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