]>
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) |
82cc1a7a | 154 | * @sk_gso_max_size: Maximum GSO segment size to build |
4dc3b16b | 155 | * @sk_lingertime: %SO_LINGER l_linger setting |
4dc3b16b PP |
156 | * @sk_backlog: always used with the per-socket spinlock held |
157 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
158 | * @sk_error_queue: rarely used | |
33c732c3 WC |
159 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
160 | * IPV6_ADDRFORM for instance) | |
4dc3b16b | 161 | * @sk_err: last error |
33c732c3 WC |
162 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
163 | * persistent failure not just 'timed out' | |
164 | * @sk_drops: raw drops counter | |
4dc3b16b PP |
165 | * @sk_ack_backlog: current listen backlog |
166 | * @sk_max_ack_backlog: listen backlog set in listen() | |
167 | * @sk_priority: %SO_PRIORITY setting | |
168 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
169 | * @sk_protocol: which protocol this socket belongs in this network family | |
170 | * @sk_peercred: %SO_PEERCRED setting | |
171 | * @sk_rcvlowat: %SO_RCVLOWAT setting | |
172 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
173 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
174 | * @sk_filter: socket filtering instructions | |
175 | * @sk_protinfo: private area, net family specific, when not using slab | |
176 | * @sk_timer: sock cleanup timer | |
177 | * @sk_stamp: time stamp of last packet received | |
178 | * @sk_socket: Identd and reporting IO signals | |
179 | * @sk_user_data: RPC layer private data | |
180 | * @sk_sndmsg_page: cached page for sendmsg | |
181 | * @sk_sndmsg_off: cached offset for sendmsg | |
182 | * @sk_send_head: front of stuff to transmit | |
67be2dd1 | 183 | * @sk_security: used by security modules |
31729363 | 184 | * @sk_mark: generic packet mark |
4dc3b16b PP |
185 | * @sk_write_pending: a write to stream socket waits to start |
186 | * @sk_state_change: callback to indicate change in the state of the sock | |
187 | * @sk_data_ready: callback to indicate there is data to be processed | |
188 | * @sk_write_space: callback to indicate there is bf sending space available | |
189 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
190 | * @sk_backlog_rcv: callback to process the backlog | |
191 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
1da177e4 LT |
192 | */ |
193 | struct sock { | |
194 | /* | |
8feaf0c0 | 195 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
196 | * don't add nothing before this first member (__sk_common) --acme |
197 | */ | |
198 | struct sock_common __sk_common; | |
199 | #define sk_family __sk_common.skc_family | |
200 | #define sk_state __sk_common.skc_state | |
201 | #define sk_reuse __sk_common.skc_reuse | |
202 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if | |
203 | #define sk_node __sk_common.skc_node | |
204 | #define sk_bind_node __sk_common.skc_bind_node | |
205 | #define sk_refcnt __sk_common.skc_refcnt | |
81c3d547 | 206 | #define sk_hash __sk_common.skc_hash |
8feaf0c0 | 207 | #define sk_prot __sk_common.skc_prot |
07feaebf | 208 | #define sk_net __sk_common.skc_net |
1da177e4 LT |
209 | unsigned char sk_shutdown : 2, |
210 | sk_no_check : 2, | |
211 | sk_userlocks : 4; | |
212 | unsigned char sk_protocol; | |
213 | unsigned short sk_type; | |
214 | int sk_rcvbuf; | |
215 | socket_lock_t sk_lock; | |
fa438ccf ED |
216 | /* |
217 | * The backlog queue is special, it is always used with | |
218 | * the per-socket spinlock held and requires low latency | |
219 | * access. Therefore we special case it's implementation. | |
220 | */ | |
221 | struct { | |
222 | struct sk_buff *head; | |
223 | struct sk_buff *tail; | |
224 | } sk_backlog; | |
1da177e4 LT |
225 | wait_queue_head_t *sk_sleep; |
226 | struct dst_entry *sk_dst_cache; | |
227 | struct xfrm_policy *sk_policy[2]; | |
228 | rwlock_t sk_dst_lock; | |
229 | atomic_t sk_rmem_alloc; | |
230 | atomic_t sk_wmem_alloc; | |
231 | atomic_t sk_omem_alloc; | |
4e07a91c | 232 | int sk_sndbuf; |
1da177e4 LT |
233 | struct sk_buff_head sk_receive_queue; |
234 | struct sk_buff_head sk_write_queue; | |
97fc2f08 | 235 | struct sk_buff_head sk_async_wait_queue; |
1da177e4 LT |
236 | int sk_wmem_queued; |
237 | int sk_forward_alloc; | |
7d877f3b | 238 | gfp_t sk_allocation; |
1da177e4 | 239 | int sk_route_caps; |
bcd76111 | 240 | int sk_gso_type; |
82cc1a7a | 241 | unsigned int sk_gso_max_size; |
9932cf95 | 242 | int sk_rcvlowat; |
1da177e4 LT |
243 | unsigned long sk_flags; |
244 | unsigned long sk_lingertime; | |
1da177e4 | 245 | struct sk_buff_head sk_error_queue; |
476e19cf | 246 | struct proto *sk_prot_creator; |
1da177e4 LT |
247 | rwlock_t sk_callback_lock; |
248 | int sk_err, | |
249 | sk_err_soft; | |
33c732c3 | 250 | atomic_t sk_drops; |
1da177e4 LT |
251 | unsigned short sk_ack_backlog; |
252 | unsigned short sk_max_ack_backlog; | |
253 | __u32 sk_priority; | |
254 | struct ucred sk_peercred; | |
1da177e4 LT |
255 | long sk_rcvtimeo; |
256 | long sk_sndtimeo; | |
257 | struct sk_filter *sk_filter; | |
258 | void *sk_protinfo; | |
259 | struct timer_list sk_timer; | |
b7aa0bf7 | 260 | ktime_t sk_stamp; |
1da177e4 LT |
261 | struct socket *sk_socket; |
262 | void *sk_user_data; | |
263 | struct page *sk_sndmsg_page; | |
264 | struct sk_buff *sk_send_head; | |
265 | __u32 sk_sndmsg_off; | |
266 | int sk_write_pending; | |
267 | void *sk_security; | |
4a19ec58 LAT |
268 | __u32 sk_mark; |
269 | /* XXX 4 bytes hole on 64 bit */ | |
1da177e4 LT |
270 | void (*sk_state_change)(struct sock *sk); |
271 | void (*sk_data_ready)(struct sock *sk, int bytes); | |
272 | void (*sk_write_space)(struct sock *sk); | |
273 | void (*sk_error_report)(struct sock *sk); | |
274 | int (*sk_backlog_rcv)(struct sock *sk, | |
275 | struct sk_buff *skb); | |
276 | void (*sk_destruct)(struct sock *sk); | |
277 | }; | |
278 | ||
279 | /* | |
280 | * Hashed lists helper routines | |
281 | */ | |
e48c414e | 282 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
283 | { |
284 | return hlist_entry(head->first, struct sock, sk_node); | |
285 | } | |
286 | ||
e48c414e | 287 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
288 | { |
289 | return hlist_empty(head) ? NULL : __sk_head(head); | |
290 | } | |
291 | ||
e48c414e | 292 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 LT |
293 | { |
294 | return sk->sk_node.next ? | |
295 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; | |
296 | } | |
297 | ||
e48c414e | 298 | static inline int sk_unhashed(const struct sock *sk) |
1da177e4 LT |
299 | { |
300 | return hlist_unhashed(&sk->sk_node); | |
301 | } | |
302 | ||
e48c414e | 303 | static inline int sk_hashed(const struct sock *sk) |
1da177e4 | 304 | { |
da753bea | 305 | return !sk_unhashed(sk); |
1da177e4 LT |
306 | } |
307 | ||
308 | static __inline__ void sk_node_init(struct hlist_node *node) | |
309 | { | |
310 | node->pprev = NULL; | |
311 | } | |
312 | ||
313 | static __inline__ void __sk_del_node(struct sock *sk) | |
314 | { | |
315 | __hlist_del(&sk->sk_node); | |
316 | } | |
317 | ||
318 | static __inline__ int __sk_del_node_init(struct sock *sk) | |
319 | { | |
320 | if (sk_hashed(sk)) { | |
321 | __sk_del_node(sk); | |
322 | sk_node_init(&sk->sk_node); | |
323 | return 1; | |
324 | } | |
325 | return 0; | |
326 | } | |
327 | ||
328 | /* Grab socket reference count. This operation is valid only | |
329 | when sk is ALREADY grabbed f.e. it is found in hash table | |
330 | or a list and the lookup is made under lock preventing hash table | |
331 | modifications. | |
332 | */ | |
333 | ||
334 | static inline void sock_hold(struct sock *sk) | |
335 | { | |
336 | atomic_inc(&sk->sk_refcnt); | |
337 | } | |
338 | ||
339 | /* Ungrab socket in the context, which assumes that socket refcnt | |
340 | cannot hit zero, f.e. it is true in context of any socketcall. | |
341 | */ | |
342 | static inline void __sock_put(struct sock *sk) | |
343 | { | |
344 | atomic_dec(&sk->sk_refcnt); | |
345 | } | |
346 | ||
347 | static __inline__ int sk_del_node_init(struct sock *sk) | |
348 | { | |
349 | int rc = __sk_del_node_init(sk); | |
350 | ||
351 | if (rc) { | |
352 | /* paranoid for a while -acme */ | |
353 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
354 | __sock_put(sk); | |
355 | } | |
356 | return rc; | |
357 | } | |
358 | ||
359 | static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) | |
360 | { | |
361 | hlist_add_head(&sk->sk_node, list); | |
362 | } | |
363 | ||
364 | static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) | |
365 | { | |
366 | sock_hold(sk); | |
367 | __sk_add_node(sk, list); | |
368 | } | |
369 | ||
370 | static __inline__ void __sk_del_bind_node(struct sock *sk) | |
371 | { | |
372 | __hlist_del(&sk->sk_bind_node); | |
373 | } | |
374 | ||
375 | static __inline__ void sk_add_bind_node(struct sock *sk, | |
376 | struct hlist_head *list) | |
377 | { | |
378 | hlist_add_head(&sk->sk_bind_node, list); | |
379 | } | |
380 | ||
381 | #define sk_for_each(__sk, node, list) \ | |
382 | hlist_for_each_entry(__sk, node, list, sk_node) | |
383 | #define sk_for_each_from(__sk, node) \ | |
384 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
385 | hlist_for_each_entry_from(__sk, node, sk_node) | |
386 | #define sk_for_each_continue(__sk, node) \ | |
387 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
388 | hlist_for_each_entry_continue(__sk, node, sk_node) | |
389 | #define sk_for_each_safe(__sk, node, tmp, list) \ | |
390 | hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) | |
391 | #define sk_for_each_bound(__sk, node, list) \ | |
392 | hlist_for_each_entry(__sk, node, list, sk_bind_node) | |
393 | ||
394 | /* Sock flags */ | |
395 | enum sock_flags { | |
396 | SOCK_DEAD, | |
397 | SOCK_DONE, | |
398 | SOCK_URGINLINE, | |
399 | SOCK_KEEPOPEN, | |
400 | SOCK_LINGER, | |
401 | SOCK_DESTROY, | |
402 | SOCK_BROADCAST, | |
403 | SOCK_TIMESTAMP, | |
404 | SOCK_ZAPPED, | |
405 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
406 | SOCK_DBG, /* %SO_DEBUG setting */ | |
407 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
92f37fd2 | 408 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
1da177e4 LT |
409 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
410 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
411 | }; | |
412 | ||
53b924b3 RB |
413 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
414 | { | |
415 | nsk->sk_flags = osk->sk_flags; | |
416 | } | |
417 | ||
1da177e4 LT |
418 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
419 | { | |
420 | __set_bit(flag, &sk->sk_flags); | |
421 | } | |
422 | ||
423 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
424 | { | |
425 | __clear_bit(flag, &sk->sk_flags); | |
426 | } | |
427 | ||
428 | static inline int sock_flag(struct sock *sk, enum sock_flags flag) | |
429 | { | |
430 | return test_bit(flag, &sk->sk_flags); | |
431 | } | |
432 | ||
433 | static inline void sk_acceptq_removed(struct sock *sk) | |
434 | { | |
435 | sk->sk_ack_backlog--; | |
436 | } | |
437 | ||
438 | static inline void sk_acceptq_added(struct sock *sk) | |
439 | { | |
440 | sk->sk_ack_backlog++; | |
441 | } | |
442 | ||
443 | static inline int sk_acceptq_is_full(struct sock *sk) | |
444 | { | |
64a14651 | 445 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
1da177e4 LT |
446 | } |
447 | ||
448 | /* | |
449 | * Compute minimal free write space needed to queue new packets. | |
450 | */ | |
451 | static inline int sk_stream_min_wspace(struct sock *sk) | |
452 | { | |
8df09ea3 | 453 | return sk->sk_wmem_queued >> 1; |
1da177e4 LT |
454 | } |
455 | ||
456 | static inline int sk_stream_wspace(struct sock *sk) | |
457 | { | |
458 | return sk->sk_sndbuf - sk->sk_wmem_queued; | |
459 | } | |
460 | ||
461 | extern void sk_stream_write_space(struct sock *sk); | |
462 | ||
463 | static inline int sk_stream_memory_free(struct sock *sk) | |
464 | { | |
465 | return sk->sk_wmem_queued < sk->sk_sndbuf; | |
466 | } | |
467 | ||
1da177e4 | 468 | /* The per-socket spinlock must be held here. */ |
9ee6b535 SH |
469 | static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
470 | { | |
471 | if (!sk->sk_backlog.tail) { | |
472 | sk->sk_backlog.head = sk->sk_backlog.tail = skb; | |
473 | } else { | |
474 | sk->sk_backlog.tail->next = skb; | |
475 | sk->sk_backlog.tail = skb; | |
476 | } | |
477 | skb->next = NULL; | |
478 | } | |
1da177e4 | 479 | |
cfcabdcc SH |
480 | #define sk_wait_event(__sk, __timeo, __condition) \ |
481 | ({ int __rc; \ | |
482 | release_sock(__sk); \ | |
483 | __rc = __condition; \ | |
484 | if (!__rc) { \ | |
485 | *(__timeo) = schedule_timeout(*(__timeo)); \ | |
486 | } \ | |
487 | lock_sock(__sk); \ | |
488 | __rc = __condition; \ | |
489 | __rc; \ | |
490 | }) | |
1da177e4 LT |
491 | |
492 | extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); | |
493 | extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
494 | extern void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
495 | extern int sk_stream_error(struct sock *sk, int flags, int err); | |
496 | extern void sk_stream_kill_queues(struct sock *sk); | |
497 | ||
498 | extern int sk_wait_data(struct sock *sk, long *timeo); | |
499 | ||
60236fdd | 500 | struct request_sock_ops; |
6d6ee43e | 501 | struct timewait_sock_ops; |
ab1e0a13 | 502 | struct inet_hashinfo; |
2e6599cb | 503 | |
1da177e4 LT |
504 | /* Networking protocol blocks we attach to sockets. |
505 | * socket layer -> transport layer interface | |
506 | * transport -> network interface is defined by struct inet_proto | |
507 | */ | |
508 | struct proto { | |
509 | void (*close)(struct sock *sk, | |
510 | long timeout); | |
511 | int (*connect)(struct sock *sk, | |
512 | struct sockaddr *uaddr, | |
513 | int addr_len); | |
514 | int (*disconnect)(struct sock *sk, int flags); | |
515 | ||
516 | struct sock * (*accept) (struct sock *sk, int flags, int *err); | |
517 | ||
518 | int (*ioctl)(struct sock *sk, int cmd, | |
519 | unsigned long arg); | |
520 | int (*init)(struct sock *sk); | |
521 | int (*destroy)(struct sock *sk); | |
522 | void (*shutdown)(struct sock *sk, int how); | |
523 | int (*setsockopt)(struct sock *sk, int level, | |
524 | int optname, char __user *optval, | |
525 | int optlen); | |
526 | int (*getsockopt)(struct sock *sk, int level, | |
527 | int optname, char __user *optval, | |
528 | int __user *option); | |
3fdadf7d DM |
529 | int (*compat_setsockopt)(struct sock *sk, |
530 | int level, | |
531 | int optname, char __user *optval, | |
532 | int optlen); | |
533 | int (*compat_getsockopt)(struct sock *sk, | |
534 | int level, | |
535 | int optname, char __user *optval, | |
536 | int __user *option); | |
1da177e4 LT |
537 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, |
538 | struct msghdr *msg, size_t len); | |
539 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, | |
540 | struct msghdr *msg, | |
541 | size_t len, int noblock, int flags, | |
542 | int *addr_len); | |
543 | int (*sendpage)(struct sock *sk, struct page *page, | |
544 | int offset, size_t size, int flags); | |
545 | int (*bind)(struct sock *sk, | |
546 | struct sockaddr *uaddr, int addr_len); | |
547 | ||
548 | int (*backlog_rcv) (struct sock *sk, | |
549 | struct sk_buff *skb); | |
550 | ||
551 | /* Keeping track of sk's, looking them up, and port selection methods. */ | |
552 | void (*hash)(struct sock *sk); | |
553 | void (*unhash)(struct sock *sk); | |
554 | int (*get_port)(struct sock *sk, unsigned short snum); | |
555 | ||
286ab3d4 | 556 | /* Keeping track of sockets in use */ |
65f76517 | 557 | #ifdef CONFIG_PROC_FS |
ebb53d75 | 558 | struct pcounter inuse; |
65f76517 | 559 | #endif |
ebb53d75 | 560 | |
1da177e4 LT |
561 | /* Memory pressure */ |
562 | void (*enter_memory_pressure)(void); | |
563 | atomic_t *memory_allocated; /* Current allocated memory. */ | |
564 | atomic_t *sockets_allocated; /* Current number of sockets. */ | |
565 | /* | |
566 | * Pressure flag: try to collapse. | |
567 | * Technical note: it is used by multiple contexts non atomically. | |
3ab224be | 568 | * All the __sk_mem_schedule() is of this nature: accounting |
1da177e4 LT |
569 | * is strict, actions are advisory and have some latency. |
570 | */ | |
571 | int *memory_pressure; | |
572 | int *sysctl_mem; | |
573 | int *sysctl_wmem; | |
574 | int *sysctl_rmem; | |
575 | int max_header; | |
576 | ||
e18b890b | 577 | struct kmem_cache *slab; |
1da177e4 LT |
578 | unsigned int obj_size; |
579 | ||
0a5578cf | 580 | atomic_t *orphan_count; |
8feaf0c0 | 581 | |
60236fdd | 582 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 583 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 584 | |
ab1e0a13 ACM |
585 | struct inet_hashinfo *hashinfo; |
586 | ||
1da177e4 LT |
587 | struct module *owner; |
588 | ||
589 | char name[32]; | |
590 | ||
591 | struct list_head node; | |
e6848976 ACM |
592 | #ifdef SOCK_REFCNT_DEBUG |
593 | atomic_t socks; | |
594 | #endif | |
1da177e4 LT |
595 | }; |
596 | ||
597 | extern int proto_register(struct proto *prot, int alloc_slab); | |
598 | extern void proto_unregister(struct proto *prot); | |
599 | ||
e6848976 ACM |
600 | #ifdef SOCK_REFCNT_DEBUG |
601 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
602 | { | |
603 | atomic_inc(&sk->sk_prot->socks); | |
604 | } | |
605 | ||
606 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
607 | { | |
608 | atomic_dec(&sk->sk_prot->socks); | |
609 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
610 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
611 | } | |
612 | ||
613 | static inline void sk_refcnt_debug_release(const struct sock *sk) | |
614 | { | |
615 | if (atomic_read(&sk->sk_refcnt) != 1) | |
616 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", | |
617 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); | |
618 | } | |
619 | #else /* SOCK_REFCNT_DEBUG */ | |
620 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
621 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
622 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
623 | #endif /* SOCK_REFCNT_DEBUG */ | |
624 | ||
65f76517 ED |
625 | |
626 | #ifdef CONFIG_PROC_FS | |
627 | # define DEFINE_PROTO_INUSE(NAME) DEFINE_PCOUNTER(NAME) | |
628 | # define REF_PROTO_INUSE(NAME) PCOUNTER_MEMBER_INITIALIZER(NAME, .inuse) | |
1da177e4 | 629 | /* Called with local bh disabled */ |
65f76517 | 630 | static inline void sock_prot_inuse_add(struct proto *prot, int inc) |
1da177e4 | 631 | { |
65f76517 | 632 | pcounter_add(&prot->inuse, inc); |
1da177e4 | 633 | } |
65f76517 | 634 | static inline int sock_prot_inuse_init(struct proto *proto) |
1da177e4 | 635 | { |
65f76517 | 636 | return pcounter_alloc(&proto->inuse); |
286ab3d4 | 637 | } |
65f76517 | 638 | static inline int sock_prot_inuse_get(struct proto *proto) |
286ab3d4 | 639 | { |
ebb53d75 | 640 | return pcounter_getval(&proto->inuse); |
1da177e4 | 641 | } |
65f76517 ED |
642 | static inline void sock_prot_inuse_free(struct proto *proto) |
643 | { | |
644 | pcounter_free(&proto->inuse); | |
645 | } | |
646 | #else | |
647 | # define DEFINE_PROTO_INUSE(NAME) | |
648 | # define REF_PROTO_INUSE(NAME) | |
649 | static void inline sock_prot_inuse_add(struct proto *prot, int inc) | |
650 | { | |
651 | } | |
652 | static int inline sock_prot_inuse_init(struct proto *proto) | |
653 | { | |
654 | return 0; | |
655 | } | |
656 | static void inline sock_prot_inuse_free(struct proto *proto) | |
657 | { | |
658 | } | |
659 | #endif | |
660 | ||
1da177e4 | 661 | |
614c6cb4 ACM |
662 | /* With per-bucket locks this operation is not-atomic, so that |
663 | * this version is not worse. | |
664 | */ | |
665 | static inline void __sk_prot_rehash(struct sock *sk) | |
666 | { | |
667 | sk->sk_prot->unhash(sk); | |
668 | sk->sk_prot->hash(sk); | |
669 | } | |
670 | ||
1da177e4 LT |
671 | /* About 10 seconds */ |
672 | #define SOCK_DESTROY_TIME (10*HZ) | |
673 | ||
674 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
675 | #define PROT_SOCK 1024 | |
676 | ||
677 | #define SHUTDOWN_MASK 3 | |
678 | #define RCV_SHUTDOWN 1 | |
679 | #define SEND_SHUTDOWN 2 | |
680 | ||
681 | #define SOCK_SNDBUF_LOCK 1 | |
682 | #define SOCK_RCVBUF_LOCK 2 | |
683 | #define SOCK_BINDADDR_LOCK 4 | |
684 | #define SOCK_BINDPORT_LOCK 8 | |
685 | ||
686 | /* sock_iocb: used to kick off async processing of socket ios */ | |
687 | struct sock_iocb { | |
688 | struct list_head list; | |
689 | ||
690 | int flags; | |
691 | int size; | |
692 | struct socket *sock; | |
693 | struct sock *sk; | |
694 | struct scm_cookie *scm; | |
695 | struct msghdr *msg, async_msg; | |
1da177e4 LT |
696 | struct kiocb *kiocb; |
697 | }; | |
698 | ||
699 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) | |
700 | { | |
701 | return (struct sock_iocb *)iocb->private; | |
702 | } | |
703 | ||
704 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) | |
705 | { | |
706 | return si->kiocb; | |
707 | } | |
708 | ||
709 | struct socket_alloc { | |
710 | struct socket socket; | |
711 | struct inode vfs_inode; | |
712 | }; | |
713 | ||
714 | static inline struct socket *SOCKET_I(struct inode *inode) | |
715 | { | |
716 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
717 | } | |
718 | ||
719 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
720 | { | |
721 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
722 | } | |
723 | ||
3ab224be HA |
724 | /* |
725 | * Functions for memory accounting | |
726 | */ | |
727 | extern int __sk_mem_schedule(struct sock *sk, int size, int kind); | |
728 | extern void __sk_mem_reclaim(struct sock *sk); | |
1da177e4 | 729 | |
3ab224be HA |
730 | #define SK_MEM_QUANTUM ((int)PAGE_SIZE) |
731 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) | |
732 | #define SK_MEM_SEND 0 | |
733 | #define SK_MEM_RECV 1 | |
1da177e4 | 734 | |
3ab224be | 735 | static inline int sk_mem_pages(int amt) |
1da177e4 | 736 | { |
3ab224be | 737 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
1da177e4 LT |
738 | } |
739 | ||
3ab224be | 740 | static inline int sk_has_account(struct sock *sk) |
1da177e4 | 741 | { |
3ab224be HA |
742 | /* return true if protocol supports memory accounting */ |
743 | return !!sk->sk_prot->memory_allocated; | |
1da177e4 LT |
744 | } |
745 | ||
3ab224be | 746 | static inline int sk_wmem_schedule(struct sock *sk, int size) |
1da177e4 | 747 | { |
3ab224be HA |
748 | if (!sk_has_account(sk)) |
749 | return 1; | |
750 | return size <= sk->sk_forward_alloc || | |
751 | __sk_mem_schedule(sk, size, SK_MEM_SEND); | |
1da177e4 LT |
752 | } |
753 | ||
3ab224be | 754 | static inline int sk_rmem_schedule(struct sock *sk, int size) |
d80d99d6 | 755 | { |
3ab224be HA |
756 | if (!sk_has_account(sk)) |
757 | return 1; | |
d80d99d6 | 758 | return size <= sk->sk_forward_alloc || |
3ab224be HA |
759 | __sk_mem_schedule(sk, size, SK_MEM_RECV); |
760 | } | |
761 | ||
762 | static inline void sk_mem_reclaim(struct sock *sk) | |
763 | { | |
764 | if (!sk_has_account(sk)) | |
765 | return; | |
766 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) | |
767 | __sk_mem_reclaim(sk); | |
768 | } | |
769 | ||
9993e7d3 DM |
770 | static inline void sk_mem_reclaim_partial(struct sock *sk) |
771 | { | |
772 | if (!sk_has_account(sk)) | |
773 | return; | |
774 | if (sk->sk_forward_alloc > SK_MEM_QUANTUM) | |
775 | __sk_mem_reclaim(sk); | |
776 | } | |
777 | ||
3ab224be HA |
778 | static inline void sk_mem_charge(struct sock *sk, int size) |
779 | { | |
780 | if (!sk_has_account(sk)) | |
781 | return; | |
782 | sk->sk_forward_alloc -= size; | |
783 | } | |
784 | ||
785 | static inline void sk_mem_uncharge(struct sock *sk, int size) | |
786 | { | |
787 | if (!sk_has_account(sk)) | |
788 | return; | |
789 | sk->sk_forward_alloc += size; | |
790 | } | |
791 | ||
792 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) | |
793 | { | |
794 | skb_truesize_check(skb); | |
795 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); | |
796 | sk->sk_wmem_queued -= skb->truesize; | |
797 | sk_mem_uncharge(sk, skb->truesize); | |
798 | __kfree_skb(skb); | |
d80d99d6 HX |
799 | } |
800 | ||
1da177e4 LT |
801 | /* Used by processes to "lock" a socket state, so that |
802 | * interrupts and bottom half handlers won't change it | |
803 | * from under us. It essentially blocks any incoming | |
804 | * packets, so that we won't get any new data or any | |
805 | * packets that change the state of the socket. | |
806 | * | |
807 | * While locked, BH processing will add new packets to | |
808 | * the backlog queue. This queue is processed by the | |
809 | * owner of the socket lock right before it is released. | |
810 | * | |
811 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
812 | * accesses from user process context. | |
813 | */ | |
d2e9117c | 814 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) |
1da177e4 | 815 | |
ed07536e PZ |
816 | /* |
817 | * Macro so as to not evaluate some arguments when | |
818 | * lockdep is not enabled. | |
819 | * | |
820 | * Mark both the sk_lock and the sk_lock.slock as a | |
821 | * per-address-family lock class. | |
822 | */ | |
823 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ | |
824 | do { \ | |
d2e9117c | 825 | sk->sk_lock.owned = 0; \ |
ed07536e PZ |
826 | init_waitqueue_head(&sk->sk_lock.wq); \ |
827 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
828 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
829 | sizeof((sk)->sk_lock)); \ | |
830 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
831 | (skey), (sname)); \ | |
832 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ | |
833 | } while (0) | |
834 | ||
41380930 | 835 | extern void lock_sock_nested(struct sock *sk, int subclass); |
fcc70d5f PZ |
836 | |
837 | static inline void lock_sock(struct sock *sk) | |
838 | { | |
839 | lock_sock_nested(sk, 0); | |
840 | } | |
841 | ||
41380930 | 842 | extern void release_sock(struct sock *sk); |
1da177e4 LT |
843 | |
844 | /* BH context may only use the following locking interface. */ | |
845 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
c6366184 IM |
846 | #define bh_lock_sock_nested(__sk) \ |
847 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
848 | SINGLE_DEPTH_NESTING) | |
1da177e4 LT |
849 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
850 | ||
1b8d7ae4 | 851 | extern struct sock *sk_alloc(struct net *net, int family, |
dd0fc66f | 852 | gfp_t priority, |
6257ff21 | 853 | struct proto *prot); |
1da177e4 | 854 | extern void sk_free(struct sock *sk); |
edf02087 | 855 | extern void sk_release_kernel(struct sock *sk); |
87d11ceb | 856 | extern struct sock *sk_clone(const struct sock *sk, |
dd0fc66f | 857 | const gfp_t priority); |
1da177e4 LT |
858 | |
859 | extern struct sk_buff *sock_wmalloc(struct sock *sk, | |
860 | unsigned long size, int force, | |
dd0fc66f | 861 | gfp_t priority); |
1da177e4 LT |
862 | extern struct sk_buff *sock_rmalloc(struct sock *sk, |
863 | unsigned long size, int force, | |
dd0fc66f | 864 | gfp_t priority); |
1da177e4 LT |
865 | extern void sock_wfree(struct sk_buff *skb); |
866 | extern void sock_rfree(struct sk_buff *skb); | |
867 | ||
868 | extern int sock_setsockopt(struct socket *sock, int level, | |
869 | int op, char __user *optval, | |
870 | int optlen); | |
871 | ||
872 | extern int sock_getsockopt(struct socket *sock, int level, | |
873 | int op, char __user *optval, | |
874 | int __user *optlen); | |
875 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, | |
876 | unsigned long size, | |
877 | int noblock, | |
878 | int *errcode); | |
86a76caf | 879 | extern void *sock_kmalloc(struct sock *sk, int size, |
dd0fc66f | 880 | gfp_t priority); |
1da177e4 LT |
881 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); |
882 | extern void sk_send_sigurg(struct sock *sk); | |
883 | ||
884 | /* | |
885 | * Functions to fill in entries in struct proto_ops when a protocol | |
886 | * does not implement a particular function. | |
887 | */ | |
888 | extern int sock_no_bind(struct socket *, | |
889 | struct sockaddr *, int); | |
890 | extern int sock_no_connect(struct socket *, | |
891 | struct sockaddr *, int, int); | |
892 | extern int sock_no_socketpair(struct socket *, | |
893 | struct socket *); | |
894 | extern int sock_no_accept(struct socket *, | |
895 | struct socket *, int); | |
896 | extern int sock_no_getname(struct socket *, | |
897 | struct sockaddr *, int *, int); | |
898 | extern unsigned int sock_no_poll(struct file *, struct socket *, | |
899 | struct poll_table_struct *); | |
900 | extern int sock_no_ioctl(struct socket *, unsigned int, | |
901 | unsigned long); | |
902 | extern int sock_no_listen(struct socket *, int); | |
903 | extern int sock_no_shutdown(struct socket *, int); | |
904 | extern int sock_no_getsockopt(struct socket *, int , int, | |
905 | char __user *, int __user *); | |
906 | extern int sock_no_setsockopt(struct socket *, int, int, | |
907 | char __user *, int); | |
908 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, | |
909 | struct msghdr *, size_t); | |
910 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, | |
911 | struct msghdr *, size_t, int); | |
912 | extern int sock_no_mmap(struct file *file, | |
913 | struct socket *sock, | |
914 | struct vm_area_struct *vma); | |
915 | extern ssize_t sock_no_sendpage(struct socket *sock, | |
916 | struct page *page, | |
917 | int offset, size_t size, | |
918 | int flags); | |
919 | ||
920 | /* | |
921 | * Functions to fill in entries in struct proto_ops when a protocol | |
922 | * uses the inet style. | |
923 | */ | |
924 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, | |
925 | char __user *optval, int __user *optlen); | |
926 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, | |
927 | struct msghdr *msg, size_t size, int flags); | |
928 | extern int sock_common_setsockopt(struct socket *sock, int level, int optname, | |
929 | char __user *optval, int optlen); | |
3fdadf7d DM |
930 | extern int compat_sock_common_getsockopt(struct socket *sock, int level, |
931 | int optname, char __user *optval, int __user *optlen); | |
932 | extern int compat_sock_common_setsockopt(struct socket *sock, int level, | |
933 | int optname, char __user *optval, int optlen); | |
1da177e4 LT |
934 | |
935 | extern void sk_common_release(struct sock *sk); | |
936 | ||
937 | /* | |
938 | * Default socket callbacks and setup code | |
939 | */ | |
940 | ||
941 | /* Initialise core socket variables */ | |
942 | extern void sock_init_data(struct socket *sock, struct sock *sk); | |
943 | ||
944 | /** | |
945 | * sk_filter - run a packet through a socket filter | |
946 | * @sk: sock associated with &sk_buff | |
947 | * @skb: buffer to filter | |
948 | * @needlock: set to 1 if the sock is not locked by caller. | |
949 | * | |
950 | * Run the filter code and then cut skb->data to correct size returned by | |
951 | * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller | |
952 | * than pkt_len we keep whole skb->data. This is the socket level | |
953 | * wrapper to sk_run_filter. It returns 0 if the packet should | |
954 | * be accepted or -EPERM if the packet should be tossed. | |
955 | * | |
956 | */ | |
957 | ||
fda9ef5d | 958 | static inline int sk_filter(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
959 | { |
960 | int err; | |
fda9ef5d | 961 | struct sk_filter *filter; |
1da177e4 LT |
962 | |
963 | err = security_sock_rcv_skb(sk, skb); | |
964 | if (err) | |
965 | return err; | |
966 | ||
fda9ef5d | 967 | rcu_read_lock_bh(); |
9d3e4442 | 968 | filter = rcu_dereference(sk->sk_filter); |
fda9ef5d DM |
969 | if (filter) { |
970 | unsigned int pkt_len = sk_run_filter(skb, filter->insns, | |
971 | filter->len); | |
972 | err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM; | |
1da177e4 | 973 | } |
fda9ef5d DM |
974 | rcu_read_unlock_bh(); |
975 | ||
1da177e4 LT |
976 | return err; |
977 | } | |
978 | ||
dc9b3346 PB |
979 | /** |
980 | * sk_filter_release: Release a socket filter | |
981 | * @sk: socket | |
982 | * @fp: filter to remove | |
983 | * | |
984 | * Remove a filter from a socket and release its resources. | |
985 | */ | |
986 | ||
309dd5fc PE |
987 | static inline void sk_filter_release(struct sk_filter *fp) |
988 | { | |
989 | if (atomic_dec_and_test(&fp->refcnt)) | |
47e958ea | 990 | kfree(fp); |
309dd5fc PE |
991 | } |
992 | ||
993 | static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp) | |
1da177e4 LT |
994 | { |
995 | unsigned int size = sk_filter_len(fp); | |
996 | ||
997 | atomic_sub(size, &sk->sk_omem_alloc); | |
309dd5fc | 998 | sk_filter_release(fp); |
1da177e4 LT |
999 | } |
1000 | ||
1001 | static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) | |
1002 | { | |
1003 | atomic_inc(&fp->refcnt); | |
1004 | atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); | |
1005 | } | |
1006 | ||
1007 | /* | |
1008 | * Socket reference counting postulates. | |
1009 | * | |
1010 | * * Each user of socket SHOULD hold a reference count. | |
1011 | * * Each access point to socket (an hash table bucket, reference from a list, | |
1012 | * running timer, skb in flight MUST hold a reference count. | |
1013 | * * When reference count hits 0, it means it will never increase back. | |
1014 | * * When reference count hits 0, it means that no references from | |
1015 | * outside exist to this socket and current process on current CPU | |
1016 | * is last user and may/should destroy this socket. | |
1017 | * * sk_free is called from any context: process, BH, IRQ. When | |
1018 | * it is called, socket has no references from outside -> sk_free | |
1019 | * may release descendant resources allocated by the socket, but | |
1020 | * to the time when it is called, socket is NOT referenced by any | |
1021 | * hash tables, lists etc. | |
1022 | * * Packets, delivered from outside (from network or from another process) | |
1023 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1024 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1025 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1026 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1027 | * (leak to backlog). Packet socket does all the processing inside | |
1028 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1029 | * use separate SMP lock, so that they are prone too. | |
1030 | */ | |
1031 | ||
1032 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1033 | static inline void sock_put(struct sock *sk) | |
1034 | { | |
1035 | if (atomic_dec_and_test(&sk->sk_refcnt)) | |
1036 | sk_free(sk); | |
1037 | } | |
1038 | ||
58a5a7b9 ACM |
1039 | extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
1040 | const int nested); | |
25995ff5 | 1041 | |
1da177e4 LT |
1042 | /* Detach socket from process context. |
1043 | * Announce socket dead, detach it from wait queue and inode. | |
1044 | * Note that parent inode held reference count on this struct sock, | |
1045 | * we do not release it in this function, because protocol | |
1046 | * probably wants some additional cleanups or even continuing | |
1047 | * to work with this socket (TCP). | |
1048 | */ | |
1049 | static inline void sock_orphan(struct sock *sk) | |
1050 | { | |
1051 | write_lock_bh(&sk->sk_callback_lock); | |
1052 | sock_set_flag(sk, SOCK_DEAD); | |
1053 | sk->sk_socket = NULL; | |
1054 | sk->sk_sleep = NULL; | |
1055 | write_unlock_bh(&sk->sk_callback_lock); | |
1056 | } | |
1057 | ||
1058 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
1059 | { | |
1060 | write_lock_bh(&sk->sk_callback_lock); | |
1061 | sk->sk_sleep = &parent->wait; | |
1062 | parent->sk = sk; | |
1063 | sk->sk_socket = parent; | |
4237c75c | 1064 | security_sock_graft(sk, parent); |
1da177e4 LT |
1065 | write_unlock_bh(&sk->sk_callback_lock); |
1066 | } | |
1067 | ||
1068 | extern int sock_i_uid(struct sock *sk); | |
1069 | extern unsigned long sock_i_ino(struct sock *sk); | |
1070 | ||
1071 | static inline struct dst_entry * | |
1072 | __sk_dst_get(struct sock *sk) | |
1073 | { | |
1074 | return sk->sk_dst_cache; | |
1075 | } | |
1076 | ||
1077 | static inline struct dst_entry * | |
1078 | sk_dst_get(struct sock *sk) | |
1079 | { | |
1080 | struct dst_entry *dst; | |
1081 | ||
1082 | read_lock(&sk->sk_dst_lock); | |
1083 | dst = sk->sk_dst_cache; | |
1084 | if (dst) | |
1085 | dst_hold(dst); | |
1086 | read_unlock(&sk->sk_dst_lock); | |
1087 | return dst; | |
1088 | } | |
1089 | ||
1090 | static inline void | |
1091 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1092 | { | |
1093 | struct dst_entry *old_dst; | |
1094 | ||
1095 | old_dst = sk->sk_dst_cache; | |
1096 | sk->sk_dst_cache = dst; | |
1097 | dst_release(old_dst); | |
1098 | } | |
1099 | ||
1100 | static inline void | |
1101 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1102 | { | |
1103 | write_lock(&sk->sk_dst_lock); | |
1104 | __sk_dst_set(sk, dst); | |
1105 | write_unlock(&sk->sk_dst_lock); | |
1106 | } | |
1107 | ||
1108 | static inline void | |
1109 | __sk_dst_reset(struct sock *sk) | |
1110 | { | |
1111 | struct dst_entry *old_dst; | |
1112 | ||
1113 | old_dst = sk->sk_dst_cache; | |
1114 | sk->sk_dst_cache = NULL; | |
1115 | dst_release(old_dst); | |
1116 | } | |
1117 | ||
1118 | static inline void | |
1119 | sk_dst_reset(struct sock *sk) | |
1120 | { | |
1121 | write_lock(&sk->sk_dst_lock); | |
1122 | __sk_dst_reset(sk); | |
1123 | write_unlock(&sk->sk_dst_lock); | |
1124 | } | |
1125 | ||
f0088a50 | 1126 | extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1127 | |
f0088a50 | 1128 | extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1129 | |
bcd76111 HX |
1130 | static inline int sk_can_gso(const struct sock *sk) |
1131 | { | |
1132 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
1133 | } | |
1134 | ||
9958089a | 1135 | extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
6cbb0df7 | 1136 | |
1da177e4 LT |
1137 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, |
1138 | struct sk_buff *skb, struct page *page, | |
1139 | int off, int copy) | |
1140 | { | |
1141 | if (skb->ip_summed == CHECKSUM_NONE) { | |
1142 | int err = 0; | |
5084205f | 1143 | __wsum csum = csum_and_copy_from_user(from, |
1da177e4 LT |
1144 | page_address(page) + off, |
1145 | copy, 0, &err); | |
1146 | if (err) | |
1147 | return err; | |
1148 | skb->csum = csum_block_add(skb->csum, csum, skb->len); | |
1149 | } else if (copy_from_user(page_address(page) + off, from, copy)) | |
1150 | return -EFAULT; | |
1151 | ||
1152 | skb->len += copy; | |
1153 | skb->data_len += copy; | |
1154 | skb->truesize += copy; | |
1155 | sk->sk_wmem_queued += copy; | |
3ab224be | 1156 | sk_mem_charge(sk, copy); |
1da177e4 LT |
1157 | return 0; |
1158 | } | |
1159 | ||
1160 | /* | |
1161 | * Queue a received datagram if it will fit. Stream and sequenced | |
1162 | * protocols can't normally use this as they need to fit buffers in | |
1163 | * and play with them. | |
1164 | * | |
1165 | * Inlined as it's very short and called for pretty much every | |
1166 | * packet ever received. | |
1167 | */ | |
1168 | ||
1169 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) | |
1170 | { | |
1171 | sock_hold(sk); | |
1172 | skb->sk = sk; | |
1173 | skb->destructor = sock_wfree; | |
1174 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); | |
1175 | } | |
1176 | ||
1177 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
1178 | { | |
1179 | skb->sk = sk; | |
1180 | skb->destructor = sock_rfree; | |
1181 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
3ab224be | 1182 | sk_mem_charge(sk, skb->truesize); |
1da177e4 LT |
1183 | } |
1184 | ||
1185 | extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, | |
1186 | unsigned long expires); | |
1187 | ||
1188 | extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); | |
1189 | ||
f0088a50 | 1190 | extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 LT |
1191 | |
1192 | static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb) | |
1193 | { | |
1194 | /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces | |
1195 | number of warnings when compiling with -W --ANK | |
1196 | */ | |
1197 | if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >= | |
1198 | (unsigned)sk->sk_rcvbuf) | |
1199 | return -ENOMEM; | |
1200 | skb_set_owner_r(skb, sk); | |
1201 | skb_queue_tail(&sk->sk_error_queue, skb); | |
1202 | if (!sock_flag(sk, SOCK_DEAD)) | |
1203 | sk->sk_data_ready(sk, skb->len); | |
1204 | return 0; | |
1205 | } | |
1206 | ||
1207 | /* | |
1208 | * Recover an error report and clear atomically | |
1209 | */ | |
1210 | ||
1211 | static inline int sock_error(struct sock *sk) | |
1212 | { | |
c1cbe4b7 BL |
1213 | int err; |
1214 | if (likely(!sk->sk_err)) | |
1215 | return 0; | |
1216 | err = xchg(&sk->sk_err, 0); | |
1da177e4 LT |
1217 | return -err; |
1218 | } | |
1219 | ||
1220 | static inline unsigned long sock_wspace(struct sock *sk) | |
1221 | { | |
1222 | int amt = 0; | |
1223 | ||
1224 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
1225 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); | |
1226 | if (amt < 0) | |
1227 | amt = 0; | |
1228 | } | |
1229 | return amt; | |
1230 | } | |
1231 | ||
1232 | static inline void sk_wake_async(struct sock *sk, int how, int band) | |
1233 | { | |
1234 | if (sk->sk_socket && sk->sk_socket->fasync_list) | |
1235 | sock_wake_async(sk->sk_socket, how, band); | |
1236 | } | |
1237 | ||
1238 | #define SOCK_MIN_SNDBUF 2048 | |
1239 | #define SOCK_MIN_RCVBUF 256 | |
1240 | ||
1241 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
1242 | { | |
1243 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { | |
8df09ea3 | 1244 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); |
1da177e4 LT |
1245 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); |
1246 | } | |
1247 | } | |
1248 | ||
df97c708 | 1249 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp); |
1da177e4 LT |
1250 | |
1251 | static inline struct page *sk_stream_alloc_page(struct sock *sk) | |
1252 | { | |
1253 | struct page *page = NULL; | |
1254 | ||
ef015786 HX |
1255 | page = alloc_pages(sk->sk_allocation, 0); |
1256 | if (!page) { | |
1da177e4 LT |
1257 | sk->sk_prot->enter_memory_pressure(); |
1258 | sk_stream_moderate_sndbuf(sk); | |
1259 | } | |
1260 | return page; | |
1261 | } | |
1262 | ||
1da177e4 LT |
1263 | /* |
1264 | * Default write policy as shown to user space via poll/select/SIGIO | |
1265 | */ | |
1266 | static inline int sock_writeable(const struct sock *sk) | |
1267 | { | |
8df09ea3 | 1268 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); |
1da177e4 LT |
1269 | } |
1270 | ||
dd0fc66f | 1271 | static inline gfp_t gfp_any(void) |
1da177e4 | 1272 | { |
4498121c | 1273 | return in_atomic() ? GFP_ATOMIC : GFP_KERNEL; |
1da177e4 LT |
1274 | } |
1275 | ||
1276 | static inline long sock_rcvtimeo(const struct sock *sk, int noblock) | |
1277 | { | |
1278 | return noblock ? 0 : sk->sk_rcvtimeo; | |
1279 | } | |
1280 | ||
1281 | static inline long sock_sndtimeo(const struct sock *sk, int noblock) | |
1282 | { | |
1283 | return noblock ? 0 : sk->sk_sndtimeo; | |
1284 | } | |
1285 | ||
1286 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
1287 | { | |
1288 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; | |
1289 | } | |
1290 | ||
1291 | /* Alas, with timeout socket operations are not restartable. | |
1292 | * Compare this to poll(). | |
1293 | */ | |
1294 | static inline int sock_intr_errno(long timeo) | |
1295 | { | |
1296 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
1297 | } | |
1298 | ||
92f37fd2 ED |
1299 | extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
1300 | struct sk_buff *skb); | |
1301 | ||
1da177e4 LT |
1302 | static __inline__ void |
1303 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) | |
1304 | { | |
b7aa0bf7 | 1305 | ktime_t kt = skb->tstamp; |
a61bbcf2 | 1306 | |
92f37fd2 ED |
1307 | if (sock_flag(sk, SOCK_RCVTSTAMP)) |
1308 | __sock_recv_timestamp(msg, sk, skb); | |
1309 | else | |
b7aa0bf7 | 1310 | sk->sk_stamp = kt; |
1da177e4 LT |
1311 | } |
1312 | ||
1313 | /** | |
1314 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
1315 | * @sk: socket to eat this skb from |
1316 | * @skb: socket buffer to eat | |
f4b8ea78 | 1317 | * @copied_early: flag indicating whether DMA operations copied this data early |
1da177e4 LT |
1318 | * |
1319 | * This routine must be called with interrupts disabled or with the socket | |
1320 | * locked so that the sk_buff queue operation is ok. | |
1321 | */ | |
624d1164 CL |
1322 | #ifdef CONFIG_NET_DMA |
1323 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) | |
1324 | { | |
1325 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1326 | if (!copied_early) | |
1327 | __kfree_skb(skb); | |
1328 | else | |
1329 | __skb_queue_tail(&sk->sk_async_wait_queue, skb); | |
1330 | } | |
1331 | #else | |
1332 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) | |
1da177e4 LT |
1333 | { |
1334 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1335 | __kfree_skb(skb); | |
1336 | } | |
624d1164 | 1337 | #endif |
1da177e4 | 1338 | |
edf02087 DL |
1339 | /* |
1340 | * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace. | |
1341 | * They should not hold a referrence to a namespace in order to allow | |
1342 | * to stop it. | |
1343 | * Sockets after sk_change_net should be released using sk_release_kernel | |
1344 | */ | |
1345 | static inline void sk_change_net(struct sock *sk, struct net *net) | |
1346 | { | |
1347 | put_net(sk->sk_net); | |
1348 | sk->sk_net = net; | |
1349 | } | |
1350 | ||
1da177e4 LT |
1351 | extern void sock_enable_timestamp(struct sock *sk); |
1352 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); | |
ae40eb1e | 1353 | extern int sock_get_timestampns(struct sock *, struct timespec __user *); |
1da177e4 LT |
1354 | |
1355 | /* | |
1356 | * Enable debug/info messages | |
1357 | */ | |
a2a316fd SH |
1358 | extern int net_msg_warn; |
1359 | #define NETDEBUG(fmt, args...) \ | |
1360 | do { if (net_msg_warn) printk(fmt,##args); } while (0) | |
1da177e4 | 1361 | |
a2a316fd SH |
1362 | #define LIMIT_NETDEBUG(fmt, args...) \ |
1363 | do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) | |
1da177e4 LT |
1364 | |
1365 | /* | |
1366 | * Macros for sleeping on a socket. Use them like this: | |
1367 | * | |
1368 | * SOCK_SLEEP_PRE(sk) | |
1369 | * if (condition) | |
1370 | * schedule(); | |
1371 | * SOCK_SLEEP_POST(sk) | |
1372 | * | |
1373 | * N.B. These are now obsolete and were, afaik, only ever used in DECnet | |
1374 | * and when the last use of them in DECnet has gone, I'm intending to | |
1375 | * remove them. | |
1376 | */ | |
1377 | ||
1378 | #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \ | |
1379 | DECLARE_WAITQUEUE(wait, tsk); \ | |
1380 | tsk->state = TASK_INTERRUPTIBLE; \ | |
1381 | add_wait_queue((sk)->sk_sleep, &wait); \ | |
1382 | release_sock(sk); | |
1383 | ||
1384 | #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \ | |
1385 | remove_wait_queue((sk)->sk_sleep, &wait); \ | |
1386 | lock_sock(sk); \ | |
1387 | } | |
1388 | ||
1da177e4 LT |
1389 | extern __u32 sysctl_wmem_max; |
1390 | extern __u32 sysctl_rmem_max; | |
1391 | ||
20380731 ACM |
1392 | extern void sk_init(void); |
1393 | ||
6baf1f41 DM |
1394 | extern int sysctl_optmem_max; |
1395 | ||
20380731 ACM |
1396 | extern __u32 sysctl_wmem_default; |
1397 | extern __u32 sysctl_rmem_default; | |
20380731 | 1398 | |
1da177e4 | 1399 | #endif /* _SOCK_H */ |