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.
5 *
6 * Definitions for the AF_INET socket handler.
7 *
8 * Version: @(#)sock.h 1.0.4 05/13/93
9 *
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>
14 *
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
33 *
34 *
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.
39 */
40 #ifndef _SOCK_H
41 #define _SOCK_H
42
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 */
52 #include <linux/mm.h>
53 #include <linux/security.h>
54
55 #include <linux/filter.h>
56 #include <linux/rculist_nulls.h>
57 #include <linux/poll.h>
58
59 #include <asm/atomic.h>
60 #include <net/dst.h>
61 #include <net/checksum.h>
62
63 /*
64 * This structure really needs to be cleaned up.
65 * Most of it is for TCP, and not used by any of
66 * the other protocols.
67 */
68
69 /* Define this to get the SOCK_DBG debugging facility. */
70 #define SOCK_DEBUGGING
71 #ifdef SOCK_DEBUGGING
72 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
73 printk(KERN_DEBUG msg); } while (0)
74 #else
75 /* Validate arguments and do nothing */
76 static void inline int __attribute__ ((format (printf, 2, 3)))
77 SOCK_DEBUG(struct sock *sk, const char *msg, ...)
78 {
79 }
80 #endif
81
82 /* This is the per-socket lock. The spinlock provides a synchronization
83 * between user contexts and software interrupt processing, whereas the
84 * mini-semaphore synchronizes multiple users amongst themselves.
85 */
86 typedef struct {
87 spinlock_t slock;
88 int owned;
89 wait_queue_head_t wq;
90 /*
91 * We express the mutex-alike socket_lock semantics
92 * to the lock validator by explicitly managing
93 * the slock as a lock variant (in addition to
94 * the slock itself):
95 */
96 #ifdef CONFIG_DEBUG_LOCK_ALLOC
97 struct lockdep_map dep_map;
98 #endif
99 } socket_lock_t;
100
101 struct sock;
102 struct proto;
103 struct net;
104
105 /**
106 * struct sock_common - minimal network layer representation of sockets
107 * @skc_node: main hash linkage for various protocol lookup tables
108 * @skc_nulls_node: main hash linkage for UDP/UDP-Lite protocol
109 * @skc_refcnt: reference count
110 * @skc_hash: hash value used with various protocol lookup tables
111 * @skc_family: network address family
112 * @skc_state: Connection state
113 * @skc_reuse: %SO_REUSEADDR setting
114 * @skc_bound_dev_if: bound device index if != 0
115 * @skc_bind_node: bind hash linkage for various protocol lookup tables
116 * @skc_prot: protocol handlers inside a network family
117 * @skc_net: reference to the network namespace of this socket
118 *
119 * This is the minimal network layer representation of sockets, the header
120 * for struct sock and struct inet_timewait_sock.
121 */
122 struct sock_common {
123 /*
124 * first fields are not copied in sock_copy()
125 */
126 union {
127 struct hlist_node skc_node;
128 struct hlist_nulls_node skc_nulls_node;
129 };
130 atomic_t skc_refcnt;
131
132 unsigned int skc_hash;
133 unsigned short skc_family;
134 volatile unsigned char skc_state;
135 unsigned char skc_reuse;
136 int skc_bound_dev_if;
137 struct hlist_node skc_bind_node;
138 struct proto *skc_prot;
139 #ifdef CONFIG_NET_NS
140 struct net *skc_net;
141 #endif
142 };
143
144 /**
145 * struct sock - network layer representation of sockets
146 * @__sk_common: shared layout with inet_timewait_sock
147 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
148 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
149 * @sk_lock: synchronizer
150 * @sk_rcvbuf: size of receive buffer in bytes
151 * @sk_sleep: sock wait queue
152 * @sk_dst_cache: destination cache
153 * @sk_dst_lock: destination cache lock
154 * @sk_policy: flow policy
155 * @sk_rmem_alloc: receive queue bytes committed
156 * @sk_receive_queue: incoming packets
157 * @sk_wmem_alloc: transmit queue bytes committed
158 * @sk_write_queue: Packet sending queue
159 * @sk_async_wait_queue: DMA copied packets
160 * @sk_omem_alloc: "o" is "option" or "other"
161 * @sk_wmem_queued: persistent queue size
162 * @sk_forward_alloc: space allocated forward
163 * @sk_allocation: allocation mode
164 * @sk_sndbuf: size of send buffer in bytes
165 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
166 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
167 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
168 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
169 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
170 * @sk_gso_max_size: Maximum GSO segment size to build
171 * @sk_lingertime: %SO_LINGER l_linger setting
172 * @sk_backlog: always used with the per-socket spinlock held
173 * @sk_callback_lock: used with the callbacks in the end of this struct
174 * @sk_error_queue: rarely used
175 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
176 * IPV6_ADDRFORM for instance)
177 * @sk_err: last error
178 * @sk_err_soft: errors that don't cause failure but are the cause of a
179 * persistent failure not just 'timed out'
180 * @sk_drops: raw/udp drops counter
181 * @sk_ack_backlog: current listen backlog
182 * @sk_max_ack_backlog: listen backlog set in listen()
183 * @sk_priority: %SO_PRIORITY setting
184 * @sk_type: socket type (%SOCK_STREAM, etc)
185 * @sk_protocol: which protocol this socket belongs in this network family
186 * @sk_peercred: %SO_PEERCRED setting
187 * @sk_rcvlowat: %SO_RCVLOWAT setting
188 * @sk_rcvtimeo: %SO_RCVTIMEO setting
189 * @sk_sndtimeo: %SO_SNDTIMEO setting
190 * @sk_filter: socket filtering instructions
191 * @sk_protinfo: private area, net family specific, when not using slab
192 * @sk_timer: sock cleanup timer
193 * @sk_stamp: time stamp of last packet received
194 * @sk_socket: Identd and reporting IO signals
195 * @sk_user_data: RPC layer private data
196 * @sk_sndmsg_page: cached page for sendmsg
197 * @sk_sndmsg_off: cached offset for sendmsg
198 * @sk_send_head: front of stuff to transmit
199 * @sk_security: used by security modules
200 * @sk_mark: generic packet mark
201 * @sk_write_pending: a write to stream socket waits to start
202 * @sk_state_change: callback to indicate change in the state of the sock
203 * @sk_data_ready: callback to indicate there is data to be processed
204 * @sk_write_space: callback to indicate there is bf sending space available
205 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
206 * @sk_backlog_rcv: callback to process the backlog
207 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
208 */
209 struct sock {
210 /*
211 * Now struct inet_timewait_sock also uses sock_common, so please just
212 * don't add nothing before this first member (__sk_common) --acme
213 */
214 struct sock_common __sk_common;
215 #define sk_node __sk_common.skc_node
216 #define sk_nulls_node __sk_common.skc_nulls_node
217 #define sk_refcnt __sk_common.skc_refcnt
218
219 #define sk_copy_start __sk_common.skc_hash
220 #define sk_hash __sk_common.skc_hash
221 #define sk_family __sk_common.skc_family
222 #define sk_state __sk_common.skc_state
223 #define sk_reuse __sk_common.skc_reuse
224 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
225 #define sk_bind_node __sk_common.skc_bind_node
226 #define sk_prot __sk_common.skc_prot
227 #define sk_net __sk_common.skc_net
228 kmemcheck_bitfield_begin(flags);
229 unsigned int sk_shutdown : 2,
230 sk_no_check : 2,
231 sk_userlocks : 4,
232 sk_protocol : 8,
233 sk_type : 16;
234 #define SK_PROTOCOL_MAX ((u8)~0U)
235 kmemcheck_bitfield_end(flags);
236 int sk_rcvbuf;
237 socket_lock_t sk_lock;
238 /*
239 * The backlog queue is special, it is always used with
240 * the per-socket spinlock held and requires low latency
241 * access. Therefore we special case it's implementation.
242 */
243 struct {
244 struct sk_buff *head;
245 struct sk_buff *tail;
246 } sk_backlog;
247 wait_queue_head_t *sk_sleep;
248 struct dst_entry *sk_dst_cache;
249 #ifdef CONFIG_XFRM
250 struct xfrm_policy *sk_policy[2];
251 #endif
252 rwlock_t sk_dst_lock;
253 atomic_t sk_rmem_alloc;
254 atomic_t sk_wmem_alloc;
255 atomic_t sk_omem_alloc;
256 int sk_sndbuf;
257 struct sk_buff_head sk_receive_queue;
258 struct sk_buff_head sk_write_queue;
259 #ifdef CONFIG_NET_DMA
260 struct sk_buff_head sk_async_wait_queue;
261 #endif
262 int sk_wmem_queued;
263 int sk_forward_alloc;
264 gfp_t sk_allocation;
265 int sk_route_caps;
266 int sk_gso_type;
267 unsigned int sk_gso_max_size;
268 int sk_rcvlowat;
269 unsigned long sk_flags;
270 unsigned long sk_lingertime;
271 struct sk_buff_head sk_error_queue;
272 struct proto *sk_prot_creator;
273 rwlock_t sk_callback_lock;
274 int sk_err,
275 sk_err_soft;
276 atomic_t sk_drops;
277 unsigned short sk_ack_backlog;
278 unsigned short sk_max_ack_backlog;
279 __u32 sk_priority;
280 struct ucred sk_peercred;
281 long sk_rcvtimeo;
282 long sk_sndtimeo;
283 struct sk_filter *sk_filter;
284 void *sk_protinfo;
285 struct timer_list sk_timer;
286 ktime_t sk_stamp;
287 struct socket *sk_socket;
288 void *sk_user_data;
289 struct page *sk_sndmsg_page;
290 struct sk_buff *sk_send_head;
291 __u32 sk_sndmsg_off;
292 int sk_write_pending;
293 #ifdef CONFIG_SECURITY
294 void *sk_security;
295 #endif
296 __u32 sk_mark;
297 /* XXX 4 bytes hole on 64 bit */
298 void (*sk_state_change)(struct sock *sk);
299 void (*sk_data_ready)(struct sock *sk, int bytes);
300 void (*sk_write_space)(struct sock *sk);
301 void (*sk_error_report)(struct sock *sk);
302 int (*sk_backlog_rcv)(struct sock *sk,
303 struct sk_buff *skb);
304 void (*sk_destruct)(struct sock *sk);
305 };
306
307 /*
308 * Hashed lists helper routines
309 */
310 static inline struct sock *__sk_head(const struct hlist_head *head)
311 {
312 return hlist_entry(head->first, struct sock, sk_node);
313 }
314
315 static inline struct sock *sk_head(const struct hlist_head *head)
316 {
317 return hlist_empty(head) ? NULL : __sk_head(head);
318 }
319
320 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
321 {
322 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
323 }
324
325 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
326 {
327 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
328 }
329
330 static inline struct sock *sk_next(const struct sock *sk)
331 {
332 return sk->sk_node.next ?
333 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
334 }
335
336 static inline struct sock *sk_nulls_next(const struct sock *sk)
337 {
338 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
339 hlist_nulls_entry(sk->sk_nulls_node.next,
340 struct sock, sk_nulls_node) :
341 NULL;
342 }
343
344 static inline int sk_unhashed(const struct sock *sk)
345 {
346 return hlist_unhashed(&sk->sk_node);
347 }
348
349 static inline int sk_hashed(const struct sock *sk)
350 {
351 return !sk_unhashed(sk);
352 }
353
354 static __inline__ void sk_node_init(struct hlist_node *node)
355 {
356 node->pprev = NULL;
357 }
358
359 static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
360 {
361 node->pprev = NULL;
362 }
363
364 static __inline__ void __sk_del_node(struct sock *sk)
365 {
366 __hlist_del(&sk->sk_node);
367 }
368
369 static __inline__ int __sk_del_node_init(struct sock *sk)
370 {
371 if (sk_hashed(sk)) {
372 __sk_del_node(sk);
373 sk_node_init(&sk->sk_node);
374 return 1;
375 }
376 return 0;
377 }
378
379 /* Grab socket reference count. This operation is valid only
380 when sk is ALREADY grabbed f.e. it is found in hash table
381 or a list and the lookup is made under lock preventing hash table
382 modifications.
383 */
384
385 static inline void sock_hold(struct sock *sk)
386 {
387 atomic_inc(&sk->sk_refcnt);
388 }
389
390 /* Ungrab socket in the context, which assumes that socket refcnt
391 cannot hit zero, f.e. it is true in context of any socketcall.
392 */
393 static inline void __sock_put(struct sock *sk)
394 {
395 atomic_dec(&sk->sk_refcnt);
396 }
397
398 static __inline__ int sk_del_node_init(struct sock *sk)
399 {
400 int rc = __sk_del_node_init(sk);
401
402 if (rc) {
403 /* paranoid for a while -acme */
404 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
405 __sock_put(sk);
406 }
407 return rc;
408 }
409
410 static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
411 {
412 if (sk_hashed(sk)) {
413 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
414 return 1;
415 }
416 return 0;
417 }
418
419 static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
420 {
421 int rc = __sk_nulls_del_node_init_rcu(sk);
422
423 if (rc) {
424 /* paranoid for a while -acme */
425 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
426 __sock_put(sk);
427 }
428 return rc;
429 }
430
431 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
432 {
433 hlist_add_head(&sk->sk_node, list);
434 }
435
436 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
437 {
438 sock_hold(sk);
439 __sk_add_node(sk, list);
440 }
441
442 static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
443 {
444 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
445 }
446
447 static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
448 {
449 sock_hold(sk);
450 __sk_nulls_add_node_rcu(sk, list);
451 }
452
453 static __inline__ void __sk_del_bind_node(struct sock *sk)
454 {
455 __hlist_del(&sk->sk_bind_node);
456 }
457
458 static __inline__ void sk_add_bind_node(struct sock *sk,
459 struct hlist_head *list)
460 {
461 hlist_add_head(&sk->sk_bind_node, list);
462 }
463
464 #define sk_for_each(__sk, node, list) \
465 hlist_for_each_entry(__sk, node, list, sk_node)
466 #define sk_nulls_for_each(__sk, node, list) \
467 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
468 #define sk_nulls_for_each_rcu(__sk, node, list) \
469 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
470 #define sk_for_each_from(__sk, node) \
471 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
472 hlist_for_each_entry_from(__sk, node, sk_node)
473 #define sk_nulls_for_each_from(__sk, node) \
474 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
475 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
476 #define sk_for_each_continue(__sk, node) \
477 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
478 hlist_for_each_entry_continue(__sk, node, sk_node)
479 #define sk_for_each_safe(__sk, node, tmp, list) \
480 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
481 #define sk_for_each_bound(__sk, node, list) \
482 hlist_for_each_entry(__sk, node, list, sk_bind_node)
483
484 /* Sock flags */
485 enum sock_flags {
486 SOCK_DEAD,
487 SOCK_DONE,
488 SOCK_URGINLINE,
489 SOCK_KEEPOPEN,
490 SOCK_LINGER,
491 SOCK_DESTROY,
492 SOCK_BROADCAST,
493 SOCK_TIMESTAMP,
494 SOCK_ZAPPED,
495 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
496 SOCK_DBG, /* %SO_DEBUG setting */
497 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
498 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
499 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
500 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
501 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
502 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
503 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
504 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
505 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
506 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
507 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
508 };
509
510 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
511
512 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
513 {
514 nsk->sk_flags = osk->sk_flags;
515 }
516
517 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
518 {
519 __set_bit(flag, &sk->sk_flags);
520 }
521
522 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
523 {
524 __clear_bit(flag, &sk->sk_flags);
525 }
526
527 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
528 {
529 return test_bit(flag, &sk->sk_flags);
530 }
531
532 static inline void sk_acceptq_removed(struct sock *sk)
533 {
534 sk->sk_ack_backlog--;
535 }
536
537 static inline void sk_acceptq_added(struct sock *sk)
538 {
539 sk->sk_ack_backlog++;
540 }
541
542 static inline int sk_acceptq_is_full(struct sock *sk)
543 {
544 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
545 }
546
547 /*
548 * Compute minimal free write space needed to queue new packets.
549 */
550 static inline int sk_stream_min_wspace(struct sock *sk)
551 {
552 return sk->sk_wmem_queued >> 1;
553 }
554
555 static inline int sk_stream_wspace(struct sock *sk)
556 {
557 return sk->sk_sndbuf - sk->sk_wmem_queued;
558 }
559
560 extern void sk_stream_write_space(struct sock *sk);
561
562 static inline int sk_stream_memory_free(struct sock *sk)
563 {
564 return sk->sk_wmem_queued < sk->sk_sndbuf;
565 }
566
567 /* The per-socket spinlock must be held here. */
568 static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb)
569 {
570 if (!sk->sk_backlog.tail) {
571 sk->sk_backlog.head = sk->sk_backlog.tail = skb;
572 } else {
573 sk->sk_backlog.tail->next = skb;
574 sk->sk_backlog.tail = skb;
575 }
576 skb->next = NULL;
577 }
578
579 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
580 {
581 return sk->sk_backlog_rcv(sk, skb);
582 }
583
584 #define sk_wait_event(__sk, __timeo, __condition) \
585 ({ int __rc; \
586 release_sock(__sk); \
587 __rc = __condition; \
588 if (!__rc) { \
589 *(__timeo) = schedule_timeout(*(__timeo)); \
590 } \
591 lock_sock(__sk); \
592 __rc = __condition; \
593 __rc; \
594 })
595
596 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
597 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
598 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
599 extern int sk_stream_error(struct sock *sk, int flags, int err);
600 extern void sk_stream_kill_queues(struct sock *sk);
601
602 extern int sk_wait_data(struct sock *sk, long *timeo);
603
604 struct request_sock_ops;
605 struct timewait_sock_ops;
606 struct inet_hashinfo;
607 struct raw_hashinfo;
608
609 /* Networking protocol blocks we attach to sockets.
610 * socket layer -> transport layer interface
611 * transport -> network interface is defined by struct inet_proto
612 */
613 struct proto {
614 void (*close)(struct sock *sk,
615 long timeout);
616 int (*connect)(struct sock *sk,
617 struct sockaddr *uaddr,
618 int addr_len);
619 int (*disconnect)(struct sock *sk, int flags);
620
621 struct sock * (*accept) (struct sock *sk, int flags, int *err);
622
623 int (*ioctl)(struct sock *sk, int cmd,
624 unsigned long arg);
625 int (*init)(struct sock *sk);
626 void (*destroy)(struct sock *sk);
627 void (*shutdown)(struct sock *sk, int how);
628 int (*setsockopt)(struct sock *sk, int level,
629 int optname, char __user *optval,
630 unsigned int optlen);
631 int (*getsockopt)(struct sock *sk, int level,
632 int optname, char __user *optval,
633 int __user *option);
634 #ifdef CONFIG_COMPAT
635 int (*compat_setsockopt)(struct sock *sk,
636 int level,
637 int optname, char __user *optval,
638 unsigned int optlen);
639 int (*compat_getsockopt)(struct sock *sk,
640 int level,
641 int optname, char __user *optval,
642 int __user *option);
643 #endif
644 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
645 struct msghdr *msg, size_t len);
646 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
647 struct msghdr *msg,
648 size_t len, int noblock, int flags,
649 int *addr_len);
650 int (*sendpage)(struct sock *sk, struct page *page,
651 int offset, size_t size, int flags);
652 int (*bind)(struct sock *sk,
653 struct sockaddr *uaddr, int addr_len);
654
655 int (*backlog_rcv) (struct sock *sk,
656 struct sk_buff *skb);
657
658 /* Keeping track of sk's, looking them up, and port selection methods. */
659 void (*hash)(struct sock *sk);
660 void (*unhash)(struct sock *sk);
661 int (*get_port)(struct sock *sk, unsigned short snum);
662
663 /* Keeping track of sockets in use */
664 #ifdef CONFIG_PROC_FS
665 unsigned int inuse_idx;
666 #endif
667
668 /* Memory pressure */
669 void (*enter_memory_pressure)(struct sock *sk);
670 atomic_t *memory_allocated; /* Current allocated memory. */
671 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
672 /*
673 * Pressure flag: try to collapse.
674 * Technical note: it is used by multiple contexts non atomically.
675 * All the __sk_mem_schedule() is of this nature: accounting
676 * is strict, actions are advisory and have some latency.
677 */
678 int *memory_pressure;
679 int *sysctl_mem;
680 int *sysctl_wmem;
681 int *sysctl_rmem;
682 int max_header;
683
684 struct kmem_cache *slab;
685 unsigned int obj_size;
686 int slab_flags;
687
688 struct percpu_counter *orphan_count;
689
690 struct request_sock_ops *rsk_prot;
691 struct timewait_sock_ops *twsk_prot;
692
693 union {
694 struct inet_hashinfo *hashinfo;
695 struct udp_table *udp_table;
696 struct raw_hashinfo *raw_hash;
697 } h;
698
699 struct module *owner;
700
701 char name[32];
702
703 struct list_head node;
704 #ifdef SOCK_REFCNT_DEBUG
705 atomic_t socks;
706 #endif
707 };
708
709 extern int proto_register(struct proto *prot, int alloc_slab);
710 extern void proto_unregister(struct proto *prot);
711
712 #ifdef SOCK_REFCNT_DEBUG
713 static inline void sk_refcnt_debug_inc(struct sock *sk)
714 {
715 atomic_inc(&sk->sk_prot->socks);
716 }
717
718 static inline void sk_refcnt_debug_dec(struct sock *sk)
719 {
720 atomic_dec(&sk->sk_prot->socks);
721 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
722 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
723 }
724
725 static inline void sk_refcnt_debug_release(const struct sock *sk)
726 {
727 if (atomic_read(&sk->sk_refcnt) != 1)
728 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
729 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
730 }
731 #else /* SOCK_REFCNT_DEBUG */
732 #define sk_refcnt_debug_inc(sk) do { } while (0)
733 #define sk_refcnt_debug_dec(sk) do { } while (0)
734 #define sk_refcnt_debug_release(sk) do { } while (0)
735 #endif /* SOCK_REFCNT_DEBUG */
736
737
738 #ifdef CONFIG_PROC_FS
739 /* Called with local bh disabled */
740 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
741 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
742 #else
743 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
744 int inc)
745 {
746 }
747 #endif
748
749
750 /* With per-bucket locks this operation is not-atomic, so that
751 * this version is not worse.
752 */
753 static inline void __sk_prot_rehash(struct sock *sk)
754 {
755 sk->sk_prot->unhash(sk);
756 sk->sk_prot->hash(sk);
757 }
758
759 /* About 10 seconds */
760 #define SOCK_DESTROY_TIME (10*HZ)
761
762 /* Sockets 0-1023 can't be bound to unless you are superuser */
763 #define PROT_SOCK 1024
764
765 #define SHUTDOWN_MASK 3
766 #define RCV_SHUTDOWN 1
767 #define SEND_SHUTDOWN 2
768
769 #define SOCK_SNDBUF_LOCK 1
770 #define SOCK_RCVBUF_LOCK 2
771 #define SOCK_BINDADDR_LOCK 4
772 #define SOCK_BINDPORT_LOCK 8
773
774 /* sock_iocb: used to kick off async processing of socket ios */
775 struct sock_iocb {
776 struct list_head list;
777
778 int flags;
779 int size;
780 struct socket *sock;
781 struct sock *sk;
782 struct scm_cookie *scm;
783 struct msghdr *msg, async_msg;
784 struct kiocb *kiocb;
785 };
786
787 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
788 {
789 return (struct sock_iocb *)iocb->private;
790 }
791
792 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
793 {
794 return si->kiocb;
795 }
796
797 struct socket_alloc {
798 struct socket socket;
799 struct inode vfs_inode;
800 };
801
802 static inline struct socket *SOCKET_I(struct inode *inode)
803 {
804 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
805 }
806
807 static inline struct inode *SOCK_INODE(struct socket *socket)
808 {
809 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
810 }
811
812 /*
813 * Functions for memory accounting
814 */
815 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
816 extern void __sk_mem_reclaim(struct sock *sk);
817
818 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
819 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
820 #define SK_MEM_SEND 0
821 #define SK_MEM_RECV 1
822
823 static inline int sk_mem_pages(int amt)
824 {
825 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
826 }
827
828 static inline int sk_has_account(struct sock *sk)
829 {
830 /* return true if protocol supports memory accounting */
831 return !!sk->sk_prot->memory_allocated;
832 }
833
834 static inline int sk_wmem_schedule(struct sock *sk, int size)
835 {
836 if (!sk_has_account(sk))
837 return 1;
838 return size <= sk->sk_forward_alloc ||
839 __sk_mem_schedule(sk, size, SK_MEM_SEND);
840 }
841
842 static inline int sk_rmem_schedule(struct sock *sk, int size)
843 {
844 if (!sk_has_account(sk))
845 return 1;
846 return size <= sk->sk_forward_alloc ||
847 __sk_mem_schedule(sk, size, SK_MEM_RECV);
848 }
849
850 static inline void sk_mem_reclaim(struct sock *sk)
851 {
852 if (!sk_has_account(sk))
853 return;
854 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
855 __sk_mem_reclaim(sk);
856 }
857
858 static inline void sk_mem_reclaim_partial(struct sock *sk)
859 {
860 if (!sk_has_account(sk))
861 return;
862 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
863 __sk_mem_reclaim(sk);
864 }
865
866 static inline void sk_mem_charge(struct sock *sk, int size)
867 {
868 if (!sk_has_account(sk))
869 return;
870 sk->sk_forward_alloc -= size;
871 }
872
873 static inline void sk_mem_uncharge(struct sock *sk, int size)
874 {
875 if (!sk_has_account(sk))
876 return;
877 sk->sk_forward_alloc += size;
878 }
879
880 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
881 {
882 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
883 sk->sk_wmem_queued -= skb->truesize;
884 sk_mem_uncharge(sk, skb->truesize);
885 __kfree_skb(skb);
886 }
887
888 /* Used by processes to "lock" a socket state, so that
889 * interrupts and bottom half handlers won't change it
890 * from under us. It essentially blocks any incoming
891 * packets, so that we won't get any new data or any
892 * packets that change the state of the socket.
893 *
894 * While locked, BH processing will add new packets to
895 * the backlog queue. This queue is processed by the
896 * owner of the socket lock right before it is released.
897 *
898 * Since ~2.3.5 it is also exclusive sleep lock serializing
899 * accesses from user process context.
900 */
901 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
902
903 /*
904 * Macro so as to not evaluate some arguments when
905 * lockdep is not enabled.
906 *
907 * Mark both the sk_lock and the sk_lock.slock as a
908 * per-address-family lock class.
909 */
910 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
911 do { \
912 sk->sk_lock.owned = 0; \
913 init_waitqueue_head(&sk->sk_lock.wq); \
914 spin_lock_init(&(sk)->sk_lock.slock); \
915 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
916 sizeof((sk)->sk_lock)); \
917 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
918 (skey), (sname)); \
919 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
920 } while (0)
921
922 extern void lock_sock_nested(struct sock *sk, int subclass);
923
924 static inline void lock_sock(struct sock *sk)
925 {
926 lock_sock_nested(sk, 0);
927 }
928
929 extern void release_sock(struct sock *sk);
930
931 /* BH context may only use the following locking interface. */
932 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
933 #define bh_lock_sock_nested(__sk) \
934 spin_lock_nested(&((__sk)->sk_lock.slock), \
935 SINGLE_DEPTH_NESTING)
936 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
937
938 extern struct sock *sk_alloc(struct net *net, int family,
939 gfp_t priority,
940 struct proto *prot);
941 extern void sk_free(struct sock *sk);
942 extern void sk_release_kernel(struct sock *sk);
943 extern struct sock *sk_clone(const struct sock *sk,
944 const gfp_t priority);
945
946 extern struct sk_buff *sock_wmalloc(struct sock *sk,
947 unsigned long size, int force,
948 gfp_t priority);
949 extern struct sk_buff *sock_rmalloc(struct sock *sk,
950 unsigned long size, int force,
951 gfp_t priority);
952 extern void sock_wfree(struct sk_buff *skb);
953 extern void sock_rfree(struct sk_buff *skb);
954
955 extern int sock_setsockopt(struct socket *sock, int level,
956 int op, char __user *optval,
957 unsigned int optlen);
958
959 extern int sock_getsockopt(struct socket *sock, int level,
960 int op, char __user *optval,
961 int __user *optlen);
962 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
963 unsigned long size,
964 int noblock,
965 int *errcode);
966 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
967 unsigned long header_len,
968 unsigned long data_len,
969 int noblock,
970 int *errcode);
971 extern void *sock_kmalloc(struct sock *sk, int size,
972 gfp_t priority);
973 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
974 extern void sk_send_sigurg(struct sock *sk);
975
976 /*
977 * Functions to fill in entries in struct proto_ops when a protocol
978 * does not implement a particular function.
979 */
980 extern int sock_no_bind(struct socket *,
981 struct sockaddr *, int);
982 extern int sock_no_connect(struct socket *,
983 struct sockaddr *, int, int);
984 extern int sock_no_socketpair(struct socket *,
985 struct socket *);
986 extern int sock_no_accept(struct socket *,
987 struct socket *, int);
988 extern int sock_no_getname(struct socket *,
989 struct sockaddr *, int *, int);
990 extern unsigned int sock_no_poll(struct file *, struct socket *,
991 struct poll_table_struct *);
992 extern int sock_no_ioctl(struct socket *, unsigned int,
993 unsigned long);
994 extern int sock_no_listen(struct socket *, int);
995 extern int sock_no_shutdown(struct socket *, int);
996 extern int sock_no_getsockopt(struct socket *, int , int,
997 char __user *, int __user *);
998 extern int sock_no_setsockopt(struct socket *, int, int,
999 char __user *, unsigned int);
1000 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1001 struct msghdr *, size_t);
1002 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1003 struct msghdr *, size_t, int);
1004 extern int sock_no_mmap(struct file *file,
1005 struct socket *sock,
1006 struct vm_area_struct *vma);
1007 extern ssize_t sock_no_sendpage(struct socket *sock,
1008 struct page *page,
1009 int offset, size_t size,
1010 int flags);
1011
1012 /*
1013 * Functions to fill in entries in struct proto_ops when a protocol
1014 * uses the inet style.
1015 */
1016 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1017 char __user *optval, int __user *optlen);
1018 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1019 struct msghdr *msg, size_t size, int flags);
1020 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1021 char __user *optval, unsigned int optlen);
1022 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1023 int optname, char __user *optval, int __user *optlen);
1024 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1025 int optname, char __user *optval, unsigned int optlen);
1026
1027 extern void sk_common_release(struct sock *sk);
1028
1029 /*
1030 * Default socket callbacks and setup code
1031 */
1032
1033 /* Initialise core socket variables */
1034 extern void sock_init_data(struct socket *sock, struct sock *sk);
1035
1036 /**
1037 * sk_filter_release: Release a socket filter
1038 * @fp: filter to remove
1039 *
1040 * Remove a filter from a socket and release its resources.
1041 */
1042
1043 static inline void sk_filter_release(struct sk_filter *fp)
1044 {
1045 if (atomic_dec_and_test(&fp->refcnt))
1046 kfree(fp);
1047 }
1048
1049 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1050 {
1051 unsigned int size = sk_filter_len(fp);
1052
1053 atomic_sub(size, &sk->sk_omem_alloc);
1054 sk_filter_release(fp);
1055 }
1056
1057 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1058 {
1059 atomic_inc(&fp->refcnt);
1060 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1061 }
1062
1063 /*
1064 * Socket reference counting postulates.
1065 *
1066 * * Each user of socket SHOULD hold a reference count.
1067 * * Each access point to socket (an hash table bucket, reference from a list,
1068 * running timer, skb in flight MUST hold a reference count.
1069 * * When reference count hits 0, it means it will never increase back.
1070 * * When reference count hits 0, it means that no references from
1071 * outside exist to this socket and current process on current CPU
1072 * is last user and may/should destroy this socket.
1073 * * sk_free is called from any context: process, BH, IRQ. When
1074 * it is called, socket has no references from outside -> sk_free
1075 * may release descendant resources allocated by the socket, but
1076 * to the time when it is called, socket is NOT referenced by any
1077 * hash tables, lists etc.
1078 * * Packets, delivered from outside (from network or from another process)
1079 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1080 * when they sit in queue. Otherwise, packets will leak to hole, when
1081 * socket is looked up by one cpu and unhasing is made by another CPU.
1082 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1083 * (leak to backlog). Packet socket does all the processing inside
1084 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1085 * use separate SMP lock, so that they are prone too.
1086 */
1087
1088 /* Ungrab socket and destroy it, if it was the last reference. */
1089 static inline void sock_put(struct sock *sk)
1090 {
1091 if (atomic_dec_and_test(&sk->sk_refcnt))
1092 sk_free(sk);
1093 }
1094
1095 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1096 const int nested);
1097
1098 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1099 {
1100 sk->sk_socket = sock;
1101 }
1102
1103 /* Detach socket from process context.
1104 * Announce socket dead, detach it from wait queue and inode.
1105 * Note that parent inode held reference count on this struct sock,
1106 * we do not release it in this function, because protocol
1107 * probably wants some additional cleanups or even continuing
1108 * to work with this socket (TCP).
1109 */
1110 static inline void sock_orphan(struct sock *sk)
1111 {
1112 write_lock_bh(&sk->sk_callback_lock);
1113 sock_set_flag(sk, SOCK_DEAD);
1114 sk_set_socket(sk, NULL);
1115 sk->sk_sleep = NULL;
1116 write_unlock_bh(&sk->sk_callback_lock);
1117 }
1118
1119 static inline void sock_graft(struct sock *sk, struct socket *parent)
1120 {
1121 write_lock_bh(&sk->sk_callback_lock);
1122 sk->sk_sleep = &parent->wait;
1123 parent->sk = sk;
1124 sk_set_socket(sk, parent);
1125 security_sock_graft(sk, parent);
1126 write_unlock_bh(&sk->sk_callback_lock);
1127 }
1128
1129 extern int sock_i_uid(struct sock *sk);
1130 extern unsigned long sock_i_ino(struct sock *sk);
1131
1132 static inline struct dst_entry *
1133 __sk_dst_get(struct sock *sk)
1134 {
1135 return sk->sk_dst_cache;
1136 }
1137
1138 static inline struct dst_entry *
1139 sk_dst_get(struct sock *sk)
1140 {
1141 struct dst_entry *dst;
1142
1143 read_lock(&sk->sk_dst_lock);
1144 dst = sk->sk_dst_cache;
1145 if (dst)
1146 dst_hold(dst);
1147 read_unlock(&sk->sk_dst_lock);
1148 return dst;
1149 }
1150
1151 static inline void
1152 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1153 {
1154 struct dst_entry *old_dst;
1155
1156 old_dst = sk->sk_dst_cache;
1157 sk->sk_dst_cache = dst;
1158 dst_release(old_dst);
1159 }
1160
1161 static inline void
1162 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1163 {
1164 write_lock(&sk->sk_dst_lock);
1165 __sk_dst_set(sk, dst);
1166 write_unlock(&sk->sk_dst_lock);
1167 }
1168
1169 static inline void
1170 __sk_dst_reset(struct sock *sk)
1171 {
1172 struct dst_entry *old_dst;
1173
1174 old_dst = sk->sk_dst_cache;
1175 sk->sk_dst_cache = NULL;
1176 dst_release(old_dst);
1177 }
1178
1179 static inline void
1180 sk_dst_reset(struct sock *sk)
1181 {
1182 write_lock(&sk->sk_dst_lock);
1183 __sk_dst_reset(sk);
1184 write_unlock(&sk->sk_dst_lock);
1185 }
1186
1187 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1188
1189 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1190
1191 static inline int sk_can_gso(const struct sock *sk)
1192 {
1193 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1194 }
1195
1196 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1197
1198 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1199 struct sk_buff *skb, struct page *page,
1200 int off, int copy)
1201 {
1202 if (skb->ip_summed == CHECKSUM_NONE) {
1203 int err = 0;
1204 __wsum csum = csum_and_copy_from_user(from,
1205 page_address(page) + off,
1206 copy, 0, &err);
1207 if (err)
1208 return err;
1209 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1210 } else if (copy_from_user(page_address(page) + off, from, copy))
1211 return -EFAULT;
1212
1213 skb->len += copy;
1214 skb->data_len += copy;
1215 skb->truesize += copy;
1216 sk->sk_wmem_queued += copy;
1217 sk_mem_charge(sk, copy);
1218 return 0;
1219 }
1220
1221 /**
1222 * sk_wmem_alloc_get - returns write allocations
1223 * @sk: socket
1224 *
1225 * Returns sk_wmem_alloc minus initial offset of one
1226 */
1227 static inline int sk_wmem_alloc_get(const struct sock *sk)
1228 {
1229 return atomic_read(&sk->sk_wmem_alloc) - 1;
1230 }
1231
1232 /**
1233 * sk_rmem_alloc_get - returns read allocations
1234 * @sk: socket
1235 *
1236 * Returns sk_rmem_alloc
1237 */
1238 static inline int sk_rmem_alloc_get(const struct sock *sk)
1239 {
1240 return atomic_read(&sk->sk_rmem_alloc);
1241 }
1242
1243 /**
1244 * sk_has_allocations - check if allocations are outstanding
1245 * @sk: socket
1246 *
1247 * Returns true if socket has write or read allocations
1248 */
1249 static inline int sk_has_allocations(const struct sock *sk)
1250 {
1251 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1252 }
1253
1254 /**
1255 * sk_has_sleeper - check if there are any waiting processes
1256 * @sk: socket
1257 *
1258 * Returns true if socket has waiting processes
1259 *
1260 * The purpose of the sk_has_sleeper and sock_poll_wait is to wrap the memory
1261 * barrier call. They were added due to the race found within the tcp code.
1262 *
1263 * Consider following tcp code paths:
1264 *
1265 * CPU1 CPU2
1266 *
1267 * sys_select receive packet
1268 * ... ...
1269 * __add_wait_queue update tp->rcv_nxt
1270 * ... ...
1271 * tp->rcv_nxt check sock_def_readable
1272 * ... {
1273 * schedule ...
1274 * if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
1275 * wake_up_interruptible(sk->sk_sleep)
1276 * ...
1277 * }
1278 *
1279 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1280 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1281 * could then endup calling schedule and sleep forever if there are no more
1282 * data on the socket.
1283 *
1284 * The sk_has_sleeper is always called right after a call to read_lock, so we
1285 * can use smp_mb__after_lock barrier.
1286 */
1287 static inline int sk_has_sleeper(struct sock *sk)
1288 {
1289 /*
1290 * We need to be sure we are in sync with the
1291 * add_wait_queue modifications to the wait queue.
1292 *
1293 * This memory barrier is paired in the sock_poll_wait.
1294 */
1295 smp_mb__after_lock();
1296 return sk->sk_sleep && waitqueue_active(sk->sk_sleep);
1297 }
1298
1299 /**
1300 * sock_poll_wait - place memory barrier behind the poll_wait call.
1301 * @filp: file
1302 * @wait_address: socket wait queue
1303 * @p: poll_table
1304 *
1305 * See the comments in the sk_has_sleeper function.
1306 */
1307 static inline void sock_poll_wait(struct file *filp,
1308 wait_queue_head_t *wait_address, poll_table *p)
1309 {
1310 if (p && wait_address) {
1311 poll_wait(filp, wait_address, p);
1312 /*
1313 * We need to be sure we are in sync with the
1314 * socket flags modification.
1315 *
1316 * This memory barrier is paired in the sk_has_sleeper.
1317 */
1318 smp_mb();
1319 }
1320 }
1321
1322 /*
1323 * Queue a received datagram if it will fit. Stream and sequenced
1324 * protocols can't normally use this as they need to fit buffers in
1325 * and play with them.
1326 *
1327 * Inlined as it's very short and called for pretty much every
1328 * packet ever received.
1329 */
1330
1331 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1332 {
1333 skb_orphan(skb);
1334 skb->sk = sk;
1335 skb->destructor = sock_wfree;
1336 /*
1337 * We used to take a refcount on sk, but following operation
1338 * is enough to guarantee sk_free() wont free this sock until
1339 * all in-flight packets are completed
1340 */
1341 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1342 }
1343
1344 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1345 {
1346 skb_orphan(skb);
1347 skb->sk = sk;
1348 skb->destructor = sock_rfree;
1349 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1350 sk_mem_charge(sk, skb->truesize);
1351 }
1352
1353 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1354 unsigned long expires);
1355
1356 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1357
1358 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1359
1360 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1361
1362 /*
1363 * Recover an error report and clear atomically
1364 */
1365
1366 static inline int sock_error(struct sock *sk)
1367 {
1368 int err;
1369 if (likely(!sk->sk_err))
1370 return 0;
1371 err = xchg(&sk->sk_err, 0);
1372 return -err;
1373 }
1374
1375 static inline unsigned long sock_wspace(struct sock *sk)
1376 {
1377 int amt = 0;
1378
1379 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1380 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1381 if (amt < 0)
1382 amt = 0;
1383 }
1384 return amt;
1385 }
1386
1387 static inline void sk_wake_async(struct sock *sk, int how, int band)
1388 {
1389 if (sk->sk_socket && sk->sk_socket->fasync_list)
1390 sock_wake_async(sk->sk_socket, how, band);
1391 }
1392
1393 #define SOCK_MIN_SNDBUF 2048
1394 #define SOCK_MIN_RCVBUF 256
1395
1396 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1397 {
1398 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1399 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1400 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1401 }
1402 }
1403
1404 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1405
1406 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1407 {
1408 struct page *page = NULL;
1409
1410 page = alloc_pages(sk->sk_allocation, 0);
1411 if (!page) {
1412 sk->sk_prot->enter_memory_pressure(sk);
1413 sk_stream_moderate_sndbuf(sk);
1414 }
1415 return page;
1416 }
1417
1418 /*
1419 * Default write policy as shown to user space via poll/select/SIGIO
1420 */
1421 static inline int sock_writeable(const struct sock *sk)
1422 {
1423 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1424 }
1425
1426 static inline gfp_t gfp_any(void)
1427 {
1428 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1429 }
1430
1431 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1432 {
1433 return noblock ? 0 : sk->sk_rcvtimeo;
1434 }
1435
1436 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1437 {
1438 return noblock ? 0 : sk->sk_sndtimeo;
1439 }
1440
1441 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1442 {
1443 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1444 }
1445
1446 /* Alas, with timeout socket operations are not restartable.
1447 * Compare this to poll().
1448 */
1449 static inline int sock_intr_errno(long timeo)
1450 {
1451 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1452 }
1453
1454 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1455 struct sk_buff *skb);
1456
1457 static __inline__ void
1458 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1459 {
1460 ktime_t kt = skb->tstamp;
1461 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
1462
1463 /*
1464 * generate control messages if
1465 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1466 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1467 * - software time stamp available and wanted
1468 * (SOCK_TIMESTAMPING_SOFTWARE)
1469 * - hardware time stamps available and wanted
1470 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1471 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1472 */
1473 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1474 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1475 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1476 (hwtstamps->hwtstamp.tv64 &&
1477 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1478 (hwtstamps->syststamp.tv64 &&
1479 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
1480 __sock_recv_timestamp(msg, sk, skb);
1481 else
1482 sk->sk_stamp = kt;
1483 }
1484
1485 /**
1486 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1487 * @msg: outgoing packet
1488 * @sk: socket sending this packet
1489 * @shtx: filled with instructions for time stamping
1490 *
1491 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1492 * parameters are invalid.
1493 */
1494 extern int sock_tx_timestamp(struct msghdr *msg,
1495 struct sock *sk,
1496 union skb_shared_tx *shtx);
1497
1498
1499 /**
1500 * sk_eat_skb - Release a skb if it is no longer needed
1501 * @sk: socket to eat this skb from
1502 * @skb: socket buffer to eat
1503 * @copied_early: flag indicating whether DMA operations copied this data early
1504 *
1505 * This routine must be called with interrupts disabled or with the socket
1506 * locked so that the sk_buff queue operation is ok.
1507 */
1508 #ifdef CONFIG_NET_DMA
1509 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1510 {
1511 __skb_unlink(skb, &sk->sk_receive_queue);
1512 if (!copied_early)
1513 __kfree_skb(skb);
1514 else
1515 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1516 }
1517 #else
1518 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1519 {
1520 __skb_unlink(skb, &sk->sk_receive_queue);
1521 __kfree_skb(skb);
1522 }
1523 #endif
1524
1525 static inline
1526 struct net *sock_net(const struct sock *sk)
1527 {
1528 #ifdef CONFIG_NET_NS
1529 return sk->sk_net;
1530 #else
1531 return &init_net;
1532 #endif
1533 }
1534
1535 static inline
1536 void sock_net_set(struct sock *sk, struct net *net)
1537 {
1538 #ifdef CONFIG_NET_NS
1539 sk->sk_net = net;
1540 #endif
1541 }
1542
1543 /*
1544 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1545 * They should not hold a referrence to a namespace in order to allow
1546 * to stop it.
1547 * Sockets after sk_change_net should be released using sk_release_kernel
1548 */
1549 static inline void sk_change_net(struct sock *sk, struct net *net)
1550 {
1551 put_net(sock_net(sk));
1552 sock_net_set(sk, hold_net(net));
1553 }
1554
1555 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
1556 {
1557 if (unlikely(skb->sk)) {
1558 struct sock *sk = skb->sk;
1559
1560 skb->destructor = NULL;
1561 skb->sk = NULL;
1562 return sk;
1563 }
1564 return NULL;
1565 }
1566
1567 extern void sock_enable_timestamp(struct sock *sk, int flag);
1568 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1569 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1570
1571 /*
1572 * Enable debug/info messages
1573 */
1574 extern int net_msg_warn;
1575 #define NETDEBUG(fmt, args...) \
1576 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1577
1578 #define LIMIT_NETDEBUG(fmt, args...) \
1579 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1580
1581 extern __u32 sysctl_wmem_max;
1582 extern __u32 sysctl_rmem_max;
1583
1584 extern void sk_init(void);
1585
1586 extern int sysctl_optmem_max;
1587
1588 extern __u32 sysctl_wmem_default;
1589 extern __u32 sysctl_rmem_default;
1590
1591 #endif /* _SOCK_H */
1592
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.