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_daddr: Foreign IPv4 addr
109 * @skc_rcv_saddr: Bound local IPv4 addr
110 * @skc_hash: hash value used with various protocol lookup tables
111 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
112 * @skc_family: network address family
113 * @skc_state: Connection state
114 * @skc_reuse: %SO_REUSEADDR setting
115 * @skc_bound_dev_if: bound device index if != 0
116 * @skc_bind_node: bind hash linkage for various protocol lookup tables
117 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
118 * @skc_prot: protocol handlers inside a network family
119 * @skc_net: reference to the network namespace of this socket
120 * @skc_node: main hash linkage for various protocol lookup tables
121 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
122 * @skc_tx_queue_mapping: tx queue number for this connection
123 * @skc_refcnt: reference count
125 * This is the minimal network layer representation of sockets, the header
126 * for struct sock and struct inet_timewait_sock.
129 /* skc_daddr and skc_rcv_saddr must be grouped :
130 * cf INET_MATCH() and INET_TW_MATCH()
133 __be32 skc_rcv_saddr;
136 unsigned int skc_hash;
137 __u16 skc_u16hashes[2];
139 unsigned short skc_family;
140 volatile unsigned char skc_state;
141 unsigned char skc_reuse;
142 int skc_bound_dev_if;
144 struct hlist_node skc_bind_node;
145 struct hlist_nulls_node skc_portaddr_node;
147 struct proto *skc_prot;
152 * fields between dontcopy_begin/dontcopy_end
153 * are not copied in sock_copy()
155 int skc_dontcopy_begin[0];
157 struct hlist_node skc_node;
158 struct hlist_nulls_node skc_nulls_node;
160 int skc_tx_queue_mapping;
162 int skc_dontcopy_end[0];
166 * struct sock - network layer representation of sockets
167 * @__sk_common: shared layout with inet_timewait_sock
168 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
169 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
170 * @sk_lock: synchronizer
171 * @sk_rcvbuf: size of receive buffer in bytes
172 * @sk_wq: sock wait queue and async head
173 * @sk_dst_cache: destination cache
174 * @sk_dst_lock: destination cache lock
175 * @sk_policy: flow policy
176 * @sk_rmem_alloc: receive queue bytes committed
177 * @sk_receive_queue: incoming packets
178 * @sk_wmem_alloc: transmit queue bytes committed
179 * @sk_write_queue: Packet sending queue
180 * @sk_async_wait_queue: DMA copied packets
181 * @sk_omem_alloc: "o" is "option" or "other"
182 * @sk_wmem_queued: persistent queue size
183 * @sk_forward_alloc: space allocated forward
184 * @sk_allocation: allocation mode
185 * @sk_sndbuf: size of send buffer in bytes
186 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
187 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
188 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
189 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
190 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
191 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
192 * @sk_gso_max_size: Maximum GSO segment size to build
193 * @sk_lingertime: %SO_LINGER l_linger setting
194 * @sk_backlog: always used with the per-socket spinlock held
195 * @sk_callback_lock: used with the callbacks in the end of this struct
196 * @sk_error_queue: rarely used
197 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
198 * IPV6_ADDRFORM for instance)
199 * @sk_err: last error
200 * @sk_err_soft: errors that don't cause failure but are the cause of a
201 * persistent failure not just 'timed out'
202 * @sk_drops: raw/udp drops counter
203 * @sk_ack_backlog: current listen backlog
204 * @sk_max_ack_backlog: listen backlog set in listen()
205 * @sk_priority: %SO_PRIORITY setting
206 * @sk_type: socket type (%SOCK_STREAM, etc)
207 * @sk_protocol: which protocol this socket belongs in this network family
208 * @sk_peer_pid: &struct pid for this socket's peer
209 * @sk_peer_cred: %SO_PEERCRED setting
210 * @sk_rcvlowat: %SO_RCVLOWAT setting
211 * @sk_rcvtimeo: %SO_RCVTIMEO setting
212 * @sk_sndtimeo: %SO_SNDTIMEO setting
213 * @sk_rxhash: flow hash received from netif layer
214 * @sk_filter: socket filtering instructions
215 * @sk_protinfo: private area, net family specific, when not using slab
216 * @sk_timer: sock cleanup timer
217 * @sk_stamp: time stamp of last packet received
218 * @sk_socket: Identd and reporting IO signals
219 * @sk_user_data: RPC layer private data
220 * @sk_sndmsg_page: cached page for sendmsg
221 * @sk_sndmsg_off: cached offset for sendmsg
222 * @sk_send_head: front of stuff to transmit
223 * @sk_security: used by security modules
224 * @sk_mark: generic packet mark
225 * @sk_classid: this socket's cgroup classid
226 * @sk_write_pending: a write to stream socket waits to start
227 * @sk_state_change: callback to indicate change in the state of the sock
228 * @sk_data_ready: callback to indicate there is data to be processed
229 * @sk_write_space: callback to indicate there is bf sending space available
230 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
231 * @sk_backlog_rcv: callback to process the backlog
232 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
236 * Now struct inet_timewait_sock also uses sock_common, so please just
237 * don't add nothing before this first member (__sk_common) --acme
239 struct sock_common __sk_common;
240 #define sk_node __sk_common.skc_node
241 #define sk_nulls_node __sk_common.skc_nulls_node
242 #define sk_refcnt __sk_common.skc_refcnt
243 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
245 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
246 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
247 #define sk_hash __sk_common.skc_hash
248 #define sk_family __sk_common.skc_family
249 #define sk_state __sk_common.skc_state
250 #define sk_reuse __sk_common.skc_reuse
251 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
252 #define sk_bind_node __sk_common.skc_bind_node
253 #define sk_prot __sk_common.skc_prot
254 #define sk_net __sk_common.skc_net
255 socket_lock_t sk_lock;
256 struct sk_buff_head sk_receive_queue;
258 * The backlog queue is special, it is always used with
259 * the per-socket spinlock held and requires low latency
260 * access. Therefore we special case it's implementation.
261 * Note : rmem_alloc is in this structure to fill a hole
262 * on 64bit arches, not because its logically part of
268 struct sk_buff *head;
269 struct sk_buff *tail;
271 #define sk_rmem_alloc sk_backlog.rmem_alloc
272 int sk_forward_alloc;
279 struct sk_filter __rcu *sk_filter;
280 struct socket_wq *sk_wq;
282 #ifdef CONFIG_NET_DMA
283 struct sk_buff_head sk_async_wait_queue;
287 struct xfrm_policy *sk_policy[2];
289 unsigned long sk_flags;
290 struct dst_entry *sk_dst_cache;
291 spinlock_t sk_dst_lock;
292 atomic_t sk_wmem_alloc;
293 atomic_t sk_omem_alloc;
295 struct sk_buff_head sk_write_queue;
296 kmemcheck_bitfield_begin(flags);
297 unsigned int sk_shutdown : 2,
302 kmemcheck_bitfield_end(flags);
308 unsigned int sk_gso_max_size;
310 unsigned long sk_lingertime;
311 struct sk_buff_head sk_error_queue;
312 struct proto *sk_prot_creator;
313 rwlock_t sk_callback_lock;
316 unsigned short sk_ack_backlog;
317 unsigned short sk_max_ack_backlog;
319 struct pid *sk_peer_pid;
320 const struct cred *sk_peer_cred;
324 struct timer_list sk_timer;
326 struct socket *sk_socket;
328 struct page *sk_sndmsg_page;
329 struct sk_buff *sk_send_head;
331 int sk_write_pending;
332 #ifdef CONFIG_SECURITY
337 void (*sk_state_change)(struct sock *sk);
338 void (*sk_data_ready)(struct sock *sk, int bytes);
339 void (*sk_write_space)(struct sock *sk);
340 void (*sk_error_report)(struct sock *sk);
341 int (*sk_backlog_rcv)(struct sock *sk,
342 struct sk_buff *skb);
343 void (*sk_destruct)(struct sock *sk);
347 * Hashed lists helper routines
349 static inline struct sock *sk_entry(const struct hlist_node *node)
351 return hlist_entry(node, struct sock, sk_node);
354 static inline struct sock *__sk_head(const struct hlist_head *head)
356 return hlist_entry(head->first, struct sock, sk_node);
359 static inline struct sock *sk_head(const struct hlist_head *head)
361 return hlist_empty(head) ? NULL : __sk_head(head);
364 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
366 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
369 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
371 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
374 static inline struct sock *sk_next(const struct sock *sk)
376 return sk->sk_node.next ?
377 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
380 static inline struct sock *sk_nulls_next(const struct sock *sk)
382 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
383 hlist_nulls_entry(sk->sk_nulls_node.next,
384 struct sock, sk_nulls_node) :
388 static inline int sk_unhashed(const struct sock *sk)
390 return hlist_unhashed(&sk->sk_node);
393 static inline int sk_hashed(const struct sock *sk)
395 return !sk_unhashed(sk);
398 static __inline__ void sk_node_init(struct hlist_node *node)
403 static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
408 static __inline__ void __sk_del_node(struct sock *sk)
410 __hlist_del(&sk->sk_node);
413 /* NB: equivalent to hlist_del_init_rcu */
414 static __inline__ int __sk_del_node_init(struct sock *sk)
418 sk_node_init(&sk->sk_node);
424 /* Grab socket reference count. This operation is valid only
425 when sk is ALREADY grabbed f.e. it is found in hash table
426 or a list and the lookup is made under lock preventing hash table
430 static inline void sock_hold(struct sock *sk)
432 atomic_inc(&sk->sk_refcnt);
435 /* Ungrab socket in the context, which assumes that socket refcnt
436 cannot hit zero, f.e. it is true in context of any socketcall.
438 static inline void __sock_put(struct sock *sk)
440 atomic_dec(&sk->sk_refcnt);
443 static __inline__ int sk_del_node_init(struct sock *sk)
445 int rc = __sk_del_node_init(sk);
448 /* paranoid for a while -acme */
449 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
454 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
456 static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
459 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
465 static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
467 int rc = __sk_nulls_del_node_init_rcu(sk);
470 /* paranoid for a while -acme */
471 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
477 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
479 hlist_add_head(&sk->sk_node, list);
482 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
485 __sk_add_node(sk, list);
488 static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
491 hlist_add_head_rcu(&sk->sk_node, list);
494 static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
496 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
499 static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
502 __sk_nulls_add_node_rcu(sk, list);
505 static __inline__ void __sk_del_bind_node(struct sock *sk)
507 __hlist_del(&sk->sk_bind_node);
510 static __inline__ void sk_add_bind_node(struct sock *sk,
511 struct hlist_head *list)
513 hlist_add_head(&sk->sk_bind_node, list);
516 #define sk_for_each(__sk, node, list) \
517 hlist_for_each_entry(__sk, node, list, sk_node)
518 #define sk_for_each_rcu(__sk, node, list) \
519 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
520 #define sk_nulls_for_each(__sk, node, list) \
521 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
522 #define sk_nulls_for_each_rcu(__sk, node, list) \
523 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
524 #define sk_for_each_from(__sk, node) \
525 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
526 hlist_for_each_entry_from(__sk, node, sk_node)
527 #define sk_nulls_for_each_from(__sk, node) \
528 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
529 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
530 #define sk_for_each_safe(__sk, node, tmp, list) \
531 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
532 #define sk_for_each_bound(__sk, node, list) \
533 hlist_for_each_entry(__sk, node, list, sk_bind_node)
546 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
547 SOCK_DBG, /* %SO_DEBUG setting */
548 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
549 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
550 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
551 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
552 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
553 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
554 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
555 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
556 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
557 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
558 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
559 SOCK_FASYNC, /* fasync() active */
563 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
565 nsk->sk_flags = osk->sk_flags;
568 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
570 __set_bit(flag, &sk->sk_flags);
573 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
575 __clear_bit(flag, &sk->sk_flags);
578 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
580 return test_bit(flag, &sk->sk_flags);
583 static inline void sk_acceptq_removed(struct sock *sk)
585 sk->sk_ack_backlog--;
588 static inline void sk_acceptq_added(struct sock *sk)
590 sk->sk_ack_backlog++;
593 static inline int sk_acceptq_is_full(struct sock *sk)
595 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
599 * Compute minimal free write space needed to queue new packets.
601 static inline int sk_stream_min_wspace(struct sock *sk)
603 return sk->sk_wmem_queued >> 1;
606 static inline int sk_stream_wspace(struct sock *sk)
608 return sk->sk_sndbuf - sk->sk_wmem_queued;
611 extern void sk_stream_write_space(struct sock *sk);
613 static inline int sk_stream_memory_free(struct sock *sk)
615 return sk->sk_wmem_queued < sk->sk_sndbuf;
618 /* OOB backlog add */
619 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
621 /* dont let skb dst not refcounted, we are going to leave rcu lock */
624 if (!sk->sk_backlog.tail)
625 sk->sk_backlog.head = skb;
627 sk->sk_backlog.tail->next = skb;
629 sk->sk_backlog.tail = skb;
634 * Take into account size of receive queue and backlog queue
636 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
638 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
640 return qsize + skb->truesize > sk->sk_rcvbuf;
643 /* The per-socket spinlock must be held here. */
644 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
646 if (sk_rcvqueues_full(sk, skb))
649 __sk_add_backlog(sk, skb);
650 sk->sk_backlog.len += skb->truesize;
654 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
656 return sk->sk_backlog_rcv(sk, skb);
659 static inline void sock_rps_record_flow(const struct sock *sk)
662 struct rps_sock_flow_table *sock_flow_table;
665 sock_flow_table = rcu_dereference(rps_sock_flow_table);
666 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
671 static inline void sock_rps_reset_flow(const struct sock *sk)
674 struct rps_sock_flow_table *sock_flow_table;
677 sock_flow_table = rcu_dereference(rps_sock_flow_table);
678 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
683 static inline void sock_rps_save_rxhash(struct sock *sk, u32 rxhash)
686 if (unlikely(sk->sk_rxhash != rxhash)) {
687 sock_rps_reset_flow(sk);
688 sk->sk_rxhash = rxhash;
693 #define sk_wait_event(__sk, __timeo, __condition) \
695 release_sock(__sk); \
696 __rc = __condition; \
698 *(__timeo) = schedule_timeout(*(__timeo)); \
701 __rc = __condition; \
705 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
706 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
707 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
708 extern int sk_stream_error(struct sock *sk, int flags, int err);
709 extern void sk_stream_kill_queues(struct sock *sk);
711 extern int sk_wait_data(struct sock *sk, long *timeo);
713 struct request_sock_ops;
714 struct timewait_sock_ops;
715 struct inet_hashinfo;
718 /* Networking protocol blocks we attach to sockets.
719 * socket layer -> transport layer interface
720 * transport -> network interface is defined by struct inet_proto
723 void (*close)(struct sock *sk,
725 int (*connect)(struct sock *sk,
726 struct sockaddr *uaddr,
728 int (*disconnect)(struct sock *sk, int flags);
730 struct sock * (*accept) (struct sock *sk, int flags, int *err);
732 int (*ioctl)(struct sock *sk, int cmd,
734 int (*init)(struct sock *sk);
735 void (*destroy)(struct sock *sk);
736 void (*shutdown)(struct sock *sk, int how);
737 int (*setsockopt)(struct sock *sk, int level,
738 int optname, char __user *optval,
739 unsigned int optlen);
740 int (*getsockopt)(struct sock *sk, int level,
741 int optname, char __user *optval,
744 int (*compat_setsockopt)(struct sock *sk,
746 int optname, char __user *optval,
747 unsigned int optlen);
748 int (*compat_getsockopt)(struct sock *sk,
750 int optname, char __user *optval,
753 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
754 struct msghdr *msg, size_t len);
755 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
757 size_t len, int noblock, int flags,
759 int (*sendpage)(struct sock *sk, struct page *page,
760 int offset, size_t size, int flags);
761 int (*bind)(struct sock *sk,
762 struct sockaddr *uaddr, int addr_len);
764 int (*backlog_rcv) (struct sock *sk,
765 struct sk_buff *skb);
767 /* Keeping track of sk's, looking them up, and port selection methods. */
768 void (*hash)(struct sock *sk);
769 void (*unhash)(struct sock *sk);
770 void (*rehash)(struct sock *sk);
771 int (*get_port)(struct sock *sk, unsigned short snum);
773 /* Keeping track of sockets in use */
774 #ifdef CONFIG_PROC_FS
775 unsigned int inuse_idx;
778 /* Memory pressure */
779 void (*enter_memory_pressure)(struct sock *sk);
780 atomic_long_t *memory_allocated; /* Current allocated memory. */
781 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
783 * Pressure flag: try to collapse.
784 * Technical note: it is used by multiple contexts non atomically.
785 * All the __sk_mem_schedule() is of this nature: accounting
786 * is strict, actions are advisory and have some latency.
788 int *memory_pressure;
795 struct kmem_cache *slab;
796 unsigned int obj_size;
799 struct percpu_counter *orphan_count;
801 struct request_sock_ops *rsk_prot;
802 struct timewait_sock_ops *twsk_prot;
805 struct inet_hashinfo *hashinfo;
806 struct udp_table *udp_table;
807 struct raw_hashinfo *raw_hash;
810 struct module *owner;
814 struct list_head node;
815 #ifdef SOCK_REFCNT_DEBUG
820 extern int proto_register(struct proto *prot, int alloc_slab);
821 extern void proto_unregister(struct proto *prot);
823 #ifdef SOCK_REFCNT_DEBUG
824 static inline void sk_refcnt_debug_inc(struct sock *sk)
826 atomic_inc(&sk->sk_prot->socks);
829 static inline void sk_refcnt_debug_dec(struct sock *sk)
831 atomic_dec(&sk->sk_prot->socks);
832 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
833 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
836 static inline void sk_refcnt_debug_release(const struct sock *sk)
838 if (atomic_read(&sk->sk_refcnt) != 1)
839 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
840 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
842 #else /* SOCK_REFCNT_DEBUG */
843 #define sk_refcnt_debug_inc(sk) do { } while (0)
844 #define sk_refcnt_debug_dec(sk) do { } while (0)
845 #define sk_refcnt_debug_release(sk) do { } while (0)
846 #endif /* SOCK_REFCNT_DEBUG */
849 #ifdef CONFIG_PROC_FS
850 /* Called with local bh disabled */
851 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
852 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
854 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
861 /* With per-bucket locks this operation is not-atomic, so that
862 * this version is not worse.
864 static inline void __sk_prot_rehash(struct sock *sk)
866 sk->sk_prot->unhash(sk);
867 sk->sk_prot->hash(sk);
870 /* About 10 seconds */
871 #define SOCK_DESTROY_TIME (10*HZ)
873 /* Sockets 0-1023 can't be bound to unless you are superuser */
874 #define PROT_SOCK 1024
876 #define SHUTDOWN_MASK 3
877 #define RCV_SHUTDOWN 1
878 #define SEND_SHUTDOWN 2
880 #define SOCK_SNDBUF_LOCK 1
881 #define SOCK_RCVBUF_LOCK 2
882 #define SOCK_BINDADDR_LOCK 4
883 #define SOCK_BINDPORT_LOCK 8
885 /* sock_iocb: used to kick off async processing of socket ios */
887 struct list_head list;
893 struct scm_cookie *scm;
894 struct msghdr *msg, async_msg;
898 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
900 return (struct sock_iocb *)iocb->private;
903 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
908 struct socket_alloc {
909 struct socket socket;
910 struct inode vfs_inode;
913 static inline struct socket *SOCKET_I(struct inode *inode)
915 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
918 static inline struct inode *SOCK_INODE(struct socket *socket)
920 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
924 * Functions for memory accounting
926 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
927 extern void __sk_mem_reclaim(struct sock *sk);
929 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
930 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
931 #define SK_MEM_SEND 0
932 #define SK_MEM_RECV 1
934 static inline int sk_mem_pages(int amt)
936 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
939 static inline int sk_has_account(struct sock *sk)
941 /* return true if protocol supports memory accounting */
942 return !!sk->sk_prot->memory_allocated;
945 static inline int sk_wmem_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_SEND);
953 static inline int sk_rmem_schedule(struct sock *sk, int size)
955 if (!sk_has_account(sk))
957 return size <= sk->sk_forward_alloc ||
958 __sk_mem_schedule(sk, size, SK_MEM_RECV);
961 static inline void sk_mem_reclaim(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_reclaim_partial(struct sock *sk)
971 if (!sk_has_account(sk))
973 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
974 __sk_mem_reclaim(sk);
977 static inline void sk_mem_charge(struct sock *sk, int size)
979 if (!sk_has_account(sk))
981 sk->sk_forward_alloc -= size;
984 static inline void sk_mem_uncharge(struct sock *sk, int size)
986 if (!sk_has_account(sk))
988 sk->sk_forward_alloc += size;
991 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
993 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
994 sk->sk_wmem_queued -= skb->truesize;
995 sk_mem_uncharge(sk, skb->truesize);
999 /* Used by processes to "lock" a socket state, so that
1000 * interrupts and bottom half handlers won't change it
1001 * from under us. It essentially blocks any incoming
1002 * packets, so that we won't get any new data or any
1003 * packets that change the state of the socket.
1005 * While locked, BH processing will add new packets to
1006 * the backlog queue. This queue is processed by the
1007 * owner of the socket lock right before it is released.
1009 * Since ~2.3.5 it is also exclusive sleep lock serializing
1010 * accesses from user process context.
1012 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1015 * Macro so as to not evaluate some arguments when
1016 * lockdep is not enabled.
1018 * Mark both the sk_lock and the sk_lock.slock as a
1019 * per-address-family lock class.
1021 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1023 sk->sk_lock.owned = 0; \
1024 init_waitqueue_head(&sk->sk_lock.wq); \
1025 spin_lock_init(&(sk)->sk_lock.slock); \
1026 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1027 sizeof((sk)->sk_lock)); \
1028 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1030 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1033 extern void lock_sock_nested(struct sock *sk, int subclass);
1035 static inline void lock_sock(struct sock *sk)
1037 lock_sock_nested(sk, 0);
1040 extern void release_sock(struct sock *sk);
1042 /* BH context may only use the following locking interface. */
1043 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1044 #define bh_lock_sock_nested(__sk) \
1045 spin_lock_nested(&((__sk)->sk_lock.slock), \
1046 SINGLE_DEPTH_NESTING)
1047 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1049 extern bool lock_sock_fast(struct sock *sk);
1051 * unlock_sock_fast - complement of lock_sock_fast
1055 * fast unlock socket for user context.
1056 * If slow mode is on, we call regular release_sock()
1058 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1063 spin_unlock_bh(&sk->sk_lock.slock);
1067 extern struct sock *sk_alloc(struct net *net, int family,
1069 struct proto *prot);
1070 extern void sk_free(struct sock *sk);
1071 extern void sk_release_kernel(struct sock *sk);
1072 extern struct sock *sk_clone(const struct sock *sk,
1073 const gfp_t priority);
1075 extern struct sk_buff *sock_wmalloc(struct sock *sk,
1076 unsigned long size, int force,
1078 extern struct sk_buff *sock_rmalloc(struct sock *sk,
1079 unsigned long size, int force,
1081 extern void sock_wfree(struct sk_buff *skb);
1082 extern void sock_rfree(struct sk_buff *skb);
1084 extern int sock_setsockopt(struct socket *sock, int level,
1085 int op, char __user *optval,
1086 unsigned int optlen);
1088 extern int sock_getsockopt(struct socket *sock, int level,
1089 int op, char __user *optval,
1090 int __user *optlen);
1091 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1095 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1096 unsigned long header_len,
1097 unsigned long data_len,
1100 extern void *sock_kmalloc(struct sock *sk, int size,
1102 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1103 extern void sk_send_sigurg(struct sock *sk);
1105 #ifdef CONFIG_CGROUPS
1106 extern void sock_update_classid(struct sock *sk);
1108 static inline void sock_update_classid(struct sock *sk)
1114 * Functions to fill in entries in struct proto_ops when a protocol
1115 * does not implement a particular function.
1117 extern int sock_no_bind(struct socket *,
1118 struct sockaddr *, int);
1119 extern int sock_no_connect(struct socket *,
1120 struct sockaddr *, int, int);
1121 extern int sock_no_socketpair(struct socket *,
1123 extern int sock_no_accept(struct socket *,
1124 struct socket *, int);
1125 extern int sock_no_getname(struct socket *,
1126 struct sockaddr *, int *, int);
1127 extern unsigned int sock_no_poll(struct file *, struct socket *,
1128 struct poll_table_struct *);
1129 extern int sock_no_ioctl(struct socket *, unsigned int,
1131 extern int sock_no_listen(struct socket *, int);
1132 extern int sock_no_shutdown(struct socket *, int);
1133 extern int sock_no_getsockopt(struct socket *, int , int,
1134 char __user *, int __user *);
1135 extern int sock_no_setsockopt(struct socket *, int, int,
1136 char __user *, unsigned int);
1137 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1138 struct msghdr *, size_t);
1139 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1140 struct msghdr *, size_t, int);
1141 extern int sock_no_mmap(struct file *file,
1142 struct socket *sock,
1143 struct vm_area_struct *vma);
1144 extern ssize_t sock_no_sendpage(struct socket *sock,
1146 int offset, size_t size,
1150 * Functions to fill in entries in struct proto_ops when a protocol
1151 * uses the inet style.
1153 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1154 char __user *optval, int __user *optlen);
1155 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1156 struct msghdr *msg, size_t size, int flags);
1157 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1158 char __user *optval, unsigned int optlen);
1159 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1160 int optname, char __user *optval, int __user *optlen);
1161 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1162 int optname, char __user *optval, unsigned int optlen);
1164 extern void sk_common_release(struct sock *sk);
1167 * Default socket callbacks and setup code
1170 /* Initialise core socket variables */
1171 extern void sock_init_data(struct socket *sock, struct sock *sk);
1173 extern void sk_filter_release_rcu(struct rcu_head *rcu);
1176 * sk_filter_release - release a socket filter
1177 * @fp: filter to remove
1179 * Remove a filter from a socket and release its resources.
1182 static inline void sk_filter_release(struct sk_filter *fp)
1184 if (atomic_dec_and_test(&fp->refcnt))
1185 call_rcu_bh(&fp->rcu, sk_filter_release_rcu);
1188 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1190 unsigned int size = sk_filter_len(fp);
1192 atomic_sub(size, &sk->sk_omem_alloc);
1193 sk_filter_release(fp);
1196 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1198 atomic_inc(&fp->refcnt);
1199 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1203 * Socket reference counting postulates.
1205 * * Each user of socket SHOULD hold a reference count.
1206 * * Each access point to socket (an hash table bucket, reference from a list,
1207 * running timer, skb in flight MUST hold a reference count.
1208 * * When reference count hits 0, it means it will never increase back.
1209 * * When reference count hits 0, it means that no references from
1210 * outside exist to this socket and current process on current CPU
1211 * is last user and may/should destroy this socket.
1212 * * sk_free is called from any context: process, BH, IRQ. When
1213 * it is called, socket has no references from outside -> sk_free
1214 * may release descendant resources allocated by the socket, but
1215 * to the time when it is called, socket is NOT referenced by any
1216 * hash tables, lists etc.
1217 * * Packets, delivered from outside (from network or from another process)
1218 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1219 * when they sit in queue. Otherwise, packets will leak to hole, when
1220 * socket is looked up by one cpu and unhasing is made by another CPU.
1221 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1222 * (leak to backlog). Packet socket does all the processing inside
1223 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1224 * use separate SMP lock, so that they are prone too.
1227 /* Ungrab socket and destroy it, if it was the last reference. */
1228 static inline void sock_put(struct sock *sk)
1230 if (atomic_dec_and_test(&sk->sk_refcnt))
1234 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1237 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1239 sk->sk_tx_queue_mapping = tx_queue;
1242 static inline void sk_tx_queue_clear(struct sock *sk)
1244 sk->sk_tx_queue_mapping = -1;
1247 static inline int sk_tx_queue_get(const struct sock *sk)
1249 return sk ? sk->sk_tx_queue_mapping : -1;
1252 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1254 sk_tx_queue_clear(sk);
1255 sk->sk_socket = sock;
1258 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1260 return &sk->sk_wq->wait;
1262 /* Detach socket from process context.
1263 * Announce socket dead, detach it from wait queue and inode.
1264 * Note that parent inode held reference count on this struct sock,
1265 * we do not release it in this function, because protocol
1266 * probably wants some additional cleanups or even continuing
1267 * to work with this socket (TCP).
1269 static inline void sock_orphan(struct sock *sk)
1271 write_lock_bh(&sk->sk_callback_lock);
1272 sock_set_flag(sk, SOCK_DEAD);
1273 sk_set_socket(sk, NULL);
1275 write_unlock_bh(&sk->sk_callback_lock);
1278 static inline void sock_graft(struct sock *sk, struct socket *parent)
1280 write_lock_bh(&sk->sk_callback_lock);
1281 rcu_assign_pointer(sk->sk_wq, parent->wq);
1283 sk_set_socket(sk, parent);
1284 security_sock_graft(sk, parent);
1285 write_unlock_bh(&sk->sk_callback_lock);
1288 extern int sock_i_uid(struct sock *sk);
1289 extern unsigned long sock_i_ino(struct sock *sk);
1291 static inline struct dst_entry *
1292 __sk_dst_get(struct sock *sk)
1294 return rcu_dereference_check(sk->sk_dst_cache, rcu_read_lock_held() ||
1295 sock_owned_by_user(sk) ||
1296 lockdep_is_held(&sk->sk_lock.slock));
1299 static inline struct dst_entry *
1300 sk_dst_get(struct sock *sk)
1302 struct dst_entry *dst;
1305 dst = rcu_dereference(sk->sk_dst_cache);
1312 extern void sk_reset_txq(struct sock *sk);
1314 static inline void dst_negative_advice(struct sock *sk)
1316 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1318 if (dst && dst->ops->negative_advice) {
1319 ndst = dst->ops->negative_advice(dst);
1322 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1329 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1331 struct dst_entry *old_dst;
1333 sk_tx_queue_clear(sk);
1335 * This can be called while sk is owned by the caller only,
1336 * with no state that can be checked in a rcu_dereference_check() cond
1338 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1339 rcu_assign_pointer(sk->sk_dst_cache, dst);
1340 dst_release(old_dst);
1344 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1346 spin_lock(&sk->sk_dst_lock);
1347 __sk_dst_set(sk, dst);
1348 spin_unlock(&sk->sk_dst_lock);
1352 __sk_dst_reset(struct sock *sk)
1354 __sk_dst_set(sk, NULL);
1358 sk_dst_reset(struct sock *sk)
1360 spin_lock(&sk->sk_dst_lock);
1362 spin_unlock(&sk->sk_dst_lock);
1365 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1367 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1369 static inline int sk_can_gso(const struct sock *sk)
1371 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1374 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1376 static inline void sk_nocaps_add(struct sock *sk, int flags)
1378 sk->sk_route_nocaps |= flags;
1379 sk->sk_route_caps &= ~flags;
1382 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1383 struct sk_buff *skb, struct page *page,
1386 if (skb->ip_summed == CHECKSUM_NONE) {
1388 __wsum csum = csum_and_copy_from_user(from,
1389 page_address(page) + off,
1393 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1394 } else if (copy_from_user(page_address(page) + off, from, copy))
1398 skb->data_len += copy;
1399 skb->truesize += copy;
1400 sk->sk_wmem_queued += copy;
1401 sk_mem_charge(sk, copy);
1406 * sk_wmem_alloc_get - returns write allocations
1409 * Returns sk_wmem_alloc minus initial offset of one
1411 static inline int sk_wmem_alloc_get(const struct sock *sk)
1413 return atomic_read(&sk->sk_wmem_alloc) - 1;
1417 * sk_rmem_alloc_get - returns read allocations
1420 * Returns sk_rmem_alloc
1422 static inline int sk_rmem_alloc_get(const struct sock *sk)
1424 return atomic_read(&sk->sk_rmem_alloc);
1428 * sk_has_allocations - check if allocations are outstanding
1431 * Returns true if socket has write or read allocations
1433 static inline int sk_has_allocations(const struct sock *sk)
1435 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1439 * wq_has_sleeper - check if there are any waiting processes
1440 * @wq: struct socket_wq
1442 * Returns true if socket_wq has waiting processes
1444 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1445 * barrier call. They were added due to the race found within the tcp code.
1447 * Consider following tcp code paths:
1451 * sys_select receive packet
1453 * __add_wait_queue update tp->rcv_nxt
1455 * tp->rcv_nxt check sock_def_readable
1457 * schedule rcu_read_lock();
1458 * wq = rcu_dereference(sk->sk_wq);
1459 * if (wq && waitqueue_active(&wq->wait))
1460 * wake_up_interruptible(&wq->wait)
1464 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1465 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1466 * could then endup calling schedule and sleep forever if there are no more
1467 * data on the socket.
1470 static inline bool wq_has_sleeper(struct socket_wq *wq)
1474 * We need to be sure we are in sync with the
1475 * add_wait_queue modifications to the wait queue.
1477 * This memory barrier is paired in the sock_poll_wait.
1480 return wq && waitqueue_active(&wq->wait);
1484 * sock_poll_wait - place memory barrier behind the poll_wait call.
1486 * @wait_address: socket wait queue
1489 * See the comments in the wq_has_sleeper function.
1491 static inline void sock_poll_wait(struct file *filp,
1492 wait_queue_head_t *wait_address, poll_table *p)
1494 if (p && wait_address) {
1495 poll_wait(filp, wait_address, p);
1497 * We need to be sure we are in sync with the
1498 * socket flags modification.
1500 * This memory barrier is paired in the wq_has_sleeper.
1507 * Queue a received datagram if it will fit. Stream and sequenced
1508 * protocols can't normally use this as they need to fit buffers in
1509 * and play with them.
1511 * Inlined as it's very short and called for pretty much every
1512 * packet ever received.
1515 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1519 skb->destructor = sock_wfree;
1521 * We used to take a refcount on sk, but following operation
1522 * is enough to guarantee sk_free() wont free this sock until
1523 * all in-flight packets are completed
1525 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1528 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1532 skb->destructor = sock_rfree;
1533 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1534 sk_mem_charge(sk, skb->truesize);
1537 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1538 unsigned long expires);
1540 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1542 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1544 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1547 * Recover an error report and clear atomically
1550 static inline int sock_error(struct sock *sk)
1553 if (likely(!sk->sk_err))
1555 err = xchg(&sk->sk_err, 0);
1559 static inline unsigned long sock_wspace(struct sock *sk)
1563 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1564 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1571 static inline void sk_wake_async(struct sock *sk, int how, int band)
1573 if (sock_flag(sk, SOCK_FASYNC))
1574 sock_wake_async(sk->sk_socket, how, band);
1577 #define SOCK_MIN_SNDBUF 2048
1579 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1580 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1582 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1584 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1586 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1587 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1588 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1592 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1594 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1596 struct page *page = NULL;
1598 page = alloc_pages(sk->sk_allocation, 0);
1600 sk->sk_prot->enter_memory_pressure(sk);
1601 sk_stream_moderate_sndbuf(sk);
1607 * Default write policy as shown to user space via poll/select/SIGIO
1609 static inline int sock_writeable(const struct sock *sk)
1611 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1614 static inline gfp_t gfp_any(void)
1616 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1619 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1621 return noblock ? 0 : sk->sk_rcvtimeo;
1624 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1626 return noblock ? 0 : sk->sk_sndtimeo;
1629 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1631 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1634 /* Alas, with timeout socket operations are not restartable.
1635 * Compare this to poll().
1637 static inline int sock_intr_errno(long timeo)
1639 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1642 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1643 struct sk_buff *skb);
1645 static __inline__ void
1646 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1648 ktime_t kt = skb->tstamp;
1649 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
1652 * generate control messages if
1653 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1654 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1655 * - software time stamp available and wanted
1656 * (SOCK_TIMESTAMPING_SOFTWARE)
1657 * - hardware time stamps available and wanted
1658 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1659 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1661 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1662 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1663 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1664 (hwtstamps->hwtstamp.tv64 &&
1665 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1666 (hwtstamps->syststamp.tv64 &&
1667 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
1668 __sock_recv_timestamp(msg, sk, skb);
1673 extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1674 struct sk_buff *skb);
1676 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1677 struct sk_buff *skb)
1679 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
1680 (1UL << SOCK_RCVTSTAMP) | \
1681 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
1682 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
1683 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
1684 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
1686 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
1687 __sock_recv_ts_and_drops(msg, sk, skb);
1689 sk->sk_stamp = skb->tstamp;
1693 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1694 * @sk: socket sending this packet
1695 * @tx_flags: filled with instructions for time stamping
1697 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1698 * parameters are invalid.
1700 extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
1703 * sk_eat_skb - Release a skb if it is no longer needed
1704 * @sk: socket to eat this skb from
1705 * @skb: socket buffer to eat
1706 * @copied_early: flag indicating whether DMA operations copied this data early
1708 * This routine must be called with interrupts disabled or with the socket
1709 * locked so that the sk_buff queue operation is ok.
1711 #ifdef CONFIG_NET_DMA
1712 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1714 __skb_unlink(skb, &sk->sk_receive_queue);
1718 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1721 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1723 __skb_unlink(skb, &sk->sk_receive_queue);
1729 struct net *sock_net(const struct sock *sk)
1731 return read_pnet(&sk->sk_net);
1735 void sock_net_set(struct sock *sk, struct net *net)
1737 write_pnet(&sk->sk_net, net);
1741 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1742 * They should not hold a referrence to a namespace in order to allow
1744 * Sockets after sk_change_net should be released using sk_release_kernel
1746 static inline void sk_change_net(struct sock *sk, struct net *net)
1748 put_net(sock_net(sk));
1749 sock_net_set(sk, hold_net(net));
1752 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
1754 if (unlikely(skb->sk)) {
1755 struct sock *sk = skb->sk;
1757 skb->destructor = NULL;
1764 extern void sock_enable_timestamp(struct sock *sk, int flag);
1765 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1766 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1769 * Enable debug/info messages
1771 extern int net_msg_warn;
1772 #define NETDEBUG(fmt, args...) \
1773 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1775 #define LIMIT_NETDEBUG(fmt, args...) \
1776 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1778 extern __u32 sysctl_wmem_max;
1779 extern __u32 sysctl_rmem_max;
1781 extern void sk_init(void);
1783 extern int sysctl_optmem_max;
1785 extern __u32 sysctl_wmem_default;
1786 extern __u32 sysctl_rmem_default;
1788 #endif /* _SOCK_H */