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