]>
Commit | Line | Data |
---|---|---|
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 | * | |
10 | * Authors: Ross Biro | |
11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> | |
12 | * Corey Minyard <wf-rch!minyard@relay.EU.net> | |
13 | * Florian La Roche <flla@stud.uni-sb.de> | |
14 | * | |
15 | * Fixes: | |
16 | * Alan Cox : Volatiles in skbuff pointers. See | |
17 | * skbuff comments. May be overdone, | |
18 | * better to prove they can be removed | |
19 | * than the reverse. | |
20 | * Alan Cox : Added a zapped field for tcp to note | |
21 | * a socket is reset and must stay shut up | |
22 | * Alan Cox : New fields for options | |
23 | * Pauline Middelink : identd support | |
24 | * Alan Cox : Eliminate low level recv/recvfrom | |
25 | * David S. Miller : New socket lookup architecture. | |
26 | * Steve Whitehouse: Default routines for sock_ops | |
27 | * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made | |
28 | * protinfo be just a void pointer, as the | |
29 | * protocol specific parts were moved to | |
30 | * respective headers and ipv4/v6, etc now | |
31 | * use private slabcaches for its socks | |
32 | * Pedro Hortas : New flags field for socket options | |
33 | * | |
34 | * | |
35 | * This program is free software; you can redistribute it and/or | |
36 | * modify it under the terms of the GNU General Public License | |
37 | * as published by the Free Software Foundation; either version | |
38 | * 2 of the License, or (at your option) any later version. | |
39 | */ | |
40 | #ifndef _SOCK_H | |
41 | #define _SOCK_H | |
42 | ||
43 | #include <linux/hardirq.h> | |
44 | #include <linux/kernel.h> | |
45 | #include <linux/list.h> | |
46 | #include <linux/list_nulls.h> | |
47 | #include <linux/timer.h> | |
48 | #include <linux/cache.h> | |
49 | #include <linux/module.h> | |
50 | #include <linux/lockdep.h> | |
51 | #include <linux/netdevice.h> | |
52 | #include <linux/skbuff.h> /* struct sk_buff */ | |
53 | #include <linux/mm.h> | |
54 | #include <linux/security.h> | |
55 | #include <linux/slab.h> | |
56 | #include <linux/uaccess.h> | |
57 | ||
58 | #include <linux/filter.h> | |
59 | #include <linux/rculist_nulls.h> | |
60 | #include <linux/poll.h> | |
61 | ||
62 | #include <linux/atomic.h> | |
63 | #include <net/dst.h> | |
64 | #include <net/checksum.h> | |
65 | ||
66 | /* | |
67 | * This structure really needs to be cleaned up. | |
68 | * Most of it is for TCP, and not used by any of | |
69 | * the other protocols. | |
70 | */ | |
71 | ||
72 | /* Define this to get the SOCK_DBG debugging facility. */ | |
73 | #define SOCK_DEBUGGING | |
74 | #ifdef SOCK_DEBUGGING | |
75 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ | |
76 | printk(KERN_DEBUG msg); } while (0) | |
77 | #else | |
78 | /* Validate arguments and do nothing */ | |
79 | static inline void __attribute__ ((format (printf, 2, 3))) | |
80 | SOCK_DEBUG(struct sock *sk, const char *msg, ...) | |
81 | { | |
82 | } | |
83 | #endif | |
84 | ||
85 | /* This is the per-socket lock. The spinlock provides a synchronization | |
86 | * between user contexts and software interrupt processing, whereas the | |
87 | * mini-semaphore synchronizes multiple users amongst themselves. | |
88 | */ | |
89 | typedef struct { | |
90 | spinlock_t slock; | |
91 | int owned; | |
92 | wait_queue_head_t wq; | |
93 | /* | |
94 | * We express the mutex-alike socket_lock semantics | |
95 | * to the lock validator by explicitly managing | |
96 | * the slock as a lock variant (in addition to | |
97 | * the slock itself): | |
98 | */ | |
99 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
100 | struct lockdep_map dep_map; | |
101 | #endif | |
102 | } socket_lock_t; | |
103 | ||
104 | struct sock; | |
105 | struct proto; | |
106 | struct net; | |
107 | ||
108 | /** | |
109 | * struct sock_common - minimal network layer representation of sockets | |
110 | * @skc_daddr: Foreign IPv4 addr | |
111 | * @skc_rcv_saddr: Bound local IPv4 addr | |
112 | * @skc_hash: hash value used with various protocol lookup tables | |
113 | * @skc_u16hashes: two u16 hash values used by UDP lookup tables | |
114 | * @skc_family: network address family | |
115 | * @skc_state: Connection state | |
116 | * @skc_reuse: %SO_REUSEADDR setting | |
117 | * @skc_bound_dev_if: bound device index if != 0 | |
118 | * @skc_bind_node: bind hash linkage for various protocol lookup tables | |
119 | * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol | |
120 | * @skc_prot: protocol handlers inside a network family | |
121 | * @skc_net: reference to the network namespace of this socket | |
122 | * @skc_node: main hash linkage for various protocol lookup tables | |
123 | * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol | |
124 | * @skc_tx_queue_mapping: tx queue number for this connection | |
125 | * @skc_refcnt: reference count | |
126 | * | |
127 | * This is the minimal network layer representation of sockets, the header | |
128 | * for struct sock and struct inet_timewait_sock. | |
129 | */ | |
130 | struct sock_common { | |
131 | /* skc_daddr and skc_rcv_saddr must be grouped : | |
132 | * cf INET_MATCH() and INET_TW_MATCH() | |
133 | */ | |
134 | __be32 skc_daddr; | |
135 | __be32 skc_rcv_saddr; | |
136 | ||
137 | union { | |
138 | unsigned int skc_hash; | |
139 | __u16 skc_u16hashes[2]; | |
140 | }; | |
141 | unsigned short skc_family; | |
142 | volatile unsigned char skc_state; | |
143 | unsigned char skc_reuse; | |
144 | int skc_bound_dev_if; | |
145 | union { | |
146 | struct hlist_node skc_bind_node; | |
147 | struct hlist_nulls_node skc_portaddr_node; | |
148 | }; | |
149 | struct proto *skc_prot; | |
150 | #ifdef CONFIG_NET_NS | |
151 | struct net *skc_net; | |
152 | #endif | |
153 | /* | |
154 | * fields between dontcopy_begin/dontcopy_end | |
155 | * are not copied in sock_copy() | |
156 | */ | |
157 | /* private: */ | |
158 | int skc_dontcopy_begin[0]; | |
159 | /* public: */ | |
160 | union { | |
161 | struct hlist_node skc_node; | |
162 | struct hlist_nulls_node skc_nulls_node; | |
163 | }; | |
164 | int skc_tx_queue_mapping; | |
165 | atomic_t skc_refcnt; | |
166 | /* private: */ | |
167 | int skc_dontcopy_end[0]; | |
168 | /* public: */ | |
169 | }; | |
170 | ||
171 | /** | |
172 | * struct sock - network layer representation of sockets | |
173 | * @__sk_common: shared layout with inet_timewait_sock | |
174 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN | |
175 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
176 | * @sk_lock: synchronizer | |
177 | * @sk_rcvbuf: size of receive buffer in bytes | |
178 | * @sk_wq: sock wait queue and async head | |
179 | * @sk_dst_cache: destination cache | |
180 | * @sk_dst_lock: destination cache lock | |
181 | * @sk_policy: flow policy | |
182 | * @sk_receive_queue: incoming packets | |
183 | * @sk_wmem_alloc: transmit queue bytes committed | |
184 | * @sk_write_queue: Packet sending queue | |
185 | * @sk_async_wait_queue: DMA copied packets | |
186 | * @sk_omem_alloc: "o" is "option" or "other" | |
187 | * @sk_wmem_queued: persistent queue size | |
188 | * @sk_forward_alloc: space allocated forward | |
189 | * @sk_allocation: allocation mode | |
190 | * @sk_sndbuf: size of send buffer in bytes | |
191 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, | |
192 | * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings | |
193 | * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets | |
194 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) | |
195 | * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) | |
196 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) | |
197 | * @sk_gso_max_size: Maximum GSO segment size to build | |
198 | * @sk_lingertime: %SO_LINGER l_linger setting | |
199 | * @sk_backlog: always used with the per-socket spinlock held | |
200 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
201 | * @sk_error_queue: rarely used | |
202 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, | |
203 | * IPV6_ADDRFORM for instance) | |
204 | * @sk_err: last error | |
205 | * @sk_err_soft: errors that don't cause failure but are the cause of a | |
206 | * persistent failure not just 'timed out' | |
207 | * @sk_drops: raw/udp drops counter | |
208 | * @sk_ack_backlog: current listen backlog | |
209 | * @sk_max_ack_backlog: listen backlog set in listen() | |
210 | * @sk_priority: %SO_PRIORITY setting | |
211 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
212 | * @sk_protocol: which protocol this socket belongs in this network family | |
213 | * @sk_peer_pid: &struct pid for this socket's peer | |
214 | * @sk_peer_cred: %SO_PEERCRED setting | |
215 | * @sk_rcvlowat: %SO_RCVLOWAT setting | |
216 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
217 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
218 | * @sk_rxhash: flow hash received from netif layer | |
219 | * @sk_filter: socket filtering instructions | |
220 | * @sk_protinfo: private area, net family specific, when not using slab | |
221 | * @sk_timer: sock cleanup timer | |
222 | * @sk_stamp: time stamp of last packet received | |
223 | * @sk_socket: Identd and reporting IO signals | |
224 | * @sk_user_data: RPC layer private data | |
225 | * @sk_sndmsg_page: cached page for sendmsg | |
226 | * @sk_sndmsg_off: cached offset for sendmsg | |
227 | * @sk_send_head: front of stuff to transmit | |
228 | * @sk_security: used by security modules | |
229 | * @sk_mark: generic packet mark | |
230 | * @sk_classid: this socket's cgroup classid | |
231 | * @sk_write_pending: a write to stream socket waits to start | |
232 | * @sk_state_change: callback to indicate change in the state of the sock | |
233 | * @sk_data_ready: callback to indicate there is data to be processed | |
234 | * @sk_write_space: callback to indicate there is bf sending space available | |
235 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
236 | * @sk_backlog_rcv: callback to process the backlog | |
237 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
238 | */ | |
239 | struct sock { | |
240 | /* | |
241 | * Now struct inet_timewait_sock also uses sock_common, so please just | |
242 | * don't add nothing before this first member (__sk_common) --acme | |
243 | */ | |
244 | struct sock_common __sk_common; | |
245 | #define sk_node __sk_common.skc_node | |
246 | #define sk_nulls_node __sk_common.skc_nulls_node | |
247 | #define sk_refcnt __sk_common.skc_refcnt | |
248 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping | |
249 | ||
250 | #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin | |
251 | #define sk_dontcopy_end __sk_common.skc_dontcopy_end | |
252 | #define sk_hash __sk_common.skc_hash | |
253 | #define sk_family __sk_common.skc_family | |
254 | #define sk_state __sk_common.skc_state | |
255 | #define sk_reuse __sk_common.skc_reuse | |
256 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if | |
257 | #define sk_bind_node __sk_common.skc_bind_node | |
258 | #define sk_prot __sk_common.skc_prot | |
259 | #define sk_net __sk_common.skc_net | |
260 | socket_lock_t sk_lock; | |
261 | struct sk_buff_head sk_receive_queue; | |
262 | /* | |
263 | * The backlog queue is special, it is always used with | |
264 | * the per-socket spinlock held and requires low latency | |
265 | * access. Therefore we special case it's implementation. | |
266 | * Note : rmem_alloc is in this structure to fill a hole | |
267 | * on 64bit arches, not because its logically part of | |
268 | * backlog. | |
269 | */ | |
270 | struct { | |
271 | atomic_t rmem_alloc; | |
272 | int len; | |
273 | struct sk_buff *head; | |
274 | struct sk_buff *tail; | |
275 | } sk_backlog; | |
276 | #define sk_rmem_alloc sk_backlog.rmem_alloc | |
277 | int sk_forward_alloc; | |
278 | #ifdef CONFIG_RPS | |
279 | __u32 sk_rxhash; | |
280 | #endif | |
281 | atomic_t sk_drops; | |
282 | int sk_rcvbuf; | |
283 | ||
284 | struct sk_filter __rcu *sk_filter; | |
285 | struct socket_wq __rcu *sk_wq; | |
286 | ||
287 | #ifdef CONFIG_NET_DMA | |
288 | struct sk_buff_head sk_async_wait_queue; | |
289 | #endif | |
290 | ||
291 | #ifdef CONFIG_XFRM | |
292 | struct xfrm_policy *sk_policy[2]; | |
293 | #endif | |
294 | unsigned long sk_flags; | |
295 | struct dst_entry *sk_dst_cache; | |
296 | spinlock_t sk_dst_lock; | |
297 | atomic_t sk_wmem_alloc; | |
298 | atomic_t sk_omem_alloc; | |
299 | int sk_sndbuf; | |
300 | struct sk_buff_head sk_write_queue; | |
301 | kmemcheck_bitfield_begin(flags); | |
302 | unsigned int sk_shutdown : 2, | |
303 | sk_no_check : 2, | |
304 | sk_userlocks : 4, | |
305 | sk_protocol : 8, | |
306 | sk_type : 16; | |
307 | kmemcheck_bitfield_end(flags); | |
308 | int sk_wmem_queued; | |
309 | gfp_t sk_allocation; | |
310 | int sk_route_caps; | |
311 | int sk_route_nocaps; | |
312 | int sk_gso_type; | |
313 | unsigned int sk_gso_max_size; | |
314 | int sk_rcvlowat; | |
315 | unsigned long sk_lingertime; | |
316 | struct sk_buff_head sk_error_queue; | |
317 | struct proto *sk_prot_creator; | |
318 | rwlock_t sk_callback_lock; | |
319 | int sk_err, | |
320 | sk_err_soft; | |
321 | unsigned short sk_ack_backlog; | |
322 | unsigned short sk_max_ack_backlog; | |
323 | __u32 sk_priority; | |
324 | struct pid *sk_peer_pid; | |
325 | const struct cred *sk_peer_cred; | |
326 | long sk_rcvtimeo; | |
327 | long sk_sndtimeo; | |
328 | void *sk_protinfo; | |
329 | struct timer_list sk_timer; | |
330 | ktime_t sk_stamp; | |
331 | struct socket *sk_socket; | |
332 | void *sk_user_data; | |
333 | struct page *sk_sndmsg_page; | |
334 | struct sk_buff *sk_send_head; | |
335 | __u32 sk_sndmsg_off; | |
336 | int sk_write_pending; | |
337 | #ifdef CONFIG_SECURITY | |
338 | void *sk_security; | |
339 | #endif | |
340 | __u32 sk_mark; | |
341 | u32 sk_classid; | |
342 | void (*sk_state_change)(struct sock *sk); | |
343 | void (*sk_data_ready)(struct sock *sk, int bytes); | |
344 | void (*sk_write_space)(struct sock *sk); | |
345 | void (*sk_error_report)(struct sock *sk); | |
346 | int (*sk_backlog_rcv)(struct sock *sk, | |
347 | struct sk_buff *skb); | |
348 | void (*sk_destruct)(struct sock *sk); | |
349 | }; | |
350 | ||
351 | /* | |
352 | * Hashed lists helper routines | |
353 | */ | |
354 | static inline struct sock *sk_entry(const struct hlist_node *node) | |
355 | { | |
356 | return hlist_entry(node, struct sock, sk_node); | |
357 | } | |
358 | ||
359 | static inline struct sock *__sk_head(const struct hlist_head *head) | |
360 | { | |
361 | return hlist_entry(head->first, struct sock, sk_node); | |
362 | } | |
363 | ||
364 | static inline struct sock *sk_head(const struct hlist_head *head) | |
365 | { | |
366 | return hlist_empty(head) ? NULL : __sk_head(head); | |
367 | } | |
368 | ||
369 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) | |
370 | { | |
371 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); | |
372 | } | |
373 | ||
374 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) | |
375 | { | |
376 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); | |
377 | } | |
378 | ||
379 | static inline struct sock *sk_next(const struct sock *sk) | |
380 | { | |
381 | return sk->sk_node.next ? | |
382 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; | |
383 | } | |
384 | ||
385 | static inline struct sock *sk_nulls_next(const struct sock *sk) | |
386 | { | |
387 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? | |
388 | hlist_nulls_entry(sk->sk_nulls_node.next, | |
389 | struct sock, sk_nulls_node) : | |
390 | NULL; | |
391 | } | |
392 | ||
393 | static inline int sk_unhashed(const struct sock *sk) | |
394 | { | |
395 | return hlist_unhashed(&sk->sk_node); | |
396 | } | |
397 | ||
398 | static inline int sk_hashed(const struct sock *sk) | |
399 | { | |
400 | return !sk_unhashed(sk); | |
401 | } | |
402 | ||
403 | static __inline__ void sk_node_init(struct hlist_node *node) | |
404 | { | |
405 | node->pprev = NULL; | |
406 | } | |
407 | ||
408 | static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node) | |
409 | { | |
410 | node->pprev = NULL; | |
411 | } | |
412 | ||
413 | static __inline__ void __sk_del_node(struct sock *sk) | |
414 | { | |
415 | __hlist_del(&sk->sk_node); | |
416 | } | |
417 | ||
418 | /* NB: equivalent to hlist_del_init_rcu */ | |
419 | static __inline__ int __sk_del_node_init(struct sock *sk) | |
420 | { | |
421 | if (sk_hashed(sk)) { | |
422 | __sk_del_node(sk); | |
423 | sk_node_init(&sk->sk_node); | |
424 | return 1; | |
425 | } | |
426 | return 0; | |
427 | } | |
428 | ||
429 | /* Grab socket reference count. This operation is valid only | |
430 | when sk is ALREADY grabbed f.e. it is found in hash table | |
431 | or a list and the lookup is made under lock preventing hash table | |
432 | modifications. | |
433 | */ | |
434 | ||
435 | static inline void sock_hold(struct sock *sk) | |
436 | { | |
437 | atomic_inc(&sk->sk_refcnt); | |
438 | } | |
439 | ||
440 | /* Ungrab socket in the context, which assumes that socket refcnt | |
441 | cannot hit zero, f.e. it is true in context of any socketcall. | |
442 | */ | |
443 | static inline void __sock_put(struct sock *sk) | |
444 | { | |
445 | atomic_dec(&sk->sk_refcnt); | |
446 | } | |
447 | ||
448 | static __inline__ int sk_del_node_init(struct sock *sk) | |
449 | { | |
450 | int rc = __sk_del_node_init(sk); | |
451 | ||
452 | if (rc) { | |
453 | /* paranoid for a while -acme */ | |
454 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
455 | __sock_put(sk); | |
456 | } | |
457 | return rc; | |
458 | } | |
459 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) | |
460 | ||
461 | static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk) | |
462 | { | |
463 | if (sk_hashed(sk)) { | |
464 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); | |
465 | return 1; | |
466 | } | |
467 | return 0; | |
468 | } | |
469 | ||
470 | static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk) | |
471 | { | |
472 | int rc = __sk_nulls_del_node_init_rcu(sk); | |
473 | ||
474 | if (rc) { | |
475 | /* paranoid for a while -acme */ | |
476 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
477 | __sock_put(sk); | |
478 | } | |
479 | return rc; | |
480 | } | |
481 | ||
482 | static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) | |
483 | { | |
484 | hlist_add_head(&sk->sk_node, list); | |
485 | } | |
486 | ||
487 | static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) | |
488 | { | |
489 | sock_hold(sk); | |
490 | __sk_add_node(sk, list); | |
491 | } | |
492 | ||
493 | static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) | |
494 | { | |
495 | sock_hold(sk); | |
496 | hlist_add_head_rcu(&sk->sk_node, list); | |
497 | } | |
498 | ||
499 | static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) | |
500 | { | |
501 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); | |
502 | } | |
503 | ||
504 | static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) | |
505 | { | |
506 | sock_hold(sk); | |
507 | __sk_nulls_add_node_rcu(sk, list); | |
508 | } | |
509 | ||
510 | static __inline__ void __sk_del_bind_node(struct sock *sk) | |
511 | { | |
512 | __hlist_del(&sk->sk_bind_node); | |
513 | } | |
514 | ||
515 | static __inline__ void sk_add_bind_node(struct sock *sk, | |
516 | struct hlist_head *list) | |
517 | { | |
518 | hlist_add_head(&sk->sk_bind_node, list); | |
519 | } | |
520 | ||
521 | #define sk_for_each(__sk, node, list) \ | |
522 | hlist_for_each_entry(__sk, node, list, sk_node) | |
523 | #define sk_for_each_rcu(__sk, node, list) \ | |
524 | hlist_for_each_entry_rcu(__sk, node, list, sk_node) | |
525 | #define sk_nulls_for_each(__sk, node, list) \ | |
526 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) | |
527 | #define sk_nulls_for_each_rcu(__sk, node, list) \ | |
528 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) | |
529 | #define sk_for_each_from(__sk, node) \ | |
530 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
531 | hlist_for_each_entry_from(__sk, node, sk_node) | |
532 | #define sk_nulls_for_each_from(__sk, node) \ | |
533 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ | |
534 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) | |
535 | #define sk_for_each_safe(__sk, node, tmp, list) \ | |
536 | hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) | |
537 | #define sk_for_each_bound(__sk, node, list) \ | |
538 | hlist_for_each_entry(__sk, node, list, sk_bind_node) | |
539 | ||
540 | /* Sock flags */ | |
541 | enum sock_flags { | |
542 | SOCK_DEAD, | |
543 | SOCK_DONE, | |
544 | SOCK_URGINLINE, | |
545 | SOCK_KEEPOPEN, | |
546 | SOCK_LINGER, | |
547 | SOCK_DESTROY, | |
548 | SOCK_BROADCAST, | |
549 | SOCK_TIMESTAMP, | |
550 | SOCK_ZAPPED, | |
551 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
552 | SOCK_DBG, /* %SO_DEBUG setting */ | |
553 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
554 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ | |
555 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ | |
556 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
557 | SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */ | |
558 | SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */ | |
559 | SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */ | |
560 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ | |
561 | SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */ | |
562 | SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */ | |
563 | SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */ | |
564 | SOCK_FASYNC, /* fasync() active */ | |
565 | SOCK_RXQ_OVFL, | |
566 | SOCK_ZEROCOPY, /* buffers from userspace */ | |
567 | }; | |
568 | ||
569 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) | |
570 | { | |
571 | nsk->sk_flags = osk->sk_flags; | |
572 | } | |
573 | ||
574 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) | |
575 | { | |
576 | __set_bit(flag, &sk->sk_flags); | |
577 | } | |
578 | ||
579 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
580 | { | |
581 | __clear_bit(flag, &sk->sk_flags); | |
582 | } | |
583 | ||
584 | static inline int sock_flag(struct sock *sk, enum sock_flags flag) | |
585 | { | |
586 | return test_bit(flag, &sk->sk_flags); | |
587 | } | |
588 | ||
589 | static inline void sk_acceptq_removed(struct sock *sk) | |
590 | { | |
591 | sk->sk_ack_backlog--; | |
592 | } | |
593 | ||
594 | static inline void sk_acceptq_added(struct sock *sk) | |
595 | { | |
596 | sk->sk_ack_backlog++; | |
597 | } | |
598 | ||
599 | static inline int sk_acceptq_is_full(struct sock *sk) | |
600 | { | |
601 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; | |
602 | } | |
603 | ||
604 | /* | |
605 | * Compute minimal free write space needed to queue new packets. | |
606 | */ | |
607 | static inline int sk_stream_min_wspace(struct sock *sk) | |
608 | { | |
609 | return sk->sk_wmem_queued >> 1; | |
610 | } | |
611 | ||
612 | static inline int sk_stream_wspace(struct sock *sk) | |
613 | { | |
614 | return sk->sk_sndbuf - sk->sk_wmem_queued; | |
615 | } | |
616 | ||
617 | extern void sk_stream_write_space(struct sock *sk); | |
618 | ||
619 | static inline int sk_stream_memory_free(struct sock *sk) | |
620 | { | |
621 | return sk->sk_wmem_queued < sk->sk_sndbuf; | |
622 | } | |
623 | ||
624 | /* OOB backlog add */ | |
625 | static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) | |
626 | { | |
627 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ | |
628 | skb_dst_force(skb); | |
629 | ||
630 | if (!sk->sk_backlog.tail) | |
631 | sk->sk_backlog.head = skb; | |
632 | else | |
633 | sk->sk_backlog.tail->next = skb; | |
634 | ||
635 | sk->sk_backlog.tail = skb; | |
636 | skb->next = NULL; | |
637 | } | |
638 | ||
639 | /* | |
640 | * Take into account size of receive queue and backlog queue | |
641 | */ | |
642 | static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb) | |
643 | { | |
644 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); | |
645 | ||
646 | return qsize + skb->truesize > sk->sk_rcvbuf; | |
647 | } | |
648 | ||
649 | /* The per-socket spinlock must be held here. */ | |
650 | static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb) | |
651 | { | |
652 | if (sk_rcvqueues_full(sk, skb)) | |
653 | return -ENOBUFS; | |
654 | ||
655 | __sk_add_backlog(sk, skb); | |
656 | sk->sk_backlog.len += skb->truesize; | |
657 | return 0; | |
658 | } | |
659 | ||
660 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) | |
661 | { | |
662 | return sk->sk_backlog_rcv(sk, skb); | |
663 | } | |
664 | ||
665 | static inline void sock_rps_record_flow(const struct sock *sk) | |
666 | { | |
667 | #ifdef CONFIG_RPS | |
668 | struct rps_sock_flow_table *sock_flow_table; | |
669 | ||
670 | rcu_read_lock(); | |
671 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
672 | rps_record_sock_flow(sock_flow_table, sk->sk_rxhash); | |
673 | rcu_read_unlock(); | |
674 | #endif | |
675 | } | |
676 | ||
677 | static inline void sock_rps_reset_flow(const struct sock *sk) | |
678 | { | |
679 | #ifdef CONFIG_RPS | |
680 | struct rps_sock_flow_table *sock_flow_table; | |
681 | ||
682 | rcu_read_lock(); | |
683 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
684 | rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash); | |
685 | rcu_read_unlock(); | |
686 | #endif | |
687 | } | |
688 | ||
689 | static inline void sock_rps_save_rxhash(struct sock *sk, | |
690 | const struct sk_buff *skb) | |
691 | { | |
692 | #ifdef CONFIG_RPS | |
693 | if (unlikely(sk->sk_rxhash != skb->rxhash)) { | |
694 | sock_rps_reset_flow(sk); | |
695 | sk->sk_rxhash = skb->rxhash; | |
696 | } | |
697 | #endif | |
698 | } | |
699 | ||
700 | static inline void sock_rps_reset_rxhash(struct sock *sk) | |
701 | { | |
702 | #ifdef CONFIG_RPS | |
703 | sock_rps_reset_flow(sk); | |
704 | sk->sk_rxhash = 0; | |
705 | #endif | |
706 | } | |
707 | ||
708 | #define sk_wait_event(__sk, __timeo, __condition) \ | |
709 | ({ int __rc; \ | |
710 | release_sock(__sk); \ | |
711 | __rc = __condition; \ | |
712 | if (!__rc) { \ | |
713 | *(__timeo) = schedule_timeout(*(__timeo)); \ | |
714 | } \ | |
715 | lock_sock(__sk); \ | |
716 | __rc = __condition; \ | |
717 | __rc; \ | |
718 | }) | |
719 | ||
720 | extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); | |
721 | extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
722 | extern void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
723 | extern int sk_stream_error(struct sock *sk, int flags, int err); | |
724 | extern void sk_stream_kill_queues(struct sock *sk); | |
725 | ||
726 | extern int sk_wait_data(struct sock *sk, long *timeo); | |
727 | ||
728 | struct request_sock_ops; | |
729 | struct timewait_sock_ops; | |
730 | struct inet_hashinfo; | |
731 | struct raw_hashinfo; | |
732 | ||
733 | /* Networking protocol blocks we attach to sockets. | |
734 | * socket layer -> transport layer interface | |
735 | * transport -> network interface is defined by struct inet_proto | |
736 | */ | |
737 | struct proto { | |
738 | void (*close)(struct sock *sk, | |
739 | long timeout); | |
740 | int (*connect)(struct sock *sk, | |
741 | struct sockaddr *uaddr, | |
742 | int addr_len); | |
743 | int (*disconnect)(struct sock *sk, int flags); | |
744 | ||
745 | struct sock * (*accept) (struct sock *sk, int flags, int *err); | |
746 | ||
747 | int (*ioctl)(struct sock *sk, int cmd, | |
748 | unsigned long arg); | |
749 | int (*init)(struct sock *sk); | |
750 | void (*destroy)(struct sock *sk); | |
751 | void (*shutdown)(struct sock *sk, int how); | |
752 | int (*setsockopt)(struct sock *sk, int level, | |
753 | int optname, char __user *optval, | |
754 | unsigned int optlen); | |
755 | int (*getsockopt)(struct sock *sk, int level, | |
756 | int optname, char __user *optval, | |
757 | int __user *option); | |
758 | #ifdef CONFIG_COMPAT | |
759 | int (*compat_setsockopt)(struct sock *sk, | |
760 | int level, | |
761 | int optname, char __user *optval, | |
762 | unsigned int optlen); | |
763 | int (*compat_getsockopt)(struct sock *sk, | |
764 | int level, | |
765 | int optname, char __user *optval, | |
766 | int __user *option); | |
767 | int (*compat_ioctl)(struct sock *sk, | |
768 | unsigned int cmd, unsigned long arg); | |
769 | #endif | |
770 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, | |
771 | struct msghdr *msg, size_t len); | |
772 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, | |
773 | struct msghdr *msg, | |
774 | size_t len, int noblock, int flags, | |
775 | int *addr_len); | |
776 | int (*sendpage)(struct sock *sk, struct page *page, | |
777 | int offset, size_t size, int flags); | |
778 | int (*bind)(struct sock *sk, | |
779 | struct sockaddr *uaddr, int addr_len); | |
780 | ||
781 | int (*backlog_rcv) (struct sock *sk, | |
782 | struct sk_buff *skb); | |
783 | ||
784 | /* Keeping track of sk's, looking them up, and port selection methods. */ | |
785 | void (*hash)(struct sock *sk); | |
786 | void (*unhash)(struct sock *sk); | |
787 | void (*rehash)(struct sock *sk); | |
788 | int (*get_port)(struct sock *sk, unsigned short snum); | |
789 | void (*clear_sk)(struct sock *sk, int size); | |
790 | ||
791 | /* Keeping track of sockets in use */ | |
792 | #ifdef CONFIG_PROC_FS | |
793 | unsigned int inuse_idx; | |
794 | #endif | |
795 | ||
796 | /* Memory pressure */ | |
797 | void (*enter_memory_pressure)(struct sock *sk); | |
798 | atomic_long_t *memory_allocated; /* Current allocated memory. */ | |
799 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ | |
800 | /* | |
801 | * Pressure flag: try to collapse. | |
802 | * Technical note: it is used by multiple contexts non atomically. | |
803 | * All the __sk_mem_schedule() is of this nature: accounting | |
804 | * is strict, actions are advisory and have some latency. | |
805 | */ | |
806 | int *memory_pressure; | |
807 | long *sysctl_mem; | |
808 | int *sysctl_wmem; | |
809 | int *sysctl_rmem; | |
810 | int max_header; | |
811 | bool no_autobind; | |
812 | ||
813 | struct kmem_cache *slab; | |
814 | unsigned int obj_size; | |
815 | int slab_flags; | |
816 | ||
817 | struct percpu_counter *orphan_count; | |
818 | ||
819 | struct request_sock_ops *rsk_prot; | |
820 | struct timewait_sock_ops *twsk_prot; | |
821 | ||
822 | union { | |
823 | struct inet_hashinfo *hashinfo; | |
824 | struct udp_table *udp_table; | |
825 | struct raw_hashinfo *raw_hash; | |
826 | } h; | |
827 | ||
828 | struct module *owner; | |
829 | ||
830 | char name[32]; | |
831 | ||
832 | struct list_head node; | |
833 | #ifdef SOCK_REFCNT_DEBUG | |
834 | atomic_t socks; | |
835 | #endif | |
836 | }; | |
837 | ||
838 | extern int proto_register(struct proto *prot, int alloc_slab); | |
839 | extern void proto_unregister(struct proto *prot); | |
840 | ||
841 | #ifdef SOCK_REFCNT_DEBUG | |
842 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
843 | { | |
844 | atomic_inc(&sk->sk_prot->socks); | |
845 | } | |
846 | ||
847 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
848 | { | |
849 | atomic_dec(&sk->sk_prot->socks); | |
850 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
851 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
852 | } | |
853 | ||
854 | static inline void sk_refcnt_debug_release(const struct sock *sk) | |
855 | { | |
856 | if (atomic_read(&sk->sk_refcnt) != 1) | |
857 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", | |
858 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); | |
859 | } | |
860 | #else /* SOCK_REFCNT_DEBUG */ | |
861 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
862 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
863 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
864 | #endif /* SOCK_REFCNT_DEBUG */ | |
865 | ||
866 | ||
867 | #ifdef CONFIG_PROC_FS | |
868 | /* Called with local bh disabled */ | |
869 | extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); | |
870 | extern int sock_prot_inuse_get(struct net *net, struct proto *proto); | |
871 | #else | |
872 | static void inline sock_prot_inuse_add(struct net *net, struct proto *prot, | |
873 | int inc) | |
874 | { | |
875 | } | |
876 | #endif | |
877 | ||
878 | ||
879 | /* With per-bucket locks this operation is not-atomic, so that | |
880 | * this version is not worse. | |
881 | */ | |
882 | static inline void __sk_prot_rehash(struct sock *sk) | |
883 | { | |
884 | sk->sk_prot->unhash(sk); | |
885 | sk->sk_prot->hash(sk); | |
886 | } | |
887 | ||
888 | void sk_prot_clear_portaddr_nulls(struct sock *sk, int size); | |
889 | ||
890 | /* About 10 seconds */ | |
891 | #define SOCK_DESTROY_TIME (10*HZ) | |
892 | ||
893 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
894 | #define PROT_SOCK 1024 | |
895 | ||
896 | #define SHUTDOWN_MASK 3 | |
897 | #define RCV_SHUTDOWN 1 | |
898 | #define SEND_SHUTDOWN 2 | |
899 | ||
900 | #define SOCK_SNDBUF_LOCK 1 | |
901 | #define SOCK_RCVBUF_LOCK 2 | |
902 | #define SOCK_BINDADDR_LOCK 4 | |
903 | #define SOCK_BINDPORT_LOCK 8 | |
904 | ||
905 | /* sock_iocb: used to kick off async processing of socket ios */ | |
906 | struct sock_iocb { | |
907 | struct list_head list; | |
908 | ||
909 | int flags; | |
910 | int size; | |
911 | struct socket *sock; | |
912 | struct sock *sk; | |
913 | struct scm_cookie *scm; | |
914 | struct msghdr *msg, async_msg; | |
915 | struct kiocb *kiocb; | |
916 | }; | |
917 | ||
918 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) | |
919 | { | |
920 | return (struct sock_iocb *)iocb->private; | |
921 | } | |
922 | ||
923 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) | |
924 | { | |
925 | return si->kiocb; | |
926 | } | |
927 | ||
928 | struct socket_alloc { | |
929 | struct socket socket; | |
930 | struct inode vfs_inode; | |
931 | }; | |
932 | ||
933 | static inline struct socket *SOCKET_I(struct inode *inode) | |
934 | { | |
935 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
936 | } | |
937 | ||
938 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
939 | { | |
940 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
941 | } | |
942 | ||
943 | /* | |
944 | * Functions for memory accounting | |
945 | */ | |
946 | extern int __sk_mem_schedule(struct sock *sk, int size, int kind); | |
947 | extern void __sk_mem_reclaim(struct sock *sk); | |
948 | ||
949 | #define SK_MEM_QUANTUM ((int)PAGE_SIZE) | |
950 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) | |
951 | #define SK_MEM_SEND 0 | |
952 | #define SK_MEM_RECV 1 | |
953 | ||
954 | static inline int sk_mem_pages(int amt) | |
955 | { | |
956 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; | |
957 | } | |
958 | ||
959 | static inline int sk_has_account(struct sock *sk) | |
960 | { | |
961 | /* return true if protocol supports memory accounting */ | |
962 | return !!sk->sk_prot->memory_allocated; | |
963 | } | |
964 | ||
965 | static inline int sk_wmem_schedule(struct sock *sk, int size) | |
966 | { | |
967 | if (!sk_has_account(sk)) | |
968 | return 1; | |
969 | return size <= sk->sk_forward_alloc || | |
970 | __sk_mem_schedule(sk, size, SK_MEM_SEND); | |
971 | } | |
972 | ||
973 | static inline int sk_rmem_schedule(struct sock *sk, int size) | |
974 | { | |
975 | if (!sk_has_account(sk)) | |
976 | return 1; | |
977 | return size <= sk->sk_forward_alloc || | |
978 | __sk_mem_schedule(sk, size, SK_MEM_RECV); | |
979 | } | |
980 | ||
981 | static inline void sk_mem_reclaim(struct sock *sk) | |
982 | { | |
983 | if (!sk_has_account(sk)) | |
984 | return; | |
985 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) | |
986 | __sk_mem_reclaim(sk); | |
987 | } | |
988 | ||
989 | static inline void sk_mem_reclaim_partial(struct sock *sk) | |
990 | { | |
991 | if (!sk_has_account(sk)) | |
992 | return; | |
993 | if (sk->sk_forward_alloc > SK_MEM_QUANTUM) | |
994 | __sk_mem_reclaim(sk); | |
995 | } | |
996 | ||
997 | static inline void sk_mem_charge(struct sock *sk, int size) | |
998 | { | |
999 | if (!sk_has_account(sk)) | |
1000 | return; | |
1001 | sk->sk_forward_alloc -= size; | |
1002 | } | |
1003 | ||
1004 | static inline void sk_mem_uncharge(struct sock *sk, int size) | |
1005 | { | |
1006 | if (!sk_has_account(sk)) | |
1007 | return; | |
1008 | sk->sk_forward_alloc += size; | |
1009 | } | |
1010 | ||
1011 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) | |
1012 | { | |
1013 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); | |
1014 | sk->sk_wmem_queued -= skb->truesize; | |
1015 | sk_mem_uncharge(sk, skb->truesize); | |
1016 | __kfree_skb(skb); | |
1017 | } | |
1018 | ||
1019 | /* Used by processes to "lock" a socket state, so that | |
1020 | * interrupts and bottom half handlers won't change it | |
1021 | * from under us. It essentially blocks any incoming | |
1022 | * packets, so that we won't get any new data or any | |
1023 | * packets that change the state of the socket. | |
1024 | * | |
1025 | * While locked, BH processing will add new packets to | |
1026 | * the backlog queue. This queue is processed by the | |
1027 | * owner of the socket lock right before it is released. | |
1028 | * | |
1029 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
1030 | * accesses from user process context. | |
1031 | */ | |
1032 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) | |
1033 | ||
1034 | /* | |
1035 | * Macro so as to not evaluate some arguments when | |
1036 | * lockdep is not enabled. | |
1037 | * | |
1038 | * Mark both the sk_lock and the sk_lock.slock as a | |
1039 | * per-address-family lock class. | |
1040 | */ | |
1041 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ | |
1042 | do { \ | |
1043 | sk->sk_lock.owned = 0; \ | |
1044 | init_waitqueue_head(&sk->sk_lock.wq); \ | |
1045 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
1046 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
1047 | sizeof((sk)->sk_lock)); \ | |
1048 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
1049 | (skey), (sname)); \ | |
1050 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ | |
1051 | } while (0) | |
1052 | ||
1053 | extern void lock_sock_nested(struct sock *sk, int subclass); | |
1054 | ||
1055 | static inline void lock_sock(struct sock *sk) | |
1056 | { | |
1057 | lock_sock_nested(sk, 0); | |
1058 | } | |
1059 | ||
1060 | extern void release_sock(struct sock *sk); | |
1061 | ||
1062 | /* BH context may only use the following locking interface. */ | |
1063 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
1064 | #define bh_lock_sock_nested(__sk) \ | |
1065 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
1066 | SINGLE_DEPTH_NESTING) | |
1067 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) | |
1068 | ||
1069 | extern bool lock_sock_fast(struct sock *sk); | |
1070 | /** | |
1071 | * unlock_sock_fast - complement of lock_sock_fast | |
1072 | * @sk: socket | |
1073 | * @slow: slow mode | |
1074 | * | |
1075 | * fast unlock socket for user context. | |
1076 | * If slow mode is on, we call regular release_sock() | |
1077 | */ | |
1078 | static inline void unlock_sock_fast(struct sock *sk, bool slow) | |
1079 | { | |
1080 | if (slow) | |
1081 | release_sock(sk); | |
1082 | else | |
1083 | spin_unlock_bh(&sk->sk_lock.slock); | |
1084 | } | |
1085 | ||
1086 | ||
1087 | extern struct sock *sk_alloc(struct net *net, int family, | |
1088 | gfp_t priority, | |
1089 | struct proto *prot); | |
1090 | extern void sk_free(struct sock *sk); | |
1091 | extern void sk_release_kernel(struct sock *sk); | |
1092 | extern struct sock *sk_clone(const struct sock *sk, | |
1093 | const gfp_t priority); | |
1094 | ||
1095 | extern struct sk_buff *sock_wmalloc(struct sock *sk, | |
1096 | unsigned long size, int force, | |
1097 | gfp_t priority); | |
1098 | extern struct sk_buff *sock_rmalloc(struct sock *sk, | |
1099 | unsigned long size, int force, | |
1100 | gfp_t priority); | |
1101 | extern void sock_wfree(struct sk_buff *skb); | |
1102 | extern void sock_rfree(struct sk_buff *skb); | |
1103 | ||
1104 | extern int sock_setsockopt(struct socket *sock, int level, | |
1105 | int op, char __user *optval, | |
1106 | unsigned int optlen); | |
1107 | ||
1108 | extern int sock_getsockopt(struct socket *sock, int level, | |
1109 | int op, char __user *optval, | |
1110 | int __user *optlen); | |
1111 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, | |
1112 | unsigned long size, | |
1113 | int noblock, | |
1114 | int *errcode); | |
1115 | extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk, | |
1116 | unsigned long header_len, | |
1117 | unsigned long data_len, | |
1118 | int noblock, | |
1119 | int *errcode); | |
1120 | extern void *sock_kmalloc(struct sock *sk, int size, | |
1121 | gfp_t priority); | |
1122 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); | |
1123 | extern void sk_send_sigurg(struct sock *sk); | |
1124 | ||
1125 | #ifdef CONFIG_CGROUPS | |
1126 | extern void sock_update_classid(struct sock *sk); | |
1127 | #else | |
1128 | static inline void sock_update_classid(struct sock *sk) | |
1129 | { | |
1130 | } | |
1131 | #endif | |
1132 | ||
1133 | /* | |
1134 | * Functions to fill in entries in struct proto_ops when a protocol | |
1135 | * does not implement a particular function. | |
1136 | */ | |
1137 | extern int sock_no_bind(struct socket *, | |
1138 | struct sockaddr *, int); | |
1139 | extern int sock_no_connect(struct socket *, | |
1140 | struct sockaddr *, int, int); | |
1141 | extern int sock_no_socketpair(struct socket *, | |
1142 | struct socket *); | |
1143 | extern int sock_no_accept(struct socket *, | |
1144 | struct socket *, int); | |
1145 | extern int sock_no_getname(struct socket *, | |
1146 | struct sockaddr *, int *, int); | |
1147 | extern unsigned int sock_no_poll(struct file *, struct socket *, | |
1148 | struct poll_table_struct *); | |
1149 | extern int sock_no_ioctl(struct socket *, unsigned int, | |
1150 | unsigned long); | |
1151 | extern int sock_no_listen(struct socket *, int); | |
1152 | extern int sock_no_shutdown(struct socket *, int); | |
1153 | extern int sock_no_getsockopt(struct socket *, int , int, | |
1154 | char __user *, int __user *); | |
1155 | extern int sock_no_setsockopt(struct socket *, int, int, | |
1156 | char __user *, unsigned int); | |
1157 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, | |
1158 | struct msghdr *, size_t); | |
1159 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, | |
1160 | struct msghdr *, size_t, int); | |
1161 | extern int sock_no_mmap(struct file *file, | |
1162 | struct socket *sock, | |
1163 | struct vm_area_struct *vma); | |
1164 | extern ssize_t sock_no_sendpage(struct socket *sock, | |
1165 | struct page *page, | |
1166 | int offset, size_t size, | |
1167 | int flags); | |
1168 | ||
1169 | /* | |
1170 | * Functions to fill in entries in struct proto_ops when a protocol | |
1171 | * uses the inet style. | |
1172 | */ | |
1173 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, | |
1174 | char __user *optval, int __user *optlen); | |
1175 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, | |
1176 | struct msghdr *msg, size_t size, int flags); | |
1177 | extern int sock_common_setsockopt(struct socket *sock, int level, int optname, | |
1178 | char __user *optval, unsigned int optlen); | |
1179 | extern int compat_sock_common_getsockopt(struct socket *sock, int level, | |
1180 | int optname, char __user *optval, int __user *optlen); | |
1181 | extern int compat_sock_common_setsockopt(struct socket *sock, int level, | |
1182 | int optname, char __user *optval, unsigned int optlen); | |
1183 | ||
1184 | extern void sk_common_release(struct sock *sk); | |
1185 | ||
1186 | /* | |
1187 | * Default socket callbacks and setup code | |
1188 | */ | |
1189 | ||
1190 | /* Initialise core socket variables */ | |
1191 | extern void sock_init_data(struct socket *sock, struct sock *sk); | |
1192 | ||
1193 | extern void sk_filter_release_rcu(struct rcu_head *rcu); | |
1194 | ||
1195 | /** | |
1196 | * sk_filter_release - release a socket filter | |
1197 | * @fp: filter to remove | |
1198 | * | |
1199 | * Remove a filter from a socket and release its resources. | |
1200 | */ | |
1201 | ||
1202 | static inline void sk_filter_release(struct sk_filter *fp) | |
1203 | { | |
1204 | if (atomic_dec_and_test(&fp->refcnt)) | |
1205 | call_rcu(&fp->rcu, sk_filter_release_rcu); | |
1206 | } | |
1207 | ||
1208 | static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp) | |
1209 | { | |
1210 | unsigned int size = sk_filter_len(fp); | |
1211 | ||
1212 | atomic_sub(size, &sk->sk_omem_alloc); | |
1213 | sk_filter_release(fp); | |
1214 | } | |
1215 | ||
1216 | static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) | |
1217 | { | |
1218 | atomic_inc(&fp->refcnt); | |
1219 | atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); | |
1220 | } | |
1221 | ||
1222 | /* | |
1223 | * Socket reference counting postulates. | |
1224 | * | |
1225 | * * Each user of socket SHOULD hold a reference count. | |
1226 | * * Each access point to socket (an hash table bucket, reference from a list, | |
1227 | * running timer, skb in flight MUST hold a reference count. | |
1228 | * * When reference count hits 0, it means it will never increase back. | |
1229 | * * When reference count hits 0, it means that no references from | |
1230 | * outside exist to this socket and current process on current CPU | |
1231 | * is last user and may/should destroy this socket. | |
1232 | * * sk_free is called from any context: process, BH, IRQ. When | |
1233 | * it is called, socket has no references from outside -> sk_free | |
1234 | * may release descendant resources allocated by the socket, but | |
1235 | * to the time when it is called, socket is NOT referenced by any | |
1236 | * hash tables, lists etc. | |
1237 | * * Packets, delivered from outside (from network or from another process) | |
1238 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1239 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1240 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1241 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1242 | * (leak to backlog). Packet socket does all the processing inside | |
1243 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1244 | * use separate SMP lock, so that they are prone too. | |
1245 | */ | |
1246 | ||
1247 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1248 | static inline void sock_put(struct sock *sk) | |
1249 | { | |
1250 | if (atomic_dec_and_test(&sk->sk_refcnt)) | |
1251 | sk_free(sk); | |
1252 | } | |
1253 | ||
1254 | extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, | |
1255 | const int nested); | |
1256 | ||
1257 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) | |
1258 | { | |
1259 | sk->sk_tx_queue_mapping = tx_queue; | |
1260 | } | |
1261 | ||
1262 | static inline void sk_tx_queue_clear(struct sock *sk) | |
1263 | { | |
1264 | sk->sk_tx_queue_mapping = -1; | |
1265 | } | |
1266 | ||
1267 | static inline int sk_tx_queue_get(const struct sock *sk) | |
1268 | { | |
1269 | return sk ? sk->sk_tx_queue_mapping : -1; | |
1270 | } | |
1271 | ||
1272 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) | |
1273 | { | |
1274 | sk_tx_queue_clear(sk); | |
1275 | sk->sk_socket = sock; | |
1276 | } | |
1277 | ||
1278 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) | |
1279 | { | |
1280 | BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); | |
1281 | return &rcu_dereference_raw(sk->sk_wq)->wait; | |
1282 | } | |
1283 | /* Detach socket from process context. | |
1284 | * Announce socket dead, detach it from wait queue and inode. | |
1285 | * Note that parent inode held reference count on this struct sock, | |
1286 | * we do not release it in this function, because protocol | |
1287 | * probably wants some additional cleanups or even continuing | |
1288 | * to work with this socket (TCP). | |
1289 | */ | |
1290 | static inline void sock_orphan(struct sock *sk) | |
1291 | { | |
1292 | write_lock_bh(&sk->sk_callback_lock); | |
1293 | sock_set_flag(sk, SOCK_DEAD); | |
1294 | sk_set_socket(sk, NULL); | |
1295 | sk->sk_wq = NULL; | |
1296 | write_unlock_bh(&sk->sk_callback_lock); | |
1297 | } | |
1298 | ||
1299 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
1300 | { | |
1301 | write_lock_bh(&sk->sk_callback_lock); | |
1302 | sk->sk_wq = parent->wq; | |
1303 | parent->sk = sk; | |
1304 | sk_set_socket(sk, parent); | |
1305 | security_sock_graft(sk, parent); | |
1306 | write_unlock_bh(&sk->sk_callback_lock); | |
1307 | } | |
1308 | ||
1309 | extern int sock_i_uid(struct sock *sk); | |
1310 | extern unsigned long sock_i_ino(struct sock *sk); | |
1311 | ||
1312 | static inline struct dst_entry * | |
1313 | __sk_dst_get(struct sock *sk) | |
1314 | { | |
1315 | return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) || | |
1316 | lockdep_is_held(&sk->sk_lock.slock)); | |
1317 | } | |
1318 | ||
1319 | static inline struct dst_entry * | |
1320 | sk_dst_get(struct sock *sk) | |
1321 | { | |
1322 | struct dst_entry *dst; | |
1323 | ||
1324 | rcu_read_lock(); | |
1325 | dst = rcu_dereference(sk->sk_dst_cache); | |
1326 | if (dst) | |
1327 | dst_hold(dst); | |
1328 | rcu_read_unlock(); | |
1329 | return dst; | |
1330 | } | |
1331 | ||
1332 | extern void sk_reset_txq(struct sock *sk); | |
1333 | ||
1334 | static inline void dst_negative_advice(struct sock *sk) | |
1335 | { | |
1336 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); | |
1337 | ||
1338 | if (dst && dst->ops->negative_advice) { | |
1339 | ndst = dst->ops->negative_advice(dst); | |
1340 | ||
1341 | if (ndst != dst) { | |
1342 | rcu_assign_pointer(sk->sk_dst_cache, ndst); | |
1343 | sk_reset_txq(sk); | |
1344 | } | |
1345 | } | |
1346 | } | |
1347 | ||
1348 | static inline void | |
1349 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1350 | { | |
1351 | struct dst_entry *old_dst; | |
1352 | ||
1353 | sk_tx_queue_clear(sk); | |
1354 | /* | |
1355 | * This can be called while sk is owned by the caller only, | |
1356 | * with no state that can be checked in a rcu_dereference_check() cond | |
1357 | */ | |
1358 | old_dst = rcu_dereference_raw(sk->sk_dst_cache); | |
1359 | rcu_assign_pointer(sk->sk_dst_cache, dst); | |
1360 | dst_release(old_dst); | |
1361 | } | |
1362 | ||
1363 | static inline void | |
1364 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1365 | { | |
1366 | spin_lock(&sk->sk_dst_lock); | |
1367 | __sk_dst_set(sk, dst); | |
1368 | spin_unlock(&sk->sk_dst_lock); | |
1369 | } | |
1370 | ||
1371 | static inline void | |
1372 | __sk_dst_reset(struct sock *sk) | |
1373 | { | |
1374 | __sk_dst_set(sk, NULL); | |
1375 | } | |
1376 | ||
1377 | static inline void | |
1378 | sk_dst_reset(struct sock *sk) | |
1379 | { | |
1380 | spin_lock(&sk->sk_dst_lock); | |
1381 | __sk_dst_reset(sk); | |
1382 | spin_unlock(&sk->sk_dst_lock); | |
1383 | } | |
1384 | ||
1385 | extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); | |
1386 | ||
1387 | extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); | |
1388 | ||
1389 | static inline int sk_can_gso(const struct sock *sk) | |
1390 | { | |
1391 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
1392 | } | |
1393 | ||
1394 | extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst); | |
1395 | ||
1396 | static inline void sk_nocaps_add(struct sock *sk, int flags) | |
1397 | { | |
1398 | sk->sk_route_nocaps |= flags; | |
1399 | sk->sk_route_caps &= ~flags; | |
1400 | } | |
1401 | ||
1402 | static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, | |
1403 | char __user *from, char *to, | |
1404 | int copy, int offset) | |
1405 | { | |
1406 | if (skb->ip_summed == CHECKSUM_NONE) { | |
1407 | int err = 0; | |
1408 | __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err); | |
1409 | if (err) | |
1410 | return err; | |
1411 | skb->csum = csum_block_add(skb->csum, csum, offset); | |
1412 | } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { | |
1413 | if (!access_ok(VERIFY_READ, from, copy) || | |
1414 | __copy_from_user_nocache(to, from, copy)) | |
1415 | return -EFAULT; | |
1416 | } else if (copy_from_user(to, from, copy)) | |
1417 | return -EFAULT; | |
1418 | ||
1419 | return 0; | |
1420 | } | |
1421 | ||
1422 | static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, | |
1423 | char __user *from, int copy) | |
1424 | { | |
1425 | int err, offset = skb->len; | |
1426 | ||
1427 | err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), | |
1428 | copy, offset); | |
1429 | if (err) | |
1430 | __skb_trim(skb, offset); | |
1431 | ||
1432 | return err; | |
1433 | } | |
1434 | ||
1435 | static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from, | |
1436 | struct sk_buff *skb, | |
1437 | struct page *page, | |
1438 | int off, int copy) | |
1439 | { | |
1440 | int err; | |
1441 | ||
1442 | err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, | |
1443 | copy, skb->len); | |
1444 | if (err) | |
1445 | return err; | |
1446 | ||
1447 | skb->len += copy; | |
1448 | skb->data_len += copy; | |
1449 | skb->truesize += copy; | |
1450 | sk->sk_wmem_queued += copy; | |
1451 | sk_mem_charge(sk, copy); | |
1452 | return 0; | |
1453 | } | |
1454 | ||
1455 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, | |
1456 | struct sk_buff *skb, struct page *page, | |
1457 | int off, int copy) | |
1458 | { | |
1459 | if (skb->ip_summed == CHECKSUM_NONE) { | |
1460 | int err = 0; | |
1461 | __wsum csum = csum_and_copy_from_user(from, | |
1462 | page_address(page) + off, | |
1463 | copy, 0, &err); | |
1464 | if (err) | |
1465 | return err; | |
1466 | skb->csum = csum_block_add(skb->csum, csum, skb->len); | |
1467 | } else if (copy_from_user(page_address(page) + off, from, copy)) | |
1468 | return -EFAULT; | |
1469 | ||
1470 | skb->len += copy; | |
1471 | skb->data_len += copy; | |
1472 | skb->truesize += copy; | |
1473 | sk->sk_wmem_queued += copy; | |
1474 | sk_mem_charge(sk, copy); | |
1475 | return 0; | |
1476 | } | |
1477 | ||
1478 | /** | |
1479 | * sk_wmem_alloc_get - returns write allocations | |
1480 | * @sk: socket | |
1481 | * | |
1482 | * Returns sk_wmem_alloc minus initial offset of one | |
1483 | */ | |
1484 | static inline int sk_wmem_alloc_get(const struct sock *sk) | |
1485 | { | |
1486 | return atomic_read(&sk->sk_wmem_alloc) - 1; | |
1487 | } | |
1488 | ||
1489 | /** | |
1490 | * sk_rmem_alloc_get - returns read allocations | |
1491 | * @sk: socket | |
1492 | * | |
1493 | * Returns sk_rmem_alloc | |
1494 | */ | |
1495 | static inline int sk_rmem_alloc_get(const struct sock *sk) | |
1496 | { | |
1497 | return atomic_read(&sk->sk_rmem_alloc); | |
1498 | } | |
1499 | ||
1500 | /** | |
1501 | * sk_has_allocations - check if allocations are outstanding | |
1502 | * @sk: socket | |
1503 | * | |
1504 | * Returns true if socket has write or read allocations | |
1505 | */ | |
1506 | static inline int sk_has_allocations(const struct sock *sk) | |
1507 | { | |
1508 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); | |
1509 | } | |
1510 | ||
1511 | /** | |
1512 | * wq_has_sleeper - check if there are any waiting processes | |
1513 | * @wq: struct socket_wq | |
1514 | * | |
1515 | * Returns true if socket_wq has waiting processes | |
1516 | * | |
1517 | * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory | |
1518 | * barrier call. They were added due to the race found within the tcp code. | |
1519 | * | |
1520 | * Consider following tcp code paths: | |
1521 | * | |
1522 | * CPU1 CPU2 | |
1523 | * | |
1524 | * sys_select receive packet | |
1525 | * ... ... | |
1526 | * __add_wait_queue update tp->rcv_nxt | |
1527 | * ... ... | |
1528 | * tp->rcv_nxt check sock_def_readable | |
1529 | * ... { | |
1530 | * schedule rcu_read_lock(); | |
1531 | * wq = rcu_dereference(sk->sk_wq); | |
1532 | * if (wq && waitqueue_active(&wq->wait)) | |
1533 | * wake_up_interruptible(&wq->wait) | |
1534 | * ... | |
1535 | * } | |
1536 | * | |
1537 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay | |
1538 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 | |
1539 | * could then endup calling schedule and sleep forever if there are no more | |
1540 | * data on the socket. | |
1541 | * | |
1542 | */ | |
1543 | static inline bool wq_has_sleeper(struct socket_wq *wq) | |
1544 | { | |
1545 | ||
1546 | /* | |
1547 | * We need to be sure we are in sync with the | |
1548 | * add_wait_queue modifications to the wait queue. | |
1549 | * | |
1550 | * This memory barrier is paired in the sock_poll_wait. | |
1551 | */ | |
1552 | smp_mb(); | |
1553 | return wq && waitqueue_active(&wq->wait); | |
1554 | } | |
1555 | ||
1556 | /** | |
1557 | * sock_poll_wait - place memory barrier behind the poll_wait call. | |
1558 | * @filp: file | |
1559 | * @wait_address: socket wait queue | |
1560 | * @p: poll_table | |
1561 | * | |
1562 | * See the comments in the wq_has_sleeper function. | |
1563 | */ | |
1564 | static inline void sock_poll_wait(struct file *filp, | |
1565 | wait_queue_head_t *wait_address, poll_table *p) | |
1566 | { | |
1567 | if (p && wait_address) { | |
1568 | poll_wait(filp, wait_address, p); | |
1569 | /* | |
1570 | * We need to be sure we are in sync with the | |
1571 | * socket flags modification. | |
1572 | * | |
1573 | * This memory barrier is paired in the wq_has_sleeper. | |
1574 | */ | |
1575 | smp_mb(); | |
1576 | } | |
1577 | } | |
1578 | ||
1579 | /* | |
1580 | * Queue a received datagram if it will fit. Stream and sequenced | |
1581 | * protocols can't normally use this as they need to fit buffers in | |
1582 | * and play with them. | |
1583 | * | |
1584 | * Inlined as it's very short and called for pretty much every | |
1585 | * packet ever received. | |
1586 | */ | |
1587 | ||
1588 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) | |
1589 | { | |
1590 | skb_orphan(skb); | |
1591 | skb->sk = sk; | |
1592 | skb->destructor = sock_wfree; | |
1593 | /* | |
1594 | * We used to take a refcount on sk, but following operation | |
1595 | * is enough to guarantee sk_free() wont free this sock until | |
1596 | * all in-flight packets are completed | |
1597 | */ | |
1598 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); | |
1599 | } | |
1600 | ||
1601 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
1602 | { | |
1603 | skb_orphan(skb); | |
1604 | skb->sk = sk; | |
1605 | skb->destructor = sock_rfree; | |
1606 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
1607 | sk_mem_charge(sk, skb->truesize); | |
1608 | } | |
1609 | ||
1610 | extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, | |
1611 | unsigned long expires); | |
1612 | ||
1613 | extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); | |
1614 | ||
1615 | extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); | |
1616 | ||
1617 | extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); | |
1618 | ||
1619 | /* | |
1620 | * Recover an error report and clear atomically | |
1621 | */ | |
1622 | ||
1623 | static inline int sock_error(struct sock *sk) | |
1624 | { | |
1625 | int err; | |
1626 | if (likely(!sk->sk_err)) | |
1627 | return 0; | |
1628 | err = xchg(&sk->sk_err, 0); | |
1629 | return -err; | |
1630 | } | |
1631 | ||
1632 | static inline unsigned long sock_wspace(struct sock *sk) | |
1633 | { | |
1634 | int amt = 0; | |
1635 | ||
1636 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
1637 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); | |
1638 | if (amt < 0) | |
1639 | amt = 0; | |
1640 | } | |
1641 | return amt; | |
1642 | } | |
1643 | ||
1644 | static inline void sk_wake_async(struct sock *sk, int how, int band) | |
1645 | { | |
1646 | if (sock_flag(sk, SOCK_FASYNC)) | |
1647 | sock_wake_async(sk->sk_socket, how, band); | |
1648 | } | |
1649 | ||
1650 | #define SOCK_MIN_SNDBUF 2048 | |
1651 | /* | |
1652 | * Since sk_rmem_alloc sums skb->truesize, even a small frame might need | |
1653 | * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak | |
1654 | */ | |
1655 | #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff)) | |
1656 | ||
1657 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
1658 | { | |
1659 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { | |
1660 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); | |
1661 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); | |
1662 | } | |
1663 | } | |
1664 | ||
1665 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp); | |
1666 | ||
1667 | static inline struct page *sk_stream_alloc_page(struct sock *sk) | |
1668 | { | |
1669 | struct page *page = NULL; | |
1670 | ||
1671 | page = alloc_pages(sk->sk_allocation, 0); | |
1672 | if (!page) { | |
1673 | sk->sk_prot->enter_memory_pressure(sk); | |
1674 | sk_stream_moderate_sndbuf(sk); | |
1675 | } | |
1676 | return page; | |
1677 | } | |
1678 | ||
1679 | /* | |
1680 | * Default write policy as shown to user space via poll/select/SIGIO | |
1681 | */ | |
1682 | static inline int sock_writeable(const struct sock *sk) | |
1683 | { | |
1684 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); | |
1685 | } | |
1686 | ||
1687 | static inline gfp_t gfp_any(void) | |
1688 | { | |
1689 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; | |
1690 | } | |
1691 | ||
1692 | static inline long sock_rcvtimeo(const struct sock *sk, int noblock) | |
1693 | { | |
1694 | return noblock ? 0 : sk->sk_rcvtimeo; | |
1695 | } | |
1696 | ||
1697 | static inline long sock_sndtimeo(const struct sock *sk, int noblock) | |
1698 | { | |
1699 | return noblock ? 0 : sk->sk_sndtimeo; | |
1700 | } | |
1701 | ||
1702 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
1703 | { | |
1704 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; | |
1705 | } | |
1706 | ||
1707 | /* Alas, with timeout socket operations are not restartable. | |
1708 | * Compare this to poll(). | |
1709 | */ | |
1710 | static inline int sock_intr_errno(long timeo) | |
1711 | { | |
1712 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
1713 | } | |
1714 | ||
1715 | extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, | |
1716 | struct sk_buff *skb); | |
1717 | ||
1718 | static __inline__ void | |
1719 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) | |
1720 | { | |
1721 | ktime_t kt = skb->tstamp; | |
1722 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); | |
1723 | ||
1724 | /* | |
1725 | * generate control messages if | |
1726 | * - receive time stamping in software requested (SOCK_RCVTSTAMP | |
1727 | * or SOCK_TIMESTAMPING_RX_SOFTWARE) | |
1728 | * - software time stamp available and wanted | |
1729 | * (SOCK_TIMESTAMPING_SOFTWARE) | |
1730 | * - hardware time stamps available and wanted | |
1731 | * (SOCK_TIMESTAMPING_SYS_HARDWARE or | |
1732 | * SOCK_TIMESTAMPING_RAW_HARDWARE) | |
1733 | */ | |
1734 | if (sock_flag(sk, SOCK_RCVTSTAMP) || | |
1735 | sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) || | |
1736 | (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) || | |
1737 | (hwtstamps->hwtstamp.tv64 && | |
1738 | sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) || | |
1739 | (hwtstamps->syststamp.tv64 && | |
1740 | sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))) | |
1741 | __sock_recv_timestamp(msg, sk, skb); | |
1742 | else | |
1743 | sk->sk_stamp = kt; | |
1744 | } | |
1745 | ||
1746 | extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, | |
1747 | struct sk_buff *skb); | |
1748 | ||
1749 | static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, | |
1750 | struct sk_buff *skb) | |
1751 | { | |
1752 | #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ | |
1753 | (1UL << SOCK_RCVTSTAMP) | \ | |
1754 | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \ | |
1755 | (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \ | |
1756 | (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \ | |
1757 | (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE)) | |
1758 | ||
1759 | if (sk->sk_flags & FLAGS_TS_OR_DROPS) | |
1760 | __sock_recv_ts_and_drops(msg, sk, skb); | |
1761 | else | |
1762 | sk->sk_stamp = skb->tstamp; | |
1763 | } | |
1764 | ||
1765 | /** | |
1766 | * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped | |
1767 | * @sk: socket sending this packet | |
1768 | * @tx_flags: filled with instructions for time stamping | |
1769 | * | |
1770 | * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if | |
1771 | * parameters are invalid. | |
1772 | */ | |
1773 | extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags); | |
1774 | ||
1775 | /** | |
1776 | * sk_eat_skb - Release a skb if it is no longer needed | |
1777 | * @sk: socket to eat this skb from | |
1778 | * @skb: socket buffer to eat | |
1779 | * @copied_early: flag indicating whether DMA operations copied this data early | |
1780 | * | |
1781 | * This routine must be called with interrupts disabled or with the socket | |
1782 | * locked so that the sk_buff queue operation is ok. | |
1783 | */ | |
1784 | #ifdef CONFIG_NET_DMA | |
1785 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) | |
1786 | { | |
1787 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1788 | if (!copied_early) | |
1789 | __kfree_skb(skb); | |
1790 | else | |
1791 | __skb_queue_tail(&sk->sk_async_wait_queue, skb); | |
1792 | } | |
1793 | #else | |
1794 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) | |
1795 | { | |
1796 | __skb_unlink(skb, &sk->sk_receive_queue); | |
1797 | __kfree_skb(skb); | |
1798 | } | |
1799 | #endif | |
1800 | ||
1801 | static inline | |
1802 | struct net *sock_net(const struct sock *sk) | |
1803 | { | |
1804 | return read_pnet(&sk->sk_net); | |
1805 | } | |
1806 | ||
1807 | static inline | |
1808 | void sock_net_set(struct sock *sk, struct net *net) | |
1809 | { | |
1810 | write_pnet(&sk->sk_net, net); | |
1811 | } | |
1812 | ||
1813 | /* | |
1814 | * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace. | |
1815 | * They should not hold a reference to a namespace in order to allow | |
1816 | * to stop it. | |
1817 | * Sockets after sk_change_net should be released using sk_release_kernel | |
1818 | */ | |
1819 | static inline void sk_change_net(struct sock *sk, struct net *net) | |
1820 | { | |
1821 | put_net(sock_net(sk)); | |
1822 | sock_net_set(sk, hold_net(net)); | |
1823 | } | |
1824 | ||
1825 | static inline struct sock *skb_steal_sock(struct sk_buff *skb) | |
1826 | { | |
1827 | if (unlikely(skb->sk)) { | |
1828 | struct sock *sk = skb->sk; | |
1829 | ||
1830 | skb->destructor = NULL; | |
1831 | skb->sk = NULL; | |
1832 | return sk; | |
1833 | } | |
1834 | return NULL; | |
1835 | } | |
1836 | ||
1837 | extern void sock_enable_timestamp(struct sock *sk, int flag); | |
1838 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); | |
1839 | extern int sock_get_timestampns(struct sock *, struct timespec __user *); | |
1840 | ||
1841 | /* | |
1842 | * Enable debug/info messages | |
1843 | */ | |
1844 | extern int net_msg_warn; | |
1845 | #define NETDEBUG(fmt, args...) \ | |
1846 | do { if (net_msg_warn) printk(fmt,##args); } while (0) | |
1847 | ||
1848 | #define LIMIT_NETDEBUG(fmt, args...) \ | |
1849 | do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) | |
1850 | ||
1851 | extern __u32 sysctl_wmem_max; | |
1852 | extern __u32 sysctl_rmem_max; | |
1853 | ||
1854 | extern void sk_init(void); | |
1855 | ||
1856 | extern int sysctl_optmem_max; | |
1857 | ||
1858 | extern __u32 sysctl_wmem_default; | |
1859 | extern __u32 sysctl_rmem_default; | |
1860 | ||
1861 | #endif /* _SOCK_H */ |