Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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 | * |
Jesper Juhl | 02c30a8 | 2005-05-05 16:16:16 -0700 | [diff] [blame] | 10 | * Authors: Ross Biro |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
| 12 | * Corey Minyard <wf-rch!minyard@relay.EU.net> |
| 13 | * Florian La Roche <flla@stud.uni-sb.de> |
| 14 | * |
| 15 | * Fixes: |
| 16 | * Alan Cox : Volatiles in skbuff pointers. See |
| 17 | * skbuff comments. May be overdone, |
| 18 | * better to prove they can be removed |
| 19 | * than the reverse. |
| 20 | * Alan Cox : Added a zapped field for tcp to note |
| 21 | * a socket is reset and must stay shut up |
| 22 | * Alan Cox : New fields for options |
| 23 | * Pauline Middelink : identd support |
| 24 | * Alan Cox : Eliminate low level recv/recvfrom |
| 25 | * David S. Miller : New socket lookup architecture. |
| 26 | * Steve Whitehouse: Default routines for sock_ops |
| 27 | * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made |
| 28 | * protinfo be just a void pointer, as the |
| 29 | * protocol specific parts were moved to |
| 30 | * respective headers and ipv4/v6, etc now |
| 31 | * use private slabcaches for its socks |
| 32 | * Pedro Hortas : New flags field for socket options |
| 33 | * |
| 34 | * |
| 35 | * This program is free software; you can redistribute it and/or |
| 36 | * modify it under the terms of the GNU General Public License |
| 37 | * as published by the Free Software Foundation; either version |
| 38 | * 2 of the License, or (at your option) any later version. |
| 39 | */ |
| 40 | #ifndef _SOCK_H |
| 41 | #define _SOCK_H |
| 42 | |
| 43 | #include <linux/config.h> |
| 44 | #include <linux/list.h> |
| 45 | #include <linux/timer.h> |
| 46 | #include <linux/cache.h> |
| 47 | #include <linux/module.h> |
| 48 | #include <linux/netdevice.h> |
| 49 | #include <linux/skbuff.h> /* struct sk_buff */ |
| 50 | #include <linux/security.h> |
| 51 | |
| 52 | #include <linux/filter.h> |
| 53 | |
| 54 | #include <asm/atomic.h> |
| 55 | #include <net/dst.h> |
| 56 | #include <net/checksum.h> |
| 57 | |
| 58 | /* |
| 59 | * This structure really needs to be cleaned up. |
| 60 | * Most of it is for TCP, and not used by any of |
| 61 | * the other protocols. |
| 62 | */ |
| 63 | |
| 64 | /* Define this to get the SOCK_DBG debugging facility. */ |
| 65 | #define SOCK_DEBUGGING |
| 66 | #ifdef SOCK_DEBUGGING |
| 67 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ |
| 68 | printk(KERN_DEBUG msg); } while (0) |
| 69 | #else |
| 70 | #define SOCK_DEBUG(sk, msg...) do { } while (0) |
| 71 | #endif |
| 72 | |
| 73 | /* This is the per-socket lock. The spinlock provides a synchronization |
| 74 | * between user contexts and software interrupt processing, whereas the |
| 75 | * mini-semaphore synchronizes multiple users amongst themselves. |
| 76 | */ |
| 77 | struct sock_iocb; |
| 78 | typedef struct { |
| 79 | spinlock_t slock; |
| 80 | struct sock_iocb *owner; |
| 81 | wait_queue_head_t wq; |
| 82 | } socket_lock_t; |
| 83 | |
| 84 | #define sock_lock_init(__sk) \ |
| 85 | do { spin_lock_init(&((__sk)->sk_lock.slock)); \ |
| 86 | (__sk)->sk_lock.owner = NULL; \ |
| 87 | init_waitqueue_head(&((__sk)->sk_lock.wq)); \ |
| 88 | } while(0) |
| 89 | |
| 90 | struct sock; |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 91 | struct proto; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 92 | |
| 93 | /** |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 94 | * struct sock_common - minimal network layer representation of sockets |
| 95 | * @skc_family: network address family |
| 96 | * @skc_state: Connection state |
| 97 | * @skc_reuse: %SO_REUSEADDR setting |
| 98 | * @skc_bound_dev_if: bound device index if != 0 |
| 99 | * @skc_node: main hash linkage for various protocol lookup tables |
| 100 | * @skc_bind_node: bind hash linkage for various protocol lookup tables |
| 101 | * @skc_refcnt: reference count |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 102 | * @skc_prot: protocol handlers inside a network family |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 103 | * |
| 104 | * This is the minimal network layer representation of sockets, the header |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 105 | * for struct sock and struct inet_timewait_sock. |
| 106 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 107 | struct sock_common { |
| 108 | unsigned short skc_family; |
| 109 | volatile unsigned char skc_state; |
| 110 | unsigned char skc_reuse; |
| 111 | int skc_bound_dev_if; |
| 112 | struct hlist_node skc_node; |
| 113 | struct hlist_node skc_bind_node; |
| 114 | atomic_t skc_refcnt; |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 115 | struct proto *skc_prot; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 116 | }; |
| 117 | |
| 118 | /** |
| 119 | * struct sock - network layer representation of sockets |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 120 | * @__sk_common: shared layout with inet_timewait_sock |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 121 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
| 122 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings |
| 123 | * @sk_lock: synchronizer |
| 124 | * @sk_rcvbuf: size of receive buffer in bytes |
| 125 | * @sk_sleep: sock wait queue |
| 126 | * @sk_dst_cache: destination cache |
| 127 | * @sk_dst_lock: destination cache lock |
| 128 | * @sk_policy: flow policy |
| 129 | * @sk_rmem_alloc: receive queue bytes committed |
| 130 | * @sk_receive_queue: incoming packets |
| 131 | * @sk_wmem_alloc: transmit queue bytes committed |
| 132 | * @sk_write_queue: Packet sending queue |
| 133 | * @sk_omem_alloc: "o" is "option" or "other" |
| 134 | * @sk_wmem_queued: persistent queue size |
| 135 | * @sk_forward_alloc: space allocated forward |
| 136 | * @sk_allocation: allocation mode |
| 137 | * @sk_sndbuf: size of send buffer in bytes |
| 138 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings |
| 139 | * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets |
| 140 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) |
| 141 | * @sk_lingertime: %SO_LINGER l_linger setting |
Arnaldo Carvalho de Melo | 6e04e02 | 2005-08-09 20:07:35 -0700 | [diff] [blame] | 142 | * @sk_hashent: hash entry in several tables (e.g. inet_hashinfo.ehash) |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 143 | * @sk_backlog: always used with the per-socket spinlock held |
| 144 | * @sk_callback_lock: used with the callbacks in the end of this struct |
| 145 | * @sk_error_queue: rarely used |
Arnaldo Carvalho de Melo | 476e19c | 2005-05-05 13:35:15 -0700 | [diff] [blame] | 146 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance) |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 147 | * @sk_err: last error |
| 148 | * @sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out' |
| 149 | * @sk_ack_backlog: current listen backlog |
| 150 | * @sk_max_ack_backlog: listen backlog set in listen() |
| 151 | * @sk_priority: %SO_PRIORITY setting |
| 152 | * @sk_type: socket type (%SOCK_STREAM, etc) |
| 153 | * @sk_protocol: which protocol this socket belongs in this network family |
| 154 | * @sk_peercred: %SO_PEERCRED setting |
| 155 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
| 156 | * @sk_rcvtimeo: %SO_RCVTIMEO setting |
| 157 | * @sk_sndtimeo: %SO_SNDTIMEO setting |
| 158 | * @sk_filter: socket filtering instructions |
| 159 | * @sk_protinfo: private area, net family specific, when not using slab |
| 160 | * @sk_timer: sock cleanup timer |
| 161 | * @sk_stamp: time stamp of last packet received |
| 162 | * @sk_socket: Identd and reporting IO signals |
| 163 | * @sk_user_data: RPC layer private data |
| 164 | * @sk_sndmsg_page: cached page for sendmsg |
| 165 | * @sk_sndmsg_off: cached offset for sendmsg |
| 166 | * @sk_send_head: front of stuff to transmit |
Martin Waitz | 67be2dd | 2005-05-01 08:59:26 -0700 | [diff] [blame] | 167 | * @sk_security: used by security modules |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 168 | * @sk_write_pending: a write to stream socket waits to start |
| 169 | * @sk_state_change: callback to indicate change in the state of the sock |
| 170 | * @sk_data_ready: callback to indicate there is data to be processed |
| 171 | * @sk_write_space: callback to indicate there is bf sending space available |
| 172 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) |
| 173 | * @sk_backlog_rcv: callback to process the backlog |
| 174 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 175 | */ |
| 176 | struct sock { |
| 177 | /* |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 178 | * Now struct inet_timewait_sock also uses sock_common, so please just |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 179 | * don't add nothing before this first member (__sk_common) --acme |
| 180 | */ |
| 181 | struct sock_common __sk_common; |
| 182 | #define sk_family __sk_common.skc_family |
| 183 | #define sk_state __sk_common.skc_state |
| 184 | #define sk_reuse __sk_common.skc_reuse |
| 185 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if |
| 186 | #define sk_node __sk_common.skc_node |
| 187 | #define sk_bind_node __sk_common.skc_bind_node |
| 188 | #define sk_refcnt __sk_common.skc_refcnt |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 189 | #define sk_prot __sk_common.skc_prot |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 190 | unsigned char sk_shutdown : 2, |
| 191 | sk_no_check : 2, |
| 192 | sk_userlocks : 4; |
| 193 | unsigned char sk_protocol; |
| 194 | unsigned short sk_type; |
| 195 | int sk_rcvbuf; |
| 196 | socket_lock_t sk_lock; |
| 197 | wait_queue_head_t *sk_sleep; |
| 198 | struct dst_entry *sk_dst_cache; |
| 199 | struct xfrm_policy *sk_policy[2]; |
| 200 | rwlock_t sk_dst_lock; |
| 201 | atomic_t sk_rmem_alloc; |
| 202 | atomic_t sk_wmem_alloc; |
| 203 | atomic_t sk_omem_alloc; |
| 204 | struct sk_buff_head sk_receive_queue; |
| 205 | struct sk_buff_head sk_write_queue; |
| 206 | int sk_wmem_queued; |
| 207 | int sk_forward_alloc; |
| 208 | unsigned int sk_allocation; |
| 209 | int sk_sndbuf; |
| 210 | int sk_route_caps; |
| 211 | int sk_hashent; |
| 212 | unsigned long sk_flags; |
| 213 | unsigned long sk_lingertime; |
| 214 | /* |
| 215 | * The backlog queue is special, it is always used with |
| 216 | * the per-socket spinlock held and requires low latency |
| 217 | * access. Therefore we special case it's implementation. |
| 218 | */ |
| 219 | struct { |
| 220 | struct sk_buff *head; |
| 221 | struct sk_buff *tail; |
| 222 | } sk_backlog; |
| 223 | struct sk_buff_head sk_error_queue; |
Arnaldo Carvalho de Melo | 476e19c | 2005-05-05 13:35:15 -0700 | [diff] [blame] | 224 | struct proto *sk_prot_creator; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 225 | rwlock_t sk_callback_lock; |
| 226 | int sk_err, |
| 227 | sk_err_soft; |
| 228 | unsigned short sk_ack_backlog; |
| 229 | unsigned short sk_max_ack_backlog; |
| 230 | __u32 sk_priority; |
| 231 | struct ucred sk_peercred; |
| 232 | int sk_rcvlowat; |
| 233 | long sk_rcvtimeo; |
| 234 | long sk_sndtimeo; |
| 235 | struct sk_filter *sk_filter; |
| 236 | void *sk_protinfo; |
| 237 | struct timer_list sk_timer; |
| 238 | struct timeval sk_stamp; |
| 239 | struct socket *sk_socket; |
| 240 | void *sk_user_data; |
| 241 | struct page *sk_sndmsg_page; |
| 242 | struct sk_buff *sk_send_head; |
| 243 | __u32 sk_sndmsg_off; |
| 244 | int sk_write_pending; |
| 245 | void *sk_security; |
| 246 | void (*sk_state_change)(struct sock *sk); |
| 247 | void (*sk_data_ready)(struct sock *sk, int bytes); |
| 248 | void (*sk_write_space)(struct sock *sk); |
| 249 | void (*sk_error_report)(struct sock *sk); |
| 250 | int (*sk_backlog_rcv)(struct sock *sk, |
| 251 | struct sk_buff *skb); |
| 252 | void (*sk_destruct)(struct sock *sk); |
| 253 | }; |
| 254 | |
| 255 | /* |
| 256 | * Hashed lists helper routines |
| 257 | */ |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 258 | static inline struct sock *__sk_head(const struct hlist_head *head) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 259 | { |
| 260 | return hlist_entry(head->first, struct sock, sk_node); |
| 261 | } |
| 262 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 263 | static inline struct sock *sk_head(const struct hlist_head *head) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 264 | { |
| 265 | return hlist_empty(head) ? NULL : __sk_head(head); |
| 266 | } |
| 267 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 268 | static inline struct sock *sk_next(const struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 269 | { |
| 270 | return sk->sk_node.next ? |
| 271 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; |
| 272 | } |
| 273 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 274 | static inline int sk_unhashed(const struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 275 | { |
| 276 | return hlist_unhashed(&sk->sk_node); |
| 277 | } |
| 278 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 279 | static inline int sk_hashed(const struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 280 | { |
| 281 | return sk->sk_node.pprev != NULL; |
| 282 | } |
| 283 | |
| 284 | static __inline__ void sk_node_init(struct hlist_node *node) |
| 285 | { |
| 286 | node->pprev = NULL; |
| 287 | } |
| 288 | |
| 289 | static __inline__ void __sk_del_node(struct sock *sk) |
| 290 | { |
| 291 | __hlist_del(&sk->sk_node); |
| 292 | } |
| 293 | |
| 294 | static __inline__ int __sk_del_node_init(struct sock *sk) |
| 295 | { |
| 296 | if (sk_hashed(sk)) { |
| 297 | __sk_del_node(sk); |
| 298 | sk_node_init(&sk->sk_node); |
| 299 | return 1; |
| 300 | } |
| 301 | return 0; |
| 302 | } |
| 303 | |
| 304 | /* Grab socket reference count. This operation is valid only |
| 305 | when sk is ALREADY grabbed f.e. it is found in hash table |
| 306 | or a list and the lookup is made under lock preventing hash table |
| 307 | modifications. |
| 308 | */ |
| 309 | |
| 310 | static inline void sock_hold(struct sock *sk) |
| 311 | { |
| 312 | atomic_inc(&sk->sk_refcnt); |
| 313 | } |
| 314 | |
| 315 | /* Ungrab socket in the context, which assumes that socket refcnt |
| 316 | cannot hit zero, f.e. it is true in context of any socketcall. |
| 317 | */ |
| 318 | static inline void __sock_put(struct sock *sk) |
| 319 | { |
| 320 | atomic_dec(&sk->sk_refcnt); |
| 321 | } |
| 322 | |
| 323 | static __inline__ int sk_del_node_init(struct sock *sk) |
| 324 | { |
| 325 | int rc = __sk_del_node_init(sk); |
| 326 | |
| 327 | if (rc) { |
| 328 | /* paranoid for a while -acme */ |
| 329 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
| 330 | __sock_put(sk); |
| 331 | } |
| 332 | return rc; |
| 333 | } |
| 334 | |
| 335 | static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) |
| 336 | { |
| 337 | hlist_add_head(&sk->sk_node, list); |
| 338 | } |
| 339 | |
| 340 | static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) |
| 341 | { |
| 342 | sock_hold(sk); |
| 343 | __sk_add_node(sk, list); |
| 344 | } |
| 345 | |
| 346 | static __inline__ void __sk_del_bind_node(struct sock *sk) |
| 347 | { |
| 348 | __hlist_del(&sk->sk_bind_node); |
| 349 | } |
| 350 | |
| 351 | static __inline__ void sk_add_bind_node(struct sock *sk, |
| 352 | struct hlist_head *list) |
| 353 | { |
| 354 | hlist_add_head(&sk->sk_bind_node, list); |
| 355 | } |
| 356 | |
| 357 | #define sk_for_each(__sk, node, list) \ |
| 358 | hlist_for_each_entry(__sk, node, list, sk_node) |
| 359 | #define sk_for_each_from(__sk, node) \ |
| 360 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ |
| 361 | hlist_for_each_entry_from(__sk, node, sk_node) |
| 362 | #define sk_for_each_continue(__sk, node) \ |
| 363 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ |
| 364 | hlist_for_each_entry_continue(__sk, node, sk_node) |
| 365 | #define sk_for_each_safe(__sk, node, tmp, list) \ |
| 366 | hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) |
| 367 | #define sk_for_each_bound(__sk, node, list) \ |
| 368 | hlist_for_each_entry(__sk, node, list, sk_bind_node) |
| 369 | |
| 370 | /* Sock flags */ |
| 371 | enum sock_flags { |
| 372 | SOCK_DEAD, |
| 373 | SOCK_DONE, |
| 374 | SOCK_URGINLINE, |
| 375 | SOCK_KEEPOPEN, |
| 376 | SOCK_LINGER, |
| 377 | SOCK_DESTROY, |
| 378 | SOCK_BROADCAST, |
| 379 | SOCK_TIMESTAMP, |
| 380 | SOCK_ZAPPED, |
| 381 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ |
| 382 | SOCK_DBG, /* %SO_DEBUG setting */ |
| 383 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ |
| 384 | SOCK_NO_LARGESEND, /* whether to sent large segments or not */ |
| 385 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
| 386 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ |
| 387 | }; |
| 388 | |
Ralf Baechle | 53b924b | 2005-08-23 10:11:30 -0700 | [diff] [blame] | 389 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
| 390 | { |
| 391 | nsk->sk_flags = osk->sk_flags; |
| 392 | } |
| 393 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 394 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
| 395 | { |
| 396 | __set_bit(flag, &sk->sk_flags); |
| 397 | } |
| 398 | |
| 399 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) |
| 400 | { |
| 401 | __clear_bit(flag, &sk->sk_flags); |
| 402 | } |
| 403 | |
| 404 | static inline int sock_flag(struct sock *sk, enum sock_flags flag) |
| 405 | { |
| 406 | return test_bit(flag, &sk->sk_flags); |
| 407 | } |
| 408 | |
| 409 | static inline void sk_acceptq_removed(struct sock *sk) |
| 410 | { |
| 411 | sk->sk_ack_backlog--; |
| 412 | } |
| 413 | |
| 414 | static inline void sk_acceptq_added(struct sock *sk) |
| 415 | { |
| 416 | sk->sk_ack_backlog++; |
| 417 | } |
| 418 | |
| 419 | static inline int sk_acceptq_is_full(struct sock *sk) |
| 420 | { |
| 421 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
| 422 | } |
| 423 | |
| 424 | /* |
| 425 | * Compute minimal free write space needed to queue new packets. |
| 426 | */ |
| 427 | static inline int sk_stream_min_wspace(struct sock *sk) |
| 428 | { |
| 429 | return sk->sk_wmem_queued / 2; |
| 430 | } |
| 431 | |
| 432 | static inline int sk_stream_wspace(struct sock *sk) |
| 433 | { |
| 434 | return sk->sk_sndbuf - sk->sk_wmem_queued; |
| 435 | } |
| 436 | |
| 437 | extern void sk_stream_write_space(struct sock *sk); |
| 438 | |
| 439 | static inline int sk_stream_memory_free(struct sock *sk) |
| 440 | { |
| 441 | return sk->sk_wmem_queued < sk->sk_sndbuf; |
| 442 | } |
| 443 | |
| 444 | extern void sk_stream_rfree(struct sk_buff *skb); |
| 445 | |
| 446 | static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk) |
| 447 | { |
| 448 | skb->sk = sk; |
| 449 | skb->destructor = sk_stream_rfree; |
| 450 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); |
| 451 | sk->sk_forward_alloc -= skb->truesize; |
| 452 | } |
| 453 | |
| 454 | static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb) |
| 455 | { |
| 456 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
| 457 | sk->sk_wmem_queued -= skb->truesize; |
| 458 | sk->sk_forward_alloc += skb->truesize; |
| 459 | __kfree_skb(skb); |
| 460 | } |
| 461 | |
| 462 | /* The per-socket spinlock must be held here. */ |
| 463 | #define sk_add_backlog(__sk, __skb) \ |
| 464 | do { if (!(__sk)->sk_backlog.tail) { \ |
| 465 | (__sk)->sk_backlog.head = \ |
| 466 | (__sk)->sk_backlog.tail = (__skb); \ |
| 467 | } else { \ |
| 468 | ((__sk)->sk_backlog.tail)->next = (__skb); \ |
| 469 | (__sk)->sk_backlog.tail = (__skb); \ |
| 470 | } \ |
| 471 | (__skb)->next = NULL; \ |
| 472 | } while(0) |
| 473 | |
| 474 | #define sk_wait_event(__sk, __timeo, __condition) \ |
| 475 | ({ int rc; \ |
| 476 | release_sock(__sk); \ |
| 477 | rc = __condition; \ |
| 478 | if (!rc) { \ |
| 479 | *(__timeo) = schedule_timeout(*(__timeo)); \ |
| 480 | rc = __condition; \ |
| 481 | } \ |
| 482 | lock_sock(__sk); \ |
| 483 | rc; \ |
| 484 | }) |
| 485 | |
| 486 | extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); |
| 487 | extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); |
| 488 | extern void sk_stream_wait_close(struct sock *sk, long timeo_p); |
| 489 | extern int sk_stream_error(struct sock *sk, int flags, int err); |
| 490 | extern void sk_stream_kill_queues(struct sock *sk); |
| 491 | |
| 492 | extern int sk_wait_data(struct sock *sk, long *timeo); |
| 493 | |
Arnaldo Carvalho de Melo | 60236fd | 2005-06-18 22:47:21 -0700 | [diff] [blame] | 494 | struct request_sock_ops; |
Arnaldo Carvalho de Melo | 2e6599c | 2005-06-18 22:46:52 -0700 | [diff] [blame] | 495 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 496 | /* Networking protocol blocks we attach to sockets. |
| 497 | * socket layer -> transport layer interface |
| 498 | * transport -> network interface is defined by struct inet_proto |
| 499 | */ |
| 500 | struct proto { |
| 501 | void (*close)(struct sock *sk, |
| 502 | long timeout); |
| 503 | int (*connect)(struct sock *sk, |
| 504 | struct sockaddr *uaddr, |
| 505 | int addr_len); |
| 506 | int (*disconnect)(struct sock *sk, int flags); |
| 507 | |
| 508 | struct sock * (*accept) (struct sock *sk, int flags, int *err); |
| 509 | |
| 510 | int (*ioctl)(struct sock *sk, int cmd, |
| 511 | unsigned long arg); |
| 512 | int (*init)(struct sock *sk); |
| 513 | int (*destroy)(struct sock *sk); |
| 514 | void (*shutdown)(struct sock *sk, int how); |
| 515 | int (*setsockopt)(struct sock *sk, int level, |
| 516 | int optname, char __user *optval, |
| 517 | int optlen); |
| 518 | int (*getsockopt)(struct sock *sk, int level, |
| 519 | int optname, char __user *optval, |
| 520 | int __user *option); |
| 521 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, |
| 522 | struct msghdr *msg, size_t len); |
| 523 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, |
| 524 | struct msghdr *msg, |
| 525 | size_t len, int noblock, int flags, |
| 526 | int *addr_len); |
| 527 | int (*sendpage)(struct sock *sk, struct page *page, |
| 528 | int offset, size_t size, int flags); |
| 529 | int (*bind)(struct sock *sk, |
| 530 | struct sockaddr *uaddr, int addr_len); |
| 531 | |
| 532 | int (*backlog_rcv) (struct sock *sk, |
| 533 | struct sk_buff *skb); |
| 534 | |
| 535 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
| 536 | void (*hash)(struct sock *sk); |
| 537 | void (*unhash)(struct sock *sk); |
| 538 | int (*get_port)(struct sock *sk, unsigned short snum); |
| 539 | |
| 540 | /* Memory pressure */ |
| 541 | void (*enter_memory_pressure)(void); |
| 542 | atomic_t *memory_allocated; /* Current allocated memory. */ |
| 543 | atomic_t *sockets_allocated; /* Current number of sockets. */ |
| 544 | /* |
| 545 | * Pressure flag: try to collapse. |
| 546 | * Technical note: it is used by multiple contexts non atomically. |
| 547 | * All the sk_stream_mem_schedule() is of this nature: accounting |
| 548 | * is strict, actions are advisory and have some latency. |
| 549 | */ |
| 550 | int *memory_pressure; |
| 551 | int *sysctl_mem; |
| 552 | int *sysctl_wmem; |
| 553 | int *sysctl_rmem; |
| 554 | int max_header; |
| 555 | |
| 556 | kmem_cache_t *slab; |
| 557 | unsigned int obj_size; |
| 558 | |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 559 | kmem_cache_t *twsk_slab; |
| 560 | unsigned int twsk_obj_size; |
Arnaldo Carvalho de Melo | 0a5578c | 2005-08-09 20:11:41 -0700 | [diff] [blame] | 561 | atomic_t *orphan_count; |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 562 | |
Arnaldo Carvalho de Melo | 60236fd | 2005-06-18 22:47:21 -0700 | [diff] [blame] | 563 | struct request_sock_ops *rsk_prot; |
Arnaldo Carvalho de Melo | 2e6599c | 2005-06-18 22:46:52 -0700 | [diff] [blame] | 564 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 565 | struct module *owner; |
| 566 | |
| 567 | char name[32]; |
| 568 | |
| 569 | struct list_head node; |
Arnaldo Carvalho de Melo | e684897 | 2005-08-09 19:45:38 -0700 | [diff] [blame] | 570 | #ifdef SOCK_REFCNT_DEBUG |
| 571 | atomic_t socks; |
| 572 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 573 | struct { |
| 574 | int inuse; |
| 575 | u8 __pad[SMP_CACHE_BYTES - sizeof(int)]; |
| 576 | } stats[NR_CPUS]; |
| 577 | }; |
| 578 | |
| 579 | extern int proto_register(struct proto *prot, int alloc_slab); |
| 580 | extern void proto_unregister(struct proto *prot); |
| 581 | |
Arnaldo Carvalho de Melo | e684897 | 2005-08-09 19:45:38 -0700 | [diff] [blame] | 582 | #ifdef SOCK_REFCNT_DEBUG |
| 583 | static inline void sk_refcnt_debug_inc(struct sock *sk) |
| 584 | { |
| 585 | atomic_inc(&sk->sk_prot->socks); |
| 586 | } |
| 587 | |
| 588 | static inline void sk_refcnt_debug_dec(struct sock *sk) |
| 589 | { |
| 590 | atomic_dec(&sk->sk_prot->socks); |
| 591 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", |
| 592 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); |
| 593 | } |
| 594 | |
| 595 | static inline void sk_refcnt_debug_release(const struct sock *sk) |
| 596 | { |
| 597 | if (atomic_read(&sk->sk_refcnt) != 1) |
| 598 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", |
| 599 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); |
| 600 | } |
| 601 | #else /* SOCK_REFCNT_DEBUG */ |
| 602 | #define sk_refcnt_debug_inc(sk) do { } while (0) |
| 603 | #define sk_refcnt_debug_dec(sk) do { } while (0) |
| 604 | #define sk_refcnt_debug_release(sk) do { } while (0) |
| 605 | #endif /* SOCK_REFCNT_DEBUG */ |
| 606 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 607 | /* Called with local bh disabled */ |
| 608 | static __inline__ void sock_prot_inc_use(struct proto *prot) |
| 609 | { |
| 610 | prot->stats[smp_processor_id()].inuse++; |
| 611 | } |
| 612 | |
| 613 | static __inline__ void sock_prot_dec_use(struct proto *prot) |
| 614 | { |
| 615 | prot->stats[smp_processor_id()].inuse--; |
| 616 | } |
| 617 | |
Arnaldo Carvalho de Melo | 614c6cb | 2005-08-09 19:47:37 -0700 | [diff] [blame] | 618 | /* With per-bucket locks this operation is not-atomic, so that |
| 619 | * this version is not worse. |
| 620 | */ |
| 621 | static inline void __sk_prot_rehash(struct sock *sk) |
| 622 | { |
| 623 | sk->sk_prot->unhash(sk); |
| 624 | sk->sk_prot->hash(sk); |
| 625 | } |
| 626 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 627 | /* About 10 seconds */ |
| 628 | #define SOCK_DESTROY_TIME (10*HZ) |
| 629 | |
| 630 | /* Sockets 0-1023 can't be bound to unless you are superuser */ |
| 631 | #define PROT_SOCK 1024 |
| 632 | |
| 633 | #define SHUTDOWN_MASK 3 |
| 634 | #define RCV_SHUTDOWN 1 |
| 635 | #define SEND_SHUTDOWN 2 |
| 636 | |
| 637 | #define SOCK_SNDBUF_LOCK 1 |
| 638 | #define SOCK_RCVBUF_LOCK 2 |
| 639 | #define SOCK_BINDADDR_LOCK 4 |
| 640 | #define SOCK_BINDPORT_LOCK 8 |
| 641 | |
| 642 | /* sock_iocb: used to kick off async processing of socket ios */ |
| 643 | struct sock_iocb { |
| 644 | struct list_head list; |
| 645 | |
| 646 | int flags; |
| 647 | int size; |
| 648 | struct socket *sock; |
| 649 | struct sock *sk; |
| 650 | struct scm_cookie *scm; |
| 651 | struct msghdr *msg, async_msg; |
| 652 | struct iovec async_iov; |
| 653 | struct kiocb *kiocb; |
| 654 | }; |
| 655 | |
| 656 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) |
| 657 | { |
| 658 | return (struct sock_iocb *)iocb->private; |
| 659 | } |
| 660 | |
| 661 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) |
| 662 | { |
| 663 | return si->kiocb; |
| 664 | } |
| 665 | |
| 666 | struct socket_alloc { |
| 667 | struct socket socket; |
| 668 | struct inode vfs_inode; |
| 669 | }; |
| 670 | |
| 671 | static inline struct socket *SOCKET_I(struct inode *inode) |
| 672 | { |
| 673 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; |
| 674 | } |
| 675 | |
| 676 | static inline struct inode *SOCK_INODE(struct socket *socket) |
| 677 | { |
| 678 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; |
| 679 | } |
| 680 | |
| 681 | extern void __sk_stream_mem_reclaim(struct sock *sk); |
| 682 | extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind); |
| 683 | |
| 684 | #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE) |
| 685 | |
| 686 | static inline int sk_stream_pages(int amt) |
| 687 | { |
| 688 | return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM; |
| 689 | } |
| 690 | |
| 691 | static inline void sk_stream_mem_reclaim(struct sock *sk) |
| 692 | { |
| 693 | if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM) |
| 694 | __sk_stream_mem_reclaim(sk); |
| 695 | } |
| 696 | |
| 697 | static inline void sk_stream_writequeue_purge(struct sock *sk) |
| 698 | { |
| 699 | struct sk_buff *skb; |
| 700 | |
| 701 | while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) |
| 702 | sk_stream_free_skb(sk, skb); |
| 703 | sk_stream_mem_reclaim(sk); |
| 704 | } |
| 705 | |
| 706 | static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb) |
| 707 | { |
| 708 | return (int)skb->truesize <= sk->sk_forward_alloc || |
| 709 | sk_stream_mem_schedule(sk, skb->truesize, 1); |
| 710 | } |
| 711 | |
| 712 | /* Used by processes to "lock" a socket state, so that |
| 713 | * interrupts and bottom half handlers won't change it |
| 714 | * from under us. It essentially blocks any incoming |
| 715 | * packets, so that we won't get any new data or any |
| 716 | * packets that change the state of the socket. |
| 717 | * |
| 718 | * While locked, BH processing will add new packets to |
| 719 | * the backlog queue. This queue is processed by the |
| 720 | * owner of the socket lock right before it is released. |
| 721 | * |
| 722 | * Since ~2.3.5 it is also exclusive sleep lock serializing |
| 723 | * accesses from user process context. |
| 724 | */ |
| 725 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owner) |
| 726 | |
| 727 | extern void FASTCALL(lock_sock(struct sock *sk)); |
| 728 | extern void FASTCALL(release_sock(struct sock *sk)); |
| 729 | |
| 730 | /* BH context may only use the following locking interface. */ |
| 731 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) |
| 732 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
| 733 | |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 734 | extern struct sock *sk_alloc(int family, |
| 735 | unsigned int __nocast priority, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 736 | struct proto *prot, int zero_it); |
| 737 | extern void sk_free(struct sock *sk); |
Arnaldo Carvalho de Melo | 87d11ce | 2005-08-09 20:10:12 -0700 | [diff] [blame] | 738 | extern struct sock *sk_clone(const struct sock *sk, |
| 739 | const unsigned int __nocast priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 740 | |
| 741 | extern struct sk_buff *sock_wmalloc(struct sock *sk, |
| 742 | unsigned long size, int force, |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 743 | unsigned int __nocast priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 744 | extern struct sk_buff *sock_rmalloc(struct sock *sk, |
| 745 | unsigned long size, int force, |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 746 | unsigned int __nocast priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 747 | extern void sock_wfree(struct sk_buff *skb); |
| 748 | extern void sock_rfree(struct sk_buff *skb); |
| 749 | |
| 750 | extern int sock_setsockopt(struct socket *sock, int level, |
| 751 | int op, char __user *optval, |
| 752 | int optlen); |
| 753 | |
| 754 | extern int sock_getsockopt(struct socket *sock, int level, |
| 755 | int op, char __user *optval, |
| 756 | int __user *optlen); |
| 757 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, |
| 758 | unsigned long size, |
| 759 | int noblock, |
| 760 | int *errcode); |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 761 | extern void *sock_kmalloc(struct sock *sk, int size, |
| 762 | unsigned int __nocast priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 763 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); |
| 764 | extern void sk_send_sigurg(struct sock *sk); |
| 765 | |
| 766 | /* |
| 767 | * Functions to fill in entries in struct proto_ops when a protocol |
| 768 | * does not implement a particular function. |
| 769 | */ |
| 770 | extern int sock_no_bind(struct socket *, |
| 771 | struct sockaddr *, int); |
| 772 | extern int sock_no_connect(struct socket *, |
| 773 | struct sockaddr *, int, int); |
| 774 | extern int sock_no_socketpair(struct socket *, |
| 775 | struct socket *); |
| 776 | extern int sock_no_accept(struct socket *, |
| 777 | struct socket *, int); |
| 778 | extern int sock_no_getname(struct socket *, |
| 779 | struct sockaddr *, int *, int); |
| 780 | extern unsigned int sock_no_poll(struct file *, struct socket *, |
| 781 | struct poll_table_struct *); |
| 782 | extern int sock_no_ioctl(struct socket *, unsigned int, |
| 783 | unsigned long); |
| 784 | extern int sock_no_listen(struct socket *, int); |
| 785 | extern int sock_no_shutdown(struct socket *, int); |
| 786 | extern int sock_no_getsockopt(struct socket *, int , int, |
| 787 | char __user *, int __user *); |
| 788 | extern int sock_no_setsockopt(struct socket *, int, int, |
| 789 | char __user *, int); |
| 790 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, |
| 791 | struct msghdr *, size_t); |
| 792 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, |
| 793 | struct msghdr *, size_t, int); |
| 794 | extern int sock_no_mmap(struct file *file, |
| 795 | struct socket *sock, |
| 796 | struct vm_area_struct *vma); |
| 797 | extern ssize_t sock_no_sendpage(struct socket *sock, |
| 798 | struct page *page, |
| 799 | int offset, size_t size, |
| 800 | int flags); |
| 801 | |
| 802 | /* |
| 803 | * Functions to fill in entries in struct proto_ops when a protocol |
| 804 | * uses the inet style. |
| 805 | */ |
| 806 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, |
| 807 | char __user *optval, int __user *optlen); |
| 808 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, |
| 809 | struct msghdr *msg, size_t size, int flags); |
| 810 | extern int sock_common_setsockopt(struct socket *sock, int level, int optname, |
| 811 | char __user *optval, int optlen); |
| 812 | |
| 813 | extern void sk_common_release(struct sock *sk); |
| 814 | |
| 815 | /* |
| 816 | * Default socket callbacks and setup code |
| 817 | */ |
| 818 | |
| 819 | /* Initialise core socket variables */ |
| 820 | extern void sock_init_data(struct socket *sock, struct sock *sk); |
| 821 | |
| 822 | /** |
| 823 | * sk_filter - run a packet through a socket filter |
| 824 | * @sk: sock associated with &sk_buff |
| 825 | * @skb: buffer to filter |
| 826 | * @needlock: set to 1 if the sock is not locked by caller. |
| 827 | * |
| 828 | * Run the filter code and then cut skb->data to correct size returned by |
| 829 | * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller |
| 830 | * than pkt_len we keep whole skb->data. This is the socket level |
| 831 | * wrapper to sk_run_filter. It returns 0 if the packet should |
| 832 | * be accepted or -EPERM if the packet should be tossed. |
| 833 | * |
| 834 | */ |
| 835 | |
| 836 | static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock) |
| 837 | { |
| 838 | int err; |
| 839 | |
| 840 | err = security_sock_rcv_skb(sk, skb); |
| 841 | if (err) |
| 842 | return err; |
| 843 | |
| 844 | if (sk->sk_filter) { |
| 845 | struct sk_filter *filter; |
| 846 | |
| 847 | if (needlock) |
| 848 | bh_lock_sock(sk); |
| 849 | |
| 850 | filter = sk->sk_filter; |
| 851 | if (filter) { |
| 852 | int pkt_len = sk_run_filter(skb, filter->insns, |
| 853 | filter->len); |
| 854 | if (!pkt_len) |
| 855 | err = -EPERM; |
| 856 | else |
| 857 | skb_trim(skb, pkt_len); |
| 858 | } |
| 859 | |
| 860 | if (needlock) |
| 861 | bh_unlock_sock(sk); |
| 862 | } |
| 863 | return err; |
| 864 | } |
| 865 | |
| 866 | /** |
| 867 | * sk_filter_release: Release a socket filter |
| 868 | * @sk: socket |
| 869 | * @fp: filter to remove |
| 870 | * |
| 871 | * Remove a filter from a socket and release its resources. |
| 872 | */ |
| 873 | |
| 874 | static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp) |
| 875 | { |
| 876 | unsigned int size = sk_filter_len(fp); |
| 877 | |
| 878 | atomic_sub(size, &sk->sk_omem_alloc); |
| 879 | |
| 880 | if (atomic_dec_and_test(&fp->refcnt)) |
| 881 | kfree(fp); |
| 882 | } |
| 883 | |
| 884 | static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) |
| 885 | { |
| 886 | atomic_inc(&fp->refcnt); |
| 887 | atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); |
| 888 | } |
| 889 | |
| 890 | /* |
| 891 | * Socket reference counting postulates. |
| 892 | * |
| 893 | * * Each user of socket SHOULD hold a reference count. |
| 894 | * * Each access point to socket (an hash table bucket, reference from a list, |
| 895 | * running timer, skb in flight MUST hold a reference count. |
| 896 | * * When reference count hits 0, it means it will never increase back. |
| 897 | * * When reference count hits 0, it means that no references from |
| 898 | * outside exist to this socket and current process on current CPU |
| 899 | * is last user and may/should destroy this socket. |
| 900 | * * sk_free is called from any context: process, BH, IRQ. When |
| 901 | * it is called, socket has no references from outside -> sk_free |
| 902 | * may release descendant resources allocated by the socket, but |
| 903 | * to the time when it is called, socket is NOT referenced by any |
| 904 | * hash tables, lists etc. |
| 905 | * * Packets, delivered from outside (from network or from another process) |
| 906 | * and enqueued on receive/error queues SHOULD NOT grab reference count, |
| 907 | * when they sit in queue. Otherwise, packets will leak to hole, when |
| 908 | * socket is looked up by one cpu and unhasing is made by another CPU. |
| 909 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp |
| 910 | * (leak to backlog). Packet socket does all the processing inside |
| 911 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets |
| 912 | * use separate SMP lock, so that they are prone too. |
| 913 | */ |
| 914 | |
| 915 | /* Ungrab socket and destroy it, if it was the last reference. */ |
| 916 | static inline void sock_put(struct sock *sk) |
| 917 | { |
| 918 | if (atomic_dec_and_test(&sk->sk_refcnt)) |
| 919 | sk_free(sk); |
| 920 | } |
| 921 | |
| 922 | /* Detach socket from process context. |
| 923 | * Announce socket dead, detach it from wait queue and inode. |
| 924 | * Note that parent inode held reference count on this struct sock, |
| 925 | * we do not release it in this function, because protocol |
| 926 | * probably wants some additional cleanups or even continuing |
| 927 | * to work with this socket (TCP). |
| 928 | */ |
| 929 | static inline void sock_orphan(struct sock *sk) |
| 930 | { |
| 931 | write_lock_bh(&sk->sk_callback_lock); |
| 932 | sock_set_flag(sk, SOCK_DEAD); |
| 933 | sk->sk_socket = NULL; |
| 934 | sk->sk_sleep = NULL; |
| 935 | write_unlock_bh(&sk->sk_callback_lock); |
| 936 | } |
| 937 | |
| 938 | static inline void sock_graft(struct sock *sk, struct socket *parent) |
| 939 | { |
| 940 | write_lock_bh(&sk->sk_callback_lock); |
| 941 | sk->sk_sleep = &parent->wait; |
| 942 | parent->sk = sk; |
| 943 | sk->sk_socket = parent; |
| 944 | write_unlock_bh(&sk->sk_callback_lock); |
| 945 | } |
| 946 | |
| 947 | extern int sock_i_uid(struct sock *sk); |
| 948 | extern unsigned long sock_i_ino(struct sock *sk); |
| 949 | |
| 950 | static inline struct dst_entry * |
| 951 | __sk_dst_get(struct sock *sk) |
| 952 | { |
| 953 | return sk->sk_dst_cache; |
| 954 | } |
| 955 | |
| 956 | static inline struct dst_entry * |
| 957 | sk_dst_get(struct sock *sk) |
| 958 | { |
| 959 | struct dst_entry *dst; |
| 960 | |
| 961 | read_lock(&sk->sk_dst_lock); |
| 962 | dst = sk->sk_dst_cache; |
| 963 | if (dst) |
| 964 | dst_hold(dst); |
| 965 | read_unlock(&sk->sk_dst_lock); |
| 966 | return dst; |
| 967 | } |
| 968 | |
| 969 | static inline void |
| 970 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) |
| 971 | { |
| 972 | struct dst_entry *old_dst; |
| 973 | |
| 974 | old_dst = sk->sk_dst_cache; |
| 975 | sk->sk_dst_cache = dst; |
| 976 | dst_release(old_dst); |
| 977 | } |
| 978 | |
| 979 | static inline void |
| 980 | sk_dst_set(struct sock *sk, struct dst_entry *dst) |
| 981 | { |
| 982 | write_lock(&sk->sk_dst_lock); |
| 983 | __sk_dst_set(sk, dst); |
| 984 | write_unlock(&sk->sk_dst_lock); |
| 985 | } |
| 986 | |
| 987 | static inline void |
| 988 | __sk_dst_reset(struct sock *sk) |
| 989 | { |
| 990 | struct dst_entry *old_dst; |
| 991 | |
| 992 | old_dst = sk->sk_dst_cache; |
| 993 | sk->sk_dst_cache = NULL; |
| 994 | dst_release(old_dst); |
| 995 | } |
| 996 | |
| 997 | static inline void |
| 998 | sk_dst_reset(struct sock *sk) |
| 999 | { |
| 1000 | write_lock(&sk->sk_dst_lock); |
| 1001 | __sk_dst_reset(sk); |
| 1002 | write_unlock(&sk->sk_dst_lock); |
| 1003 | } |
| 1004 | |
| 1005 | static inline struct dst_entry * |
| 1006 | __sk_dst_check(struct sock *sk, u32 cookie) |
| 1007 | { |
| 1008 | struct dst_entry *dst = sk->sk_dst_cache; |
| 1009 | |
| 1010 | if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) { |
| 1011 | sk->sk_dst_cache = NULL; |
| 1012 | dst_release(dst); |
| 1013 | return NULL; |
| 1014 | } |
| 1015 | |
| 1016 | return dst; |
| 1017 | } |
| 1018 | |
| 1019 | static inline struct dst_entry * |
| 1020 | sk_dst_check(struct sock *sk, u32 cookie) |
| 1021 | { |
| 1022 | struct dst_entry *dst = sk_dst_get(sk); |
| 1023 | |
| 1024 | if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) { |
| 1025 | sk_dst_reset(sk); |
| 1026 | dst_release(dst); |
| 1027 | return NULL; |
| 1028 | } |
| 1029 | |
| 1030 | return dst; |
| 1031 | } |
| 1032 | |
Arnaldo Carvalho de Melo | 6cbb0df | 2005-08-09 19:49:02 -0700 | [diff] [blame] | 1033 | static inline void sk_setup_caps(struct sock *sk, struct dst_entry *dst) |
| 1034 | { |
| 1035 | __sk_dst_set(sk, dst); |
| 1036 | sk->sk_route_caps = dst->dev->features; |
| 1037 | if (sk->sk_route_caps & NETIF_F_TSO) { |
| 1038 | if (sock_flag(sk, SOCK_NO_LARGESEND) || dst->header_len) |
| 1039 | sk->sk_route_caps &= ~NETIF_F_TSO; |
| 1040 | } |
| 1041 | } |
| 1042 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1043 | static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb) |
| 1044 | { |
| 1045 | sk->sk_wmem_queued += skb->truesize; |
| 1046 | sk->sk_forward_alloc -= skb->truesize; |
| 1047 | } |
| 1048 | |
| 1049 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, |
| 1050 | struct sk_buff *skb, struct page *page, |
| 1051 | int off, int copy) |
| 1052 | { |
| 1053 | if (skb->ip_summed == CHECKSUM_NONE) { |
| 1054 | int err = 0; |
| 1055 | unsigned int csum = csum_and_copy_from_user(from, |
| 1056 | page_address(page) + off, |
| 1057 | copy, 0, &err); |
| 1058 | if (err) |
| 1059 | return err; |
| 1060 | skb->csum = csum_block_add(skb->csum, csum, skb->len); |
| 1061 | } else if (copy_from_user(page_address(page) + off, from, copy)) |
| 1062 | return -EFAULT; |
| 1063 | |
| 1064 | skb->len += copy; |
| 1065 | skb->data_len += copy; |
| 1066 | skb->truesize += copy; |
| 1067 | sk->sk_wmem_queued += copy; |
| 1068 | sk->sk_forward_alloc -= copy; |
| 1069 | return 0; |
| 1070 | } |
| 1071 | |
| 1072 | /* |
| 1073 | * Queue a received datagram if it will fit. Stream and sequenced |
| 1074 | * protocols can't normally use this as they need to fit buffers in |
| 1075 | * and play with them. |
| 1076 | * |
| 1077 | * Inlined as it's very short and called for pretty much every |
| 1078 | * packet ever received. |
| 1079 | */ |
| 1080 | |
| 1081 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) |
| 1082 | { |
| 1083 | sock_hold(sk); |
| 1084 | skb->sk = sk; |
| 1085 | skb->destructor = sock_wfree; |
| 1086 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); |
| 1087 | } |
| 1088 | |
| 1089 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) |
| 1090 | { |
| 1091 | skb->sk = sk; |
| 1092 | skb->destructor = sock_rfree; |
| 1093 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); |
| 1094 | } |
| 1095 | |
| 1096 | extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, |
| 1097 | unsigned long expires); |
| 1098 | |
| 1099 | extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); |
| 1100 | |
| 1101 | static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
| 1102 | { |
| 1103 | int err = 0; |
| 1104 | int skb_len; |
| 1105 | |
| 1106 | /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces |
| 1107 | number of warnings when compiling with -W --ANK |
| 1108 | */ |
| 1109 | if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >= |
| 1110 | (unsigned)sk->sk_rcvbuf) { |
| 1111 | err = -ENOMEM; |
| 1112 | goto out; |
| 1113 | } |
| 1114 | |
| 1115 | /* It would be deadlock, if sock_queue_rcv_skb is used |
| 1116 | with socket lock! We assume that users of this |
| 1117 | function are lock free. |
| 1118 | */ |
| 1119 | err = sk_filter(sk, skb, 1); |
| 1120 | if (err) |
| 1121 | goto out; |
| 1122 | |
| 1123 | skb->dev = NULL; |
| 1124 | skb_set_owner_r(skb, sk); |
| 1125 | |
| 1126 | /* Cache the SKB length before we tack it onto the receive |
| 1127 | * queue. Once it is added it no longer belongs to us and |
| 1128 | * may be freed by other threads of control pulling packets |
| 1129 | * from the queue. |
| 1130 | */ |
| 1131 | skb_len = skb->len; |
| 1132 | |
| 1133 | skb_queue_tail(&sk->sk_receive_queue, skb); |
| 1134 | |
| 1135 | if (!sock_flag(sk, SOCK_DEAD)) |
| 1136 | sk->sk_data_ready(sk, skb_len); |
| 1137 | out: |
| 1138 | return err; |
| 1139 | } |
| 1140 | |
| 1141 | static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb) |
| 1142 | { |
| 1143 | /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces |
| 1144 | number of warnings when compiling with -W --ANK |
| 1145 | */ |
| 1146 | if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >= |
| 1147 | (unsigned)sk->sk_rcvbuf) |
| 1148 | return -ENOMEM; |
| 1149 | skb_set_owner_r(skb, sk); |
| 1150 | skb_queue_tail(&sk->sk_error_queue, skb); |
| 1151 | if (!sock_flag(sk, SOCK_DEAD)) |
| 1152 | sk->sk_data_ready(sk, skb->len); |
| 1153 | return 0; |
| 1154 | } |
| 1155 | |
| 1156 | /* |
| 1157 | * Recover an error report and clear atomically |
| 1158 | */ |
| 1159 | |
| 1160 | static inline int sock_error(struct sock *sk) |
| 1161 | { |
| 1162 | int err = xchg(&sk->sk_err, 0); |
| 1163 | return -err; |
| 1164 | } |
| 1165 | |
| 1166 | static inline unsigned long sock_wspace(struct sock *sk) |
| 1167 | { |
| 1168 | int amt = 0; |
| 1169 | |
| 1170 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { |
| 1171 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); |
| 1172 | if (amt < 0) |
| 1173 | amt = 0; |
| 1174 | } |
| 1175 | return amt; |
| 1176 | } |
| 1177 | |
| 1178 | static inline void sk_wake_async(struct sock *sk, int how, int band) |
| 1179 | { |
| 1180 | if (sk->sk_socket && sk->sk_socket->fasync_list) |
| 1181 | sock_wake_async(sk->sk_socket, how, band); |
| 1182 | } |
| 1183 | |
| 1184 | #define SOCK_MIN_SNDBUF 2048 |
| 1185 | #define SOCK_MIN_RCVBUF 256 |
| 1186 | |
| 1187 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) |
| 1188 | { |
| 1189 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { |
| 1190 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2); |
| 1191 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); |
| 1192 | } |
| 1193 | } |
| 1194 | |
| 1195 | static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk, |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 1196 | int size, int mem, |
| 1197 | unsigned int __nocast gfp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1198 | { |
David S. Miller | c65f7f0 | 2005-07-05 15:17:25 -0700 | [diff] [blame] | 1199 | struct sk_buff *skb; |
| 1200 | int hdr_len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1201 | |
David S. Miller | c65f7f0 | 2005-07-05 15:17:25 -0700 | [diff] [blame] | 1202 | hdr_len = SKB_DATA_ALIGN(sk->sk_prot->max_header); |
| 1203 | skb = alloc_skb(size + hdr_len, gfp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1204 | if (skb) { |
| 1205 | skb->truesize += mem; |
| 1206 | if (sk->sk_forward_alloc >= (int)skb->truesize || |
| 1207 | sk_stream_mem_schedule(sk, skb->truesize, 0)) { |
David S. Miller | c65f7f0 | 2005-07-05 15:17:25 -0700 | [diff] [blame] | 1208 | skb_reserve(skb, hdr_len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1209 | return skb; |
| 1210 | } |
| 1211 | __kfree_skb(skb); |
| 1212 | } else { |
| 1213 | sk->sk_prot->enter_memory_pressure(); |
| 1214 | sk_stream_moderate_sndbuf(sk); |
| 1215 | } |
| 1216 | return NULL; |
| 1217 | } |
| 1218 | |
| 1219 | static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk, |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 1220 | int size, |
| 1221 | unsigned int __nocast gfp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1222 | { |
| 1223 | return sk_stream_alloc_pskb(sk, size, 0, gfp); |
| 1224 | } |
| 1225 | |
| 1226 | static inline struct page *sk_stream_alloc_page(struct sock *sk) |
| 1227 | { |
| 1228 | struct page *page = NULL; |
| 1229 | |
| 1230 | if (sk->sk_forward_alloc >= (int)PAGE_SIZE || |
| 1231 | sk_stream_mem_schedule(sk, PAGE_SIZE, 0)) |
| 1232 | page = alloc_pages(sk->sk_allocation, 0); |
| 1233 | else { |
| 1234 | sk->sk_prot->enter_memory_pressure(); |
| 1235 | sk_stream_moderate_sndbuf(sk); |
| 1236 | } |
| 1237 | return page; |
| 1238 | } |
| 1239 | |
| 1240 | #define sk_stream_for_retrans_queue(skb, sk) \ |
| 1241 | for (skb = (sk)->sk_write_queue.next; \ |
| 1242 | (skb != (sk)->sk_send_head) && \ |
| 1243 | (skb != (struct sk_buff *)&(sk)->sk_write_queue); \ |
| 1244 | skb = skb->next) |
| 1245 | |
| 1246 | /* |
| 1247 | * Default write policy as shown to user space via poll/select/SIGIO |
| 1248 | */ |
| 1249 | static inline int sock_writeable(const struct sock *sk) |
| 1250 | { |
| 1251 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2); |
| 1252 | } |
| 1253 | |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 1254 | static inline unsigned int __nocast gfp_any(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1255 | { |
| 1256 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
| 1257 | } |
| 1258 | |
| 1259 | static inline long sock_rcvtimeo(const struct sock *sk, int noblock) |
| 1260 | { |
| 1261 | return noblock ? 0 : sk->sk_rcvtimeo; |
| 1262 | } |
| 1263 | |
| 1264 | static inline long sock_sndtimeo(const struct sock *sk, int noblock) |
| 1265 | { |
| 1266 | return noblock ? 0 : sk->sk_sndtimeo; |
| 1267 | } |
| 1268 | |
| 1269 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) |
| 1270 | { |
| 1271 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; |
| 1272 | } |
| 1273 | |
| 1274 | /* Alas, with timeout socket operations are not restartable. |
| 1275 | * Compare this to poll(). |
| 1276 | */ |
| 1277 | static inline int sock_intr_errno(long timeo) |
| 1278 | { |
| 1279 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; |
| 1280 | } |
| 1281 | |
| 1282 | static __inline__ void |
| 1283 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
| 1284 | { |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1285 | struct timeval stamp; |
| 1286 | |
| 1287 | skb_get_timestamp(skb, &stamp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1288 | if (sock_flag(sk, SOCK_RCVTSTAMP)) { |
| 1289 | /* Race occurred between timestamp enabling and packet |
| 1290 | receiving. Fill in the current time for now. */ |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1291 | if (stamp.tv_sec == 0) |
| 1292 | do_gettimeofday(&stamp); |
| 1293 | skb_set_timestamp(skb, &stamp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1294 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval), |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1295 | &stamp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1296 | } else |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1297 | sk->sk_stamp = stamp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1298 | } |
| 1299 | |
| 1300 | /** |
| 1301 | * sk_eat_skb - Release a skb if it is no longer needed |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 1302 | * @sk: socket to eat this skb from |
| 1303 | * @skb: socket buffer to eat |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1304 | * |
| 1305 | * This routine must be called with interrupts disabled or with the socket |
| 1306 | * locked so that the sk_buff queue operation is ok. |
| 1307 | */ |
| 1308 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb) |
| 1309 | { |
| 1310 | __skb_unlink(skb, &sk->sk_receive_queue); |
| 1311 | __kfree_skb(skb); |
| 1312 | } |
| 1313 | |
| 1314 | extern void sock_enable_timestamp(struct sock *sk); |
| 1315 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); |
| 1316 | |
| 1317 | /* |
| 1318 | * Enable debug/info messages |
| 1319 | */ |
| 1320 | |
| 1321 | #if 0 |
Patrick McHardy | 64ce207 | 2005-08-09 20:50:53 -0700 | [diff] [blame] | 1322 | #define NETDEBUG(fmt, args...) do { } while (0) |
| 1323 | #define LIMIT_NETDEBUG(fmt, args...) do { } while(0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1324 | #else |
Patrick McHardy | 64ce207 | 2005-08-09 20:50:53 -0700 | [diff] [blame] | 1325 | #define NETDEBUG(fmt, args...) printk(fmt,##args) |
| 1326 | #define LIMIT_NETDEBUG(fmt, args...) do { if (net_ratelimit()) printk(fmt,##args); } while(0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1327 | #endif |
| 1328 | |
| 1329 | /* |
| 1330 | * Macros for sleeping on a socket. Use them like this: |
| 1331 | * |
| 1332 | * SOCK_SLEEP_PRE(sk) |
| 1333 | * if (condition) |
| 1334 | * schedule(); |
| 1335 | * SOCK_SLEEP_POST(sk) |
| 1336 | * |
| 1337 | * N.B. These are now obsolete and were, afaik, only ever used in DECnet |
| 1338 | * and when the last use of them in DECnet has gone, I'm intending to |
| 1339 | * remove them. |
| 1340 | */ |
| 1341 | |
| 1342 | #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \ |
| 1343 | DECLARE_WAITQUEUE(wait, tsk); \ |
| 1344 | tsk->state = TASK_INTERRUPTIBLE; \ |
| 1345 | add_wait_queue((sk)->sk_sleep, &wait); \ |
| 1346 | release_sock(sk); |
| 1347 | |
| 1348 | #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \ |
| 1349 | remove_wait_queue((sk)->sk_sleep, &wait); \ |
| 1350 | lock_sock(sk); \ |
| 1351 | } |
| 1352 | |
| 1353 | static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool) |
| 1354 | { |
| 1355 | if (valbool) |
| 1356 | sock_set_flag(sk, bit); |
| 1357 | else |
| 1358 | sock_reset_flag(sk, bit); |
| 1359 | } |
| 1360 | |
| 1361 | extern __u32 sysctl_wmem_max; |
| 1362 | extern __u32 sysctl_rmem_max; |
| 1363 | |
| 1364 | #ifdef CONFIG_NET |
| 1365 | int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg); |
| 1366 | #else |
| 1367 | static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg) |
| 1368 | { |
| 1369 | return -ENODEV; |
| 1370 | } |
| 1371 | #endif |
| 1372 | |
Arnaldo Carvalho de Melo | 2038073 | 2005-08-16 02:18:02 -0300 | [diff] [blame^] | 1373 | extern void sk_init(void); |
| 1374 | |
| 1375 | #ifdef CONFIG_SYSCTL |
| 1376 | extern struct ctl_table core_table[]; |
| 1377 | extern int sysctl_optmem_max; |
| 1378 | #endif |
| 1379 | |
| 1380 | #ifdef CONFIG_PROC_FS |
| 1381 | extern __u32 sysctl_wmem_default; |
| 1382 | extern __u32 sysctl_rmem_default; |
| 1383 | #endif |
| 1384 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1385 | #endif /* _SOCK_H */ |