]>
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 | ||
172589cc | 43 | #include <linux/kernel.h> |
1da177e4 LT |
44 | #include <linux/list.h> |
45 | #include <linux/timer.h> | |
46 | #include <linux/cache.h> | |
47 | #include <linux/module.h> | |
a5b5bb9a | 48 | #include <linux/lockdep.h> |
1da177e4 | 49 | #include <linux/netdevice.h> |
ebb53d75 | 50 | #include <linux/pcounter.h> |
1da177e4 | 51 | #include <linux/skbuff.h> /* struct sk_buff */ |
d7fe0f24 | 52 | #include <linux/mm.h> |
1da177e4 LT |
53 | #include <linux/security.h> |
54 | ||
55 | #include <linux/filter.h> | |
56 | ||
57 | #include <asm/atomic.h> | |
58 | #include <net/dst.h> | |
59 | #include <net/checksum.h> | |
60 | ||
61 | /* | |
62 | * This structure really needs to be cleaned up. | |
63 | * Most of it is for TCP, and not used by any of | |
64 | * the other protocols. | |
65 | */ | |
66 | ||
67 | /* Define this to get the SOCK_DBG debugging facility. */ | |
68 | #define SOCK_DEBUGGING | |
69 | #ifdef SOCK_DEBUGGING | |
70 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ | |
71 | printk(KERN_DEBUG msg); } while (0) | |
72 | #else | |
73 | #define SOCK_DEBUG(sk, msg...) do { } while (0) | |
74 | #endif | |
75 | ||
76 | /* This is the per-socket lock. The spinlock provides a synchronization | |
77 | * between user contexts and software interrupt processing, whereas the | |
78 | * mini-semaphore synchronizes multiple users amongst themselves. | |
79 | */ | |
1da177e4 LT |
80 | typedef struct { |
81 | spinlock_t slock; | |
d2e9117c | 82 | int owned; |
1da177e4 | 83 | wait_queue_head_t wq; |
a5b5bb9a IM |
84 | /* |
85 | * We express the mutex-alike socket_lock semantics | |
86 | * to the lock validator by explicitly managing | |
87 | * the slock as a lock variant (in addition to | |
88 | * the slock itself): | |
89 | */ | |
90 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
91 | struct lockdep_map dep_map; | |
92 | #endif | |
1da177e4 LT |
93 | } socket_lock_t; |
94 | ||
1da177e4 | 95 | struct sock; |
8feaf0c0 | 96 | struct proto; |
0eeb8ffc | 97 | struct net; |
1da177e4 LT |
98 | |
99 | /** | |
4dc3b16b PP |
100 | * struct sock_common - minimal network layer representation of sockets |
101 | * @skc_family: network address family | |
102 | * @skc_state: Connection state | |
103 | * @skc_reuse: %SO_REUSEADDR setting | |
104 | * @skc_bound_dev_if: bound device index if != 0 | |
105 | * @skc_node: main hash linkage for various protocol lookup tables | |
106 | * @skc_bind_node: bind hash linkage for various protocol lookup tables | |
107 | * @skc_refcnt: reference count | |
81c3d547 | 108 | * @skc_hash: hash value used with various protocol lookup tables |
8feaf0c0 | 109 | * @skc_prot: protocol handlers inside a network family |
07feaebf | 110 | * @skc_net: reference to the network namespace of this socket |
4dc3b16b PP |
111 | * |
112 | * This is the minimal network layer representation of sockets, the header | |
8feaf0c0 ACM |
113 | * for struct sock and struct inet_timewait_sock. |
114 | */ | |
1da177e4 LT |
115 | struct sock_common { |
116 | unsigned short skc_family; | |
117 | volatile unsigned char skc_state; | |
118 | unsigned char skc_reuse; | |
119 | int skc_bound_dev_if; | |
120 | struct hlist_node skc_node; | |
121 | struct hlist_node skc_bind_node; | |
122 | atomic_t skc_refcnt; | |
81c3d547 | 123 | unsigned int skc_hash; |
8feaf0c0 | 124 | struct proto *skc_prot; |
07feaebf | 125 | struct net *skc_net; |
1da177e4 LT |
126 | }; |
127 | ||
128 | /** | |
129 | * struct sock - network layer representation of sockets | |
8feaf0c0 | 130 | * @__sk_common: shared layout with inet_timewait_sock |
4dc3b16b PP |
131 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
132 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
133 | * @sk_lock: synchronizer | |
134 | * @sk_rcvbuf: size of receive buffer in bytes | |
135 | * @sk_sleep: sock wait queue | |
136 | * @sk_dst_cache: destination cache | |
137 | * @sk_dst_lock: destination cache lock | |
138 | * @sk_policy: flow policy | |
139 | * @sk_rmem_alloc: receive queue bytes committed | |
140 | * @sk_receive_queue: incoming packets | |
141 | * @sk_wmem_alloc: transmit queue bytes committed | |
142 | * @sk_write_queue: Packet sending queue | |
97fc2f08 | 143 | * @sk_async_wait_queue: DMA copied packets |
4dc3b16b PP |
144 | * @sk_omem_alloc: "o" is "option" or "other" |
145 | * @sk_wmem_queued: persistent queue size | |
146 | * @sk_forward_alloc: space allocated forward | |
147 | * @sk_allocation: allocation mode | |
148 | * @sk_sndbuf: size of send buffer in bytes | |
33c732c3 WC |
149 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, |
150 | * %SO_OOBINLINE settings | |
4dc3b16b PP |
151 | * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets |
152 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) | |
bcd76111 | 153 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
4dc3b16b | 154 | * @sk_lingertime: %SO_LINGER l_linger setting |
4dc3b16b PP |
155 | * @sk_backlog: always used with the per-socket spinlock held |
156 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
157 | * @sk_error_queue: rarely used | |
33c732c3 WC |
158 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
159 | * IPV6_ADDRFORM for instance) | |
4dc3b16b | 160 | * @sk_err: last error |
33c732c3 WC |
161 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
162 | * persistent failure not just 'timed out' | |
163 | * @sk_drops: raw drops counter | |
4dc3b16b PP |
164 | * @sk_ack_backlog: current listen backlog |
165 | * @sk_max_ack_backlog: listen backlog set in listen() | |
166 | * @sk_priority: %SO_PRIORITY setting | |
167 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
168 | * @sk_protocol: which protocol this socket belongs in this network family | |
169 | * @sk_peercred: %SO_PEERCRED setting | |
170 | * @sk_rcvlowat: %SO_RCVLOWAT setting | |
171 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
172 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
173 | * @sk_filter: socket filtering instructions | |
174 | * @sk_protinfo: private area, net family specific, when not using slab | |
175 | * @sk_timer: sock cleanup timer | |
176 | * @sk_stamp: time stamp of last packet received | |
177 | * @sk_socket: Identd and reporting IO signals | |
178 | * @sk_user_data: RPC layer private data | |
179 | * @sk_sndmsg_page: cached page for sendmsg | |
180 | * @sk_sndmsg_off: cached offset for sendmsg | |
181 | * @sk_send_head: front of stuff to transmit | |
67be2dd1 | 182 | * @sk_security: used by security modules |
4dc3b16b PP |
183 | * @sk_write_pending: a write to stream socket waits to start |
184 | * @sk_state_change: callback to indicate change in the state of the sock | |
185 | * @sk_data_ready: callback to indicate there is data to be processed | |
186 | * @sk_write_space: callback to indicate there is bf sending space available | |
187 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
188 | * @sk_backlog_rcv: callback to process the backlog | |
189 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
1da177e4 LT |
190 | */ |
191 | struct sock { | |
192 | /* | |
8feaf0c0 | 193 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
194 | * don't add nothing before this first member (__sk_common) --acme |
195 | */ | |
196 | struct sock_common __sk_common; | |
197 | #define sk_family __sk_common.skc_family | |
198 | #define sk_state __sk_common.skc_state | |
199 | #define sk_reuse __sk_common.skc_reuse | |
200 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if | |
201 | #define sk_node __sk_common.skc_node | |
202 | #define sk_bind_node __sk_common.skc_bind_node | |
203 | #define sk_refcnt __sk_common.skc_refcnt | |
81c3d547 | 204 | #define sk_hash __sk_common.skc_hash |
8feaf0c0 | 205 | #define sk_prot __sk_common.skc_prot |
07feaebf | 206 | #define sk_net __sk_common.skc_net |
1da177e4 LT |
207 | unsigned char sk_shutdown : 2, |
208 | sk_no_check : 2, | |
209 | sk_userlocks : 4; | |
210 | unsigned char sk_protocol; | |
211 | unsigned short sk_type; | |
212 | int sk_rcvbuf; | |
213 | socket_lock_t sk_lock; | |
fa438ccf ED |
214 | /* |
215 | * The backlog queue is special, it is always used with | |
216 | * the per-socket spinlock held and requires low latency | |
217 | * access. Therefore we special case it's implementation. | |
218 | */ | |
219 | struct { | |
220 | struct sk_buff *head; | |
221 | struct sk_buff *tail; | |
222 | } sk_backlog; | |
1da177e4 LT |
223 | wait_queue_head_t *sk_sleep; |
224 | struct dst_entry *sk_dst_cache; | |
225 | struct xfrm_policy *sk_policy[2]; | |
226 | rwlock_t sk_dst_lock; | |
227 | atomic_t sk_rmem_alloc; | |
228 | atomic_t sk_wmem_alloc; | |
229 | atomic_t sk_omem_alloc; | |
4e07a91c | 230 | int sk_sndbuf; |
1da177e4 LT |
231 | struct sk_buff_head sk_receive_queue; |
232 | struct sk_buff_head sk_write_queue; | |
97fc2f08 | 233 | struct sk_buff_head sk_async_wait_queue; |
1da177e4 LT |
234 | int sk_wmem_queued; |
235 | int sk_forward_alloc; | |
7d877f3b | 236 | gfp_t sk_allocation; |
1da177e4 | 237 | int sk_route_caps; |
bcd76111 | 238 | int sk_gso_type; |
9932cf95 | 239 | int sk_rcvlowat; |
1da177e4 LT |
240 | unsigned long sk_flags; |
241 | unsigned long sk_lingertime; | |
1da177e4 | 242 | struct sk_buff_head sk_error_queue; |
476e19cf | 243 | struct proto *sk_prot_creator; |
1da177e4 LT |
244 | rwlock_t sk_callback_lock; |
245 | int sk_err, | |
246 | sk_err_soft; | |
33c732c3 | 247 | atomic_t sk_drops; |
1da177e4 LT |
248 | unsigned short sk_ack_backlog; |
249 | unsigned short sk_max_ack_backlog; | |
250 | __u32 sk_priority; | |
251 | struct ucred sk_peercred; | |
1da177e4 LT |
252 | long sk_rcvtimeo; |
253 | long sk_sndtimeo; | |
254 | struct sk_filter *sk_filter; | |
255 | void *sk_protinfo; | |
256 | struct timer_list sk_timer; | |
b7aa0bf7 | 257 | ktime_t sk_stamp; |
1da177e4 LT |
258 | struct socket *sk_socket; |
259 | void *sk_user_data; | |
260 | struct page *sk_sndmsg_page; | |
261 | struct sk_buff *sk_send_head; | |
262 | __u32 sk_sndmsg_off; | |
263 | int sk_write_pending; | |
264 | void *sk_security; | |
265 | void (*sk_state_change)(struct sock *sk); | |
266 | void (*sk_data_ready)(struct sock *sk, int bytes); | |
267 | void (*sk_write_space)(struct sock *sk); | |
268 | void (*sk_error_report)(struct sock *sk); | |
269 | int (*sk_backlog_rcv)(struct sock *sk, | |
270 | struct sk_buff *skb); | |
271 | void (*sk_destruct)(struct sock *sk); | |
272 | }; | |
273 | ||
274 | /* | |
275 | * Hashed lists helper routines | |
276 | */ | |
e48c414e | 277 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
278 | { |
279 | return hlist_entry(head->first, struct sock, sk_node); | |
280 | } | |
281 | ||
e48c414e | 282 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
283 | { |
284 | return hlist_empty(head) ? NULL : __sk_head(head); | |
285 | } | |
286 | ||
e48c414e | 287 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 LT |
288 | { |
289 | return sk->sk_node.next ? | |
290 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; | |
291 | } | |
292 | ||
e48c414e | 293 | static inline int sk_unhashed(const struct sock *sk) |
1da177e4 LT |
294 | { |
295 | return hlist_unhashed(&sk->sk_node); | |
296 | } | |
297 | ||
e48c414e | 298 | static inline int sk_hashed(const struct sock *sk) |
1da177e4 | 299 | { |
da753bea | 300 | return !sk_unhashed(sk); |
1da177e4 LT |
301 | } |
302 | ||
303 | static __inline__ void sk_node_init(struct hlist_node *node) | |
304 | { | |
305 | node->pprev = NULL; | |
306 | } | |
307 | ||
308 | static __inline__ void __sk_del_node(struct sock *sk) | |
309 | { | |
310 | __hlist_del(&sk->sk_node); | |
311 | } | |
312 | ||
313 | static __inline__ int __sk_del_node_init(struct sock *sk) | |
314 | { | |
315 | if (sk_hashed(sk)) { | |
316 | __sk_del_node(sk); | |
317 | sk_node_init(&sk->sk_node); | |
318 | return 1; | |
319 | } | |
320 | return 0; | |
321 | } | |
322 | ||
323 | /* Grab socket reference count. This operation is valid only | |
324 | when sk is ALREADY grabbed f.e. it is found in hash table | |
325 | or a list and the lookup is made under lock preventing hash table | |
326 | modifications. | |
327 | */ | |
328 | ||
329 | static inline void sock_hold(struct sock *sk) | |
330 | { | |
331 | atomic_inc(&sk->sk_refcnt); | |
332 | } | |
333 | ||
334 | /* Ungrab socket in the context, which assumes that socket refcnt | |
335 | cannot hit zero, f.e. it is true in context of any socketcall. | |
336 | */ | |
337 | static inline void __sock_put(struct sock *sk) | |
338 | { | |
339 | atomic_dec(&sk->sk_refcnt); | |
340 | } | |
341 | ||
342 | static __inline__ int sk_del_node_init(struct sock *sk) | |
343 | { | |
344 | int rc = __sk_del_node_init(sk); | |
345 | ||
346 | if (rc) { | |
347 | /* paranoid for a while -acme */ | |
348 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
349 | __sock_put(sk); | |
350 | } | |
351 | return rc; | |
352 | } | |
353 | ||
354 | static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) | |
355 | { | |
356 | hlist_add_head(&sk->sk_node, list); | |
357 | } | |
358 | ||
359 | static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) | |
360 | { | |
361 | sock_hold(sk); | |
362 | __sk_add_node(sk, list); | |
363 | } | |
364 | ||
365 | static __inline__ void __sk_del_bind_node(struct sock *sk) | |
366 | { | |
367 | __hlist_del(&sk->sk_bind_node); | |
368 | } | |
369 | ||
370 | static __inline__ void sk_add_bind_node(struct sock *sk, | |
371 | struct hlist_head *list) | |
372 | { | |
373 | hlist_add_head(&sk->sk_bind_node, list); | |
374 | } | |
375 | ||
376 | #define sk_for_each(__sk, node, list) \ | |
377 | hlist_for_each_entry(__sk, node, list, sk_node) | |
378 | #define sk_for_each_from(__sk, node) \ | |
379 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
380 | hlist_for_each_entry_from(__sk, node, sk_node) | |
381 | #define sk_for_each_continue(__sk, node) \ | |
382 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
383 | hlist_for_each_entry_continue(__sk, node, sk_node) | |
384 | #define sk_for_each_safe(__sk, node, tmp, list) \ | |
385 | hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) | |
386 | #define sk_for_each_bound(__sk, node, list) \ | |
387 | hlist_for_each_entry(__sk, node, list, sk_bind_node) | |
388 | ||
389 | /* Sock flags */ | |
390 | enum sock_flags { | |
391 | SOCK_DEAD, | |
392 | SOCK_DONE, | |
393 | SOCK_URGINLINE, | |
394 | SOCK_KEEPOPEN, | |
395 | SOCK_LINGER, | |
396 | SOCK_DESTROY, | |
397 | SOCK_BROADCAST, | |
398 | SOCK_TIMESTAMP, | |
399 | SOCK_ZAPPED, | |
400 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
401 | SOCK_DBG, /* %SO_DEBUG setting */ | |
402 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
92f37fd2 | 403 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
1da177e4 LT |
404 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
405 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
406 | }; | |
407 | ||
53b924b3 RB |
408 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
409 | { | |
410 | nsk->sk_flags = osk->sk_flags; | |
411 | } | |
412 | ||
1da177e4 LT |
413 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
414 | { | |
415 | __set_bit(flag, &sk->sk_flags); | |
416 | } | |
417 | ||
418 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
419 | { | |
420 | __clear_bit(flag, &sk->sk_flags); | |
421 | } | |
422 | ||
423 | static inline int sock_flag(struct sock *sk, enum sock_flags flag) | |
424 | { | |
425 | return test_bit(flag, &sk->sk_flags); | |
426 | } | |
427 | ||
428 | static inline void sk_acceptq_removed(struct sock *sk) | |
429 | { | |
430 | sk->sk_ack_backlog--; | |
431 | } | |
432 | ||
433 | static inline void sk_acceptq_added(struct sock *sk) | |
434 | { | |
435 | sk->sk_ack_backlog++; | |
436 | } | |
437 | ||
438 | static inline int sk_acceptq_is_full(struct sock *sk) | |
439 | { | |
64a14651 | 440 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
1da177e4 LT |
441 | } |
442 | ||
443 | /* | |
444 | * Compute minimal free write space needed to queue new packets. | |
445 | */ | |
446 | static inline int sk_stream_min_wspace(struct sock *sk) | |
447 | { | |
8df09ea3 | 448 | return sk->sk_wmem_queued >> 1; |
1da177e4 LT |
449 | } |
450 | ||
451 | static inline int sk_stream_wspace(struct sock *sk) | |
452 | { | |
453 | return sk->sk_sndbuf - sk->sk_wmem_queued; | |
454 | } | |
455 | ||
456 | extern void sk_stream_write_space(struct sock *sk); | |
457 | ||
458 | static inline int sk_stream_memory_free(struct sock *sk) | |
459 | { | |
460 | return sk->sk_wmem_queued < sk->sk_sndbuf; | |
461 | } | |
462 | ||
1da177e4 | 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 | 474 | |
cfcabdcc SH |
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)); \ | |
481 | } \ | |
482 | lock_sock(__sk); \ | |
483 | __rc = __condition; \ | |
484 | __rc; \ | |
485 | }) | |
1da177e4 LT |
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 | ||
286ab3d4 | 550 | /* Keeping track of sockets in use */ |
65f76517 | 551 | #ifdef CONFIG_PROC_FS |
ebb53d75 | 552 | struct pcounter inuse; |
65f76517 | 553 | #endif |
ebb53d75 | 554 | |
1da177e4 LT |
555 | /* Memory pressure */ |
556 | void (*enter_memory_pressure)(void); | |
557 | atomic_t *memory_allocated; /* Current allocated memory. */ | |
558 | atomic_t *sockets_allocated; /* Current number of sockets. */ | |
559 | /* | |
560 | * Pressure flag: try to collapse. | |
561 | * Technical note: it is used by multiple contexts non atomically. | |
3ab224be | 562 | * All the __sk_mem_schedule() is of this nature: accounting |
1da177e4 LT |
563 | * is strict, actions are advisory and have some latency. |
564 | */ | |
565 | int *memory_pressure; | |
566 | int *sysctl_mem; | |
567 | int *sysctl_wmem; | |
568 | int *sysctl_rmem; | |
569 | int max_header; | |
570 | ||
e18b890b | 571 | struct kmem_cache *slab; |
1da177e4 LT |
572 | unsigned int obj_size; |
573 | ||
0a5578cf | 574 | atomic_t *orphan_count; |
8feaf0c0 | 575 | |
60236fdd | 576 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 577 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 578 | |
1da177e4 LT |
579 | struct module *owner; |
580 | ||
581 | char name[32]; | |
582 | ||
583 | struct list_head node; | |
e6848976 ACM |
584 | #ifdef SOCK_REFCNT_DEBUG |
585 | atomic_t socks; | |
586 | #endif | |
1da177e4 LT |
587 | }; |
588 | ||
589 | extern int proto_register(struct proto *prot, int alloc_slab); | |
590 | extern void proto_unregister(struct proto *prot); | |
591 | ||
e6848976 ACM |
592 | #ifdef SOCK_REFCNT_DEBUG |
593 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
594 | { | |
595 | atomic_inc(&sk->sk_prot->socks); | |
596 | } | |
597 | ||
598 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
599 | { | |
600 | atomic_dec(&sk->sk_prot->socks); | |
601 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
602 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
603 | } | |
604 | ||
605 | static inline void sk_refcnt_debug_release(const struct sock *sk) | |
606 | { | |
607 | if (atomic_read(&sk->sk_refcnt) != 1) | |
608 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", | |
609 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); | |
610 | } | |
611 | #else /* SOCK_REFCNT_DEBUG */ | |
612 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
613 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
614 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
615 | #endif /* SOCK_REFCNT_DEBUG */ | |
616 | ||
65f76517 ED |
617 | |
618 | #ifdef CONFIG_PROC_FS | |
619 | # define DEFINE_PROTO_INUSE(NAME) DEFINE_PCOUNTER(NAME) | |
620 | # define REF_PROTO_INUSE(NAME) PCOUNTER_MEMBER_INITIALIZER(NAME, .inuse) | |
1da177e4 | 621 | /* Called with local bh disabled */ |
65f76517 | 622 | static inline void sock_prot_inuse_add(struct proto *prot, int inc) |
1da177e4 | 623 | { |
65f76517 | 624 | pcounter_add(&prot->inuse, inc); |
1da177e4 | 625 | } |
65f76517 | 626 | static inline int sock_prot_inuse_init(struct proto *proto) |
1da177e4 | 627 | { |
65f76517 | 628 | return pcounter_alloc(&proto->inuse); |
286ab3d4 | 629 | } |
65f76517 | 630 | static inline int sock_prot_inuse_get(struct proto *proto) |
286ab3d4 | 631 | { |
ebb53d75 | 632 | return pcounter_getval(&proto->inuse); |
1da177e4 | 633 | } |
65f76517 ED |
634 | static inline void sock_prot_inuse_free(struct proto *proto) |
635 | { | |
636 | pcounter_free(&proto->inuse); | |
637 | } | |
638 | #else | |
639 | # define DEFINE_PROTO_INUSE(NAME) | |
640 | # define REF_PROTO_INUSE(NAME) | |
641 | static void inline sock_prot_inuse_add(struct proto *prot, int inc) | |
642 | { | |
643 | } | |
644 | static int inline sock_prot_inuse_init(struct proto *proto) | |
645 | { | |
646 | return 0; | |
647 | } | |
648 | static void inline sock_prot_inuse_free(struct proto *proto) | |
649 | { | |
650 | } | |
651 | #endif | |
652 | ||
1da177e4 | 653 | |
614c6cb4 ACM |
654 | /* With per-bucket locks this operation is not-atomic, so that |
655 | * this version is not worse. | |
656 | */ | |
657 | static inline void __sk_prot_rehash(struct sock *sk) | |
658 | { | |
659 | sk->sk_prot->unhash(sk); | |
660 | sk->sk_prot->hash(sk); | |
661 | } | |
662 | ||
1da177e4 LT |
663 | /* About 10 seconds */ |
664 | #define SOCK_DESTROY_TIME (10*HZ) | |
665 | ||
666 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
667 | #define PROT_SOCK 1024 | |
668 | ||
669 | #define SHUTDOWN_MASK 3 | |
670 | #define RCV_SHUTDOWN 1 | |
671 | #define SEND_SHUTDOWN 2 | |
672 | ||
673 | #define SOCK_SNDBUF_LOCK 1 | |
674 | #define SOCK_RCVBUF_LOCK 2 | |
675 | #define SOCK_BINDADDR_LOCK 4 | |
676 | #define SOCK_BINDPORT_LOCK 8 | |
677 | ||
678 | /* sock_iocb: used to kick off async processing of socket ios */ | |
679 | struct sock_iocb { | |
680 | struct list_head list; | |
681 | ||
682 | int flags; | |
683 | int size; | |
684 | struct socket *sock; | |
685 | struct sock *sk; | |
686 | struct scm_cookie *scm; | |
687 | struct msghdr *msg, async_msg; | |
1da177e4 LT |
688 | struct kiocb *kiocb; |
689 | }; | |
690 | ||
691 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) | |
692 | { | |
693 | return (struct sock_iocb *)iocb->private; | |
694 | } | |
695 | ||
696 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) | |
697 | { | |
698 | return si->kiocb; | |
699 | } | |
700 | ||
701 | struct socket_alloc { | |
702 | struct socket socket; | |
703 | struct inode vfs_inode; | |
704 | }; | |
705 | ||
706 | static inline struct socket *SOCKET_I(struct inode *inode) | |
707 | { | |
708 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
709 | } | |
710 | ||
711 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
712 | { | |
713 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
714 | } | |
715 | ||
3ab224be HA |
716 | /* |
717 | * Functions for memory accounting | |
718 | */ | |
719 | extern int __sk_mem_schedule(struct sock *sk, int size, int kind); | |
720 | extern void __sk_mem_reclaim(struct sock *sk); | |
1da177e4 | 721 | |
3ab224be HA |
722 | #define SK_MEM_QUANTUM ((int)PAGE_SIZE) |
723 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) | |
724 | #define SK_MEM_SEND 0 | |
725 | #define SK_MEM_RECV 1 | |
1da177e4 | 726 | |
3ab224be | 727 | static inline int sk_mem_pages(int amt) |
1da177e4 | 728 | { |
3ab224be | 729 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
1da177e4 LT |
730 | } |
731 | ||
3ab224be | 732 | static inline int sk_has_account(struct sock *sk) |
1da177e4 | 733 | { |
3ab224be HA |
734 | /* return true if protocol supports memory accounting */ |
735 | return !!sk->sk_prot->memory_allocated; | |
1da177e4 LT |
736 | } |
737 | ||
3ab224be | 738 | static inline int sk_wmem_schedule(struct sock *sk, int size) |
1da177e4 | 739 | { |
3ab224be HA |
740 | if (!sk_has_account(sk)) |
741 | return 1; | |
742 | return size <= sk->sk_forward_alloc || | |
743 | __sk_mem_schedule(sk, size, SK_MEM_SEND); | |
1da177e4 LT |
744 | } |
745 | ||
3ab224be | 746 | static inline int sk_rmem_schedule(struct sock *sk, int size) |
d80d99d6 | 747 | { |
3ab224be HA |
748 | if (!sk_has_account(sk)) |
749 | return 1; | |
d80d99d6 | 750 | return size <= sk->sk_forward_alloc || |
3ab224be HA |
751 | __sk_mem_schedule(sk, size, SK_MEM_RECV); |
752 | } | |
753 | ||
754 | static inline void sk_mem_reclaim(struct sock *sk) | |
755 | { | |
756 | if (!sk_has_account(sk)) | |
757 | return; | |
758 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) | |
759 | __sk_mem_reclaim(sk); | |
760 | } | |
761 | ||
762 | static inline void sk_mem_charge(struct sock *sk, int size) | |
763 | { | |
764 | if (!sk_has_account(sk)) | |
765 | return; | |
766 | sk->sk_forward_alloc -= size; | |
767 | } | |
768 | ||
769 | static inline void sk_mem_uncharge(struct sock *sk, int size) | |
770 | { | |
771 | if (!sk_has_account(sk)) | |
772 | return; | |
773 | sk->sk_forward_alloc += size; | |
774 | } | |
775 | ||
776 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) | |
777 | { | |
778 | skb_truesize_check(skb); | |
779 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); | |
780 | sk->sk_wmem_queued -= skb->truesize; | |
781 | sk_mem_uncharge(sk, skb->truesize); | |
782 | __kfree_skb(skb); | |
d80d99d6 HX |
783 | } |
784 | ||
1da177e4 LT |
785 | /* Used by processes to "lock" a socket state, so that |
786 | * interrupts and bottom half handlers won't change it | |
787 | * from under us. It essentially blocks any incoming | |
788 | * packets, so that we won't get any new data or any | |
789 | * packets that change the state of the socket. | |
790 | * | |
791 | * While locked, BH processing will add new packets to | |
792 | * the backlog queue. This queue is processed by the | |
793 | * owner of the socket lock right before it is released. | |
794 | * | |
795 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
796 | * accesses from user process context. | |
797 | */ | |
d2e9117c | 798 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) |
1da177e4 | 799 | |
ed07536e PZ |
800 | /* |
801 | * Macro so as to not evaluate some arguments when | |
802 | * lockdep is not enabled. | |
803 | * | |
804 | * Mark both the sk_lock and the sk_lock.slock as a | |
805 | * per-address-family lock class. | |
806 | */ | |
807 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ | |
808 | do { \ | |
d2e9117c | 809 | sk->sk_lock.owned = 0; \ |
ed07536e PZ |
810 | init_waitqueue_head(&sk->sk_lock.wq); \ |
811 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
812 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
813 | sizeof((sk)->sk_lock)); \ | |
814 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
815 | (skey), (sname)); \ | |
816 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ | |
817 | } while (0) | |
818 | ||
41380930 | 819 | extern void lock_sock_nested(struct sock *sk, int subclass); |
fcc70d5f PZ |
820 | |
821 | static inline void lock_sock(struct sock *sk) | |
822 | { | |
823 | lock_sock_nested(sk, 0); | |
824 | } | |
825 | ||
41380930 | 826 | extern void release_sock(struct sock *sk); |
1da177e4 LT |
827 | |
828 | /* BH context may only use the following locking interface. */ | |
829 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
c6366184 IM |
830 | #define bh_lock_sock_nested(__sk) \ |
831 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
832 | SINGLE_DEPTH_NESTING) | |
1da177e4 LT |
833 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
834 | ||
1b8d7ae4 | 835 | extern struct sock *sk_alloc(struct net *net, int family, |
dd0fc66f | 836 | gfp_t priority, |
6257ff21 | 837 | struct proto *prot); |
1da177e4 | 838 | extern void sk_free(struct sock *sk); |
87d11ceb | 839 | extern struct sock *sk_clone(const struct sock *sk, |
dd0fc66f | 840 | const gfp_t priority); |
1da177e4 LT |
841 | |
842 | extern struct sk_buff *sock_wmalloc(struct sock *sk, | |
843 | unsigned long size, int force, | |
dd0fc66f | 844 | gfp_t priority); |
1da177e4 LT |
845 | extern struct sk_buff *sock_rmalloc(struct sock *sk, |
846 | unsigned long size, int force, | |
dd0fc66f | 847 | gfp_t priority); |
1da177e4 LT |
848 | extern void sock_wfree(struct sk_buff *skb); |
849 | extern void sock_rfree(struct sk_buff *skb); | |
850 | ||
851 | extern int sock_setsockopt(struct socket *sock, int level, | |
852 | int op, char __user *optval, | |
853 | int optlen); | |
854 | ||
855 | extern int sock_getsockopt(struct socket *sock, int level, | |
856 | int op, char __user *optval, | |
857 | int __user *optlen); | |
858 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, | |
859 | unsigned long size, | |
860 | int noblock, | |
861 | int *errcode); | |
86a76caf | 862 | extern void *sock_kmalloc(struct sock *sk, int size, |
dd0fc66f | 863 | gfp_t priority); |
1da177e4 LT |
864 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); |
865 | extern void sk_send_sigurg(struct sock *sk); | |
866 | ||
867 | /* | |
868 | * Functions to fill in entries in struct proto_ops when a protocol | |
869 | * does not implement a particular function. | |
870 | */ | |
871 | extern int sock_no_bind(struct socket *, | |
872 | struct sockaddr *, int); | |
873 | extern int sock_no_connect(struct socket *, | |
874 | struct sockaddr *, int, int); | |
875 | extern int sock_no_socketpair(struct socket *, | |
876 | struct socket *); | |
877 | extern int sock_no_accept(struct socket *, | |
878 | struct socket *, int); | |
879 | extern int sock_no_getname(struct socket *, | |
880 | struct sockaddr *, int *, int); | |
881 | extern unsigned int sock_no_poll(struct file *, struct socket *, | |
882 | struct poll_table_struct *); | |
883 | extern int sock_no_ioctl(struct socket *, unsigned int, | |
884 | unsigned long); | |
885 | extern int sock_no_listen(struct socket *, int); | |
886 | extern int sock_no_shutdown(struct socket *, int); | |
887 | extern int sock_no_getsockopt(struct socket *, int , int, | |
888 | char __user *, int __user *); | |
889 | extern int sock_no_setsockopt(struct socket *, int, int, | |
890 | char __user *, int); | |
891 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, | |
892 | struct msghdr *, size_t); | |
893 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, | |
894 | struct msghdr *, size_t, int); | |
895 | extern int sock_no_mmap(struct file *file, | |
896 | struct socket *sock, | |
897 | struct vm_area_struct *vma); | |
898 | extern ssize_t sock_no_sendpage(struct socket *sock, | |
899 | struct page *page, | |
900 | int offset, size_t size, | |
901 | int flags); | |
902 | ||
903 | /* | |
904 | * Functions to fill in entries in struct proto_ops when a protocol | |
905 | * uses the inet style. | |
906 | */ | |
907 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, | |
908 | char __user *optval, int __user *optlen); | |
909 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, | |
910 | struct msghdr *msg, size_t size, int flags); | |
911 | extern int sock_common_setsockopt(struct socket *sock, int level, int optname, | |
912 | char __user *optval, int optlen); | |
3fdadf7d DM |
913 | extern int compat_sock_common_getsockopt(struct socket *sock, int level, |
914 | int optname, char __user *optval, int __user *optlen); | |
915 | extern int compat_sock_common_setsockopt(struct socket *sock, int level, | |
916 | int optname, char __user *optval, int optlen); | |
1da177e4 LT |
917 | |
918 | extern void sk_common_release(struct sock *sk); | |
919 | ||
920 | /* | |
921 | * Default socket callbacks and setup code | |
922 | */ | |
923 | ||
924 | /* Initialise core socket variables */ | |
925 | extern void sock_init_data(struct socket *sock, struct sock *sk); | |
926 | ||
927 | /** | |
928 | * sk_filter - run a packet through a socket filter | |
929 | * @sk: sock associated with &sk_buff | |
930 | * @skb: buffer to filter | |
931 | * @needlock: set to 1 if the sock is not locked by caller. | |
932 | * | |
933 | * Run the filter code and then cut skb->data to correct size returned by | |
934 | * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller | |
935 | * than pkt_len we keep whole skb->data. This is the socket level | |
936 | * wrapper to sk_run_filter. It returns 0 if the packet should | |
937 | * be accepted or -EPERM if the packet should be tossed. | |
938 | * | |
939 | */ | |
940 | ||
fda9ef5d | 941 | static inline int sk_filter(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
942 | { |
943 | int err; | |
fda9ef5d | 944 | struct sk_filter *filter; |
1da177e4 LT |
945 | |
946 | err = security_sock_rcv_skb(sk, skb); | |
947 | if (err) | |
948 | return err; | |
949 | ||
fda9ef5d | 950 | rcu_read_lock_bh(); |
9d3e4442 | 951 | filter = rcu_dereference(sk->sk_filter); |
fda9ef5d DM |
952 | if (filter) { |
953 | unsigned int pkt_len = sk_run_filter(skb, filter->insns, | |
954 | filter->len); | |
955 | err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM; | |
1da177e4 | 956 | } |
fda9ef5d DM |
957 | rcu_read_unlock_bh(); |
958 | ||
1da177e4 LT |
959 | return err; |
960 | } | |
961 | ||
dc9b3346 PB |
962 | /** |
963 | * sk_filter_release: Release a socket filter | |
964 | * @sk: socket | |
965 | * @fp: filter to remove | |
966 | * | |
967 | * Remove a filter from a socket and release its resources. | |
968 | */ | |
969 | ||
309dd5fc PE |
970 | static inline void sk_filter_release(struct sk_filter *fp) |
971 | { | |
972 | if (atomic_dec_and_test(&fp->refcnt)) | |
47e958ea | 973 | kfree(fp); |
309dd5fc PE |
974 | } |
975 | ||
976 | static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp) | |
1da177e4 LT |
977 | { |
978 | unsigned int size = sk_filter_len(fp); | |
979 | ||
980 | atomic_sub(size, &sk->sk_omem_alloc); | |
309dd5fc | 981 | sk_filter_release(fp); |
1da177e4 LT |
982 | } |
983 | ||
984 | static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) | |
985 | { | |
986 | atomic_inc(&fp->refcnt); | |
987 | atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); | |
988 | } | |
989 | ||
990 | /* | |
991 | * Socket reference counting postulates. | |
992 | * | |
993 | * * Each user of socket SHOULD hold a reference count. | |
994 | * * Each access point to socket (an hash table bucket, reference from a list, | |
995 | * running timer, skb in flight MUST hold a reference count. | |
996 | * * When reference count hits 0, it means it will never increase back. | |
997 | * * When reference count hits 0, it means that no references from | |
998 | * outside exist to this socket and current process on current CPU | |
999 | * is last user and may/should destroy this socket. | |
1000 | * * sk_free is called from any context: process, BH, IRQ. When | |
1001 | * it is called, socket has no references from outside -> sk_free | |
1002 | * may release descendant resources allocated by the socket, but | |
1003 | * to the time when it is called, socket is NOT referenced by any | |
1004 | * hash tables, lists etc. | |
1005 | * * Packets, delivered from outside (from network or from another process) | |
1006 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1007 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1008 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1009 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1010 | * (leak to backlog). Packet socket does all the processing inside | |
1011 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1012 | * use separate SMP lock, so that they are prone too. | |
1013 | */ | |
1014 | ||
1015 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1016 | static inline void sock_put(struct sock *sk) | |
1017 | { | |
1018 | if (atomic_dec_and_test(&sk->sk_refcnt)) | |
1019 | sk_free(sk); | |
1020 | } | |
1021 | ||
58a5a7b9 ACM |
1022 | extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
1023 | const int nested); | |
25995ff5 | 1024 | |
1da177e4 LT |
1025 | /* Detach socket from process context. |
1026 | * Announce socket dead, detach it from wait queue and inode. | |
1027 | * Note that parent inode held reference count on this struct sock, | |
1028 | * we do not release it in this function, because protocol | |
1029 | * probably wants some additional cleanups or even continuing | |
1030 | * to work with this socket (TCP). | |
1031 | */ | |
1032 | static inline void sock_orphan(struct sock *sk) | |
1033 | { | |
1034 | write_lock_bh(&sk->sk_callback_lock); | |
1035 | sock_set_flag(sk, SOCK_DEAD); | |
1036 | sk->sk_socket = NULL; | |
1037 | sk->sk_sleep = NULL; | |
1038 | write_unlock_bh(&sk->sk_callback_lock); | |
1039 | } | |
1040 | ||
1041 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
1042 | { | |
1043 | write_lock_bh(&sk->sk_callback_lock); | |
1044 | sk->sk_sleep = &parent->wait; | |
1045 | parent->sk = sk; | |
1046 | sk->sk_socket = parent; | |
4237c75c | 1047 | security_sock_graft(sk, parent); |
1da177e4 LT |
1048 | write_unlock_bh(&sk->sk_callback_lock); |
1049 | } | |
1050 | ||
1051 | extern int sock_i_uid(struct sock *sk); | |
1052 | extern unsigned long sock_i_ino(struct sock *sk); | |
1053 | ||
1054 | static inline struct dst_entry * | |
1055 | __sk_dst_get(struct sock *sk) | |
1056 | { | |
1057 | return sk->sk_dst_cache; | |
1058 | } | |
1059 | ||
1060 | static inline struct dst_entry * | |
1061 | sk_dst_get(struct sock *sk) | |
1062 | { | |
1063 | struct dst_entry *dst; | |
1064 | ||
1065 | read_lock(&sk->sk_dst_lock); | |
1066 | dst = sk->sk_dst_cache; | |
1067 | if (dst) | |
1068 | dst_hold(dst); | |
1069 | read_unlock(&sk->sk_dst_lock); | |
1070 | return dst; | |
1071 | } | |
1072 | ||
1073 | static inline void | |
1074 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1075 | { | |
1076 | struct dst_entry *old_dst; | |
1077 | ||
1078 | old_dst = sk->sk_dst_cache; | |
1079 | sk->sk_dst_cache = dst; | |
1080 | dst_release(old_dst); | |
1081 | } | |
1082 | ||
1083 | static inline void | |
1084 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1085 | { | |
1086 | write_lock(&sk->sk_dst_lock); | |
1087 | __sk_dst_set(sk, dst); | |
1088 | write_unlock(&sk->sk_dst_lock); | |
1089 | } | |
1090 | ||
1091 | static inline void | |
1092 | __sk_dst_reset(struct sock *sk) | |
1093 | { | |
1094 | struct dst_entry *old_dst; | |
1095 | ||
1096 | old_dst = sk->sk_dst_cache; | |
1097 | sk->sk_dst_cache = NULL; | |
1098 | dst_release(old_dst); | |
1099 | } | |
1100 | ||
1101 | static inline void | |
1102 | sk_dst_reset(struct sock *sk) | |
1103 | { | |
1104 | write_lock(&sk->sk_dst_lock); | |
1105 | __sk_dst_reset(sk); | |
1106 | write_unlock(&sk->sk_dst_lock); | |
1107 | } | |
1108 | ||
f0088a50 | 1109 | extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1110 | |
f0088a50 | 1111 | extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1112 | |
bcd76111 HX |
1113 | static inline int sk_can_gso(const struct sock *sk) |
1114 | { | |
1115 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
1116 | } | |
1117 | ||
9958089a | 1118 | extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
6cbb0df7 | 1119 | |
1da177e4 LT |
1120 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, |
1121 | struct sk_buff *skb, struct page *page, | |
1122 | int off, int copy) | |
1123 | { | |
1124 | if (skb->ip_summed == CHECKSUM_NONE) { | |
1125 | int err = 0; | |
5084205f | 1126 | __wsum csum = csum_and_copy_from_user(from, |
1da177e4 LT |
1127 | page_address(page) + off, |
1128 | copy, 0, &err); | |
1129 | if (err) | |
1130 | return err; | |
1131 | skb->csum = csum_block_add(skb->csum, csum, skb->len); | |
1132 | } else if (copy_from_user(page_address(page) + off, from, copy)) | |
1133 | return -EFAULT; | |
1134 | ||
1135 | skb->len += copy; | |
1136 | skb->data_len += copy; | |
1137 | skb->truesize += copy; | |
1138 | sk->sk_wmem_queued += copy; | |
3ab224be | 1139 | sk_mem_charge(sk, copy); |
1da177e4 LT |
1140 | return 0; |
1141 | } | |
1142 | ||
1143 | /* | |
1144 | * Queue a received datagram if it will fit. Stream and sequenced | |
1145 | * protocols can't normally use this as they need to fit buffers in | |
1146 | * and play with them. | |
1147 | * | |
1148 | * Inlined as it's very short and called for pretty much every | |
1149 | * packet ever received. | |
1150 | */ | |
1151 | ||
1152 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) | |
1153 | { | |
1154 | sock_hold(sk); | |
1155 | skb->sk = sk; | |
1156 | skb->destructor = sock_wfree; | |
1157 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); | |
1158 | } | |
1159 | ||
1160 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
1161 | { | |
1162 | skb->sk = sk; | |
1163 | skb->destructor = sock_rfree; | |
1164 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
3ab224be | 1165 | sk_mem_charge(sk, skb->truesize); |
1da177e4 LT |
1166 | } |
1167 | ||
1168 | extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, | |
1169 | unsigned long expires); | |
1170 | ||
1171 | extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); | |
1172 | ||
f0088a50 | 1173 | extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 LT |
1174 | |
1175 | static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb) | |
1176 | { | |
1177 | /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces | |
1178 | number of warnings when compiling with -W --ANK | |
1179 | */ | |
1180 | if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >= | |
1181 | (unsigned)sk->sk_rcvbuf) | |
1182 | return -ENOMEM; | |
1183 | skb_set_owner_r(skb, sk); | |
1184 | skb_queue_tail(&sk->sk_error_queue, skb); | |
1185 | if (!sock_flag(sk, SOCK_DEAD)) | |
1186 | sk->sk_data_ready(sk, skb->len); | |
1187 | return 0; | |
1188 | } | |
1189 | ||
1190 | /* | |
1191 | * Recover an error report and clear atomically | |
1192 | */ | |
1193 | ||
1194 | static inline int sock_error(struct sock *sk) | |
1195 | { | |
c1cbe4b7 BL |
1196 | int err; |
1197 | if (likely(!sk->sk_err)) | |
1198 | return 0; | |
1199 | err = xchg(&sk->sk_err, 0); | |
1da177e4 LT |
1200 | return -err; |
1201 | } | |
1202 | ||
1203 | static inline unsigned long sock_wspace(struct sock *sk) | |
1204 | { | |
1205 | int amt = 0; | |
1206 | ||
1207 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
1208 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); | |
1209 | if (amt < 0) | |
1210 | amt = 0; | |
1211 | } | |
1212 | return amt; | |
1213 | } | |
1214 | ||
1215 | static inline void sk_wake_async(struct sock *sk, int how, int band) | |
1216 | { | |
1217 | if (sk->sk_socket && sk->sk_socket->fasync_list) | |
1218 | sock_wake_async(sk->sk_socket, how, band); | |
1219 | } | |
1220 | ||
1221 | #define SOCK_MIN_SNDBUF 2048 | |
1222 | #define SOCK_MIN_RCVBUF 256 | |
1223 | ||
1224 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
1225 | { | |
1226 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { | |
8df09ea3 | 1227 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); |
1da177e4 LT |
1228 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); |
1229 | } | |
1230 | } | |
1231 | ||
df97c708 | 1232 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp); |
1da177e4 LT |
1233 | |
1234 | static inline struct page *sk_stream_alloc_page(struct sock *sk) | |
1235 | { | |
1236 | struct page *page = NULL; | |
1237 | ||
ef015786 HX |
1238 | page = alloc_pages(sk->sk_allocation, 0); |
1239 | if (!page) { | |
1da177e4 LT |
1240 | sk->sk_prot->enter_memory_pressure(); |
1241 | sk_stream_moderate_sndbuf(sk); | |
1242 | } | |
1243 | return page; | |
1244 | } | |
1245 | ||
1da177e4 LT |
1246 | /* |
1247 | * Default write policy as shown to user space via poll/select/SIGIO | |
1248 | */ | |
1249 | static inline int sock_writeable(const struct sock *sk) | |
1250 | { | |
8df09ea3 | 1251 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); |
1da177e4 LT |
1252 | } |
1253 | ||
dd0fc66f | 1254 | static inline gfp_t gfp_any(void) |
1da177e4 | 1255 | { |
4498121c | 1256 | return in_atomic() ? GFP_ATOMIC : GFP_KERNEL; |
1da177e4 LT |
1257 | } |
1258 | ||
1259 | static inline long sock_rcvtimeo(const struct sock *sk, int noblock) | |
1260 | { | |
1261 | return noblock ? 0 : sk->sk_rcvtimeo; | |
1262 | } | |
1263 | ||
1264 | static inline long sock_sndtimeo(const struct sock *sk, int noblock) | |
1265 | { | |
1266 | return noblock ? 0 : sk->sk_sndtimeo; | |
1267 | } | |
1268 | ||
1269 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
1270 | { | |
1271 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; | |
1272 | } | |
1273 | ||
1274 | /* Alas, with timeout socket operations are not restartable. | |
1275 | * Compare this to poll(). | |
1276 | */ | |
1277 | static inline int sock_intr_errno(long timeo) | |
1278 | { | |
1279 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
1280 | } | |
1281 | ||
92f37fd2 ED |
1282 | extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
1283 | struct sk_buff *skb); | |
1284 | ||
1da177e4 LT |
1285 | static __inline__ void |
1286 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) | |
1287 | { | |
b7aa0bf7 | 1288 | ktime_t kt = skb->tstamp; |
a61bbcf2 | 1289 | |
92f37fd2 ED |
1290 | if (sock_flag(sk, SOCK_RCVTSTAMP)) |
1291 | __sock_recv_timestamp(msg, sk, skb); | |
1292 | else | |
b7aa0bf7 | 1293 | sk->sk_stamp = kt; |
1da177e4 LT |
1294 | } |
1295 | ||
1296 | /** | |
1297 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
1298 | * @sk: socket to eat this skb from |
1299 | * @skb: socket buffer to eat | |
f4b8ea78 | 1300 | * @copied_early: flag indicating whether DMA operations copied this data early |
1da177e4 LT |
1301 | * |
1302 | * This routine must be called with interrupts disabled or with the socket | |
1303 | * locked so that the sk_buff queue operation is ok. | |
1304 | */ | |
624d1164 CL |
1305 | #ifdef CONFIG_NET_DMA |
1306 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) | |
1307 | { | |
1308 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1309 | if (!copied_early) | |
1310 | __kfree_skb(skb); | |
1311 | else | |
1312 | __skb_queue_tail(&sk->sk_async_wait_queue, skb); | |
1313 | } | |
1314 | #else | |
1315 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) | |
1da177e4 LT |
1316 | { |
1317 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1318 | __kfree_skb(skb); | |
1319 | } | |
624d1164 | 1320 | #endif |
1da177e4 LT |
1321 | |
1322 | extern void sock_enable_timestamp(struct sock *sk); | |
1323 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); | |
ae40eb1e | 1324 | extern int sock_get_timestampns(struct sock *, struct timespec __user *); |
1da177e4 LT |
1325 | |
1326 | /* | |
1327 | * Enable debug/info messages | |
1328 | */ | |
a2a316fd SH |
1329 | extern int net_msg_warn; |
1330 | #define NETDEBUG(fmt, args...) \ | |
1331 | do { if (net_msg_warn) printk(fmt,##args); } while (0) | |
1da177e4 | 1332 | |
a2a316fd SH |
1333 | #define LIMIT_NETDEBUG(fmt, args...) \ |
1334 | do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) | |
1da177e4 LT |
1335 | |
1336 | /* | |
1337 | * Macros for sleeping on a socket. Use them like this: | |
1338 | * | |
1339 | * SOCK_SLEEP_PRE(sk) | |
1340 | * if (condition) | |
1341 | * schedule(); | |
1342 | * SOCK_SLEEP_POST(sk) | |
1343 | * | |
1344 | * N.B. These are now obsolete and were, afaik, only ever used in DECnet | |
1345 | * and when the last use of them in DECnet has gone, I'm intending to | |
1346 | * remove them. | |
1347 | */ | |
1348 | ||
1349 | #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \ | |
1350 | DECLARE_WAITQUEUE(wait, tsk); \ | |
1351 | tsk->state = TASK_INTERRUPTIBLE; \ | |
1352 | add_wait_queue((sk)->sk_sleep, &wait); \ | |
1353 | release_sock(sk); | |
1354 | ||
1355 | #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \ | |
1356 | remove_wait_queue((sk)->sk_sleep, &wait); \ | |
1357 | lock_sock(sk); \ | |
1358 | } | |
1359 | ||
1da177e4 LT |
1360 | extern __u32 sysctl_wmem_max; |
1361 | extern __u32 sysctl_rmem_max; | |
1362 | ||
20380731 ACM |
1363 | extern void sk_init(void); |
1364 | ||
6baf1f41 DM |
1365 | extern int sysctl_optmem_max; |
1366 | ||
20380731 ACM |
1367 | extern __u32 sysctl_wmem_default; |
1368 | extern __u32 sysctl_rmem_default; | |
20380731 | 1369 | |
1da177e4 | 1370 | #endif /* _SOCK_H */ |