]>
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 | |
97fc2f08 | 135 | * @sk_async_wait_queue: DMA copied packets |
4dc3b16b PP |
136 | * @sk_omem_alloc: "o" is "option" or "other" |
137 | * @sk_wmem_queued: persistent queue size | |
138 | * @sk_forward_alloc: space allocated forward | |
139 | * @sk_allocation: allocation mode | |
140 | * @sk_sndbuf: size of send buffer in bytes | |
141 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings | |
142 | * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets | |
143 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) | |
144 | * @sk_lingertime: %SO_LINGER l_linger setting | |
4dc3b16b PP |
145 | * @sk_backlog: always used with the per-socket spinlock held |
146 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
147 | * @sk_error_queue: rarely used | |
476e19cf | 148 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance) |
4dc3b16b PP |
149 | * @sk_err: last error |
150 | * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out' | |
151 | * @sk_ack_backlog: current listen backlog | |
152 | * @sk_max_ack_backlog: listen backlog set in listen() | |
153 | * @sk_priority: %SO_PRIORITY setting | |
154 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
155 | * @sk_protocol: which protocol this socket belongs in this network family | |
156 | * @sk_peercred: %SO_PEERCRED setting | |
157 | * @sk_rcvlowat: %SO_RCVLOWAT setting | |
158 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
159 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
160 | * @sk_filter: socket filtering instructions | |
161 | * @sk_protinfo: private area, net family specific, when not using slab | |
162 | * @sk_timer: sock cleanup timer | |
163 | * @sk_stamp: time stamp of last packet received | |
164 | * @sk_socket: Identd and reporting IO signals | |
165 | * @sk_user_data: RPC layer private data | |
166 | * @sk_sndmsg_page: cached page for sendmsg | |
167 | * @sk_sndmsg_off: cached offset for sendmsg | |
168 | * @sk_send_head: front of stuff to transmit | |
67be2dd1 | 169 | * @sk_security: used by security modules |
4dc3b16b PP |
170 | * @sk_write_pending: a write to stream socket waits to start |
171 | * @sk_state_change: callback to indicate change in the state of the sock | |
172 | * @sk_data_ready: callback to indicate there is data to be processed | |
173 | * @sk_write_space: callback to indicate there is bf sending space available | |
174 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
175 | * @sk_backlog_rcv: callback to process the backlog | |
176 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
1da177e4 LT |
177 | */ |
178 | struct sock { | |
179 | /* | |
8feaf0c0 | 180 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
181 | * don't add nothing before this first member (__sk_common) --acme |
182 | */ | |
183 | struct sock_common __sk_common; | |
184 | #define sk_family __sk_common.skc_family | |
185 | #define sk_state __sk_common.skc_state | |
186 | #define sk_reuse __sk_common.skc_reuse | |
187 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if | |
188 | #define sk_node __sk_common.skc_node | |
189 | #define sk_bind_node __sk_common.skc_bind_node | |
190 | #define sk_refcnt __sk_common.skc_refcnt | |
81c3d547 | 191 | #define sk_hash __sk_common.skc_hash |
8feaf0c0 | 192 | #define sk_prot __sk_common.skc_prot |
1da177e4 LT |
193 | unsigned char sk_shutdown : 2, |
194 | sk_no_check : 2, | |
195 | sk_userlocks : 4; | |
196 | unsigned char sk_protocol; | |
197 | unsigned short sk_type; | |
198 | int sk_rcvbuf; | |
199 | socket_lock_t sk_lock; | |
200 | wait_queue_head_t *sk_sleep; | |
201 | struct dst_entry *sk_dst_cache; | |
202 | struct xfrm_policy *sk_policy[2]; | |
203 | rwlock_t sk_dst_lock; | |
204 | atomic_t sk_rmem_alloc; | |
205 | atomic_t sk_wmem_alloc; | |
206 | atomic_t sk_omem_alloc; | |
207 | struct sk_buff_head sk_receive_queue; | |
208 | struct sk_buff_head sk_write_queue; | |
97fc2f08 | 209 | struct sk_buff_head sk_async_wait_queue; |
1da177e4 LT |
210 | int sk_wmem_queued; |
211 | int sk_forward_alloc; | |
7d877f3b | 212 | gfp_t sk_allocation; |
1da177e4 LT |
213 | int sk_sndbuf; |
214 | int sk_route_caps; | |
9932cf95 | 215 | int sk_rcvlowat; |
1da177e4 LT |
216 | unsigned long sk_flags; |
217 | unsigned long sk_lingertime; | |
218 | /* | |
219 | * The backlog queue is special, it is always used with | |
220 | * the per-socket spinlock held and requires low latency | |
221 | * access. Therefore we special case it's implementation. | |
222 | */ | |
223 | struct { | |
224 | struct sk_buff *head; | |
225 | struct sk_buff *tail; | |
226 | } sk_backlog; | |
227 | struct sk_buff_head sk_error_queue; | |
476e19cf | 228 | struct proto *sk_prot_creator; |
1da177e4 LT |
229 | rwlock_t sk_callback_lock; |
230 | int sk_err, | |
231 | sk_err_soft; | |
232 | unsigned short sk_ack_backlog; | |
233 | unsigned short sk_max_ack_backlog; | |
234 | __u32 sk_priority; | |
235 | struct ucred sk_peercred; | |
1da177e4 LT |
236 | long sk_rcvtimeo; |
237 | long sk_sndtimeo; | |
238 | struct sk_filter *sk_filter; | |
239 | void *sk_protinfo; | |
240 | struct timer_list sk_timer; | |
241 | struct timeval sk_stamp; | |
242 | struct socket *sk_socket; | |
243 | void *sk_user_data; | |
244 | struct page *sk_sndmsg_page; | |
245 | struct sk_buff *sk_send_head; | |
246 | __u32 sk_sndmsg_off; | |
247 | int sk_write_pending; | |
248 | void *sk_security; | |
249 | void (*sk_state_change)(struct sock *sk); | |
250 | void (*sk_data_ready)(struct sock *sk, int bytes); | |
251 | void (*sk_write_space)(struct sock *sk); | |
252 | void (*sk_error_report)(struct sock *sk); | |
253 | int (*sk_backlog_rcv)(struct sock *sk, | |
254 | struct sk_buff *skb); | |
255 | void (*sk_destruct)(struct sock *sk); | |
256 | }; | |
257 | ||
258 | /* | |
259 | * Hashed lists helper routines | |
260 | */ | |
e48c414e | 261 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
262 | { |
263 | return hlist_entry(head->first, struct sock, sk_node); | |
264 | } | |
265 | ||
e48c414e | 266 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
267 | { |
268 | return hlist_empty(head) ? NULL : __sk_head(head); | |
269 | } | |
270 | ||
e48c414e | 271 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 LT |
272 | { |
273 | return sk->sk_node.next ? | |
274 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; | |
275 | } | |
276 | ||
e48c414e | 277 | static inline int sk_unhashed(const struct sock *sk) |
1da177e4 LT |
278 | { |
279 | return hlist_unhashed(&sk->sk_node); | |
280 | } | |
281 | ||
e48c414e | 282 | static inline int sk_hashed(const struct sock *sk) |
1da177e4 | 283 | { |
da753bea | 284 | return !sk_unhashed(sk); |
1da177e4 LT |
285 | } |
286 | ||
287 | static __inline__ void sk_node_init(struct hlist_node *node) | |
288 | { | |
289 | node->pprev = NULL; | |
290 | } | |
291 | ||
292 | static __inline__ void __sk_del_node(struct sock *sk) | |
293 | { | |
294 | __hlist_del(&sk->sk_node); | |
295 | } | |
296 | ||
297 | static __inline__ int __sk_del_node_init(struct sock *sk) | |
298 | { | |
299 | if (sk_hashed(sk)) { | |
300 | __sk_del_node(sk); | |
301 | sk_node_init(&sk->sk_node); | |
302 | return 1; | |
303 | } | |
304 | return 0; | |
305 | } | |
306 | ||
307 | /* Grab socket reference count. This operation is valid only | |
308 | when sk is ALREADY grabbed f.e. it is found in hash table | |
309 | or a list and the lookup is made under lock preventing hash table | |
310 | modifications. | |
311 | */ | |
312 | ||
313 | static inline void sock_hold(struct sock *sk) | |
314 | { | |
315 | atomic_inc(&sk->sk_refcnt); | |
316 | } | |
317 | ||
318 | /* Ungrab socket in the context, which assumes that socket refcnt | |
319 | cannot hit zero, f.e. it is true in context of any socketcall. | |
320 | */ | |
321 | static inline void __sock_put(struct sock *sk) | |
322 | { | |
323 | atomic_dec(&sk->sk_refcnt); | |
324 | } | |
325 | ||
326 | static __inline__ int sk_del_node_init(struct sock *sk) | |
327 | { | |
328 | int rc = __sk_del_node_init(sk); | |
329 | ||
330 | if (rc) { | |
331 | /* paranoid for a while -acme */ | |
332 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
333 | __sock_put(sk); | |
334 | } | |
335 | return rc; | |
336 | } | |
337 | ||
338 | static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) | |
339 | { | |
340 | hlist_add_head(&sk->sk_node, list); | |
341 | } | |
342 | ||
343 | static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) | |
344 | { | |
345 | sock_hold(sk); | |
346 | __sk_add_node(sk, list); | |
347 | } | |
348 | ||
349 | static __inline__ void __sk_del_bind_node(struct sock *sk) | |
350 | { | |
351 | __hlist_del(&sk->sk_bind_node); | |
352 | } | |
353 | ||
354 | static __inline__ void sk_add_bind_node(struct sock *sk, | |
355 | struct hlist_head *list) | |
356 | { | |
357 | hlist_add_head(&sk->sk_bind_node, list); | |
358 | } | |
359 | ||
360 | #define sk_for_each(__sk, node, list) \ | |
361 | hlist_for_each_entry(__sk, node, list, sk_node) | |
362 | #define sk_for_each_from(__sk, node) \ | |
363 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
364 | hlist_for_each_entry_from(__sk, node, sk_node) | |
365 | #define sk_for_each_continue(__sk, node) \ | |
366 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
367 | hlist_for_each_entry_continue(__sk, node, sk_node) | |
368 | #define sk_for_each_safe(__sk, node, tmp, list) \ | |
369 | hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) | |
370 | #define sk_for_each_bound(__sk, node, list) \ | |
371 | hlist_for_each_entry(__sk, node, list, sk_bind_node) | |
372 | ||
373 | /* Sock flags */ | |
374 | enum sock_flags { | |
375 | SOCK_DEAD, | |
376 | SOCK_DONE, | |
377 | SOCK_URGINLINE, | |
378 | SOCK_KEEPOPEN, | |
379 | SOCK_LINGER, | |
380 | SOCK_DESTROY, | |
381 | SOCK_BROADCAST, | |
382 | SOCK_TIMESTAMP, | |
383 | SOCK_ZAPPED, | |
384 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
385 | SOCK_DBG, /* %SO_DEBUG setting */ | |
386 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
387 | SOCK_NO_LARGESEND, /* whether to sent large segments or not */ | |
388 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ | |
389 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
390 | }; | |
391 | ||
53b924b3 RB |
392 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
393 | { | |
394 | nsk->sk_flags = osk->sk_flags; | |
395 | } | |
396 | ||
1da177e4 LT |
397 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
398 | { | |
399 | __set_bit(flag, &sk->sk_flags); | |
400 | } | |
401 | ||
402 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
403 | { | |
404 | __clear_bit(flag, &sk->sk_flags); | |
405 | } | |
406 | ||
407 | static inline int sock_flag(struct sock *sk, enum sock_flags flag) | |
408 | { | |
409 | return test_bit(flag, &sk->sk_flags); | |
410 | } | |
411 | ||
412 | static inline void sk_acceptq_removed(struct sock *sk) | |
413 | { | |
414 | sk->sk_ack_backlog--; | |
415 | } | |
416 | ||
417 | static inline void sk_acceptq_added(struct sock *sk) | |
418 | { | |
419 | sk->sk_ack_backlog++; | |
420 | } | |
421 | ||
422 | static inline int sk_acceptq_is_full(struct sock *sk) | |
423 | { | |
424 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; | |
425 | } | |
426 | ||
427 | /* | |
428 | * Compute minimal free write space needed to queue new packets. | |
429 | */ | |
430 | static inline int sk_stream_min_wspace(struct sock *sk) | |
431 | { | |
432 | return sk->sk_wmem_queued / 2; | |
433 | } | |
434 | ||
435 | static inline int sk_stream_wspace(struct sock *sk) | |
436 | { | |
437 | return sk->sk_sndbuf - sk->sk_wmem_queued; | |
438 | } | |
439 | ||
440 | extern void sk_stream_write_space(struct sock *sk); | |
441 | ||
442 | static inline int sk_stream_memory_free(struct sock *sk) | |
443 | { | |
444 | return sk->sk_wmem_queued < sk->sk_sndbuf; | |
445 | } | |
446 | ||
447 | extern void sk_stream_rfree(struct sk_buff *skb); | |
448 | ||
449 | static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
450 | { | |
451 | skb->sk = sk; | |
452 | skb->destructor = sk_stream_rfree; | |
453 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
454 | sk->sk_forward_alloc -= skb->truesize; | |
455 | } | |
456 | ||
457 | static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb) | |
458 | { | |
dc6de336 | 459 | skb_truesize_check(skb); |
1da177e4 LT |
460 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
461 | sk->sk_wmem_queued -= skb->truesize; | |
462 | sk->sk_forward_alloc += skb->truesize; | |
463 | __kfree_skb(skb); | |
464 | } | |
465 | ||
466 | /* The per-socket spinlock must be held here. */ | |
9ee6b535 SH |
467 | static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
468 | { | |
469 | if (!sk->sk_backlog.tail) { | |
470 | sk->sk_backlog.head = sk->sk_backlog.tail = skb; | |
471 | } else { | |
472 | sk->sk_backlog.tail->next = skb; | |
473 | sk->sk_backlog.tail = skb; | |
474 | } | |
475 | skb->next = NULL; | |
476 | } | |
1da177e4 LT |
477 | |
478 | #define sk_wait_event(__sk, __timeo, __condition) \ | |
479 | ({ int rc; \ | |
480 | release_sock(__sk); \ | |
481 | rc = __condition; \ | |
482 | if (!rc) { \ | |
483 | *(__timeo) = schedule_timeout(*(__timeo)); \ | |
1da177e4 LT |
484 | } \ |
485 | lock_sock(__sk); \ | |
265a9285 | 486 | rc = __condition; \ |
1da177e4 LT |
487 | rc; \ |
488 | }) | |
489 | ||
490 | extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); | |
491 | extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
492 | extern void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
493 | extern int sk_stream_error(struct sock *sk, int flags, int err); | |
494 | extern void sk_stream_kill_queues(struct sock *sk); | |
495 | ||
496 | extern int sk_wait_data(struct sock *sk, long *timeo); | |
497 | ||
60236fdd | 498 | struct request_sock_ops; |
6d6ee43e | 499 | struct timewait_sock_ops; |
2e6599cb | 500 | |
1da177e4 LT |
501 | /* Networking protocol blocks we attach to sockets. |
502 | * socket layer -> transport layer interface | |
503 | * transport -> network interface is defined by struct inet_proto | |
504 | */ | |
505 | struct proto { | |
506 | void (*close)(struct sock *sk, | |
507 | long timeout); | |
508 | int (*connect)(struct sock *sk, | |
509 | struct sockaddr *uaddr, | |
510 | int addr_len); | |
511 | int (*disconnect)(struct sock *sk, int flags); | |
512 | ||
513 | struct sock * (*accept) (struct sock *sk, int flags, int *err); | |
514 | ||
515 | int (*ioctl)(struct sock *sk, int cmd, | |
516 | unsigned long arg); | |
517 | int (*init)(struct sock *sk); | |
518 | int (*destroy)(struct sock *sk); | |
519 | void (*shutdown)(struct sock *sk, int how); | |
520 | int (*setsockopt)(struct sock *sk, int level, | |
521 | int optname, char __user *optval, | |
522 | int optlen); | |
523 | int (*getsockopt)(struct sock *sk, int level, | |
524 | int optname, char __user *optval, | |
525 | int __user *option); | |
3fdadf7d DM |
526 | int (*compat_setsockopt)(struct sock *sk, |
527 | int level, | |
528 | int optname, char __user *optval, | |
529 | int optlen); | |
530 | int (*compat_getsockopt)(struct sock *sk, | |
531 | int level, | |
532 | int optname, char __user *optval, | |
533 | int __user *option); | |
1da177e4 LT |
534 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, |
535 | struct msghdr *msg, size_t len); | |
536 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, | |
537 | struct msghdr *msg, | |
538 | size_t len, int noblock, int flags, | |
539 | int *addr_len); | |
540 | int (*sendpage)(struct sock *sk, struct page *page, | |
541 | int offset, size_t size, int flags); | |
542 | int (*bind)(struct sock *sk, | |
543 | struct sockaddr *uaddr, int addr_len); | |
544 | ||
545 | int (*backlog_rcv) (struct sock *sk, | |
546 | struct sk_buff *skb); | |
547 | ||
548 | /* Keeping track of sk's, looking them up, and port selection methods. */ | |
549 | void (*hash)(struct sock *sk); | |
550 | void (*unhash)(struct sock *sk); | |
551 | int (*get_port)(struct sock *sk, unsigned short snum); | |
552 | ||
553 | /* Memory pressure */ | |
554 | void (*enter_memory_pressure)(void); | |
555 | atomic_t *memory_allocated; /* Current allocated memory. */ | |
556 | atomic_t *sockets_allocated; /* Current number of sockets. */ | |
557 | /* | |
558 | * Pressure flag: try to collapse. | |
559 | * Technical note: it is used by multiple contexts non atomically. | |
560 | * All the sk_stream_mem_schedule() is of this nature: accounting | |
561 | * is strict, actions are advisory and have some latency. | |
562 | */ | |
563 | int *memory_pressure; | |
564 | int *sysctl_mem; | |
565 | int *sysctl_wmem; | |
566 | int *sysctl_rmem; | |
567 | int max_header; | |
568 | ||
569 | kmem_cache_t *slab; | |
570 | unsigned int obj_size; | |
571 | ||
0a5578cf | 572 | atomic_t *orphan_count; |
8feaf0c0 | 573 | |
60236fdd | 574 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 575 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 576 | |
1da177e4 LT |
577 | struct module *owner; |
578 | ||
579 | char name[32]; | |
580 | ||
581 | struct list_head node; | |
e6848976 ACM |
582 | #ifdef SOCK_REFCNT_DEBUG |
583 | atomic_t socks; | |
584 | #endif | |
1da177e4 LT |
585 | struct { |
586 | int inuse; | |
587 | u8 __pad[SMP_CACHE_BYTES - sizeof(int)]; | |
588 | } stats[NR_CPUS]; | |
589 | }; | |
590 | ||
591 | extern int proto_register(struct proto *prot, int alloc_slab); | |
592 | extern void proto_unregister(struct proto *prot); | |
593 | ||
e6848976 ACM |
594 | #ifdef SOCK_REFCNT_DEBUG |
595 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
596 | { | |
597 | atomic_inc(&sk->sk_prot->socks); | |
598 | } | |
599 | ||
600 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
601 | { | |
602 | atomic_dec(&sk->sk_prot->socks); | |
603 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
604 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
605 | } | |
606 | ||
607 | static inline void sk_refcnt_debug_release(const struct sock *sk) | |
608 | { | |
609 | if (atomic_read(&sk->sk_refcnt) != 1) | |
610 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", | |
611 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); | |
612 | } | |
613 | #else /* SOCK_REFCNT_DEBUG */ | |
614 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
615 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
616 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
617 | #endif /* SOCK_REFCNT_DEBUG */ | |
618 | ||
1da177e4 LT |
619 | /* Called with local bh disabled */ |
620 | static __inline__ void sock_prot_inc_use(struct proto *prot) | |
621 | { | |
622 | prot->stats[smp_processor_id()].inuse++; | |
623 | } | |
624 | ||
625 | static __inline__ void sock_prot_dec_use(struct proto *prot) | |
626 | { | |
627 | prot->stats[smp_processor_id()].inuse--; | |
628 | } | |
629 | ||
614c6cb4 ACM |
630 | /* With per-bucket locks this operation is not-atomic, so that |
631 | * this version is not worse. | |
632 | */ | |
633 | static inline void __sk_prot_rehash(struct sock *sk) | |
634 | { | |
635 | sk->sk_prot->unhash(sk); | |
636 | sk->sk_prot->hash(sk); | |
637 | } | |
638 | ||
1da177e4 LT |
639 | /* About 10 seconds */ |
640 | #define SOCK_DESTROY_TIME (10*HZ) | |
641 | ||
642 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
643 | #define PROT_SOCK 1024 | |
644 | ||
645 | #define SHUTDOWN_MASK 3 | |
646 | #define RCV_SHUTDOWN 1 | |
647 | #define SEND_SHUTDOWN 2 | |
648 | ||
649 | #define SOCK_SNDBUF_LOCK 1 | |
650 | #define SOCK_RCVBUF_LOCK 2 | |
651 | #define SOCK_BINDADDR_LOCK 4 | |
652 | #define SOCK_BINDPORT_LOCK 8 | |
653 | ||
654 | /* sock_iocb: used to kick off async processing of socket ios */ | |
655 | struct sock_iocb { | |
656 | struct list_head list; | |
657 | ||
658 | int flags; | |
659 | int size; | |
660 | struct socket *sock; | |
661 | struct sock *sk; | |
662 | struct scm_cookie *scm; | |
663 | struct msghdr *msg, async_msg; | |
664 | struct iovec async_iov; | |
665 | struct kiocb *kiocb; | |
666 | }; | |
667 | ||
668 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) | |
669 | { | |
670 | return (struct sock_iocb *)iocb->private; | |
671 | } | |
672 | ||
673 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) | |
674 | { | |
675 | return si->kiocb; | |
676 | } | |
677 | ||
678 | struct socket_alloc { | |
679 | struct socket socket; | |
680 | struct inode vfs_inode; | |
681 | }; | |
682 | ||
683 | static inline struct socket *SOCKET_I(struct inode *inode) | |
684 | { | |
685 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
686 | } | |
687 | ||
688 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
689 | { | |
690 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
691 | } | |
692 | ||
693 | extern void __sk_stream_mem_reclaim(struct sock *sk); | |
694 | extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind); | |
695 | ||
696 | #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE) | |
697 | ||
698 | static inline int sk_stream_pages(int amt) | |
699 | { | |
700 | return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM; | |
701 | } | |
702 | ||
703 | static inline void sk_stream_mem_reclaim(struct sock *sk) | |
704 | { | |
705 | if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM) | |
706 | __sk_stream_mem_reclaim(sk); | |
707 | } | |
708 | ||
709 | static inline void sk_stream_writequeue_purge(struct sock *sk) | |
710 | { | |
711 | struct sk_buff *skb; | |
712 | ||
713 | while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) | |
714 | sk_stream_free_skb(sk, skb); | |
715 | sk_stream_mem_reclaim(sk); | |
716 | } | |
717 | ||
718 | static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb) | |
719 | { | |
720 | return (int)skb->truesize <= sk->sk_forward_alloc || | |
721 | sk_stream_mem_schedule(sk, skb->truesize, 1); | |
722 | } | |
723 | ||
d80d99d6 HX |
724 | static inline int sk_stream_wmem_schedule(struct sock *sk, int size) |
725 | { | |
726 | return size <= sk->sk_forward_alloc || | |
727 | sk_stream_mem_schedule(sk, size, 0); | |
728 | } | |
729 | ||
1da177e4 LT |
730 | /* Used by processes to "lock" a socket state, so that |
731 | * interrupts and bottom half handlers won't change it | |
732 | * from under us. It essentially blocks any incoming | |
733 | * packets, so that we won't get any new data or any | |
734 | * packets that change the state of the socket. | |
735 | * | |
736 | * While locked, BH processing will add new packets to | |
737 | * the backlog queue. This queue is processed by the | |
738 | * owner of the socket lock right before it is released. | |
739 | * | |
740 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
741 | * accesses from user process context. | |
742 | */ | |
743 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owner) | |
744 | ||
745 | extern void FASTCALL(lock_sock(struct sock *sk)); | |
746 | extern void FASTCALL(release_sock(struct sock *sk)); | |
747 | ||
748 | /* BH context may only use the following locking interface. */ | |
749 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
750 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) | |
751 | ||
86a76caf | 752 | extern struct sock *sk_alloc(int family, |
dd0fc66f | 753 | gfp_t priority, |
1da177e4 LT |
754 | struct proto *prot, int zero_it); |
755 | extern void sk_free(struct sock *sk); | |
87d11ceb | 756 | extern struct sock *sk_clone(const struct sock *sk, |
dd0fc66f | 757 | const gfp_t priority); |
1da177e4 LT |
758 | |
759 | extern struct sk_buff *sock_wmalloc(struct sock *sk, | |
760 | unsigned long size, int force, | |
dd0fc66f | 761 | gfp_t priority); |
1da177e4 LT |
762 | extern struct sk_buff *sock_rmalloc(struct sock *sk, |
763 | unsigned long size, int force, | |
dd0fc66f | 764 | gfp_t priority); |
1da177e4 LT |
765 | extern void sock_wfree(struct sk_buff *skb); |
766 | extern void sock_rfree(struct sk_buff *skb); | |
767 | ||
768 | extern int sock_setsockopt(struct socket *sock, int level, | |
769 | int op, char __user *optval, | |
770 | int optlen); | |
771 | ||
772 | extern int sock_getsockopt(struct socket *sock, int level, | |
773 | int op, char __user *optval, | |
774 | int __user *optlen); | |
775 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, | |
776 | unsigned long size, | |
777 | int noblock, | |
778 | int *errcode); | |
86a76caf | 779 | extern void *sock_kmalloc(struct sock *sk, int size, |
dd0fc66f | 780 | gfp_t priority); |
1da177e4 LT |
781 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); |
782 | extern void sk_send_sigurg(struct sock *sk); | |
783 | ||
784 | /* | |
785 | * Functions to fill in entries in struct proto_ops when a protocol | |
786 | * does not implement a particular function. | |
787 | */ | |
788 | extern int sock_no_bind(struct socket *, | |
789 | struct sockaddr *, int); | |
790 | extern int sock_no_connect(struct socket *, | |
791 | struct sockaddr *, int, int); | |
792 | extern int sock_no_socketpair(struct socket *, | |
793 | struct socket *); | |
794 | extern int sock_no_accept(struct socket *, | |
795 | struct socket *, int); | |
796 | extern int sock_no_getname(struct socket *, | |
797 | struct sockaddr *, int *, int); | |
798 | extern unsigned int sock_no_poll(struct file *, struct socket *, | |
799 | struct poll_table_struct *); | |
800 | extern int sock_no_ioctl(struct socket *, unsigned int, | |
801 | unsigned long); | |
802 | extern int sock_no_listen(struct socket *, int); | |
803 | extern int sock_no_shutdown(struct socket *, int); | |
804 | extern int sock_no_getsockopt(struct socket *, int , int, | |
805 | char __user *, int __user *); | |
806 | extern int sock_no_setsockopt(struct socket *, int, int, | |
807 | char __user *, int); | |
808 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, | |
809 | struct msghdr *, size_t); | |
810 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, | |
811 | struct msghdr *, size_t, int); | |
812 | extern int sock_no_mmap(struct file *file, | |
813 | struct socket *sock, | |
814 | struct vm_area_struct *vma); | |
815 | extern ssize_t sock_no_sendpage(struct socket *sock, | |
816 | struct page *page, | |
817 | int offset, size_t size, | |
818 | int flags); | |
819 | ||
820 | /* | |
821 | * Functions to fill in entries in struct proto_ops when a protocol | |
822 | * uses the inet style. | |
823 | */ | |
824 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, | |
825 | char __user *optval, int __user *optlen); | |
826 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, | |
827 | struct msghdr *msg, size_t size, int flags); | |
828 | extern int sock_common_setsockopt(struct socket *sock, int level, int optname, | |
829 | char __user *optval, int optlen); | |
3fdadf7d DM |
830 | extern int compat_sock_common_getsockopt(struct socket *sock, int level, |
831 | int optname, char __user *optval, int __user *optlen); | |
832 | extern int compat_sock_common_setsockopt(struct socket *sock, int level, | |
833 | int optname, char __user *optval, int optlen); | |
1da177e4 LT |
834 | |
835 | extern void sk_common_release(struct sock *sk); | |
836 | ||
837 | /* | |
838 | * Default socket callbacks and setup code | |
839 | */ | |
840 | ||
841 | /* Initialise core socket variables */ | |
842 | extern void sock_init_data(struct socket *sock, struct sock *sk); | |
843 | ||
844 | /** | |
845 | * sk_filter - run a packet through a socket filter | |
846 | * @sk: sock associated with &sk_buff | |
847 | * @skb: buffer to filter | |
848 | * @needlock: set to 1 if the sock is not locked by caller. | |
849 | * | |
850 | * Run the filter code and then cut skb->data to correct size returned by | |
851 | * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller | |
852 | * than pkt_len we keep whole skb->data. This is the socket level | |
853 | * wrapper to sk_run_filter. It returns 0 if the packet should | |
854 | * be accepted or -EPERM if the packet should be tossed. | |
855 | * | |
856 | */ | |
857 | ||
858 | static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock) | |
859 | { | |
860 | int err; | |
861 | ||
862 | err = security_sock_rcv_skb(sk, skb); | |
863 | if (err) | |
864 | return err; | |
865 | ||
866 | if (sk->sk_filter) { | |
867 | struct sk_filter *filter; | |
868 | ||
869 | if (needlock) | |
870 | bh_lock_sock(sk); | |
871 | ||
872 | filter = sk->sk_filter; | |
873 | if (filter) { | |
4bad4dc9 KK |
874 | unsigned int pkt_len = sk_run_filter(skb, filter->insns, |
875 | filter->len); | |
b38dfee3 | 876 | err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM; |
1da177e4 LT |
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 | ||
f0088a50 | 941 | extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb); |
25995ff5 | 942 | |
1da177e4 LT |
943 | /* Detach socket from process context. |
944 | * Announce socket dead, detach it from wait queue and inode. | |
945 | * Note that parent inode held reference count on this struct sock, | |
946 | * we do not release it in this function, because protocol | |
947 | * probably wants some additional cleanups or even continuing | |
948 | * to work with this socket (TCP). | |
949 | */ | |
950 | static inline void sock_orphan(struct sock *sk) | |
951 | { | |
952 | write_lock_bh(&sk->sk_callback_lock); | |
953 | sock_set_flag(sk, SOCK_DEAD); | |
954 | sk->sk_socket = NULL; | |
955 | sk->sk_sleep = NULL; | |
956 | write_unlock_bh(&sk->sk_callback_lock); | |
957 | } | |
958 | ||
959 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
960 | { | |
961 | write_lock_bh(&sk->sk_callback_lock); | |
962 | sk->sk_sleep = &parent->wait; | |
963 | parent->sk = sk; | |
964 | sk->sk_socket = parent; | |
965 | write_unlock_bh(&sk->sk_callback_lock); | |
966 | } | |
967 | ||
968 | extern int sock_i_uid(struct sock *sk); | |
969 | extern unsigned long sock_i_ino(struct sock *sk); | |
970 | ||
971 | static inline struct dst_entry * | |
972 | __sk_dst_get(struct sock *sk) | |
973 | { | |
974 | return sk->sk_dst_cache; | |
975 | } | |
976 | ||
977 | static inline struct dst_entry * | |
978 | sk_dst_get(struct sock *sk) | |
979 | { | |
980 | struct dst_entry *dst; | |
981 | ||
982 | read_lock(&sk->sk_dst_lock); | |
983 | dst = sk->sk_dst_cache; | |
984 | if (dst) | |
985 | dst_hold(dst); | |
986 | read_unlock(&sk->sk_dst_lock); | |
987 | return dst; | |
988 | } | |
989 | ||
990 | static inline void | |
991 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
992 | { | |
993 | struct dst_entry *old_dst; | |
994 | ||
995 | old_dst = sk->sk_dst_cache; | |
996 | sk->sk_dst_cache = dst; | |
997 | dst_release(old_dst); | |
998 | } | |
999 | ||
1000 | static inline void | |
1001 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1002 | { | |
1003 | write_lock(&sk->sk_dst_lock); | |
1004 | __sk_dst_set(sk, dst); | |
1005 | write_unlock(&sk->sk_dst_lock); | |
1006 | } | |
1007 | ||
1008 | static inline void | |
1009 | __sk_dst_reset(struct sock *sk) | |
1010 | { | |
1011 | struct dst_entry *old_dst; | |
1012 | ||
1013 | old_dst = sk->sk_dst_cache; | |
1014 | sk->sk_dst_cache = NULL; | |
1015 | dst_release(old_dst); | |
1016 | } | |
1017 | ||
1018 | static inline void | |
1019 | sk_dst_reset(struct sock *sk) | |
1020 | { | |
1021 | write_lock(&sk->sk_dst_lock); | |
1022 | __sk_dst_reset(sk); | |
1023 | write_unlock(&sk->sk_dst_lock); | |
1024 | } | |
1025 | ||
f0088a50 | 1026 | extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1027 | |
f0088a50 | 1028 | extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1029 | |
6cbb0df7 ACM |
1030 | static inline void sk_setup_caps(struct sock *sk, struct dst_entry *dst) |
1031 | { | |
1032 | __sk_dst_set(sk, dst); | |
1033 | sk->sk_route_caps = dst->dev->features; | |
1034 | if (sk->sk_route_caps & NETIF_F_TSO) { | |
1035 | if (sock_flag(sk, SOCK_NO_LARGESEND) || dst->header_len) | |
1036 | sk->sk_route_caps &= ~NETIF_F_TSO; | |
1037 | } | |
1038 | } | |
1039 | ||
1da177e4 LT |
1040 | static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb) |
1041 | { | |
1042 | sk->sk_wmem_queued += skb->truesize; | |
1043 | sk->sk_forward_alloc -= skb->truesize; | |
1044 | } | |
1045 | ||
1046 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, | |
1047 | struct sk_buff *skb, struct page *page, | |
1048 | int off, int copy) | |
1049 | { | |
1050 | if (skb->ip_summed == CHECKSUM_NONE) { | |
1051 | int err = 0; | |
1052 | unsigned int csum = csum_and_copy_from_user(from, | |
1053 | page_address(page) + off, | |
1054 | copy, 0, &err); | |
1055 | if (err) | |
1056 | return err; | |
1057 | skb->csum = csum_block_add(skb->csum, csum, skb->len); | |
1058 | } else if (copy_from_user(page_address(page) + off, from, copy)) | |
1059 | return -EFAULT; | |
1060 | ||
1061 | skb->len += copy; | |
1062 | skb->data_len += copy; | |
1063 | skb->truesize += copy; | |
1064 | sk->sk_wmem_queued += copy; | |
1065 | sk->sk_forward_alloc -= copy; | |
1066 | return 0; | |
1067 | } | |
1068 | ||
1069 | /* | |
1070 | * Queue a received datagram if it will fit. Stream and sequenced | |
1071 | * protocols can't normally use this as they need to fit buffers in | |
1072 | * and play with them. | |
1073 | * | |
1074 | * Inlined as it's very short and called for pretty much every | |
1075 | * packet ever received. | |
1076 | */ | |
1077 | ||
1078 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) | |
1079 | { | |
1080 | sock_hold(sk); | |
1081 | skb->sk = sk; | |
1082 | skb->destructor = sock_wfree; | |
1083 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); | |
1084 | } | |
1085 | ||
1086 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
1087 | { | |
1088 | skb->sk = sk; | |
1089 | skb->destructor = sock_rfree; | |
1090 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
1091 | } | |
1092 | ||
1093 | extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, | |
1094 | unsigned long expires); | |
1095 | ||
1096 | extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); | |
1097 | ||
f0088a50 | 1098 | extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 LT |
1099 | |
1100 | static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb) | |
1101 | { | |
1102 | /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces | |
1103 | number of warnings when compiling with -W --ANK | |
1104 | */ | |
1105 | if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >= | |
1106 | (unsigned)sk->sk_rcvbuf) | |
1107 | return -ENOMEM; | |
1108 | skb_set_owner_r(skb, sk); | |
1109 | skb_queue_tail(&sk->sk_error_queue, skb); | |
1110 | if (!sock_flag(sk, SOCK_DEAD)) | |
1111 | sk->sk_data_ready(sk, skb->len); | |
1112 | return 0; | |
1113 | } | |
1114 | ||
1115 | /* | |
1116 | * Recover an error report and clear atomically | |
1117 | */ | |
1118 | ||
1119 | static inline int sock_error(struct sock *sk) | |
1120 | { | |
c1cbe4b7 BL |
1121 | int err; |
1122 | if (likely(!sk->sk_err)) | |
1123 | return 0; | |
1124 | err = xchg(&sk->sk_err, 0); | |
1da177e4 LT |
1125 | return -err; |
1126 | } | |
1127 | ||
1128 | static inline unsigned long sock_wspace(struct sock *sk) | |
1129 | { | |
1130 | int amt = 0; | |
1131 | ||
1132 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
1133 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); | |
1134 | if (amt < 0) | |
1135 | amt = 0; | |
1136 | } | |
1137 | return amt; | |
1138 | } | |
1139 | ||
1140 | static inline void sk_wake_async(struct sock *sk, int how, int band) | |
1141 | { | |
1142 | if (sk->sk_socket && sk->sk_socket->fasync_list) | |
1143 | sock_wake_async(sk->sk_socket, how, band); | |
1144 | } | |
1145 | ||
1146 | #define SOCK_MIN_SNDBUF 2048 | |
1147 | #define SOCK_MIN_RCVBUF 256 | |
1148 | ||
1149 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
1150 | { | |
1151 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { | |
1152 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2); | |
1153 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); | |
1154 | } | |
1155 | } | |
1156 | ||
1157 | static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk, | |
86a76caf | 1158 | int size, int mem, |
dd0fc66f | 1159 | gfp_t gfp) |
1da177e4 | 1160 | { |
c65f7f00 DM |
1161 | struct sk_buff *skb; |
1162 | int hdr_len; | |
1da177e4 | 1163 | |
c65f7f00 | 1164 | hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header); |
d179cd12 | 1165 | skb = alloc_skb_fclone(size + hdr_len, gfp); |
1da177e4 LT |
1166 | if (skb) { |
1167 | skb->truesize += mem; | |
d80d99d6 | 1168 | if (sk_stream_wmem_schedule(sk, skb->truesize)) { |
c65f7f00 | 1169 | skb_reserve(skb, hdr_len); |
1da177e4 LT |
1170 | return skb; |
1171 | } | |
1172 | __kfree_skb(skb); | |
1173 | } else { | |
1174 | sk->sk_prot->enter_memory_pressure(); | |
1175 | sk_stream_moderate_sndbuf(sk); | |
1176 | } | |
1177 | return NULL; | |
1178 | } | |
1179 | ||
1180 | static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk, | |
86a76caf | 1181 | int size, |
dd0fc66f | 1182 | gfp_t gfp) |
1da177e4 LT |
1183 | { |
1184 | return sk_stream_alloc_pskb(sk, size, 0, gfp); | |
1185 | } | |
1186 | ||
1187 | static inline struct page *sk_stream_alloc_page(struct sock *sk) | |
1188 | { | |
1189 | struct page *page = NULL; | |
1190 | ||
ef015786 HX |
1191 | page = alloc_pages(sk->sk_allocation, 0); |
1192 | if (!page) { | |
1da177e4 LT |
1193 | sk->sk_prot->enter_memory_pressure(); |
1194 | sk_stream_moderate_sndbuf(sk); | |
1195 | } | |
1196 | return page; | |
1197 | } | |
1198 | ||
1199 | #define sk_stream_for_retrans_queue(skb, sk) \ | |
1200 | for (skb = (sk)->sk_write_queue.next; \ | |
1201 | (skb != (sk)->sk_send_head) && \ | |
1202 | (skb != (struct sk_buff *)&(sk)->sk_write_queue); \ | |
1203 | skb = skb->next) | |
1204 | ||
6a438bbe SH |
1205 | /*from STCP for fast SACK Process*/ |
1206 | #define sk_stream_for_retrans_queue_from(skb, sk) \ | |
1207 | for (; (skb != (sk)->sk_send_head) && \ | |
1208 | (skb != (struct sk_buff *)&(sk)->sk_write_queue); \ | |
1209 | skb = skb->next) | |
1210 | ||
1da177e4 LT |
1211 | /* |
1212 | * Default write policy as shown to user space via poll/select/SIGIO | |
1213 | */ | |
1214 | static inline int sock_writeable(const struct sock *sk) | |
1215 | { | |
1216 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2); | |
1217 | } | |
1218 | ||
dd0fc66f | 1219 | static inline gfp_t gfp_any(void) |
1da177e4 LT |
1220 | { |
1221 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; | |
1222 | } | |
1223 | ||
1224 | static inline long sock_rcvtimeo(const struct sock *sk, int noblock) | |
1225 | { | |
1226 | return noblock ? 0 : sk->sk_rcvtimeo; | |
1227 | } | |
1228 | ||
1229 | static inline long sock_sndtimeo(const struct sock *sk, int noblock) | |
1230 | { | |
1231 | return noblock ? 0 : sk->sk_sndtimeo; | |
1232 | } | |
1233 | ||
1234 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
1235 | { | |
1236 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; | |
1237 | } | |
1238 | ||
1239 | /* Alas, with timeout socket operations are not restartable. | |
1240 | * Compare this to poll(). | |
1241 | */ | |
1242 | static inline int sock_intr_errno(long timeo) | |
1243 | { | |
1244 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
1245 | } | |
1246 | ||
1247 | static __inline__ void | |
1248 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) | |
1249 | { | |
a61bbcf2 PM |
1250 | struct timeval stamp; |
1251 | ||
1252 | skb_get_timestamp(skb, &stamp); | |
1da177e4 LT |
1253 | if (sock_flag(sk, SOCK_RCVTSTAMP)) { |
1254 | /* Race occurred between timestamp enabling and packet | |
1255 | receiving. Fill in the current time for now. */ | |
a61bbcf2 PM |
1256 | if (stamp.tv_sec == 0) |
1257 | do_gettimeofday(&stamp); | |
1258 | skb_set_timestamp(skb, &stamp); | |
1da177e4 | 1259 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval), |
a61bbcf2 | 1260 | &stamp); |
1da177e4 | 1261 | } else |
a61bbcf2 | 1262 | sk->sk_stamp = stamp; |
1da177e4 LT |
1263 | } |
1264 | ||
1265 | /** | |
1266 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
1267 | * @sk: socket to eat this skb from |
1268 | * @skb: socket buffer to eat | |
1da177e4 LT |
1269 | * |
1270 | * This routine must be called with interrupts disabled or with the socket | |
1271 | * locked so that the sk_buff queue operation is ok. | |
1272 | */ | |
624d1164 CL |
1273 | #ifdef CONFIG_NET_DMA |
1274 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) | |
1275 | { | |
1276 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1277 | if (!copied_early) | |
1278 | __kfree_skb(skb); | |
1279 | else | |
1280 | __skb_queue_tail(&sk->sk_async_wait_queue, skb); | |
1281 | } | |
1282 | #else | |
1283 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) | |
1da177e4 LT |
1284 | { |
1285 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1286 | __kfree_skb(skb); | |
1287 | } | |
624d1164 | 1288 | #endif |
1da177e4 LT |
1289 | |
1290 | extern void sock_enable_timestamp(struct sock *sk); | |
1291 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); | |
1292 | ||
1293 | /* | |
1294 | * Enable debug/info messages | |
1295 | */ | |
1296 | ||
0dec456d | 1297 | #ifdef CONFIG_NETDEBUG |
64ce2073 PM |
1298 | #define NETDEBUG(fmt, args...) printk(fmt,##args) |
1299 | #define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0) | |
0dec456d SH |
1300 | #else |
1301 | #define NETDEBUG(fmt, args...) do { } while (0) | |
1302 | #define LIMIT_NETDEBUG(fmt, args...) do { } while(0) | |
1da177e4 LT |
1303 | #endif |
1304 | ||
1305 | /* | |
1306 | * Macros for sleeping on a socket. Use them like this: | |
1307 | * | |
1308 | * SOCK_SLEEP_PRE(sk) | |
1309 | * if (condition) | |
1310 | * schedule(); | |
1311 | * SOCK_SLEEP_POST(sk) | |
1312 | * | |
1313 | * N.B. These are now obsolete and were, afaik, only ever used in DECnet | |
1314 | * and when the last use of them in DECnet has gone, I'm intending to | |
1315 | * remove them. | |
1316 | */ | |
1317 | ||
1318 | #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \ | |
1319 | DECLARE_WAITQUEUE(wait, tsk); \ | |
1320 | tsk->state = TASK_INTERRUPTIBLE; \ | |
1321 | add_wait_queue((sk)->sk_sleep, &wait); \ | |
1322 | release_sock(sk); | |
1323 | ||
1324 | #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \ | |
1325 | remove_wait_queue((sk)->sk_sleep, &wait); \ | |
1326 | lock_sock(sk); \ | |
1327 | } | |
1328 | ||
1329 | static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool) | |
1330 | { | |
1331 | if (valbool) | |
1332 | sock_set_flag(sk, bit); | |
1333 | else | |
1334 | sock_reset_flag(sk, bit); | |
1335 | } | |
1336 | ||
1337 | extern __u32 sysctl_wmem_max; | |
1338 | extern __u32 sysctl_rmem_max; | |
1339 | ||
1340 | #ifdef CONFIG_NET | |
1341 | int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg); | |
1342 | #else | |
1343 | static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg) | |
1344 | { | |
1345 | return -ENODEV; | |
1346 | } | |
1347 | #endif | |
1348 | ||
20380731 ACM |
1349 | extern void sk_init(void); |
1350 | ||
1351 | #ifdef CONFIG_SYSCTL | |
1352 | extern struct ctl_table core_table[]; | |
20380731 ACM |
1353 | #endif |
1354 | ||
6baf1f41 DM |
1355 | extern int sysctl_optmem_max; |
1356 | ||
20380731 ACM |
1357 | extern __u32 sysctl_wmem_default; |
1358 | extern __u32 sysctl_rmem_default; | |
20380731 | 1359 | |
1da177e4 | 1360 | #endif /* _SOCK_H */ |