Merge remote-tracking branch 'moduleh/module.h-split'
[linux-2.6/next.git] / include / net / sock.h
blobbeb1a911acbbc3d07cd3b93799fcf3201d25d70e
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the AF_INET socket handler.
8 * Version: @(#)sock.h 1.0.4 05/13/93
10 * Authors: Ross Biro
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>
15 * Fixes:
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
19 * than the reverse.
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.
40 #ifndef _SOCK_H
41 #define _SOCK_H
43 #include <linux/hardirq.h>
44 #include <linux/kernel.h>
45 #include <linux/list.h>
46 #include <linux/list_nulls.h>
47 #include <linux/timer.h>
48 #include <linux/cache.h>
49 #include <linux/lockdep.h>
50 #include <linux/netdevice.h>
51 #include <linux/skbuff.h> /* struct sk_buff */
52 #include <linux/mm.h>
53 #include <linux/security.h>
54 #include <linux/slab.h>
55 #include <linux/uaccess.h>
57 #include <linux/filter.h>
58 #include <linux/rculist_nulls.h>
59 #include <linux/poll.h>
61 #include <linux/atomic.h>
62 #include <net/dst.h>
63 #include <net/checksum.h>
66 * This structure really needs to be cleaned up.
67 * Most of it is for TCP, and not used by any of
68 * the other protocols.
71 /* Define this to get the SOCK_DBG debugging facility. */
72 #define SOCK_DEBUGGING
73 #ifdef SOCK_DEBUGGING
74 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
75 printk(KERN_DEBUG msg); } while (0)
76 #else
77 /* Validate arguments and do nothing */
78 static inline void __attribute__ ((format (printf, 2, 3)))
79 SOCK_DEBUG(struct sock *sk, const char *msg, ...)
82 #endif
84 /* This is the per-socket lock. The spinlock provides a synchronization
85 * between user contexts and software interrupt processing, whereas the
86 * mini-semaphore synchronizes multiple users amongst themselves.
88 typedef struct {
89 spinlock_t slock;
90 int owned;
91 wait_queue_head_t wq;
93 * We express the mutex-alike socket_lock semantics
94 * to the lock validator by explicitly managing
95 * the slock as a lock variant (in addition to
96 * the slock itself):
98 #ifdef CONFIG_DEBUG_LOCK_ALLOC
99 struct lockdep_map dep_map;
100 #endif
101 } socket_lock_t;
103 struct sock;
104 struct proto;
105 struct net;
108 * struct sock_common - minimal network layer representation of sockets
109 * @skc_daddr: Foreign IPv4 addr
110 * @skc_rcv_saddr: Bound local IPv4 addr
111 * @skc_hash: hash value used with various protocol lookup tables
112 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
113 * @skc_family: network address family
114 * @skc_state: Connection state
115 * @skc_reuse: %SO_REUSEADDR setting
116 * @skc_bound_dev_if: bound device index if != 0
117 * @skc_bind_node: bind hash linkage for various protocol lookup tables
118 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
119 * @skc_prot: protocol handlers inside a network family
120 * @skc_net: reference to the network namespace of this socket
121 * @skc_node: main hash linkage for various protocol lookup tables
122 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
123 * @skc_tx_queue_mapping: tx queue number for this connection
124 * @skc_refcnt: reference count
126 * This is the minimal network layer representation of sockets, the header
127 * for struct sock and struct inet_timewait_sock.
129 struct sock_common {
130 /* skc_daddr and skc_rcv_saddr must be grouped :
131 * cf INET_MATCH() and INET_TW_MATCH()
133 __be32 skc_daddr;
134 __be32 skc_rcv_saddr;
136 union {
137 unsigned int skc_hash;
138 __u16 skc_u16hashes[2];
140 unsigned short skc_family;
141 volatile unsigned char skc_state;
142 unsigned char skc_reuse;
143 int skc_bound_dev_if;
144 union {
145 struct hlist_node skc_bind_node;
146 struct hlist_nulls_node skc_portaddr_node;
148 struct proto *skc_prot;
149 #ifdef CONFIG_NET_NS
150 struct net *skc_net;
151 #endif
153 * fields between dontcopy_begin/dontcopy_end
154 * are not copied in sock_copy()
156 /* private: */
157 int skc_dontcopy_begin[0];
158 /* public: */
159 union {
160 struct hlist_node skc_node;
161 struct hlist_nulls_node skc_nulls_node;
163 int skc_tx_queue_mapping;
164 atomic_t skc_refcnt;
165 /* private: */
166 int skc_dontcopy_end[0];
167 /* public: */
171 * struct sock - network layer representation of sockets
172 * @__sk_common: shared layout with inet_timewait_sock
173 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
174 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
175 * @sk_lock: synchronizer
176 * @sk_rcvbuf: size of receive buffer in bytes
177 * @sk_wq: sock wait queue and async head
178 * @sk_dst_cache: destination cache
179 * @sk_dst_lock: destination cache lock
180 * @sk_policy: flow policy
181 * @sk_receive_queue: incoming packets
182 * @sk_wmem_alloc: transmit queue bytes committed
183 * @sk_write_queue: Packet sending queue
184 * @sk_async_wait_queue: DMA copied packets
185 * @sk_omem_alloc: "o" is "option" or "other"
186 * @sk_wmem_queued: persistent queue size
187 * @sk_forward_alloc: space allocated forward
188 * @sk_allocation: allocation mode
189 * @sk_sndbuf: size of send buffer in bytes
190 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
191 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
192 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
193 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
194 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
195 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
196 * @sk_gso_max_size: Maximum GSO segment size to build
197 * @sk_lingertime: %SO_LINGER l_linger setting
198 * @sk_backlog: always used with the per-socket spinlock held
199 * @sk_callback_lock: used with the callbacks in the end of this struct
200 * @sk_error_queue: rarely used
201 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
202 * IPV6_ADDRFORM for instance)
203 * @sk_err: last error
204 * @sk_err_soft: errors that don't cause failure but are the cause of a
205 * persistent failure not just 'timed out'
206 * @sk_drops: raw/udp drops counter
207 * @sk_ack_backlog: current listen backlog
208 * @sk_max_ack_backlog: listen backlog set in listen()
209 * @sk_priority: %SO_PRIORITY setting
210 * @sk_type: socket type (%SOCK_STREAM, etc)
211 * @sk_protocol: which protocol this socket belongs in this network family
212 * @sk_peer_pid: &struct pid for this socket's peer
213 * @sk_peer_cred: %SO_PEERCRED setting
214 * @sk_rcvlowat: %SO_RCVLOWAT setting
215 * @sk_rcvtimeo: %SO_RCVTIMEO setting
216 * @sk_sndtimeo: %SO_SNDTIMEO setting
217 * @sk_rxhash: flow hash received from netif layer
218 * @sk_filter: socket filtering instructions
219 * @sk_protinfo: private area, net family specific, when not using slab
220 * @sk_timer: sock cleanup timer
221 * @sk_stamp: time stamp of last packet received
222 * @sk_socket: Identd and reporting IO signals
223 * @sk_user_data: RPC layer private data
224 * @sk_sndmsg_page: cached page for sendmsg
225 * @sk_sndmsg_off: cached offset for sendmsg
226 * @sk_send_head: front of stuff to transmit
227 * @sk_security: used by security modules
228 * @sk_mark: generic packet mark
229 * @sk_classid: this socket's cgroup classid
230 * @sk_write_pending: a write to stream socket waits to start
231 * @sk_state_change: callback to indicate change in the state of the sock
232 * @sk_data_ready: callback to indicate there is data to be processed
233 * @sk_write_space: callback to indicate there is bf sending space available
234 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
235 * @sk_backlog_rcv: callback to process the backlog
236 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
238 struct sock {
240 * Now struct inet_timewait_sock also uses sock_common, so please just
241 * don't add nothing before this first member (__sk_common) --acme
243 struct sock_common __sk_common;
244 #define sk_node __sk_common.skc_node
245 #define sk_nulls_node __sk_common.skc_nulls_node
246 #define sk_refcnt __sk_common.skc_refcnt
247 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
249 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
250 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
251 #define sk_hash __sk_common.skc_hash
252 #define sk_family __sk_common.skc_family
253 #define sk_state __sk_common.skc_state
254 #define sk_reuse __sk_common.skc_reuse
255 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
256 #define sk_bind_node __sk_common.skc_bind_node
257 #define sk_prot __sk_common.skc_prot
258 #define sk_net __sk_common.skc_net
259 socket_lock_t sk_lock;
260 struct sk_buff_head sk_receive_queue;
262 * The backlog queue is special, it is always used with
263 * the per-socket spinlock held and requires low latency
264 * access. Therefore we special case it's implementation.
265 * Note : rmem_alloc is in this structure to fill a hole
266 * on 64bit arches, not because its logically part of
267 * backlog.
269 struct {
270 atomic_t rmem_alloc;
271 int len;
272 struct sk_buff *head;
273 struct sk_buff *tail;
274 } sk_backlog;
275 #define sk_rmem_alloc sk_backlog.rmem_alloc
276 int sk_forward_alloc;
277 #ifdef CONFIG_RPS
278 __u32 sk_rxhash;
279 #endif
280 atomic_t sk_drops;
281 int sk_rcvbuf;
283 struct sk_filter __rcu *sk_filter;
284 struct socket_wq __rcu *sk_wq;
286 #ifdef CONFIG_NET_DMA
287 struct sk_buff_head sk_async_wait_queue;
288 #endif
290 #ifdef CONFIG_XFRM
291 struct xfrm_policy *sk_policy[2];
292 #endif
293 unsigned long sk_flags;
294 struct dst_entry *sk_dst_cache;
295 spinlock_t sk_dst_lock;
296 atomic_t sk_wmem_alloc;
297 atomic_t sk_omem_alloc;
298 int sk_sndbuf;
299 struct sk_buff_head sk_write_queue;
300 kmemcheck_bitfield_begin(flags);
301 unsigned int sk_shutdown : 2,
302 sk_no_check : 2,
303 sk_userlocks : 4,
304 sk_protocol : 8,
305 sk_type : 16;
306 kmemcheck_bitfield_end(flags);
307 int sk_wmem_queued;
308 gfp_t sk_allocation;
309 int sk_route_caps;
310 int sk_route_nocaps;
311 int sk_gso_type;
312 unsigned int sk_gso_max_size;
313 int sk_rcvlowat;
314 unsigned long sk_lingertime;
315 struct sk_buff_head sk_error_queue;
316 struct proto *sk_prot_creator;
317 rwlock_t sk_callback_lock;
318 int sk_err,
319 sk_err_soft;
320 unsigned short sk_ack_backlog;
321 unsigned short sk_max_ack_backlog;
322 __u32 sk_priority;
323 struct pid *sk_peer_pid;
324 const struct cred *sk_peer_cred;
325 long sk_rcvtimeo;
326 long sk_sndtimeo;
327 void *sk_protinfo;
328 struct timer_list sk_timer;
329 ktime_t sk_stamp;
330 struct socket *sk_socket;
331 void *sk_user_data;
332 struct page *sk_sndmsg_page;
333 struct sk_buff *sk_send_head;
334 __u32 sk_sndmsg_off;
335 int sk_write_pending;
336 #ifdef CONFIG_SECURITY
337 void *sk_security;
338 #endif
339 __u32 sk_mark;
340 u32 sk_classid;
341 void (*sk_state_change)(struct sock *sk);
342 void (*sk_data_ready)(struct sock *sk, int bytes);
343 void (*sk_write_space)(struct sock *sk);
344 void (*sk_error_report)(struct sock *sk);
345 int (*sk_backlog_rcv)(struct sock *sk,
346 struct sk_buff *skb);
347 void (*sk_destruct)(struct sock *sk);
351 * Hashed lists helper routines
353 static inline struct sock *sk_entry(const struct hlist_node *node)
355 return hlist_entry(node, struct sock, sk_node);
358 static inline struct sock *__sk_head(const struct hlist_head *head)
360 return hlist_entry(head->first, struct sock, sk_node);
363 static inline struct sock *sk_head(const struct hlist_head *head)
365 return hlist_empty(head) ? NULL : __sk_head(head);
368 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
370 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
373 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
375 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
378 static inline struct sock *sk_next(const struct sock *sk)
380 return sk->sk_node.next ?
381 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
384 static inline struct sock *sk_nulls_next(const struct sock *sk)
386 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
387 hlist_nulls_entry(sk->sk_nulls_node.next,
388 struct sock, sk_nulls_node) :
389 NULL;
392 static inline int sk_unhashed(const struct sock *sk)
394 return hlist_unhashed(&sk->sk_node);
397 static inline int sk_hashed(const struct sock *sk)
399 return !sk_unhashed(sk);
402 static __inline__ void sk_node_init(struct hlist_node *node)
404 node->pprev = NULL;
407 static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
409 node->pprev = NULL;
412 static __inline__ void __sk_del_node(struct sock *sk)
414 __hlist_del(&sk->sk_node);
417 /* NB: equivalent to hlist_del_init_rcu */
418 static __inline__ int __sk_del_node_init(struct sock *sk)
420 if (sk_hashed(sk)) {
421 __sk_del_node(sk);
422 sk_node_init(&sk->sk_node);
423 return 1;
425 return 0;
428 /* Grab socket reference count. This operation is valid only
429 when sk is ALREADY grabbed f.e. it is found in hash table
430 or a list and the lookup is made under lock preventing hash table
431 modifications.
434 static inline void sock_hold(struct sock *sk)
436 atomic_inc(&sk->sk_refcnt);
439 /* Ungrab socket in the context, which assumes that socket refcnt
440 cannot hit zero, f.e. it is true in context of any socketcall.
442 static inline void __sock_put(struct sock *sk)
444 atomic_dec(&sk->sk_refcnt);
447 static __inline__ int sk_del_node_init(struct sock *sk)
449 int rc = __sk_del_node_init(sk);
451 if (rc) {
452 /* paranoid for a while -acme */
453 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
454 __sock_put(sk);
456 return rc;
458 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
460 static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
462 if (sk_hashed(sk)) {
463 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
464 return 1;
466 return 0;
469 static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
471 int rc = __sk_nulls_del_node_init_rcu(sk);
473 if (rc) {
474 /* paranoid for a while -acme */
475 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
476 __sock_put(sk);
478 return rc;
481 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
483 hlist_add_head(&sk->sk_node, list);
486 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
488 sock_hold(sk);
489 __sk_add_node(sk, list);
492 static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
494 sock_hold(sk);
495 hlist_add_head_rcu(&sk->sk_node, list);
498 static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
500 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
503 static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
505 sock_hold(sk);
506 __sk_nulls_add_node_rcu(sk, list);
509 static __inline__ void __sk_del_bind_node(struct sock *sk)
511 __hlist_del(&sk->sk_bind_node);
514 static __inline__ void sk_add_bind_node(struct sock *sk,
515 struct hlist_head *list)
517 hlist_add_head(&sk->sk_bind_node, list);
520 #define sk_for_each(__sk, node, list) \
521 hlist_for_each_entry(__sk, node, list, sk_node)
522 #define sk_for_each_rcu(__sk, node, list) \
523 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
524 #define sk_nulls_for_each(__sk, node, list) \
525 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
526 #define sk_nulls_for_each_rcu(__sk, node, list) \
527 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
528 #define sk_for_each_from(__sk, node) \
529 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
530 hlist_for_each_entry_from(__sk, node, sk_node)
531 #define sk_nulls_for_each_from(__sk, node) \
532 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
533 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
534 #define sk_for_each_safe(__sk, node, tmp, list) \
535 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
536 #define sk_for_each_bound(__sk, node, list) \
537 hlist_for_each_entry(__sk, node, list, sk_bind_node)
539 /* Sock flags */
540 enum sock_flags {
541 SOCK_DEAD,
542 SOCK_DONE,
543 SOCK_URGINLINE,
544 SOCK_KEEPOPEN,
545 SOCK_LINGER,
546 SOCK_DESTROY,
547 SOCK_BROADCAST,
548 SOCK_TIMESTAMP,
549 SOCK_ZAPPED,
550 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
551 SOCK_DBG, /* %SO_DEBUG setting */
552 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
553 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
554 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
555 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
556 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
557 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
558 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
559 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
560 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
561 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
562 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
563 SOCK_FASYNC, /* fasync() active */
564 SOCK_RXQ_OVFL,
565 SOCK_ZEROCOPY, /* buffers from userspace */
568 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
570 nsk->sk_flags = osk->sk_flags;
573 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
575 __set_bit(flag, &sk->sk_flags);
578 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
580 __clear_bit(flag, &sk->sk_flags);
583 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
585 return test_bit(flag, &sk->sk_flags);
588 static inline void sk_acceptq_removed(struct sock *sk)
590 sk->sk_ack_backlog--;
593 static inline void sk_acceptq_added(struct sock *sk)
595 sk->sk_ack_backlog++;
598 static inline int sk_acceptq_is_full(struct sock *sk)
600 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
604 * Compute minimal free write space needed to queue new packets.
606 static inline int sk_stream_min_wspace(struct sock *sk)
608 return sk->sk_wmem_queued >> 1;
611 static inline int sk_stream_wspace(struct sock *sk)
613 return sk->sk_sndbuf - sk->sk_wmem_queued;
616 extern void sk_stream_write_space(struct sock *sk);
618 static inline int sk_stream_memory_free(struct sock *sk)
620 return sk->sk_wmem_queued < sk->sk_sndbuf;
623 /* OOB backlog add */
624 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
626 /* dont let skb dst not refcounted, we are going to leave rcu lock */
627 skb_dst_force(skb);
629 if (!sk->sk_backlog.tail)
630 sk->sk_backlog.head = skb;
631 else
632 sk->sk_backlog.tail->next = skb;
634 sk->sk_backlog.tail = skb;
635 skb->next = NULL;
639 * Take into account size of receive queue and backlog queue
641 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
643 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
645 return qsize + skb->truesize > sk->sk_rcvbuf;
648 /* The per-socket spinlock must be held here. */
649 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
651 if (sk_rcvqueues_full(sk, skb))
652 return -ENOBUFS;
654 __sk_add_backlog(sk, skb);
655 sk->sk_backlog.len += skb->truesize;
656 return 0;
659 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
661 return sk->sk_backlog_rcv(sk, skb);
664 static inline void sock_rps_record_flow(const struct sock *sk)
666 #ifdef CONFIG_RPS
667 struct rps_sock_flow_table *sock_flow_table;
669 rcu_read_lock();
670 sock_flow_table = rcu_dereference(rps_sock_flow_table);
671 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
672 rcu_read_unlock();
673 #endif
676 static inline void sock_rps_reset_flow(const struct sock *sk)
678 #ifdef CONFIG_RPS
679 struct rps_sock_flow_table *sock_flow_table;
681 rcu_read_lock();
682 sock_flow_table = rcu_dereference(rps_sock_flow_table);
683 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
684 rcu_read_unlock();
685 #endif
688 static inline void sock_rps_save_rxhash(struct sock *sk,
689 const struct sk_buff *skb)
691 #ifdef CONFIG_RPS
692 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
693 sock_rps_reset_flow(sk);
694 sk->sk_rxhash = skb->rxhash;
696 #endif
699 static inline void sock_rps_reset_rxhash(struct sock *sk)
701 #ifdef CONFIG_RPS
702 sock_rps_reset_flow(sk);
703 sk->sk_rxhash = 0;
704 #endif
707 #define sk_wait_event(__sk, __timeo, __condition) \
708 ({ int __rc; \
709 release_sock(__sk); \
710 __rc = __condition; \
711 if (!__rc) { \
712 *(__timeo) = schedule_timeout(*(__timeo)); \
714 lock_sock(__sk); \
715 __rc = __condition; \
716 __rc; \
719 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
720 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
721 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
722 extern int sk_stream_error(struct sock *sk, int flags, int err);
723 extern void sk_stream_kill_queues(struct sock *sk);
725 extern int sk_wait_data(struct sock *sk, long *timeo);
727 struct request_sock_ops;
728 struct timewait_sock_ops;
729 struct inet_hashinfo;
730 struct raw_hashinfo;
731 struct module;
733 /* Networking protocol blocks we attach to sockets.
734 * socket layer -> transport layer interface
735 * transport -> network interface is defined by struct inet_proto
737 struct proto {
738 void (*close)(struct sock *sk,
739 long timeout);
740 int (*connect)(struct sock *sk,
741 struct sockaddr *uaddr,
742 int addr_len);
743 int (*disconnect)(struct sock *sk, int flags);
745 struct sock * (*accept) (struct sock *sk, int flags, int *err);
747 int (*ioctl)(struct sock *sk, int cmd,
748 unsigned long arg);
749 int (*init)(struct sock *sk);
750 void (*destroy)(struct sock *sk);
751 void (*shutdown)(struct sock *sk, int how);
752 int (*setsockopt)(struct sock *sk, int level,
753 int optname, char __user *optval,
754 unsigned int optlen);
755 int (*getsockopt)(struct sock *sk, int level,
756 int optname, char __user *optval,
757 int __user *option);
758 #ifdef CONFIG_COMPAT
759 int (*compat_setsockopt)(struct sock *sk,
760 int level,
761 int optname, char __user *optval,
762 unsigned int optlen);
763 int (*compat_getsockopt)(struct sock *sk,
764 int level,
765 int optname, char __user *optval,
766 int __user *option);
767 int (*compat_ioctl)(struct sock *sk,
768 unsigned int cmd, unsigned long arg);
769 #endif
770 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
771 struct msghdr *msg, size_t len);
772 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
773 struct msghdr *msg,
774 size_t len, int noblock, int flags,
775 int *addr_len);
776 int (*sendpage)(struct sock *sk, struct page *page,
777 int offset, size_t size, int flags);
778 int (*bind)(struct sock *sk,
779 struct sockaddr *uaddr, int addr_len);
781 int (*backlog_rcv) (struct sock *sk,
782 struct sk_buff *skb);
784 /* Keeping track of sk's, looking them up, and port selection methods. */
785 void (*hash)(struct sock *sk);
786 void (*unhash)(struct sock *sk);
787 void (*rehash)(struct sock *sk);
788 int (*get_port)(struct sock *sk, unsigned short snum);
789 void (*clear_sk)(struct sock *sk, int size);
791 /* Keeping track of sockets in use */
792 #ifdef CONFIG_PROC_FS
793 unsigned int inuse_idx;
794 #endif
796 /* Memory pressure */
797 void (*enter_memory_pressure)(struct sock *sk);
798 atomic_long_t *memory_allocated; /* Current allocated memory. */
799 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
801 * Pressure flag: try to collapse.
802 * Technical note: it is used by multiple contexts non atomically.
803 * All the __sk_mem_schedule() is of this nature: accounting
804 * is strict, actions are advisory and have some latency.
806 int *memory_pressure;
807 long *sysctl_mem;
808 int *sysctl_wmem;
809 int *sysctl_rmem;
810 int max_header;
811 bool no_autobind;
813 struct kmem_cache *slab;
814 unsigned int obj_size;
815 int slab_flags;
817 struct percpu_counter *orphan_count;
819 struct request_sock_ops *rsk_prot;
820 struct timewait_sock_ops *twsk_prot;
822 union {
823 struct inet_hashinfo *hashinfo;
824 struct udp_table *udp_table;
825 struct raw_hashinfo *raw_hash;
826 } h;
828 struct module *owner;
830 char name[32];
832 struct list_head node;
833 #ifdef SOCK_REFCNT_DEBUG
834 atomic_t socks;
835 #endif
838 extern int proto_register(struct proto *prot, int alloc_slab);
839 extern void proto_unregister(struct proto *prot);
841 #ifdef SOCK_REFCNT_DEBUG
842 static inline void sk_refcnt_debug_inc(struct sock *sk)
844 atomic_inc(&sk->sk_prot->socks);
847 static inline void sk_refcnt_debug_dec(struct sock *sk)
849 atomic_dec(&sk->sk_prot->socks);
850 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
851 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
854 static inline void sk_refcnt_debug_release(const struct sock *sk)
856 if (atomic_read(&sk->sk_refcnt) != 1)
857 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
858 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
860 #else /* SOCK_REFCNT_DEBUG */
861 #define sk_refcnt_debug_inc(sk) do { } while (0)
862 #define sk_refcnt_debug_dec(sk) do { } while (0)
863 #define sk_refcnt_debug_release(sk) do { } while (0)
864 #endif /* SOCK_REFCNT_DEBUG */
867 #ifdef CONFIG_PROC_FS
868 /* Called with local bh disabled */
869 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
870 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
871 #else
872 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
873 int inc)
876 #endif
879 /* With per-bucket locks this operation is not-atomic, so that
880 * this version is not worse.
882 static inline void __sk_prot_rehash(struct sock *sk)
884 sk->sk_prot->unhash(sk);
885 sk->sk_prot->hash(sk);
888 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
890 /* About 10 seconds */
891 #define SOCK_DESTROY_TIME (10*HZ)
893 /* Sockets 0-1023 can't be bound to unless you are superuser */
894 #define PROT_SOCK 1024
896 #define SHUTDOWN_MASK 3
897 #define RCV_SHUTDOWN 1
898 #define SEND_SHUTDOWN 2
900 #define SOCK_SNDBUF_LOCK 1
901 #define SOCK_RCVBUF_LOCK 2
902 #define SOCK_BINDADDR_LOCK 4
903 #define SOCK_BINDPORT_LOCK 8
905 /* sock_iocb: used to kick off async processing of socket ios */
906 struct sock_iocb {
907 struct list_head list;
909 int flags;
910 int size;
911 struct socket *sock;
912 struct sock *sk;
913 struct scm_cookie *scm;
914 struct msghdr *msg, async_msg;
915 struct kiocb *kiocb;
918 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
920 return (struct sock_iocb *)iocb->private;
923 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
925 return si->kiocb;
928 struct socket_alloc {
929 struct socket socket;
930 struct inode vfs_inode;
933 static inline struct socket *SOCKET_I(struct inode *inode)
935 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
938 static inline struct inode *SOCK_INODE(struct socket *socket)
940 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
944 * Functions for memory accounting
946 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
947 extern void __sk_mem_reclaim(struct sock *sk);
949 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
950 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
951 #define SK_MEM_SEND 0
952 #define SK_MEM_RECV 1
954 static inline int sk_mem_pages(int amt)
956 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
959 static inline int sk_has_account(struct sock *sk)
961 /* return true if protocol supports memory accounting */
962 return !!sk->sk_prot->memory_allocated;
965 static inline int sk_wmem_schedule(struct sock *sk, int size)
967 if (!sk_has_account(sk))
968 return 1;
969 return size <= sk->sk_forward_alloc ||
970 __sk_mem_schedule(sk, size, SK_MEM_SEND);
973 static inline int sk_rmem_schedule(struct sock *sk, int size)
975 if (!sk_has_account(sk))
976 return 1;
977 return size <= sk->sk_forward_alloc ||
978 __sk_mem_schedule(sk, size, SK_MEM_RECV);
981 static inline void sk_mem_reclaim(struct sock *sk)
983 if (!sk_has_account(sk))
984 return;
985 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
986 __sk_mem_reclaim(sk);
989 static inline void sk_mem_reclaim_partial(struct sock *sk)
991 if (!sk_has_account(sk))
992 return;
993 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
994 __sk_mem_reclaim(sk);
997 static inline void sk_mem_charge(struct sock *sk, int size)
999 if (!sk_has_account(sk))
1000 return;
1001 sk->sk_forward_alloc -= size;
1004 static inline void sk_mem_uncharge(struct sock *sk, int size)
1006 if (!sk_has_account(sk))
1007 return;
1008 sk->sk_forward_alloc += size;
1011 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1013 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1014 sk->sk_wmem_queued -= skb->truesize;
1015 sk_mem_uncharge(sk, skb->truesize);
1016 __kfree_skb(skb);
1019 /* Used by processes to "lock" a socket state, so that
1020 * interrupts and bottom half handlers won't change it
1021 * from under us. It essentially blocks any incoming
1022 * packets, so that we won't get any new data or any
1023 * packets that change the state of the socket.
1025 * While locked, BH processing will add new packets to
1026 * the backlog queue. This queue is processed by the
1027 * owner of the socket lock right before it is released.
1029 * Since ~2.3.5 it is also exclusive sleep lock serializing
1030 * accesses from user process context.
1032 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1035 * Macro so as to not evaluate some arguments when
1036 * lockdep is not enabled.
1038 * Mark both the sk_lock and the sk_lock.slock as a
1039 * per-address-family lock class.
1041 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1042 do { \
1043 sk->sk_lock.owned = 0; \
1044 init_waitqueue_head(&sk->sk_lock.wq); \
1045 spin_lock_init(&(sk)->sk_lock.slock); \
1046 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1047 sizeof((sk)->sk_lock)); \
1048 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1049 (skey), (sname)); \
1050 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1051 } while (0)
1053 extern void lock_sock_nested(struct sock *sk, int subclass);
1055 static inline void lock_sock(struct sock *sk)
1057 lock_sock_nested(sk, 0);
1060 extern void release_sock(struct sock *sk);
1062 /* BH context may only use the following locking interface. */
1063 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1064 #define bh_lock_sock_nested(__sk) \
1065 spin_lock_nested(&((__sk)->sk_lock.slock), \
1066 SINGLE_DEPTH_NESTING)
1067 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1069 extern bool lock_sock_fast(struct sock *sk);
1071 * unlock_sock_fast - complement of lock_sock_fast
1072 * @sk: socket
1073 * @slow: slow mode
1075 * fast unlock socket for user context.
1076 * If slow mode is on, we call regular release_sock()
1078 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1080 if (slow)
1081 release_sock(sk);
1082 else
1083 spin_unlock_bh(&sk->sk_lock.slock);
1087 extern struct sock *sk_alloc(struct net *net, int family,
1088 gfp_t priority,
1089 struct proto *prot);
1090 extern void sk_free(struct sock *sk);
1091 extern void sk_release_kernel(struct sock *sk);
1092 extern struct sock *sk_clone(const struct sock *sk,
1093 const gfp_t priority);
1095 extern struct sk_buff *sock_wmalloc(struct sock *sk,
1096 unsigned long size, int force,
1097 gfp_t priority);
1098 extern struct sk_buff *sock_rmalloc(struct sock *sk,
1099 unsigned long size, int force,
1100 gfp_t priority);
1101 extern void sock_wfree(struct sk_buff *skb);
1102 extern void sock_rfree(struct sk_buff *skb);
1104 extern int sock_setsockopt(struct socket *sock, int level,
1105 int op, char __user *optval,
1106 unsigned int optlen);
1108 extern int sock_getsockopt(struct socket *sock, int level,
1109 int op, char __user *optval,
1110 int __user *optlen);
1111 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1112 unsigned long size,
1113 int noblock,
1114 int *errcode);
1115 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1116 unsigned long header_len,
1117 unsigned long data_len,
1118 int noblock,
1119 int *errcode);
1120 extern void *sock_kmalloc(struct sock *sk, int size,
1121 gfp_t priority);
1122 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1123 extern void sk_send_sigurg(struct sock *sk);
1125 #ifdef CONFIG_CGROUPS
1126 extern void sock_update_classid(struct sock *sk);
1127 #else
1128 static inline void sock_update_classid(struct sock *sk)
1131 #endif
1134 * Functions to fill in entries in struct proto_ops when a protocol
1135 * does not implement a particular function.
1137 extern int sock_no_bind(struct socket *,
1138 struct sockaddr *, int);
1139 extern int sock_no_connect(struct socket *,
1140 struct sockaddr *, int, int);
1141 extern int sock_no_socketpair(struct socket *,
1142 struct socket *);
1143 extern int sock_no_accept(struct socket *,
1144 struct socket *, int);
1145 extern int sock_no_getname(struct socket *,
1146 struct sockaddr *, int *, int);
1147 extern unsigned int sock_no_poll(struct file *, struct socket *,
1148 struct poll_table_struct *);
1149 extern int sock_no_ioctl(struct socket *, unsigned int,
1150 unsigned long);
1151 extern int sock_no_listen(struct socket *, int);
1152 extern int sock_no_shutdown(struct socket *, int);
1153 extern int sock_no_getsockopt(struct socket *, int , int,
1154 char __user *, int __user *);
1155 extern int sock_no_setsockopt(struct socket *, int, int,
1156 char __user *, unsigned int);
1157 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1158 struct msghdr *, size_t);
1159 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1160 struct msghdr *, size_t, int);
1161 extern int sock_no_mmap(struct file *file,
1162 struct socket *sock,
1163 struct vm_area_struct *vma);
1164 extern ssize_t sock_no_sendpage(struct socket *sock,
1165 struct page *page,
1166 int offset, size_t size,
1167 int flags);
1170 * Functions to fill in entries in struct proto_ops when a protocol
1171 * uses the inet style.
1173 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1174 char __user *optval, int __user *optlen);
1175 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1176 struct msghdr *msg, size_t size, int flags);
1177 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1178 char __user *optval, unsigned int optlen);
1179 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1180 int optname, char __user *optval, int __user *optlen);
1181 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1182 int optname, char __user *optval, unsigned int optlen);
1184 extern void sk_common_release(struct sock *sk);
1187 * Default socket callbacks and setup code
1190 /* Initialise core socket variables */
1191 extern void sock_init_data(struct socket *sock, struct sock *sk);
1193 extern void sk_filter_release_rcu(struct rcu_head *rcu);
1196 * sk_filter_release - release a socket filter
1197 * @fp: filter to remove
1199 * Remove a filter from a socket and release its resources.
1202 static inline void sk_filter_release(struct sk_filter *fp)
1204 if (atomic_dec_and_test(&fp->refcnt))
1205 call_rcu(&fp->rcu, sk_filter_release_rcu);
1208 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1210 unsigned int size = sk_filter_len(fp);
1212 atomic_sub(size, &sk->sk_omem_alloc);
1213 sk_filter_release(fp);
1216 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1218 atomic_inc(&fp->refcnt);
1219 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1223 * Socket reference counting postulates.
1225 * * Each user of socket SHOULD hold a reference count.
1226 * * Each access point to socket (an hash table bucket, reference from a list,
1227 * running timer, skb in flight MUST hold a reference count.
1228 * * When reference count hits 0, it means it will never increase back.
1229 * * When reference count hits 0, it means that no references from
1230 * outside exist to this socket and current process on current CPU
1231 * is last user and may/should destroy this socket.
1232 * * sk_free is called from any context: process, BH, IRQ. When
1233 * it is called, socket has no references from outside -> sk_free
1234 * may release descendant resources allocated by the socket, but
1235 * to the time when it is called, socket is NOT referenced by any
1236 * hash tables, lists etc.
1237 * * Packets, delivered from outside (from network or from another process)
1238 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1239 * when they sit in queue. Otherwise, packets will leak to hole, when
1240 * socket is looked up by one cpu and unhasing is made by another CPU.
1241 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1242 * (leak to backlog). Packet socket does all the processing inside
1243 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1244 * use separate SMP lock, so that they are prone too.
1247 /* Ungrab socket and destroy it, if it was the last reference. */
1248 static inline void sock_put(struct sock *sk)
1250 if (atomic_dec_and_test(&sk->sk_refcnt))
1251 sk_free(sk);
1254 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1255 const int nested);
1257 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1259 sk->sk_tx_queue_mapping = tx_queue;
1262 static inline void sk_tx_queue_clear(struct sock *sk)
1264 sk->sk_tx_queue_mapping = -1;
1267 static inline int sk_tx_queue_get(const struct sock *sk)
1269 return sk ? sk->sk_tx_queue_mapping : -1;
1272 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1274 sk_tx_queue_clear(sk);
1275 sk->sk_socket = sock;
1278 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1280 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1281 return &rcu_dereference_raw(sk->sk_wq)->wait;
1283 /* Detach socket from process context.
1284 * Announce socket dead, detach it from wait queue and inode.
1285 * Note that parent inode held reference count on this struct sock,
1286 * we do not release it in this function, because protocol
1287 * probably wants some additional cleanups or even continuing
1288 * to work with this socket (TCP).
1290 static inline void sock_orphan(struct sock *sk)
1292 write_lock_bh(&sk->sk_callback_lock);
1293 sock_set_flag(sk, SOCK_DEAD);
1294 sk_set_socket(sk, NULL);
1295 sk->sk_wq = NULL;
1296 write_unlock_bh(&sk->sk_callback_lock);
1299 static inline void sock_graft(struct sock *sk, struct socket *parent)
1301 write_lock_bh(&sk->sk_callback_lock);
1302 sk->sk_wq = parent->wq;
1303 parent->sk = sk;
1304 sk_set_socket(sk, parent);
1305 security_sock_graft(sk, parent);
1306 write_unlock_bh(&sk->sk_callback_lock);
1309 extern int sock_i_uid(struct sock *sk);
1310 extern unsigned long sock_i_ino(struct sock *sk);
1312 static inline struct dst_entry *
1313 __sk_dst_get(struct sock *sk)
1315 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1316 lockdep_is_held(&sk->sk_lock.slock));
1319 static inline struct dst_entry *
1320 sk_dst_get(struct sock *sk)
1322 struct dst_entry *dst;
1324 rcu_read_lock();
1325 dst = rcu_dereference(sk->sk_dst_cache);
1326 if (dst)
1327 dst_hold(dst);
1328 rcu_read_unlock();
1329 return dst;
1332 extern void sk_reset_txq(struct sock *sk);
1334 static inline void dst_negative_advice(struct sock *sk)
1336 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1338 if (dst && dst->ops->negative_advice) {
1339 ndst = dst->ops->negative_advice(dst);
1341 if (ndst != dst) {
1342 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1343 sk_reset_txq(sk);
1348 static inline void
1349 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1351 struct dst_entry *old_dst;
1353 sk_tx_queue_clear(sk);
1355 * This can be called while sk is owned by the caller only,
1356 * with no state that can be checked in a rcu_dereference_check() cond
1358 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1359 rcu_assign_pointer(sk->sk_dst_cache, dst);
1360 dst_release(old_dst);
1363 static inline void
1364 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1366 spin_lock(&sk->sk_dst_lock);
1367 __sk_dst_set(sk, dst);
1368 spin_unlock(&sk->sk_dst_lock);
1371 static inline void
1372 __sk_dst_reset(struct sock *sk)
1374 __sk_dst_set(sk, NULL);
1377 static inline void
1378 sk_dst_reset(struct sock *sk)
1380 spin_lock(&sk->sk_dst_lock);
1381 __sk_dst_reset(sk);
1382 spin_unlock(&sk->sk_dst_lock);
1385 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1387 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1389 static inline int sk_can_gso(const struct sock *sk)
1391 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1394 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1396 static inline void sk_nocaps_add(struct sock *sk, int flags)
1398 sk->sk_route_nocaps |= flags;
1399 sk->sk_route_caps &= ~flags;
1402 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1403 char __user *from, char *to,
1404 int copy, int offset)
1406 if (skb->ip_summed == CHECKSUM_NONE) {
1407 int err = 0;
1408 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1409 if (err)
1410 return err;
1411 skb->csum = csum_block_add(skb->csum, csum, offset);
1412 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1413 if (!access_ok(VERIFY_READ, from, copy) ||
1414 __copy_from_user_nocache(to, from, copy))
1415 return -EFAULT;
1416 } else if (copy_from_user(to, from, copy))
1417 return -EFAULT;
1419 return 0;
1422 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1423 char __user *from, int copy)
1425 int err, offset = skb->len;
1427 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1428 copy, offset);
1429 if (err)
1430 __skb_trim(skb, offset);
1432 return err;
1435 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1436 struct sk_buff *skb,
1437 struct page *page,
1438 int off, int copy)
1440 int err;
1442 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1443 copy, skb->len);
1444 if (err)
1445 return err;
1447 skb->len += copy;
1448 skb->data_len += copy;
1449 skb->truesize += copy;
1450 sk->sk_wmem_queued += copy;
1451 sk_mem_charge(sk, copy);
1452 return 0;
1455 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1456 struct sk_buff *skb, struct page *page,
1457 int off, int copy)
1459 if (skb->ip_summed == CHECKSUM_NONE) {
1460 int err = 0;
1461 __wsum csum = csum_and_copy_from_user(from,
1462 page_address(page) + off,
1463 copy, 0, &err);
1464 if (err)
1465 return err;
1466 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1467 } else if (copy_from_user(page_address(page) + off, from, copy))
1468 return -EFAULT;
1470 skb->len += copy;
1471 skb->data_len += copy;
1472 skb->truesize += copy;
1473 sk->sk_wmem_queued += copy;
1474 sk_mem_charge(sk, copy);
1475 return 0;
1479 * sk_wmem_alloc_get - returns write allocations
1480 * @sk: socket
1482 * Returns sk_wmem_alloc minus initial offset of one
1484 static inline int sk_wmem_alloc_get(const struct sock *sk)
1486 return atomic_read(&sk->sk_wmem_alloc) - 1;
1490 * sk_rmem_alloc_get - returns read allocations
1491 * @sk: socket
1493 * Returns sk_rmem_alloc
1495 static inline int sk_rmem_alloc_get(const struct sock *sk)
1497 return atomic_read(&sk->sk_rmem_alloc);
1501 * sk_has_allocations - check if allocations are outstanding
1502 * @sk: socket
1504 * Returns true if socket has write or read allocations
1506 static inline int sk_has_allocations(const struct sock *sk)
1508 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1512 * wq_has_sleeper - check if there are any waiting processes
1513 * @wq: struct socket_wq
1515 * Returns true if socket_wq has waiting processes
1517 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1518 * barrier call. They were added due to the race found within the tcp code.
1520 * Consider following tcp code paths:
1522 * CPU1 CPU2
1524 * sys_select receive packet
1525 * ... ...
1526 * __add_wait_queue update tp->rcv_nxt
1527 * ... ...
1528 * tp->rcv_nxt check sock_def_readable
1529 * ... {
1530 * schedule rcu_read_lock();
1531 * wq = rcu_dereference(sk->sk_wq);
1532 * if (wq && waitqueue_active(&wq->wait))
1533 * wake_up_interruptible(&wq->wait)
1534 * ...
1537 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1538 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1539 * could then endup calling schedule and sleep forever if there are no more
1540 * data on the socket.
1543 static inline bool wq_has_sleeper(struct socket_wq *wq)
1547 * We need to be sure we are in sync with the
1548 * add_wait_queue modifications to the wait queue.
1550 * This memory barrier is paired in the sock_poll_wait.
1552 smp_mb();
1553 return wq && waitqueue_active(&wq->wait);
1557 * sock_poll_wait - place memory barrier behind the poll_wait call.
1558 * @filp: file
1559 * @wait_address: socket wait queue
1560 * @p: poll_table
1562 * See the comments in the wq_has_sleeper function.
1564 static inline void sock_poll_wait(struct file *filp,
1565 wait_queue_head_t *wait_address, poll_table *p)
1567 if (p && wait_address) {
1568 poll_wait(filp, wait_address, p);
1570 * We need to be sure we are in sync with the
1571 * socket flags modification.
1573 * This memory barrier is paired in the wq_has_sleeper.
1575 smp_mb();
1580 * Queue a received datagram if it will fit. Stream and sequenced
1581 * protocols can't normally use this as they need to fit buffers in
1582 * and play with them.
1584 * Inlined as it's very short and called for pretty much every
1585 * packet ever received.
1588 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1590 skb_orphan(skb);
1591 skb->sk = sk;
1592 skb->destructor = sock_wfree;
1594 * We used to take a refcount on sk, but following operation
1595 * is enough to guarantee sk_free() wont free this sock until
1596 * all in-flight packets are completed
1598 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1601 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1603 skb_orphan(skb);
1604 skb->sk = sk;
1605 skb->destructor = sock_rfree;
1606 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1607 sk_mem_charge(sk, skb->truesize);
1610 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1611 unsigned long expires);
1613 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1615 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1617 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1620 * Recover an error report and clear atomically
1623 static inline int sock_error(struct sock *sk)
1625 int err;
1626 if (likely(!sk->sk_err))
1627 return 0;
1628 err = xchg(&sk->sk_err, 0);
1629 return -err;
1632 static inline unsigned long sock_wspace(struct sock *sk)
1634 int amt = 0;
1636 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1637 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1638 if (amt < 0)
1639 amt = 0;
1641 return amt;
1644 static inline void sk_wake_async(struct sock *sk, int how, int band)
1646 if (sock_flag(sk, SOCK_FASYNC))
1647 sock_wake_async(sk->sk_socket, how, band);
1650 #define SOCK_MIN_SNDBUF 2048
1652 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1653 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1655 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1657 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1659 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1660 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1661 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1665 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1667 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1669 struct page *page = NULL;
1671 page = alloc_pages(sk->sk_allocation, 0);
1672 if (!page) {
1673 sk->sk_prot->enter_memory_pressure(sk);
1674 sk_stream_moderate_sndbuf(sk);
1676 return page;
1680 * Default write policy as shown to user space via poll/select/SIGIO
1682 static inline int sock_writeable(const struct sock *sk)
1684 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1687 static inline gfp_t gfp_any(void)
1689 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1692 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1694 return noblock ? 0 : sk->sk_rcvtimeo;
1697 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1699 return noblock ? 0 : sk->sk_sndtimeo;
1702 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1704 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1707 /* Alas, with timeout socket operations are not restartable.
1708 * Compare this to poll().
1710 static inline int sock_intr_errno(long timeo)
1712 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1715 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1716 struct sk_buff *skb);
1718 static __inline__ void
1719 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1721 ktime_t kt = skb->tstamp;
1722 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
1725 * generate control messages if
1726 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1727 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1728 * - software time stamp available and wanted
1729 * (SOCK_TIMESTAMPING_SOFTWARE)
1730 * - hardware time stamps available and wanted
1731 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1732 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1734 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1735 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1736 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1737 (hwtstamps->hwtstamp.tv64 &&
1738 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1739 (hwtstamps->syststamp.tv64 &&
1740 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
1741 __sock_recv_timestamp(msg, sk, skb);
1742 else
1743 sk->sk_stamp = kt;
1746 extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1747 struct sk_buff *skb);
1749 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1750 struct sk_buff *skb)
1752 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
1753 (1UL << SOCK_RCVTSTAMP) | \
1754 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
1755 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
1756 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
1757 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
1759 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
1760 __sock_recv_ts_and_drops(msg, sk, skb);
1761 else
1762 sk->sk_stamp = skb->tstamp;
1766 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1767 * @sk: socket sending this packet
1768 * @tx_flags: filled with instructions for time stamping
1770 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1771 * parameters are invalid.
1773 extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
1776 * sk_eat_skb - Release a skb if it is no longer needed
1777 * @sk: socket to eat this skb from
1778 * @skb: socket buffer to eat
1779 * @copied_early: flag indicating whether DMA operations copied this data early
1781 * This routine must be called with interrupts disabled or with the socket
1782 * locked so that the sk_buff queue operation is ok.
1784 #ifdef CONFIG_NET_DMA
1785 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1787 __skb_unlink(skb, &sk->sk_receive_queue);
1788 if (!copied_early)
1789 __kfree_skb(skb);
1790 else
1791 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1793 #else
1794 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1796 __skb_unlink(skb, &sk->sk_receive_queue);
1797 __kfree_skb(skb);
1799 #endif
1801 static inline
1802 struct net *sock_net(const struct sock *sk)
1804 return read_pnet(&sk->sk_net);
1807 static inline
1808 void sock_net_set(struct sock *sk, struct net *net)
1810 write_pnet(&sk->sk_net, net);
1814 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1815 * They should not hold a reference to a namespace in order to allow
1816 * to stop it.
1817 * Sockets after sk_change_net should be released using sk_release_kernel
1819 static inline void sk_change_net(struct sock *sk, struct net *net)
1821 put_net(sock_net(sk));
1822 sock_net_set(sk, hold_net(net));
1825 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
1827 if (unlikely(skb->sk)) {
1828 struct sock *sk = skb->sk;
1830 skb->destructor = NULL;
1831 skb->sk = NULL;
1832 return sk;
1834 return NULL;
1837 extern void sock_enable_timestamp(struct sock *sk, int flag);
1838 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1839 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1842 * Enable debug/info messages
1844 extern int net_msg_warn;
1845 #define NETDEBUG(fmt, args...) \
1846 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1848 #define LIMIT_NETDEBUG(fmt, args...) \
1849 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1851 extern __u32 sysctl_wmem_max;
1852 extern __u32 sysctl_rmem_max;
1854 extern void sk_init(void);
1856 extern int sysctl_optmem_max;
1858 extern __u32 sysctl_wmem_default;
1859 extern __u32 sysctl_rmem_default;
1861 #endif /* _SOCK_H */