Commit | Line | Data |
---|---|---|
1da177e4 LT |
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 | * | |
02c30a84 | 10 | * Authors: Ross Biro |
1da177e4 LT |
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/config.h> | |
44 | #include <linux/list.h> | |
45 | #include <linux/timer.h> | |
46 | #include <linux/cache.h> | |
47 | #include <linux/module.h> | |
48 | #include <linux/netdevice.h> | |
49 | #include <linux/skbuff.h> /* struct sk_buff */ | |
50 | #include <linux/security.h> | |
51 | ||
52 | #include <linux/filter.h> | |
53 | ||
54 | #include <asm/atomic.h> | |
55 | #include <net/dst.h> | |
56 | #include <net/checksum.h> | |
57 | ||
58 | /* | |
59 | * This structure really needs to be cleaned up. | |
60 | * Most of it is for TCP, and not used by any of | |
61 | * the other protocols. | |
62 | */ | |
63 | ||
64 | /* Define this to get the SOCK_DBG debugging facility. */ | |
65 | #define SOCK_DEBUGGING | |
66 | #ifdef SOCK_DEBUGGING | |
67 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ | |
68 | printk(KERN_DEBUG msg); } while (0) | |
69 | #else | |
70 | #define SOCK_DEBUG(sk, msg...) do { } while (0) | |
71 | #endif | |
72 | ||
73 | /* This is the per-socket lock. The spinlock provides a synchronization | |
74 | * between user contexts and software interrupt processing, whereas the | |
75 | * mini-semaphore synchronizes multiple users amongst themselves. | |
76 | */ | |
77 | struct sock_iocb; | |
78 | typedef struct { | |
79 | spinlock_t slock; | |
80 | struct sock_iocb *owner; | |
81 | wait_queue_head_t wq; | |
82 | } socket_lock_t; | |
83 | ||
84 | #define sock_lock_init(__sk) \ | |
85 | do { spin_lock_init(&((__sk)->sk_lock.slock)); \ | |
86 | (__sk)->sk_lock.owner = NULL; \ | |
87 | init_waitqueue_head(&((__sk)->sk_lock.wq)); \ | |
88 | } while(0) | |
89 | ||
90 | struct sock; | |
8feaf0c0 | 91 | struct proto; |
1da177e4 LT |
92 | |
93 | /** | |
4dc3b16b PP |
94 | * struct sock_common - minimal network layer representation of sockets |
95 | * @skc_family: network address family | |
96 | * @skc_state: Connection state | |
97 | * @skc_reuse: %SO_REUSEADDR setting | |
98 | * @skc_bound_dev_if: bound device index if != 0 | |
99 | * @skc_node: main hash linkage for various protocol lookup tables | |
100 | * @skc_bind_node: bind hash linkage for various protocol lookup tables | |
101 | * @skc_refcnt: reference count | |
81c3d547 | 102 | * @skc_hash: hash value used with various protocol lookup tables |
8feaf0c0 | 103 | * @skc_prot: protocol handlers inside a network family |
4dc3b16b PP |
104 | * |
105 | * This is the minimal network layer representation of sockets, the header | |
8feaf0c0 ACM |
106 | * for struct sock and struct inet_timewait_sock. |
107 | */ | |
1da177e4 LT |
108 | struct sock_common { |
109 | unsigned short skc_family; | |
110 | volatile unsigned char skc_state; | |
111 | unsigned char skc_reuse; | |
112 | int skc_bound_dev_if; | |
113 | struct hlist_node skc_node; | |
114 | struct hlist_node skc_bind_node; | |
115 | atomic_t skc_refcnt; | |
81c3d547 | 116 | unsigned int skc_hash; |
8feaf0c0 | 117 | struct proto *skc_prot; |
1da177e4 LT |
118 | }; |
119 | ||
120 | /** | |
121 | * struct sock - network layer representation of sockets | |
8feaf0c0 | 122 | * @__sk_common: shared layout with inet_timewait_sock |
4dc3b16b PP |
123 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
124 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
125 | * @sk_lock: synchronizer | |
126 | * @sk_rcvbuf: size of receive buffer in bytes | |
127 | * @sk_sleep: sock wait queue | |
128 | * @sk_dst_cache: destination cache | |
129 | * @sk_dst_lock: destination cache lock | |
130 | * @sk_policy: flow policy | |
131 | * @sk_rmem_alloc: receive queue bytes committed | |
132 | * @sk_receive_queue: incoming packets | |
133 | * @sk_wmem_alloc: transmit queue bytes committed | |
134 | * @sk_write_queue: Packet sending queue | |
135 | * @sk_omem_alloc: "o" is "option" or "other" | |
136 | * @sk_wmem_queued: persistent queue size | |
137 | * @sk_forward_alloc: space allocated forward | |
138 | * @sk_allocation: allocation mode | |
139 | * @sk_sndbuf: size of send buffer in bytes | |
140 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings | |
141 | * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets | |
142 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) | |
143 | * @sk_lingertime: %SO_LINGER l_linger setting | |
4dc3b16b PP |
144 | * @sk_backlog: always used with the per-socket spinlock held |
145 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
146 | * @sk_error_queue: rarely used | |
476e19cf | 147 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance) |
4dc3b16b PP |
148 | * @sk_err: last error |
149 | * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out' | |
150 | * @sk_ack_backlog: current listen backlog | |
151 | * @sk_max_ack_backlog: listen backlog set in listen() | |
152 | * @sk_priority: %SO_PRIORITY setting | |
153 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
154 | * @sk_protocol: which protocol this socket belongs in this network family | |
155 | * @sk_peercred: %SO_PEERCRED setting | |
156 | * @sk_rcvlowat: %SO_RCVLOWAT setting | |
157 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
158 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
159 | * @sk_filter: socket filtering instructions | |
160 | * @sk_protinfo: private area, net family specific, when not using slab | |
161 | * @sk_timer: sock cleanup timer | |
162 | * @sk_stamp: time stamp of last packet received | |
163 | * @sk_socket: Identd and reporting IO signals | |
164 | * @sk_user_data: RPC layer private data | |
165 | * @sk_sndmsg_page: cached page for sendmsg | |
166 | * @sk_sndmsg_off: cached offset for sendmsg | |
167 | * @sk_send_head: front of stuff to transmit | |
67be2dd1 | 168 | * @sk_security: used by security modules |
4dc3b16b PP |
169 | * @sk_write_pending: a write to stream socket waits to start |
170 | * @sk_state_change: callback to indicate change in the state of the sock | |
171 | * @sk_data_ready: callback to indicate there is data to be processed | |
172 | * @sk_write_space: callback to indicate there is bf sending space available | |
173 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
174 | * @sk_backlog_rcv: callback to process the backlog | |
175 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
1da177e4 LT |
176 | */ |
177 | struct sock { | |
178 | /* | |
8feaf0c0 | 179 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
180 | * don't add nothing before this first member (__sk_common) --acme |
181 | */ | |
182 | struct sock_common __sk_common; | |
183 | #define sk_family __sk_common.skc_family | |
184 | #define sk_state __sk_common.skc_state | |
185 | #define sk_reuse __sk_common.skc_reuse | |
186 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if | |
187 | #define sk_node __sk_common.skc_node | |
188 | #define sk_bind_node __sk_common.skc_bind_node | |
189 | #define sk_refcnt __sk_common.skc_refcnt | |
81c3d547 | 190 | #define sk_hash __sk_common.skc_hash |
8feaf0c0 | 191 | #define sk_prot __sk_common.skc_prot |
1da177e4 LT |
192 | unsigned char sk_shutdown : 2, |
193 | sk_no_check : 2, | |
194 | sk_userlocks : 4; | |
195 | unsigned char sk_protocol; | |
196 | unsigned short sk_type; | |
197 | int sk_rcvbuf; | |
198 | socket_lock_t sk_lock; | |
199 | wait_queue_head_t *sk_sleep; | |
200 | struct dst_entry *sk_dst_cache; | |
201 | struct xfrm_policy *sk_policy[2]; | |
202 | rwlock_t sk_dst_lock; | |
203 | atomic_t sk_rmem_alloc; | |
204 | atomic_t sk_wmem_alloc; | |
205 | atomic_t sk_omem_alloc; | |
206 | struct sk_buff_head sk_receive_queue; | |
207 | struct sk_buff_head sk_write_queue; | |
208 | int sk_wmem_queued; | |
209 | int sk_forward_alloc; | |
7d877f3b | 210 | gfp_t sk_allocation; |
1da177e4 LT |
211 | int sk_sndbuf; |
212 | int sk_route_caps; | |
1da177e4 LT |
213 | unsigned long sk_flags; |
214 | unsigned long sk_lingertime; | |
215 | /* | |
216 | * The backlog queue is special, it is always used with | |
217 | * the per-socket spinlock held and requires low latency | |
218 | * access. Therefore we special case it's implementation. | |
219 | */ | |
220 | struct { | |
221 | struct sk_buff *head; | |
222 | struct sk_buff *tail; | |
223 | } sk_backlog; | |
224 | struct sk_buff_head sk_error_queue; | |
476e19cf | 225 | struct proto *sk_prot_creator; |
1da177e4 LT |
226 | rwlock_t sk_callback_lock; |
227 | int sk_err, | |
228 | sk_err_soft; | |
229 | unsigned short sk_ack_backlog; | |
230 | unsigned short sk_max_ack_backlog; | |
231 | __u32 sk_priority; | |
232 | struct ucred sk_peercred; | |
233 | int sk_rcvlowat; | |
234 | long sk_rcvtimeo; | |
235 | long sk_sndtimeo; | |
236 | struct sk_filter *sk_filter; | |
237 | void *sk_protinfo; | |
238 | struct timer_list sk_timer; | |
239 | struct timeval sk_stamp; | |
240 | struct socket *sk_socket; | |
241 | void *sk_user_data; | |
242 | struct page *sk_sndmsg_page; | |
243 | struct sk_buff *sk_send_head; | |
244 | __u32 sk_sndmsg_off; | |
245 | int sk_write_pending; | |
246 | void *sk_security; | |
247 | void (*sk_state_change)(struct sock *sk); | |
248 | void (*sk_data_ready)(struct sock *sk, int bytes); | |
249 | void (*sk_write_space)(struct sock *sk); | |
250 | void (*sk_error_report)(struct sock *sk); | |
251 | int (*sk_backlog_rcv)(struct sock *sk, | |
252 | struct sk_buff *skb); | |
253 | void (*sk_destruct)(struct sock *sk); | |
254 | }; | |
255 | ||
256 | /* | |
257 | * Hashed lists helper routines | |
258 | */ | |
e48c414e | 259 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
260 | { |
261 | return hlist_entry(head->first, struct sock, sk_node); | |
262 | } | |
263 | ||
e48c414e | 264 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
265 | { |
266 | return hlist_empty(head) ? NULL : __sk_head(head); | |
267 | } | |
268 | ||
e48c414e | 269 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 LT |
270 | { |
271 | return sk->sk_node.next ? | |
272 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; | |
273 | } | |
274 | ||
e48c414e | 275 | static inline int sk_unhashed(const struct sock *sk) |
1da177e4 LT |
276 | { |
277 | return hlist_unhashed(&sk->sk_node); | |
278 | } | |
279 | ||
e48c414e | 280 | static inline int sk_hashed(const struct sock *sk) |
1da177e4 LT |
281 | { |
282 | return sk->sk_node.pprev != NULL; | |
283 | } | |
284 | ||
285 | static __inline__ void sk_node_init(struct hlist_node *node) | |
286 | { | |
287 | node->pprev = NULL; | |
288 | } | |
289 | ||
290 | static __inline__ void __sk_del_node(struct sock *sk) | |
291 | { | |
292 | __hlist_del(&sk->sk_node); | |
293 | } | |
294 | ||
295 | static __inline__ int __sk_del_node_init(struct sock *sk) | |
296 | { | |
297 | if (sk_hashed(sk)) { | |
298 | __sk_del_node(sk); | |
299 | sk_node_init(&sk->sk_node); | |
300 | return 1; | |
301 | } | |
302 | return 0; | |
303 | } | |
304 | ||
305 | /* Grab socket reference count. This operation is valid only | |
306 | when sk is ALREADY grabbed f.e. it is found in hash table | |
307 | or a list and the lookup is made under lock preventing hash table | |
308 | modifications. | |
309 | */ | |
310 | ||
311 | static inline void sock_hold(struct sock *sk) | |
312 | { | |
313 | atomic_inc(&sk->sk_refcnt); | |
314 | } | |
315 | ||
316 | /* Ungrab socket in the context, which assumes that socket refcnt | |
317 | cannot hit zero, f.e. it is true in context of any socketcall. | |
318 | */ | |
319 | static inline void __sock_put(struct sock *sk) | |
320 | { | |
321 | atomic_dec(&sk->sk_refcnt); | |
322 | } | |
323 | ||
324 | static __inline__ int sk_del_node_init(struct sock *sk) | |
325 | { | |
326 | int rc = __sk_del_node_init(sk); | |
327 | ||
328 | if (rc) { | |
329 | /* paranoid for a while -acme */ | |
330 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
331 | __sock_put(sk); | |
332 | } | |
333 | return rc; | |
334 | } | |
335 | ||
336 | static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) | |
337 | { | |
338 | hlist_add_head(&sk->sk_node, list); | |
339 | } | |
340 | ||
341 | static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) | |
342 | { | |
343 | sock_hold(sk); | |
344 | __sk_add_node(sk, list); | |
345 | } | |
346 | ||
347 | static __inline__ void __sk_del_bind_node(struct sock *sk) | |
348 | { | |
349 | __hlist_del(&sk->sk_bind_node); | |
350 | } | |
351 | ||
352 | static __inline__ void sk_add_bind_node(struct sock *sk, | |
353 | struct hlist_head *list) | |
354 | { | |
355 | hlist_add_head(&sk->sk_bind_node, list); | |
356 | } | |
357 | ||
358 | #define sk_for_each(__sk, node, list) \ | |
359 | hlist_for_each_entry(__sk, node, list, sk_node) | |
360 | #define sk_for_each_from(__sk, node) \ | |
361 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
362 | hlist_for_each_entry_from(__sk, node, sk_node) | |
363 | #define sk_for_each_continue(__sk, node) \ | |
364 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
365 | hlist_for_each_entry_continue(__sk, node, sk_node) | |
366 | #define sk_for_each_safe(__sk, node, tmp, list) \ | |
367 | hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) | |
368 | #define sk_for_each_bound(__sk, node, list) \ | |
369 | hlist_for_each_entry(__sk, node, list, sk_bind_node) | |
370 | ||
371 | /* Sock flags */ | |
372 | enum sock_flags { | |
373 | SOCK_DEAD, | |
374 | SOCK_DONE, | |
375 | SOCK_URGINLINE, | |
376 | SOCK_KEEPOPEN, | |
377 | SOCK_LINGER, | |
378 | SOCK_DESTROY, | |
379 | SOCK_BROADCAST, | |
380 | SOCK_TIMESTAMP, | |
381 | SOCK_ZAPPED, | |
382 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
383 | SOCK_DBG, /* %SO_DEBUG setting */ | |
384 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
385 | SOCK_NO_LARGESEND, /* whether to sent large segments or not */ | |
386 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ | |
387 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
388 | }; | |
389 | ||
53b924b3 RB |
390 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
391 | { | |
392 | nsk->sk_flags = osk->sk_flags; | |
393 | } | |
394 | ||
1da177e4 LT |
395 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
396 | { | |
397 | __set_bit(flag, &sk->sk_flags); | |
398 | } | |
399 | ||
400 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
401 | { | |
402 | __clear_bit(flag, &sk->sk_flags); | |
403 | } | |
404 | ||
405 | static inline int sock_flag(struct sock *sk, enum sock_flags flag) | |
406 | { | |
407 | return test_bit(flag, &sk->sk_flags); | |
408 | } | |
409 | ||
410 | static inline void sk_acceptq_removed(struct sock *sk) | |
411 | { | |
412 | sk->sk_ack_backlog--; | |
413 | } | |
414 | ||
415 | static inline void sk_acceptq_added(struct sock *sk) | |
416 | { | |
417 | sk->sk_ack_backlog++; | |
418 | } | |
419 | ||
420 | static inline int sk_acceptq_is_full(struct sock *sk) | |
421 | { | |
422 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; | |
423 | } | |
424 | ||
425 | /* | |
426 | * Compute minimal free write space needed to queue new packets. | |
427 | */ | |
428 | static inline int sk_stream_min_wspace(struct sock *sk) | |
429 | { | |
430 | return sk->sk_wmem_queued / 2; | |
431 | } | |
432 | ||
433 | static inline int sk_stream_wspace(struct sock *sk) | |
434 | { | |
435 | return sk->sk_sndbuf - sk->sk_wmem_queued; | |
436 | } | |
437 | ||
438 | extern void sk_stream_write_space(struct sock *sk); | |
439 | ||
440 | static inline int sk_stream_memory_free(struct sock *sk) | |
441 | { | |
442 | return sk->sk_wmem_queued < sk->sk_sndbuf; | |
443 | } | |
444 | ||
445 | extern void sk_stream_rfree(struct sk_buff *skb); | |
446 | ||
447 | static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
448 | { | |
449 | skb->sk = sk; | |
450 | skb->destructor = sk_stream_rfree; | |
451 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
452 | sk->sk_forward_alloc -= skb->truesize; | |
453 | } | |
454 | ||
455 | static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb) | |
456 | { | |
457 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); | |
458 | sk->sk_wmem_queued -= skb->truesize; | |
459 | sk->sk_forward_alloc += skb->truesize; | |
460 | __kfree_skb(skb); | |
461 | } | |
462 | ||
463 | /* The per-socket spinlock must be held here. */ | |
9ee6b535 SH |
464 | static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
465 | { | |
466 | if (!sk->sk_backlog.tail) { | |
467 | sk->sk_backlog.head = sk->sk_backlog.tail = skb; | |
468 | } else { | |
469 | sk->sk_backlog.tail->next = skb; | |
470 | sk->sk_backlog.tail = skb; | |
471 | } | |
472 | skb->next = NULL; | |
473 | } | |
1da177e4 LT |
474 | |
475 | #define sk_wait_event(__sk, __timeo, __condition) \ | |
476 | ({ int rc; \ | |
477 | release_sock(__sk); \ | |
478 | rc = __condition; \ | |
479 | if (!rc) { \ | |
480 | *(__timeo) = schedule_timeout(*(__timeo)); \ | |
1da177e4 LT |
481 | } \ |
482 | lock_sock(__sk); \ | |
265a9285 | 483 | rc = __condition; \ |
1da177e4 LT |
484 | rc; \ |
485 | }) | |
486 | ||
487 | extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); | |
488 | extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
489 | extern void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
490 | extern int sk_stream_error(struct sock *sk, int flags, int err); | |
491 | extern void sk_stream_kill_queues(struct sock *sk); | |
492 | ||
493 | extern int sk_wait_data(struct sock *sk, long *timeo); | |
494 | ||
60236fdd | 495 | struct request_sock_ops; |
6d6ee43e | 496 | struct timewait_sock_ops; |
2e6599cb | 497 | |
1da177e4 LT |
498 | /* Networking protocol blocks we attach to sockets. |
499 | * socket layer -> transport layer interface | |
500 | * transport -> network interface is defined by struct inet_proto | |
501 | */ | |
502 | struct proto { | |
503 | void (*close)(struct sock *sk, | |
504 | long timeout); | |
505 | int (*connect)(struct sock *sk, | |
506 | struct sockaddr *uaddr, | |
507 | int addr_len); | |
508 | int (*disconnect)(struct sock *sk, int flags); | |
509 | ||
510 | struct sock * (*accept) (struct sock *sk, int flags, int *err); | |
511 | ||
512 | int (*ioctl)(struct sock *sk, int cmd, | |
513 | unsigned long arg); | |
514 | int (*init)(struct sock *sk); | |
515 | int (*destroy)(struct sock *sk); | |
516 | void (*shutdown)(struct sock *sk, int how); | |
517 | int (*setsockopt)(struct sock *sk, int level, | |
518 | int optname, char __user *optval, | |
519 | int optlen); | |
520 | int (*getsockopt)(struct sock *sk, int level, | |
521 | int optname, char __user *optval, | |
522 | int __user *option); | |
3fdadf7d DM |
523 | int (*compat_setsockopt)(struct sock *sk, |
524 | int level, | |
525 | int optname, char __user *optval, | |
526 | int optlen); | |
527 | int (*compat_getsockopt)(struct sock *sk, | |
528 | int level, | |
529 | int optname, char __user *optval, | |
530 | int __user *option); | |
1da177e4 LT |
531 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, |
532 | struct msghdr *msg, size_t len); | |
533 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, | |
534 | struct msghdr *msg, | |
535 | size_t len, int noblock, int flags, | |
536 | int *addr_len); | |
537 | int (*sendpage)(struct sock *sk, struct page *page, | |
538 | int offset, size_t size, int flags); | |
539 | int (*bind)(struct sock *sk, | |
540 | struct sockaddr *uaddr, int addr_len); | |
541 | ||
542 | int (*backlog_rcv) (struct sock *sk, | |
543 | struct sk_buff *skb); | |
544 | ||
545 | /* Keeping track of sk's, looking them up, and port selection methods. */ | |
546 | void (*hash)(struct sock *sk); | |
547 | void (*unhash)(struct sock *sk); | |
548 | int (*get_port)(struct sock *sk, unsigned short snum); | |
549 | ||
550 | /* Memory pressure */ | |
551 | void (*enter_memory_pressure)(void); | |
552 | atomic_t *memory_allocated; /* Current allocated memory. */ | |
553 | atomic_t *sockets_allocated; /* Current number of sockets. */ | |
554 | /* | |
555 | * Pressure flag: try to collapse. | |
556 | * Technical note: it is used by multiple contexts non atomically. | |
557 | * All the sk_stream_mem_schedule() is of this nature: accounting | |
558 | * is strict, actions are advisory and have some latency. | |
559 | */ | |
560 | int *memory_pressure; | |
561 | int *sysctl_mem; | |
562 | int *sysctl_wmem; | |
563 | int *sysctl_rmem; | |
564 | int max_header; | |
565 | ||
566 | kmem_cache_t *slab; | |
567 | unsigned int obj_size; | |
568 | ||
0a5578cf | 569 | atomic_t *orphan_count; |
8feaf0c0 | 570 | |
60236fdd | 571 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 572 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 573 | |
1da177e4 LT |
574 | struct module *owner; |
575 | ||
576 | char name[32]; | |
577 | ||
578 | struct list_head node; | |
e6848976 ACM |
579 | #ifdef SOCK_REFCNT_DEBUG |
580 | atomic_t socks; | |
581 | #endif | |
1da177e4 LT |
582 | struct { |
583 | int inuse; | |
584 | u8 __pad[SMP_CACHE_BYTES - sizeof(int)]; | |
585 | } stats[NR_CPUS]; | |
586 | }; | |
587 | ||
588 | extern int proto_register(struct proto *prot, int alloc_slab); | |
589 | extern void proto_unregister(struct proto *prot); | |
590 | ||
e6848976 ACM |
591 | #ifdef SOCK_REFCNT_DEBUG |
592 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
593 | { | |
594 | atomic_inc(&sk->sk_prot->socks); | |
595 | } | |
596 | ||
597 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
598 | { | |
599 | atomic_dec(&sk->sk_prot->socks); | |
600 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
601 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
602 | } | |
603 | ||
604 | static inline void sk_refcnt_debug_release(const struct sock *sk) | |
605 | { | |
606 | if (atomic_read(&sk->sk_refcnt) != 1) | |
607 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", | |
608 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); | |
609 | } | |
610 | #else /* SOCK_REFCNT_DEBUG */ | |
611 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
612 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
613 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
614 | #endif /* SOCK_REFCNT_DEBUG */ | |
615 | ||
1da177e4 LT |
616 | /* Called with local bh disabled */ |
617 | static __inline__ void sock_prot_inc_use(struct proto *prot) | |
618 | { | |
619 | prot->stats[smp_processor_id()].inuse++; | |
620 | } | |
621 | ||
622 | static __inline__ void sock_prot_dec_use(struct proto *prot) | |
623 | { | |
624 | prot->stats[smp_processor_id()].inuse--; | |
625 | } | |
626 | ||
614c6cb4 ACM |
627 | /* With per-bucket locks this operation is not-atomic, so that |
628 | * this version is not worse. | |
629 | */ | |
630 | static inline void __sk_prot_rehash(struct sock *sk) | |
631 | { | |
632 | sk->sk_prot->unhash(sk); | |
633 | sk->sk_prot->hash(sk); | |
634 | } | |
635 | ||
1da177e4 LT |
636 | /* About 10 seconds */ |
637 | #define SOCK_DESTROY_TIME (10*HZ) | |
638 | ||
639 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
640 | #define PROT_SOCK 1024 | |
641 | ||
642 | #define SHUTDOWN_MASK 3 | |
643 | #define RCV_SHUTDOWN 1 | |
644 | #define SEND_SHUTDOWN 2 | |
645 | ||
646 | #define SOCK_SNDBUF_LOCK 1 | |
647 | #define SOCK_RCVBUF_LOCK 2 | |
648 | #define SOCK_BINDADDR_LOCK 4 | |
649 | #define SOCK_BINDPORT_LOCK 8 | |
650 | ||
651 | /* sock_iocb: used to kick off async processing of socket ios */ | |
652 | struct sock_iocb { | |
653 | struct list_head list; | |
654 | ||
655 | int flags; | |
656 | int size; | |
657 | struct socket *sock; | |
658 | struct sock *sk; | |
659 | struct scm_cookie *scm; | |
660 | struct msghdr *msg, async_msg; | |
661 | struct iovec async_iov; | |
662 | struct kiocb *kiocb; | |
663 | }; | |
664 | ||
665 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) | |
666 | { | |
667 | return (struct sock_iocb *)iocb->private; | |
668 | } | |
669 | ||
670 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) | |
671 | { | |
672 | return si->kiocb; | |
673 | } | |
674 | ||
675 | struct socket_alloc { | |
676 | struct socket socket; | |
677 | struct inode vfs_inode; | |
678 | }; | |
679 | ||
680 | static inline struct socket *SOCKET_I(struct inode *inode) | |
681 | { | |
682 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
683 | } | |
684 | ||
685 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
686 | { | |
687 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
688 | } | |
689 | ||
690 | extern void __sk_stream_mem_reclaim(struct sock *sk); | |
691 | extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind); | |
692 | ||
693 | #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE) | |
694 | ||
695 | static inline int sk_stream_pages(int amt) | |
696 | { | |
697 | return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM; | |
698 | } | |
699 | ||
700 | static inline void sk_stream_mem_reclaim(struct sock *sk) | |
701 | { | |
702 | if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM) | |
703 | __sk_stream_mem_reclaim(sk); | |
704 | } | |
705 | ||
706 | static inline void sk_stream_writequeue_purge(struct sock *sk) | |
707 | { | |
708 | struct sk_buff *skb; | |
709 | ||
710 | while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) | |
711 | sk_stream_free_skb(sk, skb); | |
712 | sk_stream_mem_reclaim(sk); | |
713 | } | |
714 | ||
715 | static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb) | |
716 | { | |
717 | return (int)skb->truesize <= sk->sk_forward_alloc || | |
718 | sk_stream_mem_schedule(sk, skb->truesize, 1); | |
719 | } | |
720 | ||
d80d99d6 HX |
721 | static inline int sk_stream_wmem_schedule(struct sock *sk, int size) |
722 | { | |
723 | return size <= sk->sk_forward_alloc || | |
724 | sk_stream_mem_schedule(sk, size, 0); | |
725 | } | |
726 | ||
1da177e4 LT |
727 | /* Used by processes to "lock" a socket state, so that |
728 | * interrupts and bottom half handlers won't change it | |
729 | * from under us. It essentially blocks any incoming | |
730 | * packets, so that we won't get any new data or any | |
731 | * packets that change the state of the socket. | |
732 | * | |
733 | * While locked, BH processing will add new packets to | |
734 | * the backlog queue. This queue is processed by the | |
735 | * owner of the socket lock right before it is released. | |
736 | * | |
737 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
738 | * accesses from user process context. | |
739 | */ | |
740 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owner) | |
741 | ||
742 | extern void FASTCALL(lock_sock(struct sock *sk)); | |
743 | extern void FASTCALL(release_sock(struct sock *sk)); | |
744 | ||
745 | /* BH context may only use the following locking interface. */ | |
746 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
747 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) | |
748 | ||
86a76caf | 749 | extern struct sock *sk_alloc(int family, |
dd0fc66f | 750 | gfp_t priority, |
1da177e4 LT |
751 | struct proto *prot, int zero_it); |
752 | extern void sk_free(struct sock *sk); | |
87d11ceb | 753 | extern struct sock *sk_clone(const struct sock *sk, |
dd0fc66f | 754 | const gfp_t priority); |
1da177e4 LT |
755 | |
756 | extern struct sk_buff *sock_wmalloc(struct sock *sk, | |
757 | unsigned long size, int force, | |
dd0fc66f | 758 | gfp_t priority); |
1da177e4 LT |
759 | extern struct sk_buff *sock_rmalloc(struct sock *sk, |
760 | unsigned long size, int force, | |
dd0fc66f | 761 | gfp_t priority); |
1da177e4 LT |
762 | extern void sock_wfree(struct sk_buff *skb); |
763 | extern void sock_rfree(struct sk_buff *skb); | |
764 | ||
765 | extern int sock_setsockopt(struct socket *sock, int level, | |
766 | int op, char __user *optval, | |
767 | int optlen); | |
768 | ||
769 | extern int sock_getsockopt(struct socket *sock, int level, | |
770 | int op, char __user *optval, | |
771 | int __user *optlen); | |
772 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, | |
773 | unsigned long size, | |
774 | int noblock, | |
775 | int *errcode); | |
86a76caf | 776 | extern void *sock_kmalloc(struct sock *sk, int size, |
dd0fc66f | 777 | gfp_t priority); |
1da177e4 LT |
778 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); |
779 | extern void sk_send_sigurg(struct sock *sk); | |
780 | ||
781 | /* | |
782 | * Functions to fill in entries in struct proto_ops when a protocol | |
783 | * does not implement a particular function. | |
784 | */ | |
785 | extern int sock_no_bind(struct socket *, | |
786 | struct sockaddr *, int); | |
787 | extern int sock_no_connect(struct socket *, | |
788 | struct sockaddr *, int, int); | |
789 | extern int sock_no_socketpair(struct socket *, | |
790 | struct socket *); | |
791 | extern int sock_no_accept(struct socket *, | |
792 | struct socket *, int); | |
793 | extern int sock_no_getname(struct socket *, | |
794 | struct sockaddr *, int *, int); | |
795 | extern unsigned int sock_no_poll(struct file *, struct socket *, | |
796 | struct poll_table_struct *); | |
797 | extern int sock_no_ioctl(struct socket *, unsigned int, | |
798 | unsigned long); | |
799 | extern int sock_no_listen(struct socket *, int); | |
800 | extern int sock_no_shutdown(struct socket *, int); | |
801 | extern int sock_no_getsockopt(struct socket *, int , int, | |
802 | char __user *, int __user *); | |
803 | extern int sock_no_setsockopt(struct socket *, int, int, | |
804 | char __user *, int); | |
805 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, | |
806 | struct msghdr *, size_t); | |
807 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, | |
808 | struct msghdr *, size_t, int); | |
809 | extern int sock_no_mmap(struct file *file, | |
810 | struct socket *sock, | |
811 | struct vm_area_struct *vma); | |
812 | extern ssize_t sock_no_sendpage(struct socket *sock, | |
813 | struct page *page, | |
814 | int offset, size_t size, | |
815 | int flags); | |
816 | ||
817 | /* | |
818 | * Functions to fill in entries in struct proto_ops when a protocol | |
819 | * uses the inet style. | |
820 | */ | |
821 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, | |
822 | char __user *optval, int __user *optlen); | |
823 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, | |
824 | struct msghdr *msg, size_t size, int flags); | |
825 | extern int sock_common_setsockopt(struct socket *sock, int level, int optname, | |
826 | char __user *optval, int optlen); | |
3fdadf7d DM |
827 | extern int compat_sock_common_getsockopt(struct socket *sock, int level, |
828 | int optname, char __user *optval, int __user *optlen); | |
829 | extern int compat_sock_common_setsockopt(struct socket *sock, int level, | |
830 | int optname, char __user *optval, int optlen); | |
1da177e4 LT |
831 | |
832 | extern void sk_common_release(struct sock *sk); | |
833 | ||
834 | /* | |
835 | * Default socket callbacks and setup code | |
836 | */ | |
837 | ||
838 | /* Initialise core socket variables */ | |
839 | extern void sock_init_data(struct socket *sock, struct sock *sk); | |
840 | ||
841 | /** | |
842 | * sk_filter - run a packet through a socket filter | |
843 | * @sk: sock associated with &sk_buff | |
844 | * @skb: buffer to filter | |
845 | * @needlock: set to 1 if the sock is not locked by caller. | |
846 | * | |
847 | * Run the filter code and then cut skb->data to correct size returned by | |
848 | * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller | |
849 | * than pkt_len we keep whole skb->data. This is the socket level | |
850 | * wrapper to sk_run_filter. It returns 0 if the packet should | |
851 | * be accepted or -EPERM if the packet should be tossed. | |
852 | * | |
853 | */ | |
854 | ||
855 | static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock) | |
856 | { | |
857 | int err; | |
858 | ||
859 | err = security_sock_rcv_skb(sk, skb); | |
860 | if (err) | |
861 | return err; | |
862 | ||
863 | if (sk->sk_filter) { | |
864 | struct sk_filter *filter; | |
865 | ||
866 | if (needlock) | |
867 | bh_lock_sock(sk); | |
868 | ||
869 | filter = sk->sk_filter; | |
870 | if (filter) { | |
4bad4dc9 KK |
871 | unsigned int pkt_len = sk_run_filter(skb, filter->insns, |
872 | filter->len); | |
1da177e4 LT |
873 | if (!pkt_len) |
874 | err = -EPERM; | |
875 | else | |
876 | skb_trim(skb, pkt_len); | |
877 | } | |
878 | ||
879 | if (needlock) | |
880 | bh_unlock_sock(sk); | |
881 | } | |
882 | return err; | |
883 | } | |
884 | ||
885 | /** | |
886 | * sk_filter_release: Release a socket filter | |
887 | * @sk: socket | |
888 | * @fp: filter to remove | |
889 | * | |
890 | * Remove a filter from a socket and release its resources. | |
891 | */ | |
892 | ||
893 | static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp) | |
894 | { | |
895 | unsigned int size = sk_filter_len(fp); | |
896 | ||
897 | atomic_sub(size, &sk->sk_omem_alloc); | |
898 | ||
899 | if (atomic_dec_and_test(&fp->refcnt)) | |
900 | kfree(fp); | |
901 | } | |
902 | ||
903 | static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) | |
904 | { | |
905 | atomic_inc(&fp->refcnt); | |
906 | atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); | |
907 | } | |
908 | ||
909 | /* | |
910 | * Socket reference counting postulates. | |
911 | * | |
912 | * * Each user of socket SHOULD hold a reference count. | |
913 | * * Each access point to socket (an hash table bucket, reference from a list, | |
914 | * running timer, skb in flight MUST hold a reference count. | |
915 | * * When reference count hits 0, it means it will never increase back. | |
916 | * * When reference count hits 0, it means that no references from | |
917 | * outside exist to this socket and current process on current CPU | |
918 | * is last user and may/should destroy this socket. | |
919 | * * sk_free is called from any context: process, BH, IRQ. When | |
920 | * it is called, socket has no references from outside -> sk_free | |
921 | * may release descendant resources allocated by the socket, but | |
922 | * to the time when it is called, socket is NOT referenced by any | |
923 | * hash tables, lists etc. | |
924 | * * Packets, delivered from outside (from network or from another process) | |
925 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
926 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
927 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
928 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
929 | * (leak to backlog). Packet socket does all the processing inside | |
930 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
931 | * use separate SMP lock, so that they are prone too. | |
932 | */ | |
933 | ||
934 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
935 | static inline void sock_put(struct sock *sk) | |
936 | { | |
937 | if (atomic_dec_and_test(&sk->sk_refcnt)) | |
938 | sk_free(sk); | |
939 | } | |
940 | ||
25995ff5 ACM |
941 | static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb) |
942 | { | |
943 | int rc = NET_RX_SUCCESS; | |
944 | ||
945 | if (sk_filter(sk, skb, 0)) | |
946 | goto discard_and_relse; | |
947 | ||
948 | skb->dev = NULL; | |
949 | ||
950 | bh_lock_sock(sk); | |
951 | if (!sock_owned_by_user(sk)) | |
952 | rc = sk->sk_backlog_rcv(sk, skb); | |
953 | else | |
954 | sk_add_backlog(sk, skb); | |
955 | bh_unlock_sock(sk); | |
956 | out: | |
957 | sock_put(sk); | |
958 | return rc; | |
959 | discard_and_relse: | |
960 | kfree_skb(skb); | |
961 | goto out; | |
962 | } | |
963 | ||
1da177e4 LT |
964 | /* Detach socket from process context. |
965 | * Announce socket dead, detach it from wait queue and inode. | |
966 | * Note that parent inode held reference count on this struct sock, | |
967 | * we do not release it in this function, because protocol | |
968 | * probably wants some additional cleanups or even continuing | |
969 | * to work with this socket (TCP). | |
970 | */ | |
971 | static inline void sock_orphan(struct sock *sk) | |
972 | { | |
973 | write_lock_bh(&sk->sk_callback_lock); | |
974 | sock_set_flag(sk, SOCK_DEAD); | |
975 | sk->sk_socket = NULL; | |
976 | sk->sk_sleep = NULL; | |
977 | write_unlock_bh(&sk->sk_callback_lock); | |
978 | } | |
979 | ||
980 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
981 | { | |
982 | write_lock_bh(&sk->sk_callback_lock); | |
983 | sk->sk_sleep = &parent->wait; | |
984 | parent->sk = sk; | |
985 | sk->sk_socket = parent; | |
986 | write_unlock_bh(&sk->sk_callback_lock); | |
987 | } | |
988 | ||
989 | extern int sock_i_uid(struct sock *sk); | |
990 | extern unsigned long sock_i_ino(struct sock *sk); | |
991 | ||
992 | static inline struct dst_entry * | |
993 | __sk_dst_get(struct sock *sk) | |
994 | { | |
995 | return sk->sk_dst_cache; | |
996 | } | |
997 | ||
998 | static inline struct dst_entry * | |
999 | sk_dst_get(struct sock *sk) | |
1000 | { | |
1001 | struct dst_entry *dst; | |
1002 | ||
1003 | read_lock(&sk->sk_dst_lock); | |
1004 | dst = sk->sk_dst_cache; | |
1005 | if (dst) | |
1006 | dst_hold(dst); | |
1007 | read_unlock(&sk->sk_dst_lock); | |
1008 | return dst; | |
1009 | } | |
1010 | ||
1011 | static inline void | |
1012 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1013 | { | |
1014 | struct dst_entry *old_dst; | |
1015 | ||
1016 | old_dst = sk->sk_dst_cache; | |
1017 | sk->sk_dst_cache = dst; | |
1018 | dst_release(old_dst); | |
1019 | } | |
1020 | ||
1021 | static inline void | |
1022 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1023 | { | |
1024 | write_lock(&sk->sk_dst_lock); | |
1025 | __sk_dst_set(sk, dst); | |
1026 | write_unlock(&sk->sk_dst_lock); | |
1027 | } | |
1028 | ||
1029 | static inline void | |
1030 | __sk_dst_reset(struct sock *sk) | |
1031 | { | |
1032 | struct dst_entry *old_dst; | |
1033 | ||
1034 | old_dst = sk->sk_dst_cache; | |
1035 | sk->sk_dst_cache = NULL; | |
1036 | dst_release(old_dst); | |
1037 | } | |
1038 | ||
1039 | static inline void | |
1040 | sk_dst_reset(struct sock *sk) | |
1041 | { | |
1042 | write_lock(&sk->sk_dst_lock); | |
1043 | __sk_dst_reset(sk); | |
1044 | write_unlock(&sk->sk_dst_lock); | |
1045 | } | |
1046 | ||
1047 | static inline struct dst_entry * | |
1048 | __sk_dst_check(struct sock *sk, u32 cookie) | |
1049 | { | |
1050 | struct dst_entry *dst = sk->sk_dst_cache; | |
1051 | ||
1052 | if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) { | |
1053 | sk->sk_dst_cache = NULL; | |
1054 | dst_release(dst); | |
1055 | return NULL; | |
1056 | } | |
1057 | ||
1058 | return dst; | |
1059 | } | |
1060 | ||
1061 | static inline struct dst_entry * | |
1062 | sk_dst_check(struct sock *sk, u32 cookie) | |
1063 | { | |
1064 | struct dst_entry *dst = sk_dst_get(sk); | |
1065 | ||
1066 | if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) { | |
1067 | sk_dst_reset(sk); | |
1068 | dst_release(dst); | |
1069 | return NULL; | |
1070 | } | |
1071 | ||
1072 | return dst; | |
1073 | } | |
1074 | ||
6cbb0df7 ACM |
1075 | static inline void sk_setup_caps(struct sock *sk, struct dst_entry *dst) |
1076 | { | |
1077 | __sk_dst_set(sk, dst); | |
1078 | sk->sk_route_caps = dst->dev->features; | |
1079 | if (sk->sk_route_caps & NETIF_F_TSO) { | |
1080 | if (sock_flag(sk, SOCK_NO_LARGESEND) || dst->header_len) | |
1081 | sk->sk_route_caps &= ~NETIF_F_TSO; | |
1082 | } | |
1083 | } | |
1084 | ||
1da177e4 LT |
1085 | static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb) |
1086 | { | |
1087 | sk->sk_wmem_queued += skb->truesize; | |
1088 | sk->sk_forward_alloc -= skb->truesize; | |
1089 | } | |
1090 | ||
1091 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, | |
1092 | struct sk_buff *skb, struct page *page, | |
1093 | int off, int copy) | |
1094 | { | |
1095 | if (skb->ip_summed == CHECKSUM_NONE) { | |
1096 | int err = 0; | |
1097 | unsigned int csum = csum_and_copy_from_user(from, | |
1098 | page_address(page) + off, | |
1099 | copy, 0, &err); | |
1100 | if (err) | |
1101 | return err; | |
1102 | skb->csum = csum_block_add(skb->csum, csum, skb->len); | |
1103 | } else if (copy_from_user(page_address(page) + off, from, copy)) | |
1104 | return -EFAULT; | |
1105 | ||
1106 | skb->len += copy; | |
1107 | skb->data_len += copy; | |
1108 | skb->truesize += copy; | |
1109 | sk->sk_wmem_queued += copy; | |
1110 | sk->sk_forward_alloc -= copy; | |
1111 | return 0; | |
1112 | } | |
1113 | ||
1114 | /* | |
1115 | * Queue a received datagram if it will fit. Stream and sequenced | |
1116 | * protocols can't normally use this as they need to fit buffers in | |
1117 | * and play with them. | |
1118 | * | |
1119 | * Inlined as it's very short and called for pretty much every | |
1120 | * packet ever received. | |
1121 | */ | |
1122 | ||
1123 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) | |
1124 | { | |
1125 | sock_hold(sk); | |
1126 | skb->sk = sk; | |
1127 | skb->destructor = sock_wfree; | |
1128 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); | |
1129 | } | |
1130 | ||
1131 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
1132 | { | |
1133 | skb->sk = sk; | |
1134 | skb->destructor = sock_rfree; | |
1135 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
1136 | } | |
1137 | ||
1138 | extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, | |
1139 | unsigned long expires); | |
1140 | ||
1141 | extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); | |
1142 | ||
1143 | static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) | |
1144 | { | |
1145 | int err = 0; | |
1146 | int skb_len; | |
1147 | ||
1148 | /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces | |
1149 | number of warnings when compiling with -W --ANK | |
1150 | */ | |
1151 | if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >= | |
1152 | (unsigned)sk->sk_rcvbuf) { | |
1153 | err = -ENOMEM; | |
1154 | goto out; | |
1155 | } | |
1156 | ||
1157 | /* It would be deadlock, if sock_queue_rcv_skb is used | |
1158 | with socket lock! We assume that users of this | |
1159 | function are lock free. | |
1160 | */ | |
1161 | err = sk_filter(sk, skb, 1); | |
1162 | if (err) | |
1163 | goto out; | |
1164 | ||
1165 | skb->dev = NULL; | |
1166 | skb_set_owner_r(skb, sk); | |
1167 | ||
1168 | /* Cache the SKB length before we tack it onto the receive | |
1169 | * queue. Once it is added it no longer belongs to us and | |
1170 | * may be freed by other threads of control pulling packets | |
1171 | * from the queue. | |
1172 | */ | |
1173 | skb_len = skb->len; | |
1174 | ||
1175 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
1176 | ||
1177 | if (!sock_flag(sk, SOCK_DEAD)) | |
1178 | sk->sk_data_ready(sk, skb_len); | |
1179 | out: | |
1180 | return err; | |
1181 | } | |
1182 | ||
1183 | static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb) | |
1184 | { | |
1185 | /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces | |
1186 | number of warnings when compiling with -W --ANK | |
1187 | */ | |
1188 | if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >= | |
1189 | (unsigned)sk->sk_rcvbuf) | |
1190 | return -ENOMEM; | |
1191 | skb_set_owner_r(skb, sk); | |
1192 | skb_queue_tail(&sk->sk_error_queue, skb); | |
1193 | if (!sock_flag(sk, SOCK_DEAD)) | |
1194 | sk->sk_data_ready(sk, skb->len); | |
1195 | return 0; | |
1196 | } | |
1197 | ||
1198 | /* | |
1199 | * Recover an error report and clear atomically | |
1200 | */ | |
1201 | ||
1202 | static inline int sock_error(struct sock *sk) | |
1203 | { | |
c1cbe4b7 BL |
1204 | int err; |
1205 | if (likely(!sk->sk_err)) | |
1206 | return 0; | |
1207 | err = xchg(&sk->sk_err, 0); | |
1da177e4 LT |
1208 | return -err; |
1209 | } | |
1210 | ||
1211 | static inline unsigned long sock_wspace(struct sock *sk) | |
1212 | { | |
1213 | int amt = 0; | |
1214 | ||
1215 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
1216 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); | |
1217 | if (amt < 0) | |
1218 | amt = 0; | |
1219 | } | |
1220 | return amt; | |
1221 | } | |
1222 | ||
1223 | static inline void sk_wake_async(struct sock *sk, int how, int band) | |
1224 | { | |
1225 | if (sk->sk_socket && sk->sk_socket->fasync_list) | |
1226 | sock_wake_async(sk->sk_socket, how, band); | |
1227 | } | |
1228 | ||
1229 | #define SOCK_MIN_SNDBUF 2048 | |
1230 | #define SOCK_MIN_RCVBUF 256 | |
1231 | ||
1232 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
1233 | { | |
1234 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { | |
1235 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2); | |
1236 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); | |
1237 | } | |
1238 | } | |
1239 | ||
1240 | static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk, | |
86a76caf | 1241 | int size, int mem, |
dd0fc66f | 1242 | gfp_t gfp) |
1da177e4 | 1243 | { |
c65f7f00 DM |
1244 | struct sk_buff *skb; |
1245 | int hdr_len; | |
1da177e4 | 1246 | |
c65f7f00 | 1247 | hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header); |
d179cd12 | 1248 | skb = alloc_skb_fclone(size + hdr_len, gfp); |
1da177e4 LT |
1249 | if (skb) { |
1250 | skb->truesize += mem; | |
d80d99d6 | 1251 | if (sk_stream_wmem_schedule(sk, skb->truesize)) { |
c65f7f00 | 1252 | skb_reserve(skb, hdr_len); |
1da177e4 LT |
1253 | return skb; |
1254 | } | |
1255 | __kfree_skb(skb); | |
1256 | } else { | |
1257 | sk->sk_prot->enter_memory_pressure(); | |
1258 | sk_stream_moderate_sndbuf(sk); | |
1259 | } | |
1260 | return NULL; | |
1261 | } | |
1262 | ||
1263 | static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk, | |
86a76caf | 1264 | int size, |
dd0fc66f | 1265 | gfp_t gfp) |
1da177e4 LT |
1266 | { |
1267 | return sk_stream_alloc_pskb(sk, size, 0, gfp); | |
1268 | } | |
1269 | ||
1270 | static inline struct page *sk_stream_alloc_page(struct sock *sk) | |
1271 | { | |
1272 | struct page *page = NULL; | |
1273 | ||
ef015786 HX |
1274 | page = alloc_pages(sk->sk_allocation, 0); |
1275 | if (!page) { | |
1da177e4 LT |
1276 | sk->sk_prot->enter_memory_pressure(); |
1277 | sk_stream_moderate_sndbuf(sk); | |
1278 | } | |
1279 | return page; | |
1280 | } | |
1281 | ||
1282 | #define sk_stream_for_retrans_queue(skb, sk) \ | |
1283 | for (skb = (sk)->sk_write_queue.next; \ | |
1284 | (skb != (sk)->sk_send_head) && \ | |
1285 | (skb != (struct sk_buff *)&(sk)->sk_write_queue); \ | |
1286 | skb = skb->next) | |
1287 | ||
6a438bbe SH |
1288 | /*from STCP for fast SACK Process*/ |
1289 | #define sk_stream_for_retrans_queue_from(skb, sk) \ | |
1290 | for (; (skb != (sk)->sk_send_head) && \ | |
1291 | (skb != (struct sk_buff *)&(sk)->sk_write_queue); \ | |
1292 | skb = skb->next) | |
1293 | ||
1da177e4 LT |
1294 | /* |
1295 | * Default write policy as shown to user space via poll/select/SIGIO | |
1296 | */ | |
1297 | static inline int sock_writeable(const struct sock *sk) | |
1298 | { | |
1299 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2); | |
1300 | } | |
1301 | ||
dd0fc66f | 1302 | static inline gfp_t gfp_any(void) |
1da177e4 LT |
1303 | { |
1304 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; | |
1305 | } | |
1306 | ||
1307 | static inline long sock_rcvtimeo(const struct sock *sk, int noblock) | |
1308 | { | |
1309 | return noblock ? 0 : sk->sk_rcvtimeo; | |
1310 | } | |
1311 | ||
1312 | static inline long sock_sndtimeo(const struct sock *sk, int noblock) | |
1313 | { | |
1314 | return noblock ? 0 : sk->sk_sndtimeo; | |
1315 | } | |
1316 | ||
1317 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
1318 | { | |
1319 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; | |
1320 | } | |
1321 | ||
1322 | /* Alas, with timeout socket operations are not restartable. | |
1323 | * Compare this to poll(). | |
1324 | */ | |
1325 | static inline int sock_intr_errno(long timeo) | |
1326 | { | |
1327 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
1328 | } | |
1329 | ||
1330 | static __inline__ void | |
1331 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) | |
1332 | { | |
a61bbcf2 PM |
1333 | struct timeval stamp; |
1334 | ||
1335 | skb_get_timestamp(skb, &stamp); | |
1da177e4 LT |
1336 | if (sock_flag(sk, SOCK_RCVTSTAMP)) { |
1337 | /* Race occurred between timestamp enabling and packet | |
1338 | receiving. Fill in the current time for now. */ | |
a61bbcf2 PM |
1339 | if (stamp.tv_sec == 0) |
1340 | do_gettimeofday(&stamp); | |
1341 | skb_set_timestamp(skb, &stamp); | |
1da177e4 | 1342 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval), |
a61bbcf2 | 1343 | &stamp); |
1da177e4 | 1344 | } else |
a61bbcf2 | 1345 | sk->sk_stamp = stamp; |
1da177e4 LT |
1346 | } |
1347 | ||
1348 | /** | |
1349 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
1350 | * @sk: socket to eat this skb from |
1351 | * @skb: socket buffer to eat | |
1da177e4 LT |
1352 | * |
1353 | * This routine must be called with interrupts disabled or with the socket | |
1354 | * locked so that the sk_buff queue operation is ok. | |
1355 | */ | |
1356 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb) | |
1357 | { | |
1358 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1359 | __kfree_skb(skb); | |
1360 | } | |
1361 | ||
1362 | extern void sock_enable_timestamp(struct sock *sk); | |
1363 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); | |
1364 | ||
1365 | /* | |
1366 | * Enable debug/info messages | |
1367 | */ | |
1368 | ||
0dec456d | 1369 | #ifdef CONFIG_NETDEBUG |
64ce2073 PM |
1370 | #define NETDEBUG(fmt, args...) printk(fmt,##args) |
1371 | #define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0) | |
0dec456d SH |
1372 | #else |
1373 | #define NETDEBUG(fmt, args...) do { } while (0) | |
1374 | #define LIMIT_NETDEBUG(fmt, args...) do { } while(0) | |
1da177e4 LT |
1375 | #endif |
1376 | ||
1377 | /* | |
1378 | * Macros for sleeping on a socket. Use them like this: | |
1379 | * | |
1380 | * SOCK_SLEEP_PRE(sk) | |
1381 | * if (condition) | |
1382 | * schedule(); | |
1383 | * SOCK_SLEEP_POST(sk) | |
1384 | * | |
1385 | * N.B. These are now obsolete and were, afaik, only ever used in DECnet | |
1386 | * and when the last use of them in DECnet has gone, I'm intending to | |
1387 | * remove them. | |
1388 | */ | |
1389 | ||
1390 | #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \ | |
1391 | DECLARE_WAITQUEUE(wait, tsk); \ | |
1392 | tsk->state = TASK_INTERRUPTIBLE; \ | |
1393 | add_wait_queue((sk)->sk_sleep, &wait); \ | |
1394 | release_sock(sk); | |
1395 | ||
1396 | #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \ | |
1397 | remove_wait_queue((sk)->sk_sleep, &wait); \ | |
1398 | lock_sock(sk); \ | |
1399 | } | |
1400 | ||
1401 | static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool) | |
1402 | { | |
1403 | if (valbool) | |
1404 | sock_set_flag(sk, bit); | |
1405 | else | |
1406 | sock_reset_flag(sk, bit); | |
1407 | } | |
1408 | ||
1409 | extern __u32 sysctl_wmem_max; | |
1410 | extern __u32 sysctl_rmem_max; | |
1411 | ||
1412 | #ifdef CONFIG_NET | |
1413 | int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg); | |
1414 | #else | |
1415 | static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg) | |
1416 | { | |
1417 | return -ENODEV; | |
1418 | } | |
1419 | #endif | |
1420 | ||
20380731 ACM |
1421 | extern void sk_init(void); |
1422 | ||
1423 | #ifdef CONFIG_SYSCTL | |
1424 | extern struct ctl_table core_table[]; | |
20380731 ACM |
1425 | #endif |
1426 | ||
6baf1f41 DM |
1427 | extern int sysctl_optmem_max; |
1428 | ||
20380731 ACM |
1429 | extern __u32 sysctl_wmem_default; |
1430 | extern __u32 sysctl_rmem_default; | |
20380731 | 1431 | |
1da177e4 | 1432 | #endif /* _SOCK_H */ |