2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the AF_INET socket handler.
8 * Version: @(#)sock.h 1.0.4 05/13/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
43 #include <linux/kernel.h>
44 #include <linux/list.h>
45 #include <linux/list_nulls.h>
46 #include <linux/timer.h>
47 #include <linux/cache.h>
48 #include <linux/module.h>
49 #include <linux/lockdep.h>
50 #include <linux/netdevice.h>
51 #include <linux/skbuff.h> /* struct sk_buff */
53 #include <linux/security.h>
54 #include <linux/slab.h>
56 #include <linux/filter.h>
57 #include <linux/rculist_nulls.h>
58 #include <linux/poll.h>
60 #include <linux/atomic.h>
62 #include <net/checksum.h>
65 * This structure really needs to be cleaned up.
66 * Most of it is for TCP, and not used by any of
67 * the other protocols.
70 /* Define this to get the SOCK_DBG debugging facility. */
71 #define SOCK_DEBUGGING
73 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
74 printk(KERN_DEBUG msg); } while (0)
76 /* Validate arguments and do nothing */
77 static inline void __attribute__ ((format (printf, 2, 3)))
78 SOCK_DEBUG(struct sock *sk, const char *msg, ...)
83 /* This is the per-socket lock. The spinlock provides a synchronization
84 * between user contexts and software interrupt processing, whereas the
85 * mini-semaphore synchronizes multiple users amongst themselves.
92 * We express the mutex-alike socket_lock semantics
93 * to the lock validator by explicitly managing
94 * the slock as a lock variant (in addition to
97 #ifdef CONFIG_DEBUG_LOCK_ALLOC
98 struct lockdep_map dep_map;
107 * struct sock_common - minimal network layer representation of sockets
108 * @skc_node: main hash linkage for various protocol lookup tables
109 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
110 * @skc_refcnt: reference count
111 * @skc_tx_queue_mapping: tx queue number for this connection
112 * @skc_hash: hash value used with various protocol lookup tables
113 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
114 * @skc_family: network address family
115 * @skc_state: Connection state
116 * @skc_reuse: %SO_REUSEADDR setting
117 * @skc_bound_dev_if: bound device index if != 0
118 * @skc_bind_node: bind hash linkage for various protocol lookup tables
119 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
120 * @skc_prot: protocol handlers inside a network family
121 * @skc_net: reference to the network namespace of this socket
123 * This is the minimal network layer representation of sockets, the header
124 * for struct sock and struct inet_timewait_sock.
128 * first fields are not copied in sock_copy()
131 struct hlist_node skc_node;
132 struct hlist_nulls_node skc_nulls_node;
135 int skc_tx_queue_mapping;
138 unsigned int skc_hash;
139 __u16 skc_u16hashes[2];
141 unsigned short skc_family;
142 volatile unsigned char skc_state;
143 unsigned char skc_reuse;
144 int skc_bound_dev_if;
146 struct hlist_node skc_bind_node;
147 struct hlist_nulls_node skc_portaddr_node;
149 struct proto *skc_prot;
156 * struct sock - network layer representation of sockets
157 * @__sk_common: shared layout with inet_timewait_sock
158 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
159 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
160 * @sk_lock: synchronizer
161 * @sk_rcvbuf: size of receive buffer in bytes
162 * @sk_wq: sock wait queue and async head
163 * @sk_dst_cache: destination cache
164 * @sk_dst_lock: destination cache lock
165 * @sk_policy: flow policy
166 * @sk_rmem_alloc: receive queue bytes committed
167 * @sk_receive_queue: incoming packets
168 * @sk_wmem_alloc: transmit queue bytes committed
169 * @sk_write_queue: Packet sending queue
170 * @sk_async_wait_queue: DMA copied packets
171 * @sk_omem_alloc: "o" is "option" or "other"
172 * @sk_wmem_queued: persistent queue size
173 * @sk_forward_alloc: space allocated forward
174 * @sk_allocation: allocation mode
175 * @sk_sndbuf: size of send buffer in bytes
176 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
177 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
178 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
179 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
180 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
181 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
182 * @sk_gso_max_size: Maximum GSO segment size to build
183 * @sk_lingertime: %SO_LINGER l_linger setting
184 * @sk_backlog: always used with the per-socket spinlock held
185 * @sk_callback_lock: used with the callbacks in the end of this struct
186 * @sk_error_queue: rarely used
187 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
188 * IPV6_ADDRFORM for instance)
189 * @sk_err: last error
190 * @sk_err_soft: errors that don't cause failure but are the cause of a
191 * persistent failure not just 'timed out'
192 * @sk_drops: raw/udp drops counter
193 * @sk_ack_backlog: current listen backlog
194 * @sk_max_ack_backlog: listen backlog set in listen()
195 * @sk_priority: %SO_PRIORITY setting
196 * @sk_type: socket type (%SOCK_STREAM, etc)
197 * @sk_protocol: which protocol this socket belongs in this network family
198 * @sk_peer_pid: &struct pid for this socket's peer
199 * @sk_peer_cred: %SO_PEERCRED setting
200 * @sk_rcvlowat: %SO_RCVLOWAT setting
201 * @sk_rcvtimeo: %SO_RCVTIMEO setting
202 * @sk_sndtimeo: %SO_SNDTIMEO setting
203 * @sk_rxhash: flow hash received from netif layer
204 * @sk_filter: socket filtering instructions
205 * @sk_protinfo: private area, net family specific, when not using slab
206 * @sk_timer: sock cleanup timer
207 * @sk_stamp: time stamp of last packet received
208 * @sk_socket: Identd and reporting IO signals
209 * @sk_user_data: RPC layer private data
210 * @sk_sndmsg_page: cached page for sendmsg
211 * @sk_sndmsg_off: cached offset for sendmsg
212 * @sk_send_head: front of stuff to transmit
213 * @sk_security: used by security modules
214 * @sk_mark: generic packet mark
215 * @sk_classid: this socket's cgroup classid
216 * @sk_write_pending: a write to stream socket waits to start
217 * @sk_state_change: callback to indicate change in the state of the sock
218 * @sk_data_ready: callback to indicate there is data to be processed
219 * @sk_write_space: callback to indicate there is bf sending space available
220 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
221 * @sk_backlog_rcv: callback to process the backlog
222 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
226 * Now struct inet_timewait_sock also uses sock_common, so please just
227 * don't add nothing before this first member (__sk_common) --acme
229 struct sock_common __sk_common;
230 #define sk_node __sk_common.skc_node
231 #define sk_nulls_node __sk_common.skc_nulls_node
232 #define sk_refcnt __sk_common.skc_refcnt
233 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
235 #define sk_copy_start __sk_common.skc_hash
236 #define sk_hash __sk_common.skc_hash
237 #define sk_family __sk_common.skc_family
238 #define sk_state __sk_common.skc_state
239 #define sk_reuse __sk_common.skc_reuse
240 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
241 #define sk_bind_node __sk_common.skc_bind_node
242 #define sk_prot __sk_common.skc_prot
243 #define sk_net __sk_common.skc_net
244 socket_lock_t sk_lock;
245 struct sk_buff_head sk_receive_queue;
247 * The backlog queue is special, it is always used with
248 * the per-socket spinlock held and requires low latency
249 * access. Therefore we special case it's implementation.
250 * Note : rmem_alloc is in this structure to fill a hole
251 * on 64bit arches, not because its logically part of
257 struct sk_buff *head;
258 struct sk_buff *tail;
260 #define sk_rmem_alloc sk_backlog.rmem_alloc
261 int sk_forward_alloc;
268 struct sk_filter __rcu *sk_filter;
269 struct socket_wq *sk_wq;
271 #ifdef CONFIG_NET_DMA
272 struct sk_buff_head sk_async_wait_queue;
276 struct xfrm_policy *sk_policy[2];
278 unsigned long sk_flags;
279 struct dst_entry *sk_dst_cache;
280 spinlock_t sk_dst_lock;
281 atomic_t sk_wmem_alloc;
282 atomic_t sk_omem_alloc;
284 struct sk_buff_head sk_write_queue;
285 kmemcheck_bitfield_begin(flags);
286 unsigned int sk_shutdown : 2,
291 kmemcheck_bitfield_end(flags);
297 unsigned int sk_gso_max_size;
299 unsigned long sk_lingertime;
300 struct sk_buff_head sk_error_queue;
301 struct proto *sk_prot_creator;
302 rwlock_t sk_callback_lock;
305 unsigned short sk_ack_backlog;
306 unsigned short sk_max_ack_backlog;
308 struct pid *sk_peer_pid;
309 const struct cred *sk_peer_cred;
313 struct timer_list sk_timer;
315 struct socket *sk_socket;
317 struct page *sk_sndmsg_page;
318 struct sk_buff *sk_send_head;
320 int sk_write_pending;
321 #ifdef CONFIG_SECURITY
326 void (*sk_state_change)(struct sock *sk);
327 void (*sk_data_ready)(struct sock *sk, int bytes);
328 void (*sk_write_space)(struct sock *sk);
329 void (*sk_error_report)(struct sock *sk);
330 int (*sk_backlog_rcv)(struct sock *sk,
331 struct sk_buff *skb);
332 void (*sk_destruct)(struct sock *sk);
336 * Hashed lists helper routines
338 static inline struct sock *sk_entry(const struct hlist_node *node)
340 return hlist_entry(node, struct sock, sk_node);
343 static inline struct sock *__sk_head(const struct hlist_head *head)
345 return hlist_entry(head->first, struct sock, sk_node);
348 static inline struct sock *sk_head(const struct hlist_head *head)
350 return hlist_empty(head) ? NULL : __sk_head(head);
353 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
355 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
358 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
360 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
363 static inline struct sock *sk_next(const struct sock *sk)
365 return sk->sk_node.next ?
366 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
369 static inline struct sock *sk_nulls_next(const struct sock *sk)
371 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
372 hlist_nulls_entry(sk->sk_nulls_node.next,
373 struct sock, sk_nulls_node) :
377 static inline int sk_unhashed(const struct sock *sk)
379 return hlist_unhashed(&sk->sk_node);
382 static inline int sk_hashed(const struct sock *sk)
384 return !sk_unhashed(sk);
387 static __inline__ void sk_node_init(struct hlist_node *node)
392 static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
397 static __inline__ void __sk_del_node(struct sock *sk)
399 __hlist_del(&sk->sk_node);
402 /* NB: equivalent to hlist_del_init_rcu */
403 static __inline__ int __sk_del_node_init(struct sock *sk)
407 sk_node_init(&sk->sk_node);
413 /* Grab socket reference count. This operation is valid only
414 when sk is ALREADY grabbed f.e. it is found in hash table
415 or a list and the lookup is made under lock preventing hash table
419 static inline void sock_hold(struct sock *sk)
421 atomic_inc(&sk->sk_refcnt);
424 /* Ungrab socket in the context, which assumes that socket refcnt
425 cannot hit zero, f.e. it is true in context of any socketcall.
427 static inline void __sock_put(struct sock *sk)
429 atomic_dec(&sk->sk_refcnt);
432 static __inline__ int sk_del_node_init(struct sock *sk)
434 int rc = __sk_del_node_init(sk);
437 /* paranoid for a while -acme */
438 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
443 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
445 static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
448 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
454 static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
456 int rc = __sk_nulls_del_node_init_rcu(sk);
459 /* paranoid for a while -acme */
460 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
466 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
468 hlist_add_head(&sk->sk_node, list);
471 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
474 __sk_add_node(sk, list);
477 static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
480 hlist_add_head_rcu(&sk->sk_node, list);
483 static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
485 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
488 static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
491 __sk_nulls_add_node_rcu(sk, list);
494 static __inline__ void __sk_del_bind_node(struct sock *sk)
496 __hlist_del(&sk->sk_bind_node);
499 static __inline__ void sk_add_bind_node(struct sock *sk,
500 struct hlist_head *list)
502 hlist_add_head(&sk->sk_bind_node, list);
505 #define sk_for_each(__sk, node, list) \
506 hlist_for_each_entry(__sk, node, list, sk_node)
507 #define sk_for_each_rcu(__sk, node, list) \
508 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
509 #define sk_nulls_for_each(__sk, node, list) \
510 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
511 #define sk_nulls_for_each_rcu(__sk, node, list) \
512 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
513 #define sk_for_each_from(__sk, node) \
514 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
515 hlist_for_each_entry_from(__sk, node, sk_node)
516 #define sk_nulls_for_each_from(__sk, node) \
517 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
518 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
519 #define sk_for_each_continue(__sk, node) \
520 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
521 hlist_for_each_entry_continue(__sk, node, sk_node)
522 #define sk_for_each_safe(__sk, node, tmp, list) \
523 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
524 #define sk_for_each_bound(__sk, node, list) \
525 hlist_for_each_entry(__sk, node, list, sk_bind_node)
538 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
539 SOCK_DBG, /* %SO_DEBUG setting */
540 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
541 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
542 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
543 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
544 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
545 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
546 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
547 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
548 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
549 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
550 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
551 SOCK_FASYNC, /* fasync() active */
555 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
557 nsk->sk_flags = osk->sk_flags;
560 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
562 __set_bit(flag, &sk->sk_flags);
565 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
567 __clear_bit(flag, &sk->sk_flags);
570 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
572 return test_bit(flag, &sk->sk_flags);
575 static inline void sk_acceptq_removed(struct sock *sk)
577 sk->sk_ack_backlog--;
580 static inline void sk_acceptq_added(struct sock *sk)
582 sk->sk_ack_backlog++;
585 static inline int sk_acceptq_is_full(struct sock *sk)
587 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
591 * Compute minimal free write space needed to queue new packets.
593 static inline int sk_stream_min_wspace(struct sock *sk)
595 return sk->sk_wmem_queued >> 1;
598 static inline int sk_stream_wspace(struct sock *sk)
600 return sk->sk_sndbuf - sk->sk_wmem_queued;
603 extern void sk_stream_write_space(struct sock *sk);
605 static inline int sk_stream_memory_free(struct sock *sk)
607 return sk->sk_wmem_queued < sk->sk_sndbuf;
610 /* OOB backlog add */
611 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
613 /* dont let skb dst not refcounted, we are going to leave rcu lock */
616 if (!sk->sk_backlog.tail)
617 sk->sk_backlog.head = skb;
619 sk->sk_backlog.tail->next = skb;
621 sk->sk_backlog.tail = skb;
626 * Take into account size of receive queue and backlog queue
628 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
630 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
632 return qsize + skb->truesize > sk->sk_rcvbuf;
635 /* The per-socket spinlock must be held here. */
636 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
638 if (sk_rcvqueues_full(sk, skb))
641 __sk_add_backlog(sk, skb);
642 sk->sk_backlog.len += skb->truesize;
646 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
648 return sk->sk_backlog_rcv(sk, skb);
651 static inline void sock_rps_record_flow(const struct sock *sk)
654 struct rps_sock_flow_table *sock_flow_table;
657 sock_flow_table = rcu_dereference(rps_sock_flow_table);
658 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
663 static inline void sock_rps_reset_flow(const struct sock *sk)
666 struct rps_sock_flow_table *sock_flow_table;
669 sock_flow_table = rcu_dereference(rps_sock_flow_table);
670 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
675 static inline void sock_rps_save_rxhash(struct sock *sk, u32 rxhash)
678 if (unlikely(sk->sk_rxhash != rxhash)) {
679 sock_rps_reset_flow(sk);
680 sk->sk_rxhash = rxhash;
685 #define sk_wait_event(__sk, __timeo, __condition) \
687 release_sock(__sk); \
688 __rc = __condition; \
690 *(__timeo) = schedule_timeout(*(__timeo)); \
693 __rc = __condition; \
697 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
698 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
699 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
700 extern int sk_stream_error(struct sock *sk, int flags, int err);
701 extern void sk_stream_kill_queues(struct sock *sk);
703 extern int sk_wait_data(struct sock *sk, long *timeo);
705 struct request_sock_ops;
706 struct timewait_sock_ops;
707 struct inet_hashinfo;
710 /* Networking protocol blocks we attach to sockets.
711 * socket layer -> transport layer interface
712 * transport -> network interface is defined by struct inet_proto
715 void (*close)(struct sock *sk,
717 int (*connect)(struct sock *sk,
718 struct sockaddr *uaddr,
720 int (*disconnect)(struct sock *sk, int flags);
722 struct sock * (*accept) (struct sock *sk, int flags, int *err);
724 int (*ioctl)(struct sock *sk, int cmd,
726 int (*init)(struct sock *sk);
727 void (*destroy)(struct sock *sk);
728 void (*shutdown)(struct sock *sk, int how);
729 int (*setsockopt)(struct sock *sk, int level,
730 int optname, char __user *optval,
731 unsigned int optlen);
732 int (*getsockopt)(struct sock *sk, int level,
733 int optname, char __user *optval,
736 int (*compat_setsockopt)(struct sock *sk,
738 int optname, char __user *optval,
739 unsigned int optlen);
740 int (*compat_getsockopt)(struct sock *sk,
742 int optname, char __user *optval,
745 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
746 struct msghdr *msg, size_t len);
747 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
749 size_t len, int noblock, int flags,
751 int (*sendpage)(struct sock *sk, struct page *page,
752 int offset, size_t size, int flags);
753 int (*bind)(struct sock *sk,
754 struct sockaddr *uaddr, int addr_len);
756 int (*backlog_rcv) (struct sock *sk,
757 struct sk_buff *skb);
759 /* Keeping track of sk's, looking them up, and port selection methods. */
760 void (*hash)(struct sock *sk);
761 void (*unhash)(struct sock *sk);
762 void (*rehash)(struct sock *sk);
763 int (*get_port)(struct sock *sk, unsigned short snum);
765 /* Keeping track of sockets in use */
766 #ifdef CONFIG_PROC_FS
767 unsigned int inuse_idx;
770 /* Memory pressure */
771 void (*enter_memory_pressure)(struct sock *sk);
772 atomic_long_t *memory_allocated; /* Current allocated memory. */
773 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
775 * Pressure flag: try to collapse.
776 * Technical note: it is used by multiple contexts non atomically.
777 * All the __sk_mem_schedule() is of this nature: accounting
778 * is strict, actions are advisory and have some latency.
780 int *memory_pressure;
787 struct kmem_cache *slab;
788 unsigned int obj_size;
791 struct percpu_counter *orphan_count;
793 struct request_sock_ops *rsk_prot;
794 struct timewait_sock_ops *twsk_prot;
797 struct inet_hashinfo *hashinfo;
798 struct udp_table *udp_table;
799 struct raw_hashinfo *raw_hash;
802 struct module *owner;
806 struct list_head node;
807 #ifdef SOCK_REFCNT_DEBUG
812 extern int proto_register(struct proto *prot, int alloc_slab);
813 extern void proto_unregister(struct proto *prot);
815 #ifdef SOCK_REFCNT_DEBUG
816 static inline void sk_refcnt_debug_inc(struct sock *sk)
818 atomic_inc(&sk->sk_prot->socks);
821 static inline void sk_refcnt_debug_dec(struct sock *sk)
823 atomic_dec(&sk->sk_prot->socks);
824 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
825 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
828 static inline void sk_refcnt_debug_release(const struct sock *sk)
830 if (atomic_read(&sk->sk_refcnt) != 1)
831 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
832 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
834 #else /* SOCK_REFCNT_DEBUG */
835 #define sk_refcnt_debug_inc(sk) do { } while (0)
836 #define sk_refcnt_debug_dec(sk) do { } while (0)
837 #define sk_refcnt_debug_release(sk) do { } while (0)
838 #endif /* SOCK_REFCNT_DEBUG */
841 #ifdef CONFIG_PROC_FS
842 /* Called with local bh disabled */
843 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
844 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
846 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
853 /* With per-bucket locks this operation is not-atomic, so that
854 * this version is not worse.
856 static inline void __sk_prot_rehash(struct sock *sk)
858 sk->sk_prot->unhash(sk);
859 sk->sk_prot->hash(sk);
862 /* About 10 seconds */
863 #define SOCK_DESTROY_TIME (10*HZ)
865 /* Sockets 0-1023 can't be bound to unless you are superuser */
866 #define PROT_SOCK 1024
868 #define SHUTDOWN_MASK 3
869 #define RCV_SHUTDOWN 1
870 #define SEND_SHUTDOWN 2
872 #define SOCK_SNDBUF_LOCK 1
873 #define SOCK_RCVBUF_LOCK 2
874 #define SOCK_BINDADDR_LOCK 4
875 #define SOCK_BINDPORT_LOCK 8
877 /* sock_iocb: used to kick off async processing of socket ios */
879 struct list_head list;
885 struct scm_cookie *scm;
886 struct msghdr *msg, async_msg;
890 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
892 return (struct sock_iocb *)iocb->private;
895 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
900 struct socket_alloc {
901 struct socket socket;
902 struct inode vfs_inode;
905 static inline struct socket *SOCKET_I(struct inode *inode)
907 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
910 static inline struct inode *SOCK_INODE(struct socket *socket)
912 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
916 * Functions for memory accounting
918 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
919 extern void __sk_mem_reclaim(struct sock *sk);
921 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
922 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
923 #define SK_MEM_SEND 0
924 #define SK_MEM_RECV 1
926 static inline int sk_mem_pages(int amt)
928 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
931 static inline int sk_has_account(struct sock *sk)
933 /* return true if protocol supports memory accounting */
934 return !!sk->sk_prot->memory_allocated;
937 static inline int sk_wmem_schedule(struct sock *sk, int size)
939 if (!sk_has_account(sk))
941 return size <= sk->sk_forward_alloc ||
942 __sk_mem_schedule(sk, size, SK_MEM_SEND);
945 static inline int sk_rmem_schedule(struct sock *sk, int size)
947 if (!sk_has_account(sk))
949 return size <= sk->sk_forward_alloc ||
950 __sk_mem_schedule(sk, size, SK_MEM_RECV);
953 static inline void sk_mem_reclaim(struct sock *sk)
955 if (!sk_has_account(sk))
957 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
958 __sk_mem_reclaim(sk);
961 static inline void sk_mem_reclaim_partial(struct sock *sk)
963 if (!sk_has_account(sk))
965 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
966 __sk_mem_reclaim(sk);
969 static inline void sk_mem_charge(struct sock *sk, int size)
971 if (!sk_has_account(sk))
973 sk->sk_forward_alloc -= size;
976 static inline void sk_mem_uncharge(struct sock *sk, int size)
978 if (!sk_has_account(sk))
980 sk->sk_forward_alloc += size;
983 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
985 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
986 sk->sk_wmem_queued -= skb->truesize;
987 sk_mem_uncharge(sk, skb->truesize);
991 /* Used by processes to "lock" a socket state, so that
992 * interrupts and bottom half handlers won't change it
993 * from under us. It essentially blocks any incoming
994 * packets, so that we won't get any new data or any
995 * packets that change the state of the socket.
997 * While locked, BH processing will add new packets to
998 * the backlog queue. This queue is processed by the
999 * owner of the socket lock right before it is released.
1001 * Since ~2.3.5 it is also exclusive sleep lock serializing
1002 * accesses from user process context.
1004 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1007 * Macro so as to not evaluate some arguments when
1008 * lockdep is not enabled.
1010 * Mark both the sk_lock and the sk_lock.slock as a
1011 * per-address-family lock class.
1013 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1015 sk->sk_lock.owned = 0; \
1016 init_waitqueue_head(&sk->sk_lock.wq); \
1017 spin_lock_init(&(sk)->sk_lock.slock); \
1018 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1019 sizeof((sk)->sk_lock)); \
1020 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1022 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1025 extern void lock_sock_nested(struct sock *sk, int subclass);
1027 static inline void lock_sock(struct sock *sk)
1029 lock_sock_nested(sk, 0);
1032 extern void release_sock(struct sock *sk);
1034 /* BH context may only use the following locking interface. */
1035 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1036 #define bh_lock_sock_nested(__sk) \
1037 spin_lock_nested(&((__sk)->sk_lock.slock), \
1038 SINGLE_DEPTH_NESTING)
1039 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1041 extern bool lock_sock_fast(struct sock *sk);
1043 * unlock_sock_fast - complement of lock_sock_fast
1047 * fast unlock socket for user context.
1048 * If slow mode is on, we call regular release_sock()
1050 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1055 spin_unlock_bh(&sk->sk_lock.slock);
1059 extern struct sock *sk_alloc(struct net *net, int family,
1061 struct proto *prot);
1062 extern void sk_free(struct sock *sk);
1063 extern void sk_release_kernel(struct sock *sk);
1064 extern struct sock *sk_clone(const struct sock *sk,
1065 const gfp_t priority);
1067 extern struct sk_buff *sock_wmalloc(struct sock *sk,
1068 unsigned long size, int force,
1070 extern struct sk_buff *sock_rmalloc(struct sock *sk,
1071 unsigned long size, int force,
1073 extern void sock_wfree(struct sk_buff *skb);
1074 extern void sock_rfree(struct sk_buff *skb);
1076 extern int sock_setsockopt(struct socket *sock, int level,
1077 int op, char __user *optval,
1078 unsigned int optlen);
1080 extern int sock_getsockopt(struct socket *sock, int level,
1081 int op, char __user *optval,
1082 int __user *optlen);
1083 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1087 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1088 unsigned long header_len,
1089 unsigned long data_len,
1092 extern void *sock_kmalloc(struct sock *sk, int size,
1094 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1095 extern void sk_send_sigurg(struct sock *sk);
1097 #ifdef CONFIG_CGROUPS
1098 extern void sock_update_classid(struct sock *sk);
1100 static inline void sock_update_classid(struct sock *sk)
1106 * Functions to fill in entries in struct proto_ops when a protocol
1107 * does not implement a particular function.
1109 extern int sock_no_bind(struct socket *,
1110 struct sockaddr *, int);
1111 extern int sock_no_connect(struct socket *,
1112 struct sockaddr *, int, int);
1113 extern int sock_no_socketpair(struct socket *,
1115 extern int sock_no_accept(struct socket *,
1116 struct socket *, int);
1117 extern int sock_no_getname(struct socket *,
1118 struct sockaddr *, int *, int);
1119 extern unsigned int sock_no_poll(struct file *, struct socket *,
1120 struct poll_table_struct *);
1121 extern int sock_no_ioctl(struct socket *, unsigned int,
1123 extern int sock_no_listen(struct socket *, int);
1124 extern int sock_no_shutdown(struct socket *, int);
1125 extern int sock_no_getsockopt(struct socket *, int , int,
1126 char __user *, int __user *);
1127 extern int sock_no_setsockopt(struct socket *, int, int,
1128 char __user *, unsigned int);
1129 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1130 struct msghdr *, size_t);
1131 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1132 struct msghdr *, size_t, int);
1133 extern int sock_no_mmap(struct file *file,
1134 struct socket *sock,
1135 struct vm_area_struct *vma);
1136 extern ssize_t sock_no_sendpage(struct socket *sock,
1138 int offset, size_t size,
1142 * Functions to fill in entries in struct proto_ops when a protocol
1143 * uses the inet style.
1145 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1146 char __user *optval, int __user *optlen);
1147 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1148 struct msghdr *msg, size_t size, int flags);
1149 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1150 char __user *optval, unsigned int optlen);
1151 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1152 int optname, char __user *optval, int __user *optlen);
1153 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1154 int optname, char __user *optval, unsigned int optlen);
1156 extern void sk_common_release(struct sock *sk);
1159 * Default socket callbacks and setup code
1162 /* Initialise core socket variables */
1163 extern void sock_init_data(struct socket *sock, struct sock *sk);
1166 * sk_filter_release - release a socket filter
1167 * @fp: filter to remove
1169 * Remove a filter from a socket and release its resources.
1172 static inline void sk_filter_release(struct sk_filter *fp)
1174 if (atomic_dec_and_test(&fp->refcnt))
1178 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1180 unsigned int size = sk_filter_len(fp);
1182 atomic_sub(size, &sk->sk_omem_alloc);
1183 sk_filter_release(fp);
1186 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1188 atomic_inc(&fp->refcnt);
1189 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1193 * Socket reference counting postulates.
1195 * * Each user of socket SHOULD hold a reference count.
1196 * * Each access point to socket (an hash table bucket, reference from a list,
1197 * running timer, skb in flight MUST hold a reference count.
1198 * * When reference count hits 0, it means it will never increase back.
1199 * * When reference count hits 0, it means that no references from
1200 * outside exist to this socket and current process on current CPU
1201 * is last user and may/should destroy this socket.
1202 * * sk_free is called from any context: process, BH, IRQ. When
1203 * it is called, socket has no references from outside -> sk_free
1204 * may release descendant resources allocated by the socket, but
1205 * to the time when it is called, socket is NOT referenced by any
1206 * hash tables, lists etc.
1207 * * Packets, delivered from outside (from network or from another process)
1208 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1209 * when they sit in queue. Otherwise, packets will leak to hole, when
1210 * socket is looked up by one cpu and unhasing is made by another CPU.
1211 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1212 * (leak to backlog). Packet socket does all the processing inside
1213 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1214 * use separate SMP lock, so that they are prone too.
1217 /* Ungrab socket and destroy it, if it was the last reference. */
1218 static inline void sock_put(struct sock *sk)
1220 if (atomic_dec_and_test(&sk->sk_refcnt))
1224 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1227 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1229 sk->sk_tx_queue_mapping = tx_queue;
1232 static inline void sk_tx_queue_clear(struct sock *sk)
1234 sk->sk_tx_queue_mapping = -1;
1237 static inline int sk_tx_queue_get(const struct sock *sk)
1239 return sk ? sk->sk_tx_queue_mapping : -1;
1242 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1244 sk_tx_queue_clear(sk);
1245 sk->sk_socket = sock;
1248 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1250 return &sk->sk_wq->wait;
1252 /* Detach socket from process context.
1253 * Announce socket dead, detach it from wait queue and inode.
1254 * Note that parent inode held reference count on this struct sock,
1255 * we do not release it in this function, because protocol
1256 * probably wants some additional cleanups or even continuing
1257 * to work with this socket (TCP).
1259 static inline void sock_orphan(struct sock *sk)
1261 write_lock_bh(&sk->sk_callback_lock);
1262 sock_set_flag(sk, SOCK_DEAD);
1263 sk_set_socket(sk, NULL);
1265 write_unlock_bh(&sk->sk_callback_lock);
1268 static inline void sock_graft(struct sock *sk, struct socket *parent)
1270 write_lock_bh(&sk->sk_callback_lock);
1271 rcu_assign_pointer(sk->sk_wq, parent->wq);
1273 sk_set_socket(sk, parent);
1274 security_sock_graft(sk, parent);
1275 write_unlock_bh(&sk->sk_callback_lock);
1278 extern int sock_i_uid(struct sock *sk);
1279 extern unsigned long sock_i_ino(struct sock *sk);
1281 static inline struct dst_entry *
1282 __sk_dst_get(struct sock *sk)
1284 return rcu_dereference_check(sk->sk_dst_cache, rcu_read_lock_held() ||
1285 sock_owned_by_user(sk) ||
1286 lockdep_is_held(&sk->sk_lock.slock));
1289 static inline struct dst_entry *
1290 sk_dst_get(struct sock *sk)
1292 struct dst_entry *dst;
1295 dst = rcu_dereference(sk->sk_dst_cache);
1302 extern void sk_reset_txq(struct sock *sk);
1304 static inline void dst_negative_advice(struct sock *sk)
1306 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1308 if (dst && dst->ops->negative_advice) {
1309 ndst = dst->ops->negative_advice(dst);
1312 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1319 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1321 struct dst_entry *old_dst;
1323 sk_tx_queue_clear(sk);
1325 * This can be called while sk is owned by the caller only,
1326 * with no state that can be checked in a rcu_dereference_check() cond
1328 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1329 rcu_assign_pointer(sk->sk_dst_cache, dst);
1330 dst_release(old_dst);
1334 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1336 spin_lock(&sk->sk_dst_lock);
1337 __sk_dst_set(sk, dst);
1338 spin_unlock(&sk->sk_dst_lock);
1342 __sk_dst_reset(struct sock *sk)
1344 __sk_dst_set(sk, NULL);
1348 sk_dst_reset(struct sock *sk)
1350 spin_lock(&sk->sk_dst_lock);
1352 spin_unlock(&sk->sk_dst_lock);
1355 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1357 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1359 static inline int sk_can_gso(const struct sock *sk)
1361 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1364 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1366 static inline void sk_nocaps_add(struct sock *sk, int flags)
1368 sk->sk_route_nocaps |= flags;
1369 sk->sk_route_caps &= ~flags;
1372 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1373 struct sk_buff *skb, struct page *page,
1376 if (skb->ip_summed == CHECKSUM_NONE) {
1378 __wsum csum = csum_and_copy_from_user(from,
1379 page_address(page) + off,
1383 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1384 } else if (copy_from_user(page_address(page) + off, from, copy))
1388 skb->data_len += copy;
1389 skb->truesize += copy;
1390 sk->sk_wmem_queued += copy;
1391 sk_mem_charge(sk, copy);
1396 * sk_wmem_alloc_get - returns write allocations
1399 * Returns sk_wmem_alloc minus initial offset of one
1401 static inline int sk_wmem_alloc_get(const struct sock *sk)
1403 return atomic_read(&sk->sk_wmem_alloc) - 1;
1407 * sk_rmem_alloc_get - returns read allocations
1410 * Returns sk_rmem_alloc
1412 static inline int sk_rmem_alloc_get(const struct sock *sk)
1414 return atomic_read(&sk->sk_rmem_alloc);
1418 * sk_has_allocations - check if allocations are outstanding
1421 * Returns true if socket has write or read allocations
1423 static inline int sk_has_allocations(const struct sock *sk)
1425 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1429 * wq_has_sleeper - check if there are any waiting processes
1430 * @wq: struct socket_wq
1432 * Returns true if socket_wq has waiting processes
1434 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1435 * barrier call. They were added due to the race found within the tcp code.
1437 * Consider following tcp code paths:
1441 * sys_select receive packet
1443 * __add_wait_queue update tp->rcv_nxt
1445 * tp->rcv_nxt check sock_def_readable
1447 * schedule rcu_read_lock();
1448 * wq = rcu_dereference(sk->sk_wq);
1449 * if (wq && waitqueue_active(&wq->wait))
1450 * wake_up_interruptible(&wq->wait)
1454 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1455 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1456 * could then endup calling schedule and sleep forever if there are no more
1457 * data on the socket.
1460 static inline bool wq_has_sleeper(struct socket_wq *wq)
1464 * We need to be sure we are in sync with the
1465 * add_wait_queue modifications to the wait queue.
1467 * This memory barrier is paired in the sock_poll_wait.
1470 return wq && waitqueue_active(&wq->wait);
1474 * sock_poll_wait - place memory barrier behind the poll_wait call.
1476 * @wait_address: socket wait queue
1479 * See the comments in the wq_has_sleeper function.
1481 static inline void sock_poll_wait(struct file *filp,
1482 wait_queue_head_t *wait_address, poll_table *p)
1484 if (p && wait_address) {
1485 poll_wait(filp, wait_address, p);
1487 * We need to be sure we are in sync with the
1488 * socket flags modification.
1490 * This memory barrier is paired in the wq_has_sleeper.
1497 * Queue a received datagram if it will fit. Stream and sequenced
1498 * protocols can't normally use this as they need to fit buffers in
1499 * and play with them.
1501 * Inlined as it's very short and called for pretty much every
1502 * packet ever received.
1505 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1509 skb->destructor = sock_wfree;
1511 * We used to take a refcount on sk, but following operation
1512 * is enough to guarantee sk_free() wont free this sock until
1513 * all in-flight packets are completed
1515 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1518 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1522 skb->destructor = sock_rfree;
1523 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1524 sk_mem_charge(sk, skb->truesize);
1527 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1528 unsigned long expires);
1530 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1532 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1534 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1537 * Recover an error report and clear atomically
1540 static inline int sock_error(struct sock *sk)
1543 if (likely(!sk->sk_err))
1545 err = xchg(&sk->sk_err, 0);
1549 static inline unsigned long sock_wspace(struct sock *sk)
1553 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1554 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1561 static inline void sk_wake_async(struct sock *sk, int how, int band)
1563 if (sock_flag(sk, SOCK_FASYNC))
1564 sock_wake_async(sk->sk_socket, how, band);
1567 #define SOCK_MIN_SNDBUF 2048
1569 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1570 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1572 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1574 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1576 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1577 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1578 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1582 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1584 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1586 struct page *page = NULL;
1588 page = alloc_pages(sk->sk_allocation, 0);
1590 sk->sk_prot->enter_memory_pressure(sk);
1591 sk_stream_moderate_sndbuf(sk);
1597 * Default write policy as shown to user space via poll/select/SIGIO
1599 static inline int sock_writeable(const struct sock *sk)
1601 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1604 static inline gfp_t gfp_any(void)
1606 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1609 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1611 return noblock ? 0 : sk->sk_rcvtimeo;
1614 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1616 return noblock ? 0 : sk->sk_sndtimeo;
1619 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1621 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1624 /* Alas, with timeout socket operations are not restartable.
1625 * Compare this to poll().
1627 static inline int sock_intr_errno(long timeo)
1629 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1632 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1633 struct sk_buff *skb);
1635 static __inline__ void
1636 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1638 ktime_t kt = skb->tstamp;
1639 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
1642 * generate control messages if
1643 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1644 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1645 * - software time stamp available and wanted
1646 * (SOCK_TIMESTAMPING_SOFTWARE)
1647 * - hardware time stamps available and wanted
1648 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1649 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1651 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1652 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1653 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1654 (hwtstamps->hwtstamp.tv64 &&
1655 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1656 (hwtstamps->syststamp.tv64 &&
1657 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
1658 __sock_recv_timestamp(msg, sk, skb);
1663 extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1664 struct sk_buff *skb);
1666 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1667 struct sk_buff *skb)
1669 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
1670 (1UL << SOCK_RCVTSTAMP) | \
1671 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
1672 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
1673 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
1674 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
1676 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
1677 __sock_recv_ts_and_drops(msg, sk, skb);
1679 sk->sk_stamp = skb->tstamp;
1683 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1684 * @sk: socket sending this packet
1685 * @tx_flags: filled with instructions for time stamping
1687 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1688 * parameters are invalid.
1690 extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
1693 * sk_eat_skb - Release a skb if it is no longer needed
1694 * @sk: socket to eat this skb from
1695 * @skb: socket buffer to eat
1696 * @copied_early: flag indicating whether DMA operations copied this data early
1698 * This routine must be called with interrupts disabled or with the socket
1699 * locked so that the sk_buff queue operation is ok.
1701 #ifdef CONFIG_NET_DMA
1702 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1704 __skb_unlink(skb, &sk->sk_receive_queue);
1708 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1711 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1713 __skb_unlink(skb, &sk->sk_receive_queue);
1719 struct net *sock_net(const struct sock *sk)
1721 return read_pnet(&sk->sk_net);
1725 void sock_net_set(struct sock *sk, struct net *net)
1727 write_pnet(&sk->sk_net, net);
1731 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1732 * They should not hold a referrence to a namespace in order to allow
1734 * Sockets after sk_change_net should be released using sk_release_kernel
1736 static inline void sk_change_net(struct sock *sk, struct net *net)
1738 put_net(sock_net(sk));
1739 sock_net_set(sk, hold_net(net));
1742 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
1744 if (unlikely(skb->sk)) {
1745 struct sock *sk = skb->sk;
1747 skb->destructor = NULL;
1754 extern void sock_enable_timestamp(struct sock *sk, int flag);
1755 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1756 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1759 * Enable debug/info messages
1761 extern int net_msg_warn;
1762 #define NETDEBUG(fmt, args...) \
1763 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1765 #define LIMIT_NETDEBUG(fmt, args...) \
1766 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1768 extern __u32 sysctl_wmem_max;
1769 extern __u32 sysctl_rmem_max;
1771 extern void sk_init(void);
1773 extern int sysctl_optmem_max;
1775 extern __u32 sysctl_wmem_default;
1776 extern __u32 sysctl_rmem_default;
1778 #endif /* _SOCK_H */