157 * @sk_max_ack_backlog - listen backlog set in listen()
158 * @sk_priority - %SO_PRIORITY setting
159 * @sk_type - socket type (%SOCK_STREAM, etc)
160 * @sk_localroute - route locally only, %SO_DONTROUTE setting
161 * @sk_protocol - which protocol this socket belongs in this network family
162 * @sk_peercred - %SO_PEERCRED setting
163 * @sk_rcvlowat - %SO_RCVLOWAT setting
164 * @sk_rcvtimeo - %SO_RCVTIMEO setting
165 * @sk_sndtimeo - %SO_SNDTIMEO setting
166 * @sk_filter - socket filtering instructions
167 * @sk_protinfo - private area, net family specific, when not using slab
168 * @sk_slab - the slabcache this instance was allocated from
169 * @sk_timer - sock cleanup timer
170 * @sk_stamp - time stamp of last packet received
171 * @sk_socket - Identd and reporting IO signals
172 * @sk_user_data - RPC layer private data
173 * @sk_owner - module that owns this socket
174 * @sk_state_change - callback to indicate change in the state of the sock
175 * @sk_data_ready - callback to indicate there is data to be processed
176 * @sk_write_space - callback to indicate there is bf sending space available
177 * @sk_error_report - callback to indicate errors (e.g. %MSG_ERRQUEUE)
178 * @sk_backlog_rcv - callback to process the backlog
179 * @sk_destruct - called at sock freeing time, i.e. when all refcnt == 0
183 * Now struct tcp_tw_bucket also uses sock_common, so please just
184 * don't add nothing before this first member (__sk_common) --acme
186 struct sock_common __sk_common;
187 #define sk_family __sk_common.skc_family
188 #define sk_state __sk_common.skc_state
189 #define sk_reuse __sk_common.skc_reuse
190 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
191 #define sk_node __sk_common.skc_node
192 #define sk_bind_node __sk_common.skc_bind_node
193 #define sk_refcnt __sk_common.skc_refcnt
194 #define sk_xid __sk_common.skc_xid
195 #define sk_vx_info __sk_common.skc_vx_info
196 #define sk_nid __sk_common.skc_nid
197 #define sk_nx_info __sk_common.skc_nx_info
198 volatile unsigned char sk_zapped;
199 unsigned char sk_shutdown;
200 unsigned char sk_use_write_queue;
201 unsigned char sk_userlocks;
202 socket_lock_t sk_lock;
204 wait_queue_head_t *sk_sleep;
205 struct dst_entry *sk_dst_cache;
206 rwlock_t sk_dst_lock;
207 struct xfrm_policy *sk_policy[2];
208 atomic_t sk_rmem_alloc;
209 struct sk_buff_head sk_receive_queue;
210 atomic_t sk_wmem_alloc;
211 struct sk_buff_head sk_write_queue;
212 atomic_t sk_omem_alloc;
214 int sk_forward_alloc;
215 unsigned int sk_allocation;
217 unsigned long sk_flags;
219 unsigned char sk_debug;
220 unsigned char sk_rcvtstamp;
221 unsigned char sk_no_largesend;
223 unsigned long sk_lingertime;
225 struct sock *sk_pair;
227 * The backlog queue is special, it is always used with
228 * the per-socket spinlock held and requires low latency
229 * access. Therefore we special case it's implementation.
232 struct sk_buff *head;
233 struct sk_buff *tail;
235 rwlock_t sk_callback_lock;
236 struct sk_buff_head sk_error_queue;
237 struct proto *sk_prot;
240 unsigned short sk_ack_backlog;
241 unsigned short sk_max_ack_backlog;
243 unsigned short sk_type;
244 unsigned char sk_localroute;
245 unsigned char sk_protocol;
246 struct ucred sk_peercred;
250 struct sk_filter *sk_filter;
252 kmem_cache_t *sk_slab;
253 struct timer_list sk_timer;
254 struct timeval sk_stamp;
255 struct socket *sk_socket;
257 struct module *sk_owner;
259 void (*sk_state_change)(struct sock *sk);
260 void (*sk_data_ready)(struct sock *sk, int bytes);
261 void (*sk_write_space)(struct sock *sk);
262 void (*sk_error_report)(struct sock *sk);
263 int (*sk_backlog_rcv)(struct sock *sk,
264 struct sk_buff *skb);
265 void (*sk_destruct)(struct sock *sk);
269 * Hashed lists helper routines
271 static inline struct sock *__sk_head(struct hlist_head *head)
273 return hlist_entry(head->first, struct sock, sk_node);
276 static inline struct sock *sk_head(struct hlist_head *head)
278 return hlist_empty(head) ? NULL : __sk_head(head);
281 static inline struct sock *sk_next(struct sock *sk)
283 return sk->sk_node.next ?
284 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
287 static inline int sk_unhashed(struct sock *sk)
289 return hlist_unhashed(&sk->sk_node);
292 static inline int sk_hashed(struct sock *sk)
294 return sk->sk_node.pprev != NULL;
297 static __inline__ void sk_node_init(struct hlist_node *node)
302 static __inline__ void __sk_del_node(struct sock *sk)
304 __hlist_del(&sk->sk_node);
307 static __inline__ int __sk_del_node_init(struct sock *sk)
311 sk_node_init(&sk->sk_node);
317 /* Grab socket reference count. This operation is valid only
318 when sk is ALREADY grabbed f.e. it is found in hash table
319 or a list and the lookup is made under lock preventing hash table
323 static inline void sock_hold(struct sock *sk)
325 atomic_inc(&sk->sk_refcnt);
328 /* Ungrab socket in the context, which assumes that socket refcnt
329 cannot hit zero, f.e. it is true in context of any socketcall.
331 static inline void __sock_put(struct sock *sk)
333 atomic_dec(&sk->sk_refcnt);
336 static __inline__ int sk_del_node_init(struct sock *sk)
338 int rc = __sk_del_node_init(sk);
341 /* paranoid for a while -acme */
342 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
348 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
350 hlist_add_head(&sk->sk_node, list);
353 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
356 __sk_add_node(sk, list);
359 static __inline__ void __sk_del_bind_node(struct sock *sk)
361 __hlist_del(&sk->sk_bind_node);
364 static __inline__ void sk_add_bind_node(struct sock *sk,
365 struct hlist_head *list)
367 hlist_add_head(&sk->sk_bind_node, list);
370 #define sk_for_each(__sk, node, list) \
371 hlist_for_each_entry(__sk, node, list, sk_node)
372 #define sk_for_each_from(__sk, node) \
373 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
374 hlist_for_each_entry_from(__sk, node, sk_node)
375 #define sk_for_each_continue(__sk, node) \
376 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
377 hlist_for_each_entry_continue(__sk, node, sk_node)
378 #define sk_for_each_safe(__sk, node, tmp, list) \
379 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
380 #define sk_for_each_bound(__sk, node, list) \
381 hlist_for_each_entry(__sk, node, list, sk_bind_node)
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 /* The per-socket spinlock must be held here. */
411 #define sk_add_backlog(__sk, __skb) \
412 do { if (!(__sk)->sk_backlog.tail) { \
413 (__sk)->sk_backlog.head = \
414 (__sk)->sk_backlog.tail = (__skb); \
416 ((__sk)->sk_backlog.tail)->next = (__skb); \
417 (__sk)->sk_backlog.tail = (__skb); \
419 (__skb)->next = NULL; \
422 /* IP protocol blocks we attach to sockets.
423 * socket layer -> transport layer interface
424 * transport -> network interface is defined by struct inet_proto
427 void (*close)(struct sock *sk,
429 int (*connect)(struct sock *sk,
430 struct sockaddr *uaddr,
432 int (*disconnect)(struct sock *sk, int flags);
434 struct sock * (*accept) (struct sock *sk, int flags, int *err);
436 int (*ioctl)(struct sock *sk, int cmd,
438 int (*init)(struct sock *sk);
439 int (*destroy)(struct sock *sk);
440 void (*shutdown)(struct sock *sk, int how);
441 int (*setsockopt)(struct sock *sk, int level,
442 int optname, char *optval, int optlen);
443 int (*getsockopt)(struct sock *sk, int level,
444 int optname, char *optval,
446 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
447 struct msghdr *msg, size_t len);
448 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
450 size_t len, int noblock, int flags,
452 int (*sendpage)(struct sock *sk, struct page *page,
453 int offset, size_t size, int flags);
454 int (*bind)(struct sock *sk,
455 struct sockaddr *uaddr, int addr_len);
457 int (*backlog_rcv) (struct sock *sk,
458 struct sk_buff *skb);
460 /* Keeping track of sk's, looking them up, and port selection methods. */
461 void (*hash)(struct sock *sk);
462 void (*unhash)(struct sock *sk);
463 int (*get_port)(struct sock *sk, unsigned short snum);
469 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
473 static __inline__ void sk_set_owner(struct sock *sk, struct module *owner)
476 * One should use sk_set_owner just once, after struct sock creation,
477 * be it shortly after sk_alloc or after a function that returns a new
478 * struct sock (and that down the call chain called sk_alloc), e.g. the
479 * IPv4 and IPv6 modules share tcp_create_openreq_child, so if
480 * tcp_create_openreq_child called sk_set_owner IPv6 would have to
481 * change the ownership of this struct sock, with one not needed
482 * transient sk_set_owner call.
484 BUG_ON(sk->sk_owner != NULL);
486 sk->sk_owner = owner;
490 /* Called with local bh disabled */
491 static __inline__ void sock_prot_inc_use(struct proto *prot)
493 prot->stats[smp_processor_id()].inuse++;
496 static __inline__ void sock_prot_dec_use(struct proto *prot)
498 prot->stats[smp_processor_id()].inuse--;
501 /* About 10 seconds */
502 #define SOCK_DESTROY_TIME (10*HZ)
504 /* Sockets 0-1023 can't be bound to unless you are superuser */
505 #define PROT_SOCK 1024
507 #define SHUTDOWN_MASK 3
508 #define RCV_SHUTDOWN 1
509 #define SEND_SHUTDOWN 2
511 #define SOCK_SNDBUF_LOCK 1
512 #define SOCK_RCVBUF_LOCK 2
513 #define SOCK_BINDADDR_LOCK 4
514 #define SOCK_BINDPORT_LOCK 8
516 /* sock_iocb: used to kick off async processing of socket ios */
518 struct list_head list;
524 struct scm_cookie *scm;
525 struct msghdr *msg, async_msg;
526 struct iovec async_iov;
529 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
531 BUG_ON(sizeof(struct sock_iocb) > KIOCB_PRIVATE_SIZE);
532 return (struct sock_iocb *)iocb->private;
535 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
537 return container_of((void *)si, struct kiocb, private);
540 struct socket_alloc {
541 struct socket socket;
542 struct inode vfs_inode;
545 static inline struct socket *SOCKET_I(struct inode *inode)
547 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
550 static inline struct inode *SOCK_INODE(struct socket *socket)
552 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
555 /* Used by processes to "lock" a socket state, so that
556 * interrupts and bottom half handlers won't change it
557 * from under us. It essentially blocks any incoming
558 * packets, so that we won't get any new data or any
559 * packets that change the state of the socket.
561 * While locked, BH processing will add new packets to
562 * the backlog queue. This queue is processed by the
563 * owner of the socket lock right before it is released.
565 * Since ~2.3.5 it is also exclusive sleep lock serializing
566 * accesses from user process context.
568 extern void __lock_sock(struct sock *sk);
569 extern void __release_sock(struct sock *sk);
570 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
572 extern void FASTCALL(lock_sock(struct sock *sk));
573 extern void FASTCALL(release_sock(struct sock *sk));
575 /* BH context may only use the following locking interface. */
576 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
577 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
579 extern struct sock * sk_alloc(int family, int priority, int zero_it,
581 extern void sk_free(struct sock *sk);
583 extern struct sk_buff *sock_wmalloc(struct sock *sk,
584 unsigned long size, int force,
586 extern struct sk_buff *sock_rmalloc(struct sock *sk,
587 unsigned long size, int force,
589 extern void sock_wfree(struct sk_buff *skb);
590 extern void sock_rfree(struct sk_buff *skb);
592 extern int sock_setsockopt(struct socket *sock, int level,
593 int op, char __user *optval,
596 extern int sock_getsockopt(struct socket *sock, int level,
597 int op, char __user *optval,
599 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
603 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
604 unsigned long header_len,
605 unsigned long data_len,
608 extern void *sock_kmalloc(struct sock *sk, int size, int priority);
609 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
610 extern void sk_send_sigurg(struct sock *sk);
613 * Functions to fill in entries in struct proto_ops when a protocol
614 * does not implement a particular function.
616 extern int sock_no_release(struct socket *);
617 extern int sock_no_bind(struct socket *,
618 struct sockaddr *, int);
619 extern int sock_no_connect(struct socket *,
620 struct sockaddr *, int, int);
621 extern int sock_no_socketpair(struct socket *,
623 extern int sock_no_accept(struct socket *,
624 struct socket *, int);
625 extern int sock_no_getname(struct socket *,
626 struct sockaddr *, int *, int);
627 extern unsigned int sock_no_poll(struct file *, struct socket *,
628 struct poll_table_struct *);
629 extern int sock_no_ioctl(struct socket *, unsigned int,
631 extern int sock_no_listen(struct socket *, int);
632 extern int sock_no_shutdown(struct socket *, int);
633 extern int sock_no_getsockopt(struct socket *, int , int,
634 char __user *, int __user *);
635 extern int sock_no_setsockopt(struct socket *, int, int,
637 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
638 struct msghdr *, size_t);
639 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
640 struct msghdr *, size_t, int);
641 extern int sock_no_mmap(struct file *file,
643 struct vm_area_struct *vma);
644 extern ssize_t sock_no_sendpage(struct socket *sock,
646 int offset, size_t size,
650 * Default socket callbacks and setup code
653 extern void sock_def_destruct(struct sock *);
655 /* Initialise core socket variables */
656 extern void sock_init_data(struct socket *sock, struct sock *sk);
659 * sk_filter - run a packet through a socket filter
660 * @sk: sock associated with &sk_buff
661 * @skb: buffer to filter
662 * @needlock: set to 1 if the sock is not locked by caller.
664 * Run the filter code and then cut skb->data to correct size returned by
665 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
666 * than pkt_len we keep whole skb->data. This is the socket level
667 * wrapper to sk_run_filter. It returns 0 if the packet should
668 * be accepted or -EPERM if the packet should be tossed.
672 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
676 err = security_sock_rcv_skb(sk, skb);
681 struct sk_filter *filter;
686 filter = sk->sk_filter;
688 int pkt_len = sk_run_filter(skb, filter->insns,
693 skb_trim(skb, pkt_len);
703 * sk_filter_release: Release a socket filter
705 * @fp: filter to remove
707 * Remove a filter from a socket and release its resources.
710 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
712 unsigned int size = sk_filter_len(fp);
714 atomic_sub(size, &sk->sk_omem_alloc);
716 if (atomic_dec_and_test(&fp->refcnt))
720 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
722 atomic_inc(&fp->refcnt);
723 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
727 * Socket reference counting postulates.
729 * * Each user of socket SHOULD hold a reference count.
730 * * Each access point to socket (an hash table bucket, reference from a list,
731 * running timer, skb in flight MUST hold a reference count.
732 * * When reference count hits 0, it means it will never increase back.
733 * * When reference count hits 0, it means that no references from
734 * outside exist to this socket and current process on current CPU
735 * is last user and may/should destroy this socket.
736 * * sk_free is called from any context: process, BH, IRQ. When
737 * it is called, socket has no references from outside -> sk_free
738 * may release descendant resources allocated by the socket, but
739 * to the time when it is called, socket is NOT referenced by any
740 * hash tables, lists etc.
741 * * Packets, delivered from outside (from network or from another process)
742 * and enqueued on receive/error queues SHOULD NOT grab reference count,
743 * when they sit in queue. Otherwise, packets will leak to hole, when
744 * socket is looked up by one cpu and unhasing is made by another CPU.
745 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
746 * (leak to backlog). Packet socket does all the processing inside
747 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
748 * use separate SMP lock, so that they are prone too.
751 /* Ungrab socket and destroy it, if it was the last reference. */
752 static inline void sock_put(struct sock *sk)
754 if (atomic_dec_and_test(&sk->sk_refcnt))
758 /* Detach socket from process context.
759 * Announce socket dead, detach it from wait queue and inode.
760 * Note that parent inode held reference count on this struct sock,
761 * we do not release it in this function, because protocol
762 * probably wants some additional cleanups or even continuing
763 * to work with this socket (TCP).
765 static inline void sock_orphan(struct sock *sk)
767 write_lock_bh(&sk->sk_callback_lock);
768 sock_set_flag(sk, SOCK_DEAD);
769 sk->sk_socket = NULL;
771 write_unlock_bh(&sk->sk_callback_lock);
774 static inline void sock_graft(struct sock *sk, struct socket *parent)
776 write_lock_bh(&sk->sk_callback_lock);
777 sk->sk_sleep = &parent->wait;
779 sk->sk_socket = parent;
780 write_unlock_bh(&sk->sk_callback_lock);
783 static inline int sock_i_uid(struct sock *sk)
787 read_lock(&sk->sk_callback_lock);
788 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
789 read_unlock(&sk->sk_callback_lock);
793 static inline unsigned long sock_i_ino(struct sock *sk)
797 read_lock(&sk->sk_callback_lock);
798 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
799 read_unlock(&sk->sk_callback_lock);
803 static inline struct dst_entry *
804 __sk_dst_get(struct sock *sk)
806 return sk->sk_dst_cache;
809 static inline struct dst_entry *
810 sk_dst_get(struct sock *sk)
812 struct dst_entry *dst;
814 read_lock(&sk->sk_dst_lock);
815 dst = sk->sk_dst_cache;
818 read_unlock(&sk->sk_dst_lock);
823 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
825 struct dst_entry *old_dst;
827 old_dst = sk->sk_dst_cache;
828 sk->sk_dst_cache = dst;
829 dst_release(old_dst);
833 sk_dst_set(struct sock *sk, struct dst_entry *dst)
835 write_lock(&sk->sk_dst_lock);
836 __sk_dst_set(sk, dst);
837 write_unlock(&sk->sk_dst_lock);
841 __sk_dst_reset(struct sock *sk)
843 struct dst_entry *old_dst;
845 old_dst = sk->sk_dst_cache;
846 sk->sk_dst_cache = NULL;
847 dst_release(old_dst);
851 sk_dst_reset(struct sock *sk)
853 write_lock(&sk->sk_dst_lock);
855 write_unlock(&sk->sk_dst_lock);
858 static inline struct dst_entry *
859 __sk_dst_check(struct sock *sk, u32 cookie)
861 struct dst_entry *dst = sk->sk_dst_cache;
863 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
864 sk->sk_dst_cache = NULL;
871 static inline struct dst_entry *
872 sk_dst_check(struct sock *sk, u32 cookie)
874 struct dst_entry *dst = sk_dst_get(sk);
876 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
886 * Queue a received datagram if it will fit. Stream and sequenced
887 * protocols can't normally use this as they need to fit buffers in
888 * and play with them.
890 * Inlined as it's very short and called for pretty much every
891 * packet ever received.
894 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
898 skb->destructor = sock_wfree;
899 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
902 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
905 skb->destructor = sock_rfree;
906 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
909 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
914 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
915 number of warnings when compiling with -W --ANK
917 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
918 (unsigned)sk->sk_rcvbuf) {
923 /* It would be deadlock, if sock_queue_rcv_skb is used
924 with socket lock! We assume that users of this
925 function are lock free.
927 err = sk_filter(sk, skb, 1);
932 skb_set_owner_r(skb, sk);
934 /* Cache the SKB length before we tack it onto the receive
935 * queue. Once it is added it no longer belongs to us and
936 * may be freed by other threads of control pulling packets
941 skb_queue_tail(&sk->sk_receive_queue, skb);
943 if (!sock_flag(sk, SOCK_DEAD))
944 sk->sk_data_ready(sk, skb_len);
949 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
951 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
952 number of warnings when compiling with -W --ANK
954 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
955 (unsigned)sk->sk_rcvbuf)
957 skb_set_owner_r(skb, sk);
958 skb_queue_tail(&sk->sk_error_queue, skb);
959 if (!sock_flag(sk, SOCK_DEAD))
960 sk->sk_data_ready(sk, skb->len);
965 * Recover an error report and clear atomically
968 static inline int sock_error(struct sock *sk)
970 int err = xchg(&sk->sk_err, 0);
974 static inline unsigned long sock_wspace(struct sock *sk)
978 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
979 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
986 static inline void sk_wake_async(struct sock *sk, int how, int band)
988 if (sk->sk_socket && sk->sk_socket->fasync_list)
989 sock_wake_async(sk->sk_socket, how, band);
992 #define SOCK_MIN_SNDBUF 2048
993 #define SOCK_MIN_RCVBUF 256
996 * Default write policy as shown to user space via poll/select/SIGIO
998 static inline int sock_writeable(struct sock *sk)
1000 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1003 static inline int gfp_any(void)
1005 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1008 static inline long sock_rcvtimeo(struct sock *sk, int noblock)
1010 return noblock ? 0 : sk->sk_rcvtimeo;
1013 static inline long sock_sndtimeo(struct sock *sk, int noblock)
1015 return noblock ? 0 : sk->sk_sndtimeo;
1018 static inline int sock_rcvlowat(struct sock *sk, int waitall, int len)
1020 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1023 /* Alas, with timeout socket operations are not restartable.
1024 * Compare this to poll().
1026 static inline int sock_intr_errno(long timeo)
1028 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1031 static __inline__ void
1032 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1034 struct timeval *stamp = &skb->stamp;
1035 if (sk->sk_rcvtstamp) {
1036 /* Race occurred between timestamp enabling and packet
1037 receiving. Fill in the current time for now. */
1038 if (stamp->tv_sec == 0)
1039 do_gettimeofday(stamp);
1040 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1043 sk->sk_stamp = *stamp;
1046 extern atomic_t netstamp_needed;
1047 extern void sock_enable_timestamp(struct sock *sk);
1048 extern void sock_disable_timestamp(struct sock *sk);
1050 static inline void net_timestamp(struct timeval *stamp)
1052 if (atomic_read(&netstamp_needed))
1053 do_gettimeofday(stamp);
1060 extern int sock_get_timestamp(struct sock *, struct timeval *);
1063 * Enable debug/info messages
1067 #define NETDEBUG(x) do { } while (0)
1068 #define LIMIT_NETDEBUG(x) do {} while(0)
1070 #define NETDEBUG(x) do { x; } while (0)
1071 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1075 * Macros for sleeping on a socket. Use them like this:
1077 * SOCK_SLEEP_PRE(sk)
1080 * SOCK_SLEEP_POST(sk)
1082 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1083 * and when the last use of them in DECnet has gone, I'm intending to
1087 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1088 DECLARE_WAITQUEUE(wait, tsk); \
1089 tsk->state = TASK_INTERRUPTIBLE; \
1090 add_wait_queue((sk)->sk_sleep, &wait); \
1093 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1094 remove_wait_queue((sk)->sk_sleep, &wait); \
1098 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1101 sock_set_flag(sk, bit);
1103 sock_reset_flag(sk, bit);
1106 extern __u32 sysctl_wmem_max;
1107 extern __u32 sysctl_rmem_max;
1109 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1111 #endif /* _SOCK_H */