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 LT |
49 | #include <linux/netdevice.h> |
50 | #include <linux/skbuff.h> /* struct sk_buff */ | |
d7fe0f24 | 51 | #include <linux/mm.h> |
1da177e4 LT |
52 | #include <linux/security.h> |
53 | ||
54 | #include <linux/filter.h> | |
55 | ||
56 | #include <asm/atomic.h> | |
57 | #include <net/dst.h> | |
58 | #include <net/checksum.h> | |
1b8d7ae4 | 59 | #include <net/net_namespace.h> |
1da177e4 LT |
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; |
1da177e4 LT |
97 | |
98 | /** | |
4dc3b16b PP |
99 | * struct sock_common - minimal network layer representation of sockets |
100 | * @skc_family: network address family | |
101 | * @skc_state: Connection state | |
102 | * @skc_reuse: %SO_REUSEADDR setting | |
103 | * @skc_bound_dev_if: bound device index if != 0 | |
104 | * @skc_node: main hash linkage for various protocol lookup tables | |
105 | * @skc_bind_node: bind hash linkage for various protocol lookup tables | |
106 | * @skc_refcnt: reference count | |
81c3d547 | 107 | * @skc_hash: hash value used with various protocol lookup tables |
8feaf0c0 | 108 | * @skc_prot: protocol handlers inside a network family |
07feaebf | 109 | * @skc_net: reference to the network namespace of this socket |
4dc3b16b PP |
110 | * |
111 | * This is the minimal network layer representation of sockets, the header | |
8feaf0c0 ACM |
112 | * for struct sock and struct inet_timewait_sock. |
113 | */ | |
1da177e4 LT |
114 | struct sock_common { |
115 | unsigned short skc_family; | |
116 | volatile unsigned char skc_state; | |
117 | unsigned char skc_reuse; | |
118 | int skc_bound_dev_if; | |
119 | struct hlist_node skc_node; | |
120 | struct hlist_node skc_bind_node; | |
121 | atomic_t skc_refcnt; | |
81c3d547 | 122 | unsigned int skc_hash; |
8feaf0c0 | 123 | struct proto *skc_prot; |
07feaebf | 124 | struct net *skc_net; |
1da177e4 LT |
125 | }; |
126 | ||
127 | /** | |
128 | * struct sock - network layer representation of sockets | |
8feaf0c0 | 129 | * @__sk_common: shared layout with inet_timewait_sock |
4dc3b16b PP |
130 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
131 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
132 | * @sk_lock: synchronizer | |
133 | * @sk_rcvbuf: size of receive buffer in bytes | |
134 | * @sk_sleep: sock wait queue | |
135 | * @sk_dst_cache: destination cache | |
136 | * @sk_dst_lock: destination cache lock | |
137 | * @sk_policy: flow policy | |
138 | * @sk_rmem_alloc: receive queue bytes committed | |
139 | * @sk_receive_queue: incoming packets | |
140 | * @sk_wmem_alloc: transmit queue bytes committed | |
141 | * @sk_write_queue: Packet sending queue | |
97fc2f08 | 142 | * @sk_async_wait_queue: DMA copied packets |
4dc3b16b PP |
143 | * @sk_omem_alloc: "o" is "option" or "other" |
144 | * @sk_wmem_queued: persistent queue size | |
145 | * @sk_forward_alloc: space allocated forward | |
146 | * @sk_allocation: allocation mode | |
147 | * @sk_sndbuf: size of send buffer in bytes | |
33c732c3 WC |
148 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, |
149 | * %SO_OOBINLINE settings | |
4dc3b16b PP |
150 | * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets |
151 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) | |
bcd76111 | 152 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
4dc3b16b | 153 | * @sk_lingertime: %SO_LINGER l_linger setting |
4dc3b16b PP |
154 | * @sk_backlog: always used with the per-socket spinlock held |
155 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
156 | * @sk_error_queue: rarely used | |
33c732c3 WC |
157 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
158 | * IPV6_ADDRFORM for instance) | |
4dc3b16b | 159 | * @sk_err: last error |
33c732c3 WC |
160 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
161 | * persistent failure not just 'timed out' | |
162 | * @sk_drops: raw drops counter | |
4dc3b16b PP |
163 | * @sk_ack_backlog: current listen backlog |
164 | * @sk_max_ack_backlog: listen backlog set in listen() | |
165 | * @sk_priority: %SO_PRIORITY setting | |
166 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
167 | * @sk_protocol: which protocol this socket belongs in this network family | |
168 | * @sk_peercred: %SO_PEERCRED setting | |
169 | * @sk_rcvlowat: %SO_RCVLOWAT setting | |
170 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
171 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
172 | * @sk_filter: socket filtering instructions | |
173 | * @sk_protinfo: private area, net family specific, when not using slab | |
174 | * @sk_timer: sock cleanup timer | |
175 | * @sk_stamp: time stamp of last packet received | |
176 | * @sk_socket: Identd and reporting IO signals | |
177 | * @sk_user_data: RPC layer private data | |
178 | * @sk_sndmsg_page: cached page for sendmsg | |
179 | * @sk_sndmsg_off: cached offset for sendmsg | |
180 | * @sk_send_head: front of stuff to transmit | |
67be2dd1 | 181 | * @sk_security: used by security modules |
4dc3b16b PP |
182 | * @sk_write_pending: a write to stream socket waits to start |
183 | * @sk_state_change: callback to indicate change in the state of the sock | |
184 | * @sk_data_ready: callback to indicate there is data to be processed | |
185 | * @sk_write_space: callback to indicate there is bf sending space available | |
186 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
187 | * @sk_backlog_rcv: callback to process the backlog | |
188 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
1da177e4 LT |
189 | */ |
190 | struct sock { | |
191 | /* | |
8feaf0c0 | 192 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
193 | * don't add nothing before this first member (__sk_common) --acme |
194 | */ | |
195 | struct sock_common __sk_common; | |
196 | #define sk_family __sk_common.skc_family | |
197 | #define sk_state __sk_common.skc_state | |
198 | #define sk_reuse __sk_common.skc_reuse | |
199 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if | |
200 | #define sk_node __sk_common.skc_node | |
201 | #define sk_bind_node __sk_common.skc_bind_node | |
202 | #define sk_refcnt __sk_common.skc_refcnt | |
81c3d547 | 203 | #define sk_hash __sk_common.skc_hash |
8feaf0c0 | 204 | #define sk_prot __sk_common.skc_prot |
07feaebf | 205 | #define sk_net __sk_common.skc_net |
1da177e4 LT |
206 | unsigned char sk_shutdown : 2, |
207 | sk_no_check : 2, | |
208 | sk_userlocks : 4; | |
209 | unsigned char sk_protocol; | |
210 | unsigned short sk_type; | |
211 | int sk_rcvbuf; | |
212 | socket_lock_t sk_lock; | |
fa438ccf ED |
213 | /* |
214 | * The backlog queue is special, it is always used with | |
215 | * the per-socket spinlock held and requires low latency | |
216 | * access. Therefore we special case it's implementation. | |
217 | */ | |
218 | struct { | |
219 | struct sk_buff *head; | |
220 | struct sk_buff *tail; | |
221 | } sk_backlog; | |
1da177e4 LT |
222 | wait_queue_head_t *sk_sleep; |
223 | struct dst_entry *sk_dst_cache; | |
224 | struct xfrm_policy *sk_policy[2]; | |
225 | rwlock_t sk_dst_lock; | |
226 | atomic_t sk_rmem_alloc; | |
227 | atomic_t sk_wmem_alloc; | |
228 | atomic_t sk_omem_alloc; | |
4e07a91c | 229 | int sk_sndbuf; |
1da177e4 LT |
230 | struct sk_buff_head sk_receive_queue; |
231 | struct sk_buff_head sk_write_queue; | |
97fc2f08 | 232 | struct sk_buff_head sk_async_wait_queue; |
1da177e4 LT |
233 | int sk_wmem_queued; |
234 | int sk_forward_alloc; | |
7d877f3b | 235 | gfp_t sk_allocation; |
1da177e4 | 236 | int sk_route_caps; |
bcd76111 | 237 | int sk_gso_type; |
9932cf95 | 238 | int sk_rcvlowat; |
1da177e4 LT |
239 | unsigned long sk_flags; |
240 | unsigned long sk_lingertime; | |
1da177e4 | 241 | struct sk_buff_head sk_error_queue; |
476e19cf | 242 | struct proto *sk_prot_creator; |
1da177e4 LT |
243 | rwlock_t sk_callback_lock; |
244 | int sk_err, | |
245 | sk_err_soft; | |
33c732c3 | 246 | atomic_t sk_drops; |
1da177e4 LT |
247 | unsigned short sk_ack_backlog; |
248 | unsigned short sk_max_ack_backlog; | |
249 | __u32 sk_priority; | |
250 | struct ucred sk_peercred; | |
1da177e4 LT |
251 | long sk_rcvtimeo; |
252 | long sk_sndtimeo; | |
253 | struct sk_filter *sk_filter; | |
254 | void *sk_protinfo; | |
255 | struct timer_list sk_timer; | |
b7aa0bf7 | 256 | ktime_t sk_stamp; |
1da177e4 LT |
257 | struct socket *sk_socket; |
258 | void *sk_user_data; | |
259 | struct page *sk_sndmsg_page; | |
260 | struct sk_buff *sk_send_head; | |
261 | __u32 sk_sndmsg_off; | |
262 | int sk_write_pending; | |
263 | void *sk_security; | |
264 | void (*sk_state_change)(struct sock *sk); | |
265 | void (*sk_data_ready)(struct sock *sk, int bytes); | |
266 | void (*sk_write_space)(struct sock *sk); | |
267 | void (*sk_error_report)(struct sock *sk); | |
268 | int (*sk_backlog_rcv)(struct sock *sk, | |
269 | struct sk_buff *skb); | |
270 | void (*sk_destruct)(struct sock *sk); | |
271 | }; | |
272 | ||
273 | /* | |
274 | * Hashed lists helper routines | |
275 | */ | |
e48c414e | 276 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
277 | { |
278 | return hlist_entry(head->first, struct sock, sk_node); | |
279 | } | |
280 | ||
e48c414e | 281 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
282 | { |
283 | return hlist_empty(head) ? NULL : __sk_head(head); | |
284 | } | |
285 | ||
e48c414e | 286 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 LT |
287 | { |
288 | return sk->sk_node.next ? | |
289 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; | |
290 | } | |
291 | ||
e48c414e | 292 | static inline int sk_unhashed(const struct sock *sk) |
1da177e4 LT |
293 | { |
294 | return hlist_unhashed(&sk->sk_node); | |
295 | } | |
296 | ||
e48c414e | 297 | static inline int sk_hashed(const struct sock *sk) |
1da177e4 | 298 | { |
da753bea | 299 | return !sk_unhashed(sk); |
1da177e4 LT |
300 | } |
301 | ||
302 | static __inline__ void sk_node_init(struct hlist_node *node) | |
303 | { | |
304 | node->pprev = NULL; | |
305 | } | |
306 | ||
307 | static __inline__ void __sk_del_node(struct sock *sk) | |
308 | { | |
309 | __hlist_del(&sk->sk_node); | |
310 | } | |
311 | ||
312 | static __inline__ int __sk_del_node_init(struct sock *sk) | |
313 | { | |
314 | if (sk_hashed(sk)) { | |
315 | __sk_del_node(sk); | |
316 | sk_node_init(&sk->sk_node); | |
317 | return 1; | |
318 | } | |
319 | return 0; | |
320 | } | |
321 | ||
322 | /* Grab socket reference count. This operation is valid only | |
323 | when sk is ALREADY grabbed f.e. it is found in hash table | |
324 | or a list and the lookup is made under lock preventing hash table | |
325 | modifications. | |
326 | */ | |
327 | ||
328 | static inline void sock_hold(struct sock *sk) | |
329 | { | |
330 | atomic_inc(&sk->sk_refcnt); | |
331 | } | |
332 | ||
333 | /* Ungrab socket in the context, which assumes that socket refcnt | |
334 | cannot hit zero, f.e. it is true in context of any socketcall. | |
335 | */ | |
336 | static inline void __sock_put(struct sock *sk) | |
337 | { | |
338 | atomic_dec(&sk->sk_refcnt); | |
339 | } | |
340 | ||
341 | static __inline__ int sk_del_node_init(struct sock *sk) | |
342 | { | |
343 | int rc = __sk_del_node_init(sk); | |
344 | ||
345 | if (rc) { | |
346 | /* paranoid for a while -acme */ | |
347 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
348 | __sock_put(sk); | |
349 | } | |
350 | return rc; | |
351 | } | |
352 | ||
353 | static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) | |
354 | { | |
355 | hlist_add_head(&sk->sk_node, list); | |
356 | } | |
357 | ||
358 | static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) | |
359 | { | |
360 | sock_hold(sk); | |
361 | __sk_add_node(sk, list); | |
362 | } | |
363 | ||
364 | static __inline__ void __sk_del_bind_node(struct sock *sk) | |
365 | { | |
366 | __hlist_del(&sk->sk_bind_node); | |
367 | } | |
368 | ||
369 | static __inline__ void sk_add_bind_node(struct sock *sk, | |
370 | struct hlist_head *list) | |
371 | { | |
372 | hlist_add_head(&sk->sk_bind_node, list); | |
373 | } | |
374 | ||
375 | #define sk_for_each(__sk, node, list) \ | |
376 | hlist_for_each_entry(__sk, node, list, sk_node) | |
377 | #define sk_for_each_from(__sk, node) \ | |
378 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
379 | hlist_for_each_entry_from(__sk, node, sk_node) | |
380 | #define sk_for_each_continue(__sk, node) \ | |
381 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
382 | hlist_for_each_entry_continue(__sk, node, sk_node) | |
383 | #define sk_for_each_safe(__sk, node, tmp, list) \ | |
384 | hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) | |
385 | #define sk_for_each_bound(__sk, node, list) \ | |
386 | hlist_for_each_entry(__sk, node, list, sk_bind_node) | |
387 | ||
388 | /* Sock flags */ | |
389 | enum sock_flags { | |
390 | SOCK_DEAD, | |
391 | SOCK_DONE, | |
392 | SOCK_URGINLINE, | |
393 | SOCK_KEEPOPEN, | |
394 | SOCK_LINGER, | |
395 | SOCK_DESTROY, | |
396 | SOCK_BROADCAST, | |
397 | SOCK_TIMESTAMP, | |
398 | SOCK_ZAPPED, | |
399 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
400 | SOCK_DBG, /* %SO_DEBUG setting */ | |
401 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
92f37fd2 | 402 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
1da177e4 LT |
403 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
404 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
405 | }; | |
406 | ||
53b924b3 RB |
407 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
408 | { | |
409 | nsk->sk_flags = osk->sk_flags; | |
410 | } | |
411 | ||
1da177e4 LT |
412 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
413 | { | |
414 | __set_bit(flag, &sk->sk_flags); | |
415 | } | |
416 | ||
417 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
418 | { | |
419 | __clear_bit(flag, &sk->sk_flags); | |
420 | } | |
421 | ||
422 | static inline int sock_flag(struct sock *sk, enum sock_flags flag) | |
423 | { | |
424 | return test_bit(flag, &sk->sk_flags); | |
425 | } | |
426 | ||
427 | static inline void sk_acceptq_removed(struct sock *sk) | |
428 | { | |
429 | sk->sk_ack_backlog--; | |
430 | } | |
431 | ||
432 | static inline void sk_acceptq_added(struct sock *sk) | |
433 | { | |
434 | sk->sk_ack_backlog++; | |
435 | } | |
436 | ||
437 | static inline int sk_acceptq_is_full(struct sock *sk) | |
438 | { | |
64a14651 | 439 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
1da177e4 LT |
440 | } |
441 | ||
442 | /* | |
443 | * Compute minimal free write space needed to queue new packets. | |
444 | */ | |
445 | static inline int sk_stream_min_wspace(struct sock *sk) | |
446 | { | |
447 | return sk->sk_wmem_queued / 2; | |
448 | } | |
449 | ||
450 | static inline int sk_stream_wspace(struct sock *sk) | |
451 | { | |
452 | return sk->sk_sndbuf - sk->sk_wmem_queued; | |
453 | } | |
454 | ||
455 | extern void sk_stream_write_space(struct sock *sk); | |
456 | ||
457 | static inline int sk_stream_memory_free(struct sock *sk) | |
458 | { | |
459 | return sk->sk_wmem_queued < sk->sk_sndbuf; | |
460 | } | |
461 | ||
462 | extern void sk_stream_rfree(struct sk_buff *skb); | |
463 | ||
464 | static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
465 | { | |
466 | skb->sk = sk; | |
467 | skb->destructor = sk_stream_rfree; | |
468 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
469 | sk->sk_forward_alloc -= skb->truesize; | |
470 | } | |
471 | ||
472 | static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb) | |
473 | { | |
dc6de336 | 474 | skb_truesize_check(skb); |
1da177e4 LT |
475 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
476 | sk->sk_wmem_queued -= skb->truesize; | |
477 | sk->sk_forward_alloc += skb->truesize; | |
478 | __kfree_skb(skb); | |
479 | } | |
480 | ||
481 | /* The per-socket spinlock must be held here. */ | |
9ee6b535 SH |
482 | static inline void sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
483 | { | |
484 | if (!sk->sk_backlog.tail) { | |
485 | sk->sk_backlog.head = sk->sk_backlog.tail = skb; | |
486 | } else { | |
487 | sk->sk_backlog.tail->next = skb; | |
488 | sk->sk_backlog.tail = skb; | |
489 | } | |
490 | skb->next = NULL; | |
491 | } | |
1da177e4 | 492 | |
cfcabdcc SH |
493 | #define sk_wait_event(__sk, __timeo, __condition) \ |
494 | ({ int __rc; \ | |
495 | release_sock(__sk); \ | |
496 | __rc = __condition; \ | |
497 | if (!__rc) { \ | |
498 | *(__timeo) = schedule_timeout(*(__timeo)); \ | |
499 | } \ | |
500 | lock_sock(__sk); \ | |
501 | __rc = __condition; \ | |
502 | __rc; \ | |
503 | }) | |
1da177e4 LT |
504 | |
505 | extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); | |
506 | extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
507 | extern void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
508 | extern int sk_stream_error(struct sock *sk, int flags, int err); | |
509 | extern void sk_stream_kill_queues(struct sock *sk); | |
510 | ||
511 | extern int sk_wait_data(struct sock *sk, long *timeo); | |
512 | ||
60236fdd | 513 | struct request_sock_ops; |
6d6ee43e | 514 | struct timewait_sock_ops; |
2e6599cb | 515 | |
1da177e4 LT |
516 | /* Networking protocol blocks we attach to sockets. |
517 | * socket layer -> transport layer interface | |
518 | * transport -> network interface is defined by struct inet_proto | |
519 | */ | |
520 | struct proto { | |
521 | void (*close)(struct sock *sk, | |
522 | long timeout); | |
523 | int (*connect)(struct sock *sk, | |
524 | struct sockaddr *uaddr, | |
525 | int addr_len); | |
526 | int (*disconnect)(struct sock *sk, int flags); | |
527 | ||
528 | struct sock * (*accept) (struct sock *sk, int flags, int *err); | |
529 | ||
530 | int (*ioctl)(struct sock *sk, int cmd, | |
531 | unsigned long arg); | |
532 | int (*init)(struct sock *sk); | |
533 | int (*destroy)(struct sock *sk); | |
534 | void (*shutdown)(struct sock *sk, int how); | |
535 | int (*setsockopt)(struct sock *sk, int level, | |
536 | int optname, char __user *optval, | |
537 | int optlen); | |
538 | int (*getsockopt)(struct sock *sk, int level, | |
539 | int optname, char __user *optval, | |
540 | int __user *option); | |
3fdadf7d DM |
541 | int (*compat_setsockopt)(struct sock *sk, |
542 | int level, | |
543 | int optname, char __user *optval, | |
544 | int optlen); | |
545 | int (*compat_getsockopt)(struct sock *sk, | |
546 | int level, | |
547 | int optname, char __user *optval, | |
548 | int __user *option); | |
1da177e4 LT |
549 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, |
550 | struct msghdr *msg, size_t len); | |
551 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, | |
552 | struct msghdr *msg, | |
553 | size_t len, int noblock, int flags, | |
554 | int *addr_len); | |
555 | int (*sendpage)(struct sock *sk, struct page *page, | |
556 | int offset, size_t size, int flags); | |
557 | int (*bind)(struct sock *sk, | |
558 | struct sockaddr *uaddr, int addr_len); | |
559 | ||
560 | int (*backlog_rcv) (struct sock *sk, | |
561 | struct sk_buff *skb); | |
562 | ||
563 | /* Keeping track of sk's, looking them up, and port selection methods. */ | |
564 | void (*hash)(struct sock *sk); | |
565 | void (*unhash)(struct sock *sk); | |
566 | int (*get_port)(struct sock *sk, unsigned short snum); | |
567 | ||
286ab3d4 ED |
568 | #ifdef CONFIG_SMP |
569 | /* Keeping track of sockets in use */ | |
570 | void (*inuse_add)(struct proto *prot, int inc); | |
571 | int (*inuse_getval)(const struct proto *prot); | |
572 | int *inuse_ptr; | |
573 | #else | |
574 | int inuse; | |
575 | #endif | |
1da177e4 LT |
576 | /* Memory pressure */ |
577 | void (*enter_memory_pressure)(void); | |
578 | atomic_t *memory_allocated; /* Current allocated memory. */ | |
579 | atomic_t *sockets_allocated; /* Current number of sockets. */ | |
580 | /* | |
581 | * Pressure flag: try to collapse. | |
582 | * Technical note: it is used by multiple contexts non atomically. | |
583 | * All the sk_stream_mem_schedule() is of this nature: accounting | |
584 | * is strict, actions are advisory and have some latency. | |
585 | */ | |
586 | int *memory_pressure; | |
587 | int *sysctl_mem; | |
588 | int *sysctl_wmem; | |
589 | int *sysctl_rmem; | |
590 | int max_header; | |
591 | ||
e18b890b | 592 | struct kmem_cache *slab; |
1da177e4 LT |
593 | unsigned int obj_size; |
594 | ||
0a5578cf | 595 | atomic_t *orphan_count; |
8feaf0c0 | 596 | |
60236fdd | 597 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 598 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 599 | |
1da177e4 LT |
600 | struct module *owner; |
601 | ||
602 | char name[32]; | |
603 | ||
604 | struct list_head node; | |
e6848976 ACM |
605 | #ifdef SOCK_REFCNT_DEBUG |
606 | atomic_t socks; | |
607 | #endif | |
1da177e4 LT |
608 | }; |
609 | ||
286ab3d4 ED |
610 | /* |
611 | * Special macros to let protos use a fast version of inuse{get|add} | |
612 | * using a static percpu variable per proto instead of an allocated one, | |
613 | * saving one dereference. | |
614 | * This might be changed if/when dynamic percpu vars become fast. | |
615 | */ | |
616 | #ifdef CONFIG_SMP | |
617 | # define DEFINE_PROTO_INUSE(NAME) \ | |
618 | static DEFINE_PER_CPU(int, NAME##_inuse); \ | |
619 | static void NAME##_inuse_add(struct proto *prot, int inc) \ | |
620 | { \ | |
621 | __get_cpu_var(NAME##_inuse) += inc; \ | |
622 | } \ | |
623 | \ | |
624 | static int NAME##_inuse_getval(const struct proto *prot)\ | |
625 | { \ | |
626 | int res = 0, cpu; \ | |
627 | \ | |
628 | for_each_possible_cpu(cpu) \ | |
629 | res += per_cpu(NAME##_inuse, cpu); \ | |
630 | return res; \ | |
631 | } | |
632 | # define REF_PROTO_INUSE(NAME) \ | |
633 | .inuse_add = NAME##_inuse_add, \ | |
634 | .inuse_getval = NAME##_inuse_getval, | |
635 | #else | |
636 | # define DEFINE_PROTO_INUSE(NAME) | |
637 | # define REF_PROTO_INUSE(NAME) | |
638 | #endif | |
639 | ||
1da177e4 LT |
640 | extern int proto_register(struct proto *prot, int alloc_slab); |
641 | extern void proto_unregister(struct proto *prot); | |
642 | ||
e6848976 ACM |
643 | #ifdef SOCK_REFCNT_DEBUG |
644 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
645 | { | |
646 | atomic_inc(&sk->sk_prot->socks); | |
647 | } | |
648 | ||
649 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
650 | { | |
651 | atomic_dec(&sk->sk_prot->socks); | |
652 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
653 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
654 | } | |
655 | ||
656 | static inline void sk_refcnt_debug_release(const struct sock *sk) | |
657 | { | |
658 | if (atomic_read(&sk->sk_refcnt) != 1) | |
659 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", | |
660 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); | |
661 | } | |
662 | #else /* SOCK_REFCNT_DEBUG */ | |
663 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
664 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
665 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
666 | #endif /* SOCK_REFCNT_DEBUG */ | |
667 | ||
1da177e4 LT |
668 | /* Called with local bh disabled */ |
669 | static __inline__ void sock_prot_inc_use(struct proto *prot) | |
670 | { | |
286ab3d4 ED |
671 | #ifdef CONFIG_SMP |
672 | prot->inuse_add(prot, 1); | |
673 | #else | |
674 | prot->inuse++; | |
675 | #endif | |
1da177e4 LT |
676 | } |
677 | ||
678 | static __inline__ void sock_prot_dec_use(struct proto *prot) | |
679 | { | |
286ab3d4 ED |
680 | #ifdef CONFIG_SMP |
681 | prot->inuse_add(prot, -1); | |
682 | #else | |
683 | prot->inuse--; | |
684 | #endif | |
685 | } | |
686 | ||
687 | static __inline__ int sock_prot_inuse(struct proto *proto) | |
688 | { | |
689 | #ifdef CONFIG_SMP | |
690 | return proto->inuse_getval(proto); | |
691 | #else | |
692 | return proto->inuse; | |
693 | #endif | |
1da177e4 LT |
694 | } |
695 | ||
614c6cb4 ACM |
696 | /* With per-bucket locks this operation is not-atomic, so that |
697 | * this version is not worse. | |
698 | */ | |
699 | static inline void __sk_prot_rehash(struct sock *sk) | |
700 | { | |
701 | sk->sk_prot->unhash(sk); | |
702 | sk->sk_prot->hash(sk); | |
703 | } | |
704 | ||
1da177e4 LT |
705 | /* About 10 seconds */ |
706 | #define SOCK_DESTROY_TIME (10*HZ) | |
707 | ||
708 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
709 | #define PROT_SOCK 1024 | |
710 | ||
711 | #define SHUTDOWN_MASK 3 | |
712 | #define RCV_SHUTDOWN 1 | |
713 | #define SEND_SHUTDOWN 2 | |
714 | ||
715 | #define SOCK_SNDBUF_LOCK 1 | |
716 | #define SOCK_RCVBUF_LOCK 2 | |
717 | #define SOCK_BINDADDR_LOCK 4 | |
718 | #define SOCK_BINDPORT_LOCK 8 | |
719 | ||
720 | /* sock_iocb: used to kick off async processing of socket ios */ | |
721 | struct sock_iocb { | |
722 | struct list_head list; | |
723 | ||
724 | int flags; | |
725 | int size; | |
726 | struct socket *sock; | |
727 | struct sock *sk; | |
728 | struct scm_cookie *scm; | |
729 | struct msghdr *msg, async_msg; | |
1da177e4 LT |
730 | struct kiocb *kiocb; |
731 | }; | |
732 | ||
733 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) | |
734 | { | |
735 | return (struct sock_iocb *)iocb->private; | |
736 | } | |
737 | ||
738 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) | |
739 | { | |
740 | return si->kiocb; | |
741 | } | |
742 | ||
743 | struct socket_alloc { | |
744 | struct socket socket; | |
745 | struct inode vfs_inode; | |
746 | }; | |
747 | ||
748 | static inline struct socket *SOCKET_I(struct inode *inode) | |
749 | { | |
750 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
751 | } | |
752 | ||
753 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
754 | { | |
755 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
756 | } | |
757 | ||
758 | extern void __sk_stream_mem_reclaim(struct sock *sk); | |
759 | extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind); | |
760 | ||
761 | #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE) | |
762 | ||
763 | static inline int sk_stream_pages(int amt) | |
764 | { | |
172589cc | 765 | return DIV_ROUND_UP(amt, SK_STREAM_MEM_QUANTUM); |
1da177e4 LT |
766 | } |
767 | ||
768 | static inline void sk_stream_mem_reclaim(struct sock *sk) | |
769 | { | |
770 | if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM) | |
771 | __sk_stream_mem_reclaim(sk); | |
772 | } | |
773 | ||
1da177e4 LT |
774 | static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb) |
775 | { | |
776 | return (int)skb->truesize <= sk->sk_forward_alloc || | |
777 | sk_stream_mem_schedule(sk, skb->truesize, 1); | |
778 | } | |
779 | ||
d80d99d6 HX |
780 | static inline int sk_stream_wmem_schedule(struct sock *sk, int size) |
781 | { | |
782 | return size <= sk->sk_forward_alloc || | |
783 | sk_stream_mem_schedule(sk, size, 0); | |
784 | } | |
785 | ||
1da177e4 LT |
786 | /* Used by processes to "lock" a socket state, so that |
787 | * interrupts and bottom half handlers won't change it | |
788 | * from under us. It essentially blocks any incoming | |
789 | * packets, so that we won't get any new data or any | |
790 | * packets that change the state of the socket. | |
791 | * | |
792 | * While locked, BH processing will add new packets to | |
793 | * the backlog queue. This queue is processed by the | |
794 | * owner of the socket lock right before it is released. | |
795 | * | |
796 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
797 | * accesses from user process context. | |
798 | */ | |
d2e9117c | 799 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) |
1da177e4 | 800 | |
ed07536e PZ |
801 | /* |
802 | * Macro so as to not evaluate some arguments when | |
803 | * lockdep is not enabled. | |
804 | * | |
805 | * Mark both the sk_lock and the sk_lock.slock as a | |
806 | * per-address-family lock class. | |
807 | */ | |
808 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ | |
809 | do { \ | |
d2e9117c | 810 | sk->sk_lock.owned = 0; \ |
ed07536e PZ |
811 | init_waitqueue_head(&sk->sk_lock.wq); \ |
812 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
813 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
814 | sizeof((sk)->sk_lock)); \ | |
815 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
816 | (skey), (sname)); \ | |
817 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ | |
818 | } while (0) | |
819 | ||
fcc70d5f PZ |
820 | extern void FASTCALL(lock_sock_nested(struct sock *sk, int subclass)); |
821 | ||
822 | static inline void lock_sock(struct sock *sk) | |
823 | { | |
824 | lock_sock_nested(sk, 0); | |
825 | } | |
826 | ||
1da177e4 LT |
827 | extern void FASTCALL(release_sock(struct sock *sk)); |
828 | ||
829 | /* BH context may only use the following locking interface. */ | |
830 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
c6366184 IM |
831 | #define bh_lock_sock_nested(__sk) \ |
832 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
833 | SINGLE_DEPTH_NESTING) | |
1da177e4 LT |
834 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
835 | ||
1b8d7ae4 | 836 | extern struct sock *sk_alloc(struct net *net, int family, |
dd0fc66f | 837 | gfp_t priority, |
6257ff21 | 838 | struct proto *prot); |
1da177e4 | 839 | extern void sk_free(struct sock *sk); |
87d11ceb | 840 | extern struct sock *sk_clone(const struct sock *sk, |
dd0fc66f | 841 | const gfp_t priority); |
1da177e4 LT |
842 | |
843 | extern struct sk_buff *sock_wmalloc(struct sock *sk, | |
844 | unsigned long size, int force, | |
dd0fc66f | 845 | gfp_t priority); |
1da177e4 LT |
846 | extern struct sk_buff *sock_rmalloc(struct sock *sk, |
847 | unsigned long size, int force, | |
dd0fc66f | 848 | gfp_t priority); |
1da177e4 LT |
849 | extern void sock_wfree(struct sk_buff *skb); |
850 | extern void sock_rfree(struct sk_buff *skb); | |
851 | ||
852 | extern int sock_setsockopt(struct socket *sock, int level, | |
853 | int op, char __user *optval, | |
854 | int optlen); | |
855 | ||
856 | extern int sock_getsockopt(struct socket *sock, int level, | |
857 | int op, char __user *optval, | |
858 | int __user *optlen); | |
859 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, | |
860 | unsigned long size, | |
861 | int noblock, | |
862 | int *errcode); | |
86a76caf | 863 | extern void *sock_kmalloc(struct sock *sk, int size, |
dd0fc66f | 864 | gfp_t priority); |
1da177e4 LT |
865 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); |
866 | extern void sk_send_sigurg(struct sock *sk); | |
867 | ||
868 | /* | |
869 | * Functions to fill in entries in struct proto_ops when a protocol | |
870 | * does not implement a particular function. | |
871 | */ | |
872 | extern int sock_no_bind(struct socket *, | |
873 | struct sockaddr *, int); | |
874 | extern int sock_no_connect(struct socket *, | |
875 | struct sockaddr *, int, int); | |
876 | extern int sock_no_socketpair(struct socket *, | |
877 | struct socket *); | |
878 | extern int sock_no_accept(struct socket *, | |
879 | struct socket *, int); | |
880 | extern int sock_no_getname(struct socket *, | |
881 | struct sockaddr *, int *, int); | |
882 | extern unsigned int sock_no_poll(struct file *, struct socket *, | |
883 | struct poll_table_struct *); | |
884 | extern int sock_no_ioctl(struct socket *, unsigned int, | |
885 | unsigned long); | |
886 | extern int sock_no_listen(struct socket *, int); | |
887 | extern int sock_no_shutdown(struct socket *, int); | |
888 | extern int sock_no_getsockopt(struct socket *, int , int, | |
889 | char __user *, int __user *); | |
890 | extern int sock_no_setsockopt(struct socket *, int, int, | |
891 | char __user *, int); | |
892 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, | |
893 | struct msghdr *, size_t); | |
894 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, | |
895 | struct msghdr *, size_t, int); | |
896 | extern int sock_no_mmap(struct file *file, | |
897 | struct socket *sock, | |
898 | struct vm_area_struct *vma); | |
899 | extern ssize_t sock_no_sendpage(struct socket *sock, | |
900 | struct page *page, | |
901 | int offset, size_t size, | |
902 | int flags); | |
903 | ||
904 | /* | |
905 | * Functions to fill in entries in struct proto_ops when a protocol | |
906 | * uses the inet style. | |
907 | */ | |
908 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, | |
909 | char __user *optval, int __user *optlen); | |
910 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, | |
911 | struct msghdr *msg, size_t size, int flags); | |
912 | extern int sock_common_setsockopt(struct socket *sock, int level, int optname, | |
913 | char __user *optval, int optlen); | |
3fdadf7d DM |
914 | extern int compat_sock_common_getsockopt(struct socket *sock, int level, |
915 | int optname, char __user *optval, int __user *optlen); | |
916 | extern int compat_sock_common_setsockopt(struct socket *sock, int level, | |
917 | int optname, char __user *optval, int optlen); | |
1da177e4 LT |
918 | |
919 | extern void sk_common_release(struct sock *sk); | |
920 | ||
921 | /* | |
922 | * Default socket callbacks and setup code | |
923 | */ | |
924 | ||
925 | /* Initialise core socket variables */ | |
926 | extern void sock_init_data(struct socket *sock, struct sock *sk); | |
927 | ||
928 | /** | |
929 | * sk_filter - run a packet through a socket filter | |
930 | * @sk: sock associated with &sk_buff | |
931 | * @skb: buffer to filter | |
932 | * @needlock: set to 1 if the sock is not locked by caller. | |
933 | * | |
934 | * Run the filter code and then cut skb->data to correct size returned by | |
935 | * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller | |
936 | * than pkt_len we keep whole skb->data. This is the socket level | |
937 | * wrapper to sk_run_filter. It returns 0 if the packet should | |
938 | * be accepted or -EPERM if the packet should be tossed. | |
939 | * | |
940 | */ | |
941 | ||
fda9ef5d | 942 | static inline int sk_filter(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
943 | { |
944 | int err; | |
fda9ef5d | 945 | struct sk_filter *filter; |
1da177e4 LT |
946 | |
947 | err = security_sock_rcv_skb(sk, skb); | |
948 | if (err) | |
949 | return err; | |
950 | ||
fda9ef5d | 951 | rcu_read_lock_bh(); |
9d3e4442 | 952 | filter = rcu_dereference(sk->sk_filter); |
fda9ef5d DM |
953 | if (filter) { |
954 | unsigned int pkt_len = sk_run_filter(skb, filter->insns, | |
955 | filter->len); | |
956 | err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM; | |
1da177e4 | 957 | } |
fda9ef5d DM |
958 | rcu_read_unlock_bh(); |
959 | ||
1da177e4 LT |
960 | return err; |
961 | } | |
962 | ||
dc9b3346 PB |
963 | /** |
964 | * sk_filter_release: Release a socket filter | |
965 | * @sk: socket | |
966 | * @fp: filter to remove | |
967 | * | |
968 | * Remove a filter from a socket and release its resources. | |
969 | */ | |
970 | ||
309dd5fc PE |
971 | static inline void sk_filter_release(struct sk_filter *fp) |
972 | { | |
973 | if (atomic_dec_and_test(&fp->refcnt)) | |
47e958ea | 974 | kfree(fp); |
309dd5fc PE |
975 | } |
976 | ||
977 | static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp) | |
1da177e4 LT |
978 | { |
979 | unsigned int size = sk_filter_len(fp); | |
980 | ||
981 | atomic_sub(size, &sk->sk_omem_alloc); | |
309dd5fc | 982 | sk_filter_release(fp); |
1da177e4 LT |
983 | } |
984 | ||
985 | static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) | |
986 | { | |
987 | atomic_inc(&fp->refcnt); | |
988 | atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); | |
989 | } | |
990 | ||
991 | /* | |
992 | * Socket reference counting postulates. | |
993 | * | |
994 | * * Each user of socket SHOULD hold a reference count. | |
995 | * * Each access point to socket (an hash table bucket, reference from a list, | |
996 | * running timer, skb in flight MUST hold a reference count. | |
997 | * * When reference count hits 0, it means it will never increase back. | |
998 | * * When reference count hits 0, it means that no references from | |
999 | * outside exist to this socket and current process on current CPU | |
1000 | * is last user and may/should destroy this socket. | |
1001 | * * sk_free is called from any context: process, BH, IRQ. When | |
1002 | * it is called, socket has no references from outside -> sk_free | |
1003 | * may release descendant resources allocated by the socket, but | |
1004 | * to the time when it is called, socket is NOT referenced by any | |
1005 | * hash tables, lists etc. | |
1006 | * * Packets, delivered from outside (from network or from another process) | |
1007 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1008 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1009 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1010 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1011 | * (leak to backlog). Packet socket does all the processing inside | |
1012 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1013 | * use separate SMP lock, so that they are prone too. | |
1014 | */ | |
1015 | ||
1016 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1017 | static inline void sock_put(struct sock *sk) | |
1018 | { | |
1019 | if (atomic_dec_and_test(&sk->sk_refcnt)) | |
1020 | sk_free(sk); | |
1021 | } | |
1022 | ||
58a5a7b9 ACM |
1023 | extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
1024 | const int nested); | |
25995ff5 | 1025 | |
1da177e4 LT |
1026 | /* Detach socket from process context. |
1027 | * Announce socket dead, detach it from wait queue and inode. | |
1028 | * Note that parent inode held reference count on this struct sock, | |
1029 | * we do not release it in this function, because protocol | |
1030 | * probably wants some additional cleanups or even continuing | |
1031 | * to work with this socket (TCP). | |
1032 | */ | |
1033 | static inline void sock_orphan(struct sock *sk) | |
1034 | { | |
1035 | write_lock_bh(&sk->sk_callback_lock); | |
1036 | sock_set_flag(sk, SOCK_DEAD); | |
1037 | sk->sk_socket = NULL; | |
1038 | sk->sk_sleep = NULL; | |
1039 | write_unlock_bh(&sk->sk_callback_lock); | |
1040 | } | |
1041 | ||
1042 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
1043 | { | |
1044 | write_lock_bh(&sk->sk_callback_lock); | |
1045 | sk->sk_sleep = &parent->wait; | |
1046 | parent->sk = sk; | |
1047 | sk->sk_socket = parent; | |
4237c75c | 1048 | security_sock_graft(sk, parent); |
1da177e4 LT |
1049 | write_unlock_bh(&sk->sk_callback_lock); |
1050 | } | |
1051 | ||
1052 | extern int sock_i_uid(struct sock *sk); | |
1053 | extern unsigned long sock_i_ino(struct sock *sk); | |
1054 | ||
1055 | static inline struct dst_entry * | |
1056 | __sk_dst_get(struct sock *sk) | |
1057 | { | |
1058 | return sk->sk_dst_cache; | |
1059 | } | |
1060 | ||
1061 | static inline struct dst_entry * | |
1062 | sk_dst_get(struct sock *sk) | |
1063 | { | |
1064 | struct dst_entry *dst; | |
1065 | ||
1066 | read_lock(&sk->sk_dst_lock); | |
1067 | dst = sk->sk_dst_cache; | |
1068 | if (dst) | |
1069 | dst_hold(dst); | |
1070 | read_unlock(&sk->sk_dst_lock); | |
1071 | return dst; | |
1072 | } | |
1073 | ||
1074 | static inline void | |
1075 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1076 | { | |
1077 | struct dst_entry *old_dst; | |
1078 | ||
1079 | old_dst = sk->sk_dst_cache; | |
1080 | sk->sk_dst_cache = dst; | |
1081 | dst_release(old_dst); | |
1082 | } | |
1083 | ||
1084 | static inline void | |
1085 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1086 | { | |
1087 | write_lock(&sk->sk_dst_lock); | |
1088 | __sk_dst_set(sk, dst); | |
1089 | write_unlock(&sk->sk_dst_lock); | |
1090 | } | |
1091 | ||
1092 | static inline void | |
1093 | __sk_dst_reset(struct sock *sk) | |
1094 | { | |
1095 | struct dst_entry *old_dst; | |
1096 | ||
1097 | old_dst = sk->sk_dst_cache; | |
1098 | sk->sk_dst_cache = NULL; | |
1099 | dst_release(old_dst); | |
1100 | } | |
1101 | ||
1102 | static inline void | |
1103 | sk_dst_reset(struct sock *sk) | |
1104 | { | |
1105 | write_lock(&sk->sk_dst_lock); | |
1106 | __sk_dst_reset(sk); | |
1107 | write_unlock(&sk->sk_dst_lock); | |
1108 | } | |
1109 | ||
f0088a50 | 1110 | extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1111 | |
f0088a50 | 1112 | extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1113 | |
bcd76111 HX |
1114 | static inline int sk_can_gso(const struct sock *sk) |
1115 | { | |
1116 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
1117 | } | |
1118 | ||
9958089a | 1119 | extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
6cbb0df7 | 1120 | |
1da177e4 LT |
1121 | static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb) |
1122 | { | |
1123 | sk->sk_wmem_queued += skb->truesize; | |
1124 | sk->sk_forward_alloc -= skb->truesize; | |
1125 | } | |
1126 | ||
1127 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, | |
1128 | struct sk_buff *skb, struct page *page, | |
1129 | int off, int copy) | |
1130 | { | |
1131 | if (skb->ip_summed == CHECKSUM_NONE) { | |
1132 | int err = 0; | |
5084205f | 1133 | __wsum csum = csum_and_copy_from_user(from, |
1da177e4 LT |
1134 | page_address(page) + off, |
1135 | copy, 0, &err); | |
1136 | if (err) | |
1137 | return err; | |
1138 | skb->csum = csum_block_add(skb->csum, csum, skb->len); | |
1139 | } else if (copy_from_user(page_address(page) + off, from, copy)) | |
1140 | return -EFAULT; | |
1141 | ||
1142 | skb->len += copy; | |
1143 | skb->data_len += copy; | |
1144 | skb->truesize += copy; | |
1145 | sk->sk_wmem_queued += copy; | |
1146 | sk->sk_forward_alloc -= copy; | |
1147 | return 0; | |
1148 | } | |
1149 | ||
1150 | /* | |
1151 | * Queue a received datagram if it will fit. Stream and sequenced | |
1152 | * protocols can't normally use this as they need to fit buffers in | |
1153 | * and play with them. | |
1154 | * | |
1155 | * Inlined as it's very short and called for pretty much every | |
1156 | * packet ever received. | |
1157 | */ | |
1158 | ||
1159 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) | |
1160 | { | |
1161 | sock_hold(sk); | |
1162 | skb->sk = sk; | |
1163 | skb->destructor = sock_wfree; | |
1164 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); | |
1165 | } | |
1166 | ||
1167 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
1168 | { | |
1169 | skb->sk = sk; | |
1170 | skb->destructor = sock_rfree; | |
1171 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
1172 | } | |
1173 | ||
1174 | extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, | |
1175 | unsigned long expires); | |
1176 | ||
1177 | extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); | |
1178 | ||
f0088a50 | 1179 | extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 LT |
1180 | |
1181 | static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb) | |
1182 | { | |
1183 | /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces | |
1184 | number of warnings when compiling with -W --ANK | |
1185 | */ | |
1186 | if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >= | |
1187 | (unsigned)sk->sk_rcvbuf) | |
1188 | return -ENOMEM; | |
1189 | skb_set_owner_r(skb, sk); | |
1190 | skb_queue_tail(&sk->sk_error_queue, skb); | |
1191 | if (!sock_flag(sk, SOCK_DEAD)) | |
1192 | sk->sk_data_ready(sk, skb->len); | |
1193 | return 0; | |
1194 | } | |
1195 | ||
1196 | /* | |
1197 | * Recover an error report and clear atomically | |
1198 | */ | |
1199 | ||
1200 | static inline int sock_error(struct sock *sk) | |
1201 | { | |
c1cbe4b7 BL |
1202 | int err; |
1203 | if (likely(!sk->sk_err)) | |
1204 | return 0; | |
1205 | err = xchg(&sk->sk_err, 0); | |
1da177e4 LT |
1206 | return -err; |
1207 | } | |
1208 | ||
1209 | static inline unsigned long sock_wspace(struct sock *sk) | |
1210 | { | |
1211 | int amt = 0; | |
1212 | ||
1213 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
1214 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); | |
1215 | if (amt < 0) | |
1216 | amt = 0; | |
1217 | } | |
1218 | return amt; | |
1219 | } | |
1220 | ||
1221 | static inline void sk_wake_async(struct sock *sk, int how, int band) | |
1222 | { | |
1223 | if (sk->sk_socket && sk->sk_socket->fasync_list) | |
1224 | sock_wake_async(sk->sk_socket, how, band); | |
1225 | } | |
1226 | ||
1227 | #define SOCK_MIN_SNDBUF 2048 | |
1228 | #define SOCK_MIN_RCVBUF 256 | |
1229 | ||
1230 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
1231 | { | |
1232 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { | |
1233 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2); | |
1234 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); | |
1235 | } | |
1236 | } | |
1237 | ||
1238 | static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk, | |
86a76caf | 1239 | int size, int mem, |
dd0fc66f | 1240 | gfp_t gfp) |
1da177e4 | 1241 | { |
c65f7f00 | 1242 | struct sk_buff *skb; |
1da177e4 | 1243 | |
21df56c6 HX |
1244 | /* The TCP header must be at least 32-bit aligned. */ |
1245 | size = ALIGN(size, 4); | |
1246 | ||
fb93134d | 1247 | skb = alloc_skb_fclone(size + sk->sk_prot->max_header, gfp); |
1da177e4 LT |
1248 | if (skb) { |
1249 | skb->truesize += mem; | |
d80d99d6 | 1250 | if (sk_stream_wmem_schedule(sk, skb->truesize)) { |
fb93134d HX |
1251 | /* |
1252 | * Make sure that we have exactly size bytes | |
1253 | * available to the caller, no more, no less. | |
1254 | */ | |
1255 | skb_reserve(skb, skb_tailroom(skb) - size); | |
1da177e4 LT |
1256 | return skb; |
1257 | } | |
1258 | __kfree_skb(skb); | |
1259 | } else { | |
1260 | sk->sk_prot->enter_memory_pressure(); | |
1261 | sk_stream_moderate_sndbuf(sk); | |
1262 | } | |
1263 | return NULL; | |
1264 | } | |
1265 | ||
1266 | static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk, | |
86a76caf | 1267 | int size, |
dd0fc66f | 1268 | gfp_t gfp) |
1da177e4 LT |
1269 | { |
1270 | return sk_stream_alloc_pskb(sk, size, 0, gfp); | |
1271 | } | |
1272 | ||
1273 | static inline struct page *sk_stream_alloc_page(struct sock *sk) | |
1274 | { | |
1275 | struct page *page = NULL; | |
1276 | ||
ef015786 HX |
1277 | page = alloc_pages(sk->sk_allocation, 0); |
1278 | if (!page) { | |
1da177e4 LT |
1279 | sk->sk_prot->enter_memory_pressure(); |
1280 | sk_stream_moderate_sndbuf(sk); | |
1281 | } | |
1282 | return page; | |
1283 | } | |
1284 | ||
1da177e4 LT |
1285 | /* |
1286 | * Default write policy as shown to user space via poll/select/SIGIO | |
1287 | */ | |
1288 | static inline int sock_writeable(const struct sock *sk) | |
1289 | { | |
1290 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2); | |
1291 | } | |
1292 | ||
dd0fc66f | 1293 | static inline gfp_t gfp_any(void) |
1da177e4 | 1294 | { |
4498121c | 1295 | return in_atomic() ? GFP_ATOMIC : GFP_KERNEL; |
1da177e4 LT |
1296 | } |
1297 | ||
1298 | static inline long sock_rcvtimeo(const struct sock *sk, int noblock) | |
1299 | { | |
1300 | return noblock ? 0 : sk->sk_rcvtimeo; | |
1301 | } | |
1302 | ||
1303 | static inline long sock_sndtimeo(const struct sock *sk, int noblock) | |
1304 | { | |
1305 | return noblock ? 0 : sk->sk_sndtimeo; | |
1306 | } | |
1307 | ||
1308 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
1309 | { | |
1310 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; | |
1311 | } | |
1312 | ||
1313 | /* Alas, with timeout socket operations are not restartable. | |
1314 | * Compare this to poll(). | |
1315 | */ | |
1316 | static inline int sock_intr_errno(long timeo) | |
1317 | { | |
1318 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
1319 | } | |
1320 | ||
92f37fd2 ED |
1321 | extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
1322 | struct sk_buff *skb); | |
1323 | ||
1da177e4 LT |
1324 | static __inline__ void |
1325 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) | |
1326 | { | |
b7aa0bf7 | 1327 | ktime_t kt = skb->tstamp; |
a61bbcf2 | 1328 | |
92f37fd2 ED |
1329 | if (sock_flag(sk, SOCK_RCVTSTAMP)) |
1330 | __sock_recv_timestamp(msg, sk, skb); | |
1331 | else | |
b7aa0bf7 | 1332 | sk->sk_stamp = kt; |
1da177e4 LT |
1333 | } |
1334 | ||
1335 | /** | |
1336 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
1337 | * @sk: socket to eat this skb from |
1338 | * @skb: socket buffer to eat | |
f4b8ea78 | 1339 | * @copied_early: flag indicating whether DMA operations copied this data early |
1da177e4 LT |
1340 | * |
1341 | * This routine must be called with interrupts disabled or with the socket | |
1342 | * locked so that the sk_buff queue operation is ok. | |
1343 | */ | |
624d1164 CL |
1344 | #ifdef CONFIG_NET_DMA |
1345 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) | |
1346 | { | |
1347 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1348 | if (!copied_early) | |
1349 | __kfree_skb(skb); | |
1350 | else | |
1351 | __skb_queue_tail(&sk->sk_async_wait_queue, skb); | |
1352 | } | |
1353 | #else | |
1354 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) | |
1da177e4 LT |
1355 | { |
1356 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1357 | __kfree_skb(skb); | |
1358 | } | |
624d1164 | 1359 | #endif |
1da177e4 LT |
1360 | |
1361 | extern void sock_enable_timestamp(struct sock *sk); | |
1362 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); | |
ae40eb1e | 1363 | extern int sock_get_timestampns(struct sock *, struct timespec __user *); |
1da177e4 LT |
1364 | |
1365 | /* | |
1366 | * Enable debug/info messages | |
1367 | */ | |
a2a316fd SH |
1368 | extern int net_msg_warn; |
1369 | #define NETDEBUG(fmt, args...) \ | |
1370 | do { if (net_msg_warn) printk(fmt,##args); } while (0) | |
1da177e4 | 1371 | |
a2a316fd SH |
1372 | #define LIMIT_NETDEBUG(fmt, args...) \ |
1373 | do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) | |
1da177e4 LT |
1374 | |
1375 | /* | |
1376 | * Macros for sleeping on a socket. Use them like this: | |
1377 | * | |
1378 | * SOCK_SLEEP_PRE(sk) | |
1379 | * if (condition) | |
1380 | * schedule(); | |
1381 | * SOCK_SLEEP_POST(sk) | |
1382 | * | |
1383 | * N.B. These are now obsolete and were, afaik, only ever used in DECnet | |
1384 | * and when the last use of them in DECnet has gone, I'm intending to | |
1385 | * remove them. | |
1386 | */ | |
1387 | ||
1388 | #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \ | |
1389 | DECLARE_WAITQUEUE(wait, tsk); \ | |
1390 | tsk->state = TASK_INTERRUPTIBLE; \ | |
1391 | add_wait_queue((sk)->sk_sleep, &wait); \ | |
1392 | release_sock(sk); | |
1393 | ||
1394 | #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \ | |
1395 | remove_wait_queue((sk)->sk_sleep, &wait); \ | |
1396 | lock_sock(sk); \ | |
1397 | } | |
1398 | ||
1da177e4 LT |
1399 | extern __u32 sysctl_wmem_max; |
1400 | extern __u32 sysctl_rmem_max; | |
1401 | ||
20380731 ACM |
1402 | extern void sk_init(void); |
1403 | ||
1404 | #ifdef CONFIG_SYSCTL | |
1405 | extern struct ctl_table core_table[]; | |
20380731 ACM |
1406 | #endif |
1407 | ||
6baf1f41 DM |
1408 | extern int sysctl_optmem_max; |
1409 | ||
20380731 ACM |
1410 | extern __u32 sysctl_wmem_default; |
1411 | extern __u32 sysctl_rmem_default; | |
20380731 | 1412 | |
1da177e4 | 1413 | #endif /* _SOCK_H */ |