1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the AF_INET socket handler. 7 * 8 * Version: @(#)sock.h 1.0.4 05/13/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Corey Minyard <wf-rch!minyard@relay.EU.net> 13 * Florian La Roche <flla@stud.uni-sb.de> 14 * 15 * Fixes: 16 * Alan Cox : Volatiles in skbuff pointers. See 17 * skbuff comments. May be overdone, 18 * better to prove they can be removed 19 * than the reverse. 20 * Alan Cox : Added a zapped field for tcp to note 21 * a socket is reset and must stay shut up 22 * Alan Cox : New fields for options 23 * Pauline Middelink : identd support 24 * Alan Cox : Eliminate low level recv/recvfrom 25 * David S. Miller : New socket lookup architecture. 26 * Steve Whitehouse: Default routines for sock_ops 27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made 28 * protinfo be just a void pointer, as the 29 * protocol specific parts were moved to 30 * respective headers and ipv4/v6, etc now 31 * use private slabcaches for its socks 32 * Pedro Hortas : New flags field for socket options 33 * 34 * 35 * This program is free software; you can redistribute it and/or 36 * modify it under the terms of the GNU General Public License 37 * as published by the Free Software Foundation; either version 38 * 2 of the License, or (at your option) any later version. 39 */ 40 #ifndef _SOCK_H 41 #define _SOCK_H 42 43 #include <linux/hardirq.h> 44 #include <linux/kernel.h> 45 #include <linux/list.h> 46 #include <linux/list_nulls.h> 47 #include <linux/timer.h> 48 #include <linux/cache.h> 49 #include <linux/module.h> 50 #include <linux/lockdep.h> 51 #include <linux/netdevice.h> 52 #include <linux/skbuff.h> /* struct sk_buff */ 53 #include <linux/mm.h> 54 #include <linux/security.h> 55 #include <linux/slab.h> 56 #include <linux/uaccess.h> 57 58 #include <linux/filter.h> 59 #include <linux/rculist_nulls.h> 60 #include <linux/poll.h> 61 62 #include <linux/atomic.h> 63 #include <net/dst.h> 64 #include <net/checksum.h> 65 66 /* 67 * This structure really needs to be cleaned up. 68 * Most of it is for TCP, and not used by any of 69 * the other protocols. 70 */ 71 72 /* Define this to get the SOCK_DBG debugging facility. */ 73 #define SOCK_DEBUGGING 74 #ifdef SOCK_DEBUGGING 75 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ 76 printk(KERN_DEBUG msg); } while (0) 77 #else 78 /* Validate arguments and do nothing */ 79 static inline void __attribute__ ((format (printf, 2, 3))) 80 SOCK_DEBUG(struct sock *sk, const char *msg, ...) 81 { 82 } 83 #endif 84 85 /* This is the per-socket lock. The spinlock provides a synchronization 86 * between user contexts and software interrupt processing, whereas the 87 * mini-semaphore synchronizes multiple users amongst themselves. 88 */ 89 typedef struct { 90 spinlock_t slock; 91 int owned; 92 wait_queue_head_t wq; 93 /* 94 * We express the mutex-alike socket_lock semantics 95 * to the lock validator by explicitly managing 96 * the slock as a lock variant (in addition to 97 * the slock itself): 98 */ 99 #ifdef CONFIG_DEBUG_LOCK_ALLOC 100 struct lockdep_map dep_map; 101 #endif 102 } socket_lock_t; 103 104 struct sock; 105 struct proto; 106 struct net; 107 108 /** 109 * struct sock_common - minimal network layer representation of sockets 110 * @skc_daddr: Foreign IPv4 addr 111 * @skc_rcv_saddr: Bound local IPv4 addr 112 * @skc_hash: hash value used with various protocol lookup tables 113 * @skc_u16hashes: two u16 hash values used by UDP lookup tables 114 * @skc_family: network address family 115 * @skc_state: Connection state 116 * @skc_reuse: %SO_REUSEADDR setting 117 * @skc_bound_dev_if: bound device index if != 0 118 * @skc_bind_node: bind hash linkage for various protocol lookup tables 119 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol 120 * @skc_prot: protocol handlers inside a network family 121 * @skc_net: reference to the network namespace of this socket 122 * @skc_node: main hash linkage for various protocol lookup tables 123 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol 124 * @skc_tx_queue_mapping: tx queue number for this connection 125 * @skc_refcnt: reference count 126 * 127 * This is the minimal network layer representation of sockets, the header 128 * for struct sock and struct inet_timewait_sock. 129 */ 130 struct sock_common { 131 /* skc_daddr and skc_rcv_saddr must be grouped : 132 * cf INET_MATCH() and INET_TW_MATCH() 133 */ 134 __be32 skc_daddr; 135 __be32 skc_rcv_saddr; 136 137 union { 138 unsigned int skc_hash; 139 __u16 skc_u16hashes[2]; 140 }; 141 unsigned short skc_family; 142 volatile unsigned char skc_state; 143 unsigned char skc_reuse; 144 int skc_bound_dev_if; 145 union { 146 struct hlist_node skc_bind_node; 147 struct hlist_nulls_node skc_portaddr_node; 148 }; 149 struct proto *skc_prot; 150 #ifdef CONFIG_NET_NS 151 struct net *skc_net; 152 #endif 153 /* 154 * fields between dontcopy_begin/dontcopy_end 155 * are not copied in sock_copy() 156 */ 157 /* private: */ 158 int skc_dontcopy_begin[0]; 159 /* public: */ 160 union { 161 struct hlist_node skc_node; 162 struct hlist_nulls_node skc_nulls_node; 163 }; 164 int skc_tx_queue_mapping; 165 atomic_t skc_refcnt; 166 /* private: */ 167 int skc_dontcopy_end[0]; 168 /* public: */ 169 }; 170 171 /** 172 * struct sock - network layer representation of sockets 173 * @__sk_common: shared layout with inet_timewait_sock 174 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN 175 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings 176 * @sk_lock: synchronizer 177 * @sk_rcvbuf: size of receive buffer in bytes 178 * @sk_wq: sock wait queue and async head 179 * @sk_dst_cache: destination cache 180 * @sk_dst_lock: destination cache lock 181 * @sk_policy: flow policy 182 * @sk_receive_queue: incoming packets 183 * @sk_wmem_alloc: transmit queue bytes committed 184 * @sk_write_queue: Packet sending queue 185 * @sk_async_wait_queue: DMA copied packets 186 * @sk_omem_alloc: "o" is "option" or "other" 187 * @sk_wmem_queued: persistent queue size 188 * @sk_forward_alloc: space allocated forward 189 * @sk_allocation: allocation mode 190 * @sk_sndbuf: size of send buffer in bytes 191 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, 192 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings 193 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets 194 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) 195 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) 196 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) 197 * @sk_gso_max_size: Maximum GSO segment size to build 198 * @sk_lingertime: %SO_LINGER l_linger setting 199 * @sk_backlog: always used with the per-socket spinlock held 200 * @sk_callback_lock: used with the callbacks in the end of this struct 201 * @sk_error_queue: rarely used 202 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, 203 * IPV6_ADDRFORM for instance) 204 * @sk_err: last error 205 * @sk_err_soft: errors that don't cause failure but are the cause of a 206 * persistent failure not just 'timed out' 207 * @sk_drops: raw/udp drops counter 208 * @sk_ack_backlog: current listen backlog 209 * @sk_max_ack_backlog: listen backlog set in listen() 210 * @sk_priority: %SO_PRIORITY setting 211 * @sk_type: socket type (%SOCK_STREAM, etc) 212 * @sk_protocol: which protocol this socket belongs in this network family 213 * @sk_peer_pid: &struct pid for this socket's peer 214 * @sk_peer_cred: %SO_PEERCRED setting 215 * @sk_rcvlowat: %SO_RCVLOWAT setting 216 * @sk_rcvtimeo: %SO_RCVTIMEO setting 217 * @sk_sndtimeo: %SO_SNDTIMEO setting 218 * @sk_rxhash: flow hash received from netif layer 219 * @sk_filter: socket filtering instructions 220 * @sk_protinfo: private area, net family specific, when not using slab 221 * @sk_timer: sock cleanup timer 222 * @sk_stamp: time stamp of last packet received 223 * @sk_socket: Identd and reporting IO signals 224 * @sk_user_data: RPC layer private data 225 * @sk_sndmsg_page: cached page for sendmsg 226 * @sk_sndmsg_off: cached offset for sendmsg 227 * @sk_send_head: front of stuff to transmit 228 * @sk_security: used by security modules 229 * @sk_mark: generic packet mark 230 * @sk_classid: this socket's cgroup classid 231 * @sk_write_pending: a write to stream socket waits to start 232 * @sk_state_change: callback to indicate change in the state of the sock 233 * @sk_data_ready: callback to indicate there is data to be processed 234 * @sk_write_space: callback to indicate there is bf sending space available 235 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) 236 * @sk_backlog_rcv: callback to process the backlog 237 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 238 */ 239 struct sock { 240 /* 241 * Now struct inet_timewait_sock also uses sock_common, so please just 242 * don't add nothing before this first member (__sk_common) --acme 243 */ 244 struct sock_common __sk_common; 245 #define sk_node __sk_common.skc_node 246 #define sk_nulls_node __sk_common.skc_nulls_node 247 #define sk_refcnt __sk_common.skc_refcnt 248 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping 249 250 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin 251 #define sk_dontcopy_end __sk_common.skc_dontcopy_end 252 #define sk_hash __sk_common.skc_hash 253 #define sk_family __sk_common.skc_family 254 #define sk_state __sk_common.skc_state 255 #define sk_reuse __sk_common.skc_reuse 256 #define sk_bound_dev_if __sk_common.skc_bound_dev_if 257 #define sk_bind_node __sk_common.skc_bind_node 258 #define sk_prot __sk_common.skc_prot 259 #define sk_net __sk_common.skc_net 260 socket_lock_t sk_lock; 261 struct sk_buff_head sk_receive_queue; 262 /* 263 * The backlog queue is special, it is always used with 264 * the per-socket spinlock held and requires low latency 265 * access. Therefore we special case it's implementation. 266 * Note : rmem_alloc is in this structure to fill a hole 267 * on 64bit arches, not because its logically part of 268 * backlog. 269 */ 270 struct { 271 atomic_t rmem_alloc; 272 int len; 273 struct sk_buff *head; 274 struct sk_buff *tail; 275 } sk_backlog; 276 #define sk_rmem_alloc sk_backlog.rmem_alloc 277 int sk_forward_alloc; 278 #ifdef CONFIG_RPS 279 __u32 sk_rxhash; 280 #endif 281 atomic_t sk_drops; 282 int sk_rcvbuf; 283 284 struct sk_filter __rcu *sk_filter; 285 struct socket_wq __rcu *sk_wq; 286 287 #ifdef CONFIG_NET_DMA 288 struct sk_buff_head sk_async_wait_queue; 289 #endif 290 291 #ifdef CONFIG_XFRM 292 struct xfrm_policy *sk_policy[2]; 293 #endif 294 unsigned long sk_flags; 295 struct dst_entry *sk_dst_cache; 296 spinlock_t sk_dst_lock; 297 atomic_t sk_wmem_alloc; 298 atomic_t sk_omem_alloc; 299 int sk_sndbuf; 300 struct sk_buff_head sk_write_queue; 301 kmemcheck_bitfield_begin(flags); 302 unsigned int sk_shutdown : 2, 303 sk_no_check : 2, 304 sk_userlocks : 4, 305 sk_protocol : 8, 306 sk_type : 16; 307 kmemcheck_bitfield_end(flags); 308 int sk_wmem_queued; 309 gfp_t sk_allocation; 310 int sk_route_caps; 311 int sk_route_nocaps; 312 int sk_gso_type; 313 unsigned int sk_gso_max_size; 314 int sk_rcvlowat; 315 unsigned long sk_lingertime; 316 struct sk_buff_head sk_error_queue; 317 struct proto *sk_prot_creator; 318 rwlock_t sk_callback_lock; 319 int sk_err, 320 sk_err_soft; 321 unsigned short sk_ack_backlog; 322 unsigned short sk_max_ack_backlog; 323 __u32 sk_priority; 324 struct pid *sk_peer_pid; 325 const struct cred *sk_peer_cred; 326 long sk_rcvtimeo; 327 long sk_sndtimeo; 328 void *sk_protinfo; 329 struct timer_list sk_timer; 330 ktime_t sk_stamp; 331 struct socket *sk_socket; 332 void *sk_user_data; 333 struct page *sk_sndmsg_page; 334 struct sk_buff *sk_send_head; 335 __u32 sk_sndmsg_off; 336 int sk_write_pending; 337 #ifdef CONFIG_SECURITY 338 void *sk_security; 339 #endif 340 __u32 sk_mark; 341 u32 sk_classid; 342 void (*sk_state_change)(struct sock *sk); 343 void (*sk_data_ready)(struct sock *sk, int bytes); 344 void (*sk_write_space)(struct sock *sk); 345 void (*sk_error_report)(struct sock *sk); 346 int (*sk_backlog_rcv)(struct sock *sk, 347 struct sk_buff *skb); 348 void (*sk_destruct)(struct sock *sk); 349 }; 350 351 /* 352 * Hashed lists helper routines 353 */ 354 static inline struct sock *sk_entry(const struct hlist_node *node) 355 { 356 return hlist_entry(node, struct sock, sk_node); 357 } 358 359 static inline struct sock *__sk_head(const struct hlist_head *head) 360 { 361 return hlist_entry(head->first, struct sock, sk_node); 362 } 363 364 static inline struct sock *sk_head(const struct hlist_head *head) 365 { 366 return hlist_empty(head) ? NULL : __sk_head(head); 367 } 368 369 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) 370 { 371 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); 372 } 373 374 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) 375 { 376 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); 377 } 378 379 static inline struct sock *sk_next(const struct sock *sk) 380 { 381 return sk->sk_node.next ? 382 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; 383 } 384 385 static inline struct sock *sk_nulls_next(const struct sock *sk) 386 { 387 return (!is_a_nulls(sk->sk_nulls_node.next)) ? 388 hlist_nulls_entry(sk->sk_nulls_node.next, 389 struct sock, sk_nulls_node) : 390 NULL; 391 } 392 393 static inline int sk_unhashed(const struct sock *sk) 394 { 395 return hlist_unhashed(&sk->sk_node); 396 } 397 398 static inline int sk_hashed(const struct sock *sk) 399 { 400 return !sk_unhashed(sk); 401 } 402 403 static __inline__ void sk_node_init(struct hlist_node *node) 404 { 405 node->pprev = NULL; 406 } 407 408 static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node) 409 { 410 node->pprev = NULL; 411 } 412 413 static __inline__ void __sk_del_node(struct sock *sk) 414 { 415 __hlist_del(&sk->sk_node); 416 } 417 418 /* NB: equivalent to hlist_del_init_rcu */ 419 static __inline__ int __sk_del_node_init(struct sock *sk) 420 { 421 if (sk_hashed(sk)) { 422 __sk_del_node(sk); 423 sk_node_init(&sk->sk_node); 424 return 1; 425 } 426 return 0; 427 } 428 429 /* Grab socket reference count. This operation is valid only 430 when sk is ALREADY grabbed f.e. it is found in hash table 431 or a list and the lookup is made under lock preventing hash table 432 modifications. 433 */ 434 435 static inline void sock_hold(struct sock *sk) 436 { 437 atomic_inc(&sk->sk_refcnt); 438 } 439 440 /* Ungrab socket in the context, which assumes that socket refcnt 441 cannot hit zero, f.e. it is true in context of any socketcall. 442 */ 443 static inline void __sock_put(struct sock *sk) 444 { 445 atomic_dec(&sk->sk_refcnt); 446 } 447 448 static __inline__ int sk_del_node_init(struct sock *sk) 449 { 450 int rc = __sk_del_node_init(sk); 451 452 if (rc) { 453 /* paranoid for a while -acme */ 454 WARN_ON(atomic_read(&sk->sk_refcnt) == 1); 455 __sock_put(sk); 456 } 457 return rc; 458 } 459 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) 460 461 static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk) 462 { 463 if (sk_hashed(sk)) { 464 hlist_nulls_del_init_rcu(&sk->sk_nulls_node); 465 return 1; 466 } 467 return 0; 468 } 469 470 static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk) 471 { 472 int rc = __sk_nulls_del_node_init_rcu(sk); 473 474 if (rc) { 475 /* paranoid for a while -acme */ 476 WARN_ON(atomic_read(&sk->sk_refcnt) == 1); 477 __sock_put(sk); 478 } 479 return rc; 480 } 481 482 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) 483 { 484 hlist_add_head(&sk->sk_node, list); 485 } 486 487 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) 488 { 489 sock_hold(sk); 490 __sk_add_node(sk, list); 491 } 492 493 static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) 494 { 495 sock_hold(sk); 496 hlist_add_head_rcu(&sk->sk_node, list); 497 } 498 499 static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) 500 { 501 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); 502 } 503 504 static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) 505 { 506 sock_hold(sk); 507 __sk_nulls_add_node_rcu(sk, list); 508 } 509 510 static __inline__ void __sk_del_bind_node(struct sock *sk) 511 { 512 __hlist_del(&sk->sk_bind_node); 513 } 514 515 static __inline__ void sk_add_bind_node(struct sock *sk, 516 struct hlist_head *list) 517 { 518 hlist_add_head(&sk->sk_bind_node, list); 519 } 520 521 #define sk_for_each(__sk, node, list) \ 522 hlist_for_each_entry(__sk, node, list, sk_node) 523 #define sk_for_each_rcu(__sk, node, list) \ 524 hlist_for_each_entry_rcu(__sk, node, list, sk_node) 525 #define sk_nulls_for_each(__sk, node, list) \ 526 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) 527 #define sk_nulls_for_each_rcu(__sk, node, list) \ 528 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) 529 #define sk_for_each_from(__sk, node) \ 530 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ 531 hlist_for_each_entry_from(__sk, node, sk_node) 532 #define sk_nulls_for_each_from(__sk, node) \ 533 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ 534 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) 535 #define sk_for_each_safe(__sk, node, tmp, list) \ 536 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) 537 #define sk_for_each_bound(__sk, node, list) \ 538 hlist_for_each_entry(__sk, node, list, sk_bind_node) 539 540 /* Sock flags */ 541 enum sock_flags { 542 SOCK_DEAD, 543 SOCK_DONE, 544 SOCK_URGINLINE, 545 SOCK_KEEPOPEN, 546 SOCK_LINGER, 547 SOCK_DESTROY, 548 SOCK_BROADCAST, 549 SOCK_TIMESTAMP, 550 SOCK_ZAPPED, 551 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ 552 SOCK_DBG, /* %SO_DEBUG setting */ 553 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ 554 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ 555 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ 556 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ 557 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */ 558 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */ 559 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */ 560 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ 561 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */ 562 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */ 563 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */ 564 SOCK_FASYNC, /* fasync() active */ 565 SOCK_RXQ_OVFL, 566 SOCK_ZEROCOPY, /* buffers from userspace */ 567 }; 568 569 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) 570 { 571 nsk->sk_flags = osk->sk_flags; 572 } 573 574 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) 575 { 576 __set_bit(flag, &sk->sk_flags); 577 } 578 579 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) 580 { 581 __clear_bit(flag, &sk->sk_flags); 582 } 583 584 static inline int sock_flag(struct sock *sk, enum sock_flags flag) 585 { 586 return test_bit(flag, &sk->sk_flags); 587 } 588 589 static inline void sk_acceptq_removed(struct sock *sk) 590 { 591 sk->sk_ack_backlog--; 592 } 593 594 static inline void sk_acceptq_added(struct sock *sk) 595 { 596 sk->sk_ack_backlog++; 597 } 598 599 static inline int sk_acceptq_is_full(struct sock *sk) 600 { 601 return sk->sk_ack_backlog > sk->sk_max_ack_backlog; 602 } 603 604 /* 605 * Compute minimal free write space needed to queue new packets. 606 */ 607 static inline int sk_stream_min_wspace(struct sock *sk) 608 { 609 return sk->sk_wmem_queued >> 1; 610 } 611 612 static inline int sk_stream_wspace(struct sock *sk) 613 { 614 return sk->sk_sndbuf - sk->sk_wmem_queued; 615 } 616 617 extern void sk_stream_write_space(struct sock *sk); 618 619 static inline int sk_stream_memory_free(struct sock *sk) 620 { 621 return sk->sk_wmem_queued < sk->sk_sndbuf; 622 } 623 624 /* OOB backlog add */ 625 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) 626 { 627 /* dont let skb dst not refcounted, we are going to leave rcu lock */ 628 skb_dst_force(skb); 629 630 if (!sk->sk_backlog.tail) 631 sk->sk_backlog.head = skb; 632 else 633 sk->sk_backlog.tail->next = skb; 634 635 sk->sk_backlog.tail = skb; 636 skb->next = NULL; 637 } 638 639 /* 640 * Take into account size of receive queue and backlog queue 641 */ 642 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb) 643 { 644 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); 645 646 return qsize + skb->truesize > sk->sk_rcvbuf; 647 } 648 649 /* The per-socket spinlock must be held here. */ 650 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb) 651 { 652 if (sk_rcvqueues_full(sk, skb)) 653 return -ENOBUFS; 654 655 __sk_add_backlog(sk, skb); 656 sk->sk_backlog.len += skb->truesize; 657 return 0; 658 } 659 660 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) 661 { 662 return sk->sk_backlog_rcv(sk, skb); 663 } 664 665 static inline void sock_rps_record_flow(const struct sock *sk) 666 { 667 #ifdef CONFIG_RPS 668 struct rps_sock_flow_table *sock_flow_table; 669 670 rcu_read_lock(); 671 sock_flow_table = rcu_dereference(rps_sock_flow_table); 672 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash); 673 rcu_read_unlock(); 674 #endif 675 } 676 677 static inline void sock_rps_reset_flow(const struct sock *sk) 678 { 679 #ifdef CONFIG_RPS 680 struct rps_sock_flow_table *sock_flow_table; 681 682 rcu_read_lock(); 683 sock_flow_table = rcu_dereference(rps_sock_flow_table); 684 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash); 685 rcu_read_unlock(); 686 #endif 687 } 688 689 static inline void sock_rps_save_rxhash(struct sock *sk, u32 rxhash) 690 { 691 #ifdef CONFIG_RPS 692 if (unlikely(sk->sk_rxhash != rxhash)) { 693 sock_rps_reset_flow(sk); 694 sk->sk_rxhash = rxhash; 695 } 696 #endif 697 } 698 699 #define sk_wait_event(__sk, __timeo, __condition) \ 700 ({ int __rc; \ 701 release_sock(__sk); \ 702 __rc = __condition; \ 703 if (!__rc) { \ 704 *(__timeo) = schedule_timeout(*(__timeo)); \ 705 } \ 706 lock_sock(__sk); \ 707 __rc = __condition; \ 708 __rc; \ 709 }) 710 711 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); 712 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); 713 extern void sk_stream_wait_close(struct sock *sk, long timeo_p); 714 extern int sk_stream_error(struct sock *sk, int flags, int err); 715 extern void sk_stream_kill_queues(struct sock *sk); 716 717 extern int sk_wait_data(struct sock *sk, long *timeo); 718 719 struct request_sock_ops; 720 struct timewait_sock_ops; 721 struct inet_hashinfo; 722 struct raw_hashinfo; 723 724 /* Networking protocol blocks we attach to sockets. 725 * socket layer -> transport layer interface 726 * transport -> network interface is defined by struct inet_proto 727 */ 728 struct proto { 729 void (*close)(struct sock *sk, 730 long timeout); 731 int (*connect)(struct sock *sk, 732 struct sockaddr *uaddr, 733 int addr_len); 734 int (*disconnect)(struct sock *sk, int flags); 735 736 struct sock * (*accept) (struct sock *sk, int flags, int *err); 737 738 int (*ioctl)(struct sock *sk, int cmd, 739 unsigned long arg); 740 int (*init)(struct sock *sk); 741 void (*destroy)(struct sock *sk); 742 void (*shutdown)(struct sock *sk, int how); 743 int (*setsockopt)(struct sock *sk, int level, 744 int optname, char __user *optval, 745 unsigned int optlen); 746 int (*getsockopt)(struct sock *sk, int level, 747 int optname, char __user *optval, 748 int __user *option); 749 #ifdef CONFIG_COMPAT 750 int (*compat_setsockopt)(struct sock *sk, 751 int level, 752 int optname, char __user *optval, 753 unsigned int optlen); 754 int (*compat_getsockopt)(struct sock *sk, 755 int level, 756 int optname, char __user *optval, 757 int __user *option); 758 int (*compat_ioctl)(struct sock *sk, 759 unsigned int cmd, unsigned long arg); 760 #endif 761 int (*sendmsg)(struct kiocb *iocb, struct sock *sk, 762 struct msghdr *msg, size_t len); 763 int (*recvmsg)(struct kiocb *iocb, struct sock *sk, 764 struct msghdr *msg, 765 size_t len, int noblock, int flags, 766 int *addr_len); 767 int (*sendpage)(struct sock *sk, struct page *page, 768 int offset, size_t size, int flags); 769 int (*bind)(struct sock *sk, 770 struct sockaddr *uaddr, int addr_len); 771 772 int (*backlog_rcv) (struct sock *sk, 773 struct sk_buff *skb); 774 775 /* Keeping track of sk's, looking them up, and port selection methods. */ 776 void (*hash)(struct sock *sk); 777 void (*unhash)(struct sock *sk); 778 void (*rehash)(struct sock *sk); 779 int (*get_port)(struct sock *sk, unsigned short snum); 780 void (*clear_sk)(struct sock *sk, int size); 781 782 /* Keeping track of sockets in use */ 783 #ifdef CONFIG_PROC_FS 784 unsigned int inuse_idx; 785 #endif 786 787 /* Memory pressure */ 788 void (*enter_memory_pressure)(struct sock *sk); 789 atomic_long_t *memory_allocated; /* Current allocated memory. */ 790 struct percpu_counter *sockets_allocated; /* Current number of sockets. */ 791 /* 792 * Pressure flag: try to collapse. 793 * Technical note: it is used by multiple contexts non atomically. 794 * All the __sk_mem_schedule() is of this nature: accounting 795 * is strict, actions are advisory and have some latency. 796 */ 797 int *memory_pressure; 798 long *sysctl_mem; 799 int *sysctl_wmem; 800 int *sysctl_rmem; 801 int max_header; 802 bool no_autobind; 803 804 struct kmem_cache *slab; 805 unsigned int obj_size; 806 int slab_flags; 807 808 struct percpu_counter *orphan_count; 809 810 struct request_sock_ops *rsk_prot; 811 struct timewait_sock_ops *twsk_prot; 812 813 union { 814 struct inet_hashinfo *hashinfo; 815 struct udp_table *udp_table; 816 struct raw_hashinfo *raw_hash; 817 } h; 818 819 struct module *owner; 820 821 char name[32]; 822 823 struct list_head node; 824 #ifdef SOCK_REFCNT_DEBUG 825 atomic_t socks; 826 #endif 827 }; 828 829 extern int proto_register(struct proto *prot, int alloc_slab); 830 extern void proto_unregister(struct proto *prot); 831 832 #ifdef SOCK_REFCNT_DEBUG 833 static inline void sk_refcnt_debug_inc(struct sock *sk) 834 { 835 atomic_inc(&sk->sk_prot->socks); 836 } 837 838 static inline void sk_refcnt_debug_dec(struct sock *sk) 839 { 840 atomic_dec(&sk->sk_prot->socks); 841 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", 842 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); 843 } 844 845 static inline void sk_refcnt_debug_release(const struct sock *sk) 846 { 847 if (atomic_read(&sk->sk_refcnt) != 1) 848 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", 849 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); 850 } 851 #else /* SOCK_REFCNT_DEBUG */ 852 #define sk_refcnt_debug_inc(sk) do { } while (0) 853 #define sk_refcnt_debug_dec(sk) do { } while (0) 854 #define sk_refcnt_debug_release(sk) do { } while (0) 855 #endif /* SOCK_REFCNT_DEBUG */ 856 857 858 #ifdef CONFIG_PROC_FS 859 /* Called with local bh disabled */ 860 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); 861 extern int sock_prot_inuse_get(struct net *net, struct proto *proto); 862 #else 863 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot, 864 int inc) 865 { 866 } 867 #endif 868 869 870 /* With per-bucket locks this operation is not-atomic, so that 871 * this version is not worse. 872 */ 873 static inline void __sk_prot_rehash(struct sock *sk) 874 { 875 sk->sk_prot->unhash(sk); 876 sk->sk_prot->hash(sk); 877 } 878 879 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size); 880 881 /* About 10 seconds */ 882 #define SOCK_DESTROY_TIME (10*HZ) 883 884 /* Sockets 0-1023 can't be bound to unless you are superuser */ 885 #define PROT_SOCK 1024 886 887 #define SHUTDOWN_MASK 3 888 #define RCV_SHUTDOWN 1 889 #define SEND_SHUTDOWN 2 890 891 #define SOCK_SNDBUF_LOCK 1 892 #define SOCK_RCVBUF_LOCK 2 893 #define SOCK_BINDADDR_LOCK 4 894 #define SOCK_BINDPORT_LOCK 8 895 896 /* sock_iocb: used to kick off async processing of socket ios */ 897 struct sock_iocb { 898 struct list_head list; 899 900 int flags; 901 int size; 902 struct socket *sock; 903 struct sock *sk; 904 struct scm_cookie *scm; 905 struct msghdr *msg, async_msg; 906 struct kiocb *kiocb; 907 }; 908 909 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) 910 { 911 return (struct sock_iocb *)iocb->private; 912 } 913 914 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) 915 { 916 return si->kiocb; 917 } 918 919 struct socket_alloc { 920 struct socket socket; 921 struct inode vfs_inode; 922 }; 923 924 static inline struct socket *SOCKET_I(struct inode *inode) 925 { 926 return &container_of(inode, struct socket_alloc, vfs_inode)->socket; 927 } 928 929 static inline struct inode *SOCK_INODE(struct socket *socket) 930 { 931 return &container_of(socket, struct socket_alloc, socket)->vfs_inode; 932 } 933 934 /* 935 * Functions for memory accounting 936 */ 937 extern int __sk_mem_schedule(struct sock *sk, int size, int kind); 938 extern void __sk_mem_reclaim(struct sock *sk); 939 940 #define SK_MEM_QUANTUM ((int)PAGE_SIZE) 941 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) 942 #define SK_MEM_SEND 0 943 #define SK_MEM_RECV 1 944 945 static inline int sk_mem_pages(int amt) 946 { 947 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; 948 } 949 950 static inline int sk_has_account(struct sock *sk) 951 { 952 /* return true if protocol supports memory accounting */ 953 return !!sk->sk_prot->memory_allocated; 954 } 955 956 static inline int sk_wmem_schedule(struct sock *sk, int size) 957 { 958 if (!sk_has_account(sk)) 959 return 1; 960 return size <= sk->sk_forward_alloc || 961 __sk_mem_schedule(sk, size, SK_MEM_SEND); 962 } 963 964 static inline int sk_rmem_schedule(struct sock *sk, int size) 965 { 966 if (!sk_has_account(sk)) 967 return 1; 968 return size <= sk->sk_forward_alloc || 969 __sk_mem_schedule(sk, size, SK_MEM_RECV); 970 } 971 972 static inline void sk_mem_reclaim(struct sock *sk) 973 { 974 if (!sk_has_account(sk)) 975 return; 976 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) 977 __sk_mem_reclaim(sk); 978 } 979 980 static inline void sk_mem_reclaim_partial(struct sock *sk) 981 { 982 if (!sk_has_account(sk)) 983 return; 984 if (sk->sk_forward_alloc > SK_MEM_QUANTUM) 985 __sk_mem_reclaim(sk); 986 } 987 988 static inline void sk_mem_charge(struct sock *sk, int size) 989 { 990 if (!sk_has_account(sk)) 991 return; 992 sk->sk_forward_alloc -= size; 993 } 994 995 static inline void sk_mem_uncharge(struct sock *sk, int size) 996 { 997 if (!sk_has_account(sk)) 998 return; 999 sk->sk_forward_alloc += size; 1000 } 1001 1002 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) 1003 { 1004 sock_set_flag(sk, SOCK_QUEUE_SHRUNK); 1005 sk->sk_wmem_queued -= skb->truesize; 1006 sk_mem_uncharge(sk, skb->truesize); 1007 __kfree_skb(skb); 1008 } 1009 1010 /* Used by processes to "lock" a socket state, so that 1011 * interrupts and bottom half handlers won't change it 1012 * from under us. It essentially blocks any incoming 1013 * packets, so that we won't get any new data or any 1014 * packets that change the state of the socket. 1015 * 1016 * While locked, BH processing will add new packets to 1017 * the backlog queue. This queue is processed by the 1018 * owner of the socket lock right before it is released. 1019 * 1020 * Since ~2.3.5 it is also exclusive sleep lock serializing 1021 * accesses from user process context. 1022 */ 1023 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) 1024 1025 /* 1026 * Macro so as to not evaluate some arguments when 1027 * lockdep is not enabled. 1028 * 1029 * Mark both the sk_lock and the sk_lock.slock as a 1030 * per-address-family lock class. 1031 */ 1032 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ 1033 do { \ 1034 sk->sk_lock.owned = 0; \ 1035 init_waitqueue_head(&sk->sk_lock.wq); \ 1036 spin_lock_init(&(sk)->sk_lock.slock); \ 1037 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ 1038 sizeof((sk)->sk_lock)); \ 1039 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ 1040 (skey), (sname)); \ 1041 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ 1042 } while (0) 1043 1044 extern void lock_sock_nested(struct sock *sk, int subclass); 1045 1046 static inline void lock_sock(struct sock *sk) 1047 { 1048 lock_sock_nested(sk, 0); 1049 } 1050 1051 extern void release_sock(struct sock *sk); 1052 1053 /* BH context may only use the following locking interface. */ 1054 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) 1055 #define bh_lock_sock_nested(__sk) \ 1056 spin_lock_nested(&((__sk)->sk_lock.slock), \ 1057 SINGLE_DEPTH_NESTING) 1058 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) 1059 1060 extern bool lock_sock_fast(struct sock *sk); 1061 /** 1062 * unlock_sock_fast - complement of lock_sock_fast 1063 * @sk: socket 1064 * @slow: slow mode 1065 * 1066 * fast unlock socket for user context. 1067 * If slow mode is on, we call regular release_sock() 1068 */ 1069 static inline void unlock_sock_fast(struct sock *sk, bool slow) 1070 { 1071 if (slow) 1072 release_sock(sk); 1073 else 1074 spin_unlock_bh(&sk->sk_lock.slock); 1075 } 1076 1077 1078 extern struct sock *sk_alloc(struct net *net, int family, 1079 gfp_t priority, 1080 struct proto *prot); 1081 extern void sk_free(struct sock *sk); 1082 extern void sk_release_kernel(struct sock *sk); 1083 extern struct sock *sk_clone(const struct sock *sk, 1084 const gfp_t priority); 1085 1086 extern struct sk_buff *sock_wmalloc(struct sock *sk, 1087 unsigned long size, int force, 1088 gfp_t priority); 1089 extern struct sk_buff *sock_rmalloc(struct sock *sk, 1090 unsigned long size, int force, 1091 gfp_t priority); 1092 extern void sock_wfree(struct sk_buff *skb); 1093 extern void sock_rfree(struct sk_buff *skb); 1094 1095 extern int sock_setsockopt(struct socket *sock, int level, 1096 int op, char __user *optval, 1097 unsigned int optlen); 1098 1099 extern int sock_getsockopt(struct socket *sock, int level, 1100 int op, char __user *optval, 1101 int __user *optlen); 1102 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, 1103 unsigned long size, 1104 int noblock, 1105 int *errcode); 1106 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk, 1107 unsigned long header_len, 1108 unsigned long data_len, 1109 int noblock, 1110 int *errcode); 1111 extern void *sock_kmalloc(struct sock *sk, int size, 1112 gfp_t priority); 1113 extern void sock_kfree_s(struct sock *sk, void *mem, int size); 1114 extern void sk_send_sigurg(struct sock *sk); 1115 1116 #ifdef CONFIG_CGROUPS 1117 extern void sock_update_classid(struct sock *sk); 1118 #else 1119 static inline void sock_update_classid(struct sock *sk) 1120 { 1121 } 1122 #endif 1123 1124 /* 1125 * Functions to fill in entries in struct proto_ops when a protocol 1126 * does not implement a particular function. 1127 */ 1128 extern int sock_no_bind(struct socket *, 1129 struct sockaddr *, int); 1130 extern int sock_no_connect(struct socket *, 1131 struct sockaddr *, int, int); 1132 extern int sock_no_socketpair(struct socket *, 1133 struct socket *); 1134 extern int sock_no_accept(struct socket *, 1135 struct socket *, int); 1136 extern int sock_no_getname(struct socket *, 1137 struct sockaddr *, int *, int); 1138 extern unsigned int sock_no_poll(struct file *, struct socket *, 1139 struct poll_table_struct *); 1140 extern int sock_no_ioctl(struct socket *, unsigned int, 1141 unsigned long); 1142 extern int sock_no_listen(struct socket *, int); 1143 extern int sock_no_shutdown(struct socket *, int); 1144 extern int sock_no_getsockopt(struct socket *, int , int, 1145 char __user *, int __user *); 1146 extern int sock_no_setsockopt(struct socket *, int, int, 1147 char __user *, unsigned int); 1148 extern int sock_no_sendmsg(struct kiocb *, struct socket *, 1149 struct msghdr *, size_t); 1150 extern int sock_no_recvmsg(struct kiocb *, struct socket *, 1151 struct msghdr *, size_t, int); 1152 extern int sock_no_mmap(struct file *file, 1153 struct socket *sock, 1154 struct vm_area_struct *vma); 1155 extern ssize_t sock_no_sendpage(struct socket *sock, 1156 struct page *page, 1157 int offset, size_t size, 1158 int flags); 1159 1160 /* 1161 * Functions to fill in entries in struct proto_ops when a protocol 1162 * uses the inet style. 1163 */ 1164 extern int sock_common_getsockopt(struct socket *sock, int level, int optname, 1165 char __user *optval, int __user *optlen); 1166 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, 1167 struct msghdr *msg, size_t size, int flags); 1168 extern int sock_common_setsockopt(struct socket *sock, int level, int optname, 1169 char __user *optval, unsigned int optlen); 1170 extern int compat_sock_common_getsockopt(struct socket *sock, int level, 1171 int optname, char __user *optval, int __user *optlen); 1172 extern int compat_sock_common_setsockopt(struct socket *sock, int level, 1173 int optname, char __user *optval, unsigned int optlen); 1174 1175 extern void sk_common_release(struct sock *sk); 1176 1177 /* 1178 * Default socket callbacks and setup code 1179 */ 1180 1181 /* Initialise core socket variables */ 1182 extern void sock_init_data(struct socket *sock, struct sock *sk); 1183 1184 extern void sk_filter_release_rcu(struct rcu_head *rcu); 1185 1186 /** 1187 * sk_filter_release - release a socket filter 1188 * @fp: filter to remove 1189 * 1190 * Remove a filter from a socket and release its resources. 1191 */ 1192 1193 static inline void sk_filter_release(struct sk_filter *fp) 1194 { 1195 if (atomic_dec_and_test(&fp->refcnt)) 1196 call_rcu(&fp->rcu, sk_filter_release_rcu); 1197 } 1198 1199 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp) 1200 { 1201 unsigned int size = sk_filter_len(fp); 1202 1203 atomic_sub(size, &sk->sk_omem_alloc); 1204 sk_filter_release(fp); 1205 } 1206 1207 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) 1208 { 1209 atomic_inc(&fp->refcnt); 1210 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); 1211 } 1212 1213 /* 1214 * Socket reference counting postulates. 1215 * 1216 * * Each user of socket SHOULD hold a reference count. 1217 * * Each access point to socket (an hash table bucket, reference from a list, 1218 * running timer, skb in flight MUST hold a reference count. 1219 * * When reference count hits 0, it means it will never increase back. 1220 * * When reference count hits 0, it means that no references from 1221 * outside exist to this socket and current process on current CPU 1222 * is last user and may/should destroy this socket. 1223 * * sk_free is called from any context: process, BH, IRQ. When 1224 * it is called, socket has no references from outside -> sk_free 1225 * may release descendant resources allocated by the socket, but 1226 * to the time when it is called, socket is NOT referenced by any 1227 * hash tables, lists etc. 1228 * * Packets, delivered from outside (from network or from another process) 1229 * and enqueued on receive/error queues SHOULD NOT grab reference count, 1230 * when they sit in queue. Otherwise, packets will leak to hole, when 1231 * socket is looked up by one cpu and unhasing is made by another CPU. 1232 * It is true for udp/raw, netlink (leak to receive and error queues), tcp 1233 * (leak to backlog). Packet socket does all the processing inside 1234 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets 1235 * use separate SMP lock, so that they are prone too. 1236 */ 1237 1238 /* Ungrab socket and destroy it, if it was the last reference. */ 1239 static inline void sock_put(struct sock *sk) 1240 { 1241 if (atomic_dec_and_test(&sk->sk_refcnt)) 1242 sk_free(sk); 1243 } 1244 1245 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, 1246 const int nested); 1247 1248 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) 1249 { 1250 sk->sk_tx_queue_mapping = tx_queue; 1251 } 1252 1253 static inline void sk_tx_queue_clear(struct sock *sk) 1254 { 1255 sk->sk_tx_queue_mapping = -1; 1256 } 1257 1258 static inline int sk_tx_queue_get(const struct sock *sk) 1259 { 1260 return sk ? sk->sk_tx_queue_mapping : -1; 1261 } 1262 1263 static inline void sk_set_socket(struct sock *sk, struct socket *sock) 1264 { 1265 sk_tx_queue_clear(sk); 1266 sk->sk_socket = sock; 1267 } 1268 1269 static inline wait_queue_head_t *sk_sleep(struct sock *sk) 1270 { 1271 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); 1272 return &rcu_dereference_raw(sk->sk_wq)->wait; 1273 } 1274 /* Detach socket from process context. 1275 * Announce socket dead, detach it from wait queue and inode. 1276 * Note that parent inode held reference count on this struct sock, 1277 * we do not release it in this function, because protocol 1278 * probably wants some additional cleanups or even continuing 1279 * to work with this socket (TCP). 1280 */ 1281 static inline void sock_orphan(struct sock *sk) 1282 { 1283 write_lock_bh(&sk->sk_callback_lock); 1284 sock_set_flag(sk, SOCK_DEAD); 1285 sk_set_socket(sk, NULL); 1286 sk->sk_wq = NULL; 1287 write_unlock_bh(&sk->sk_callback_lock); 1288 } 1289 1290 static inline void sock_graft(struct sock *sk, struct socket *parent) 1291 { 1292 write_lock_bh(&sk->sk_callback_lock); 1293 sk->sk_wq = parent->wq; 1294 parent->sk = sk; 1295 sk_set_socket(sk, parent); 1296 security_sock_graft(sk, parent); 1297 write_unlock_bh(&sk->sk_callback_lock); 1298 } 1299 1300 extern int sock_i_uid(struct sock *sk); 1301 extern unsigned long sock_i_ino(struct sock *sk); 1302 1303 static inline struct dst_entry * 1304 __sk_dst_get(struct sock *sk) 1305 { 1306 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) || 1307 lockdep_is_held(&sk->sk_lock.slock)); 1308 } 1309 1310 static inline struct dst_entry * 1311 sk_dst_get(struct sock *sk) 1312 { 1313 struct dst_entry *dst; 1314 1315 rcu_read_lock(); 1316 dst = rcu_dereference(sk->sk_dst_cache); 1317 if (dst) 1318 dst_hold(dst); 1319 rcu_read_unlock(); 1320 return dst; 1321 } 1322 1323 extern void sk_reset_txq(struct sock *sk); 1324 1325 static inline void dst_negative_advice(struct sock *sk) 1326 { 1327 struct dst_entry *ndst, *dst = __sk_dst_get(sk); 1328 1329 if (dst && dst->ops->negative_advice) { 1330 ndst = dst->ops->negative_advice(dst); 1331 1332 if (ndst != dst) { 1333 rcu_assign_pointer(sk->sk_dst_cache, ndst); 1334 sk_reset_txq(sk); 1335 } 1336 } 1337 } 1338 1339 static inline void 1340 __sk_dst_set(struct sock *sk, struct dst_entry *dst) 1341 { 1342 struct dst_entry *old_dst; 1343 1344 sk_tx_queue_clear(sk); 1345 /* 1346 * This can be called while sk is owned by the caller only, 1347 * with no state that can be checked in a rcu_dereference_check() cond 1348 */ 1349 old_dst = rcu_dereference_raw(sk->sk_dst_cache); 1350 rcu_assign_pointer(sk->sk_dst_cache, dst); 1351 dst_release(old_dst); 1352 } 1353 1354 static inline void 1355 sk_dst_set(struct sock *sk, struct dst_entry *dst) 1356 { 1357 spin_lock(&sk->sk_dst_lock); 1358 __sk_dst_set(sk, dst); 1359 spin_unlock(&sk->sk_dst_lock); 1360 } 1361 1362 static inline void 1363 __sk_dst_reset(struct sock *sk) 1364 { 1365 __sk_dst_set(sk, NULL); 1366 } 1367 1368 static inline void 1369 sk_dst_reset(struct sock *sk) 1370 { 1371 spin_lock(&sk->sk_dst_lock); 1372 __sk_dst_reset(sk); 1373 spin_unlock(&sk->sk_dst_lock); 1374 } 1375 1376 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); 1377 1378 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); 1379 1380 static inline int sk_can_gso(const struct sock *sk) 1381 { 1382 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); 1383 } 1384 1385 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst); 1386 1387 static inline void sk_nocaps_add(struct sock *sk, int flags) 1388 { 1389 sk->sk_route_nocaps |= flags; 1390 sk->sk_route_caps &= ~flags; 1391 } 1392 1393 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, 1394 char __user *from, char *to, 1395 int copy, int offset) 1396 { 1397 if (skb->ip_summed == CHECKSUM_NONE) { 1398 int err = 0; 1399 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err); 1400 if (err) 1401 return err; 1402 skb->csum = csum_block_add(skb->csum, csum, offset); 1403 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { 1404 if (!access_ok(VERIFY_READ, from, copy) || 1405 __copy_from_user_nocache(to, from, copy)) 1406 return -EFAULT; 1407 } else if (copy_from_user(to, from, copy)) 1408 return -EFAULT; 1409 1410 return 0; 1411 } 1412 1413 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, 1414 char __user *from, int copy) 1415 { 1416 int err, offset = skb->len; 1417 1418 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), 1419 copy, offset); 1420 if (err) 1421 __skb_trim(skb, offset); 1422 1423 return err; 1424 } 1425 1426 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from, 1427 struct sk_buff *skb, 1428 struct page *page, 1429 int off, int copy) 1430 { 1431 int err; 1432 1433 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, 1434 copy, skb->len); 1435 if (err) 1436 return err; 1437 1438 skb->len += copy; 1439 skb->data_len += copy; 1440 skb->truesize += copy; 1441 sk->sk_wmem_queued += copy; 1442 sk_mem_charge(sk, copy); 1443 return 0; 1444 } 1445 1446 static inline int skb_copy_to_page(struct sock *sk, char __user *from, 1447 struct sk_buff *skb, struct page *page, 1448 int off, int copy) 1449 { 1450 if (skb->ip_summed == CHECKSUM_NONE) { 1451 int err = 0; 1452 __wsum csum = csum_and_copy_from_user(from, 1453 page_address(page) + off, 1454 copy, 0, &err); 1455 if (err) 1456 return err; 1457 skb->csum = csum_block_add(skb->csum, csum, skb->len); 1458 } else if (copy_from_user(page_address(page) + off, from, copy)) 1459 return -EFAULT; 1460 1461 skb->len += copy; 1462 skb->data_len += copy; 1463 skb->truesize += copy; 1464 sk->sk_wmem_queued += copy; 1465 sk_mem_charge(sk, copy); 1466 return 0; 1467 } 1468 1469 /** 1470 * sk_wmem_alloc_get - returns write allocations 1471 * @sk: socket 1472 * 1473 * Returns sk_wmem_alloc minus initial offset of one 1474 */ 1475 static inline int sk_wmem_alloc_get(const struct sock *sk) 1476 { 1477 return atomic_read(&sk->sk_wmem_alloc) - 1; 1478 } 1479 1480 /** 1481 * sk_rmem_alloc_get - returns read allocations 1482 * @sk: socket 1483 * 1484 * Returns sk_rmem_alloc 1485 */ 1486 static inline int sk_rmem_alloc_get(const struct sock *sk) 1487 { 1488 return atomic_read(&sk->sk_rmem_alloc); 1489 } 1490 1491 /** 1492 * sk_has_allocations - check if allocations are outstanding 1493 * @sk: socket 1494 * 1495 * Returns true if socket has write or read allocations 1496 */ 1497 static inline int sk_has_allocations(const struct sock *sk) 1498 { 1499 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); 1500 } 1501 1502 /** 1503 * wq_has_sleeper - check if there are any waiting processes 1504 * @wq: struct socket_wq 1505 * 1506 * Returns true if socket_wq has waiting processes 1507 * 1508 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory 1509 * barrier call. They were added due to the race found within the tcp code. 1510 * 1511 * Consider following tcp code paths: 1512 * 1513 * CPU1 CPU2 1514 * 1515 * sys_select receive packet 1516 * ... ... 1517 * __add_wait_queue update tp->rcv_nxt 1518 * ... ... 1519 * tp->rcv_nxt check sock_def_readable 1520 * ... { 1521 * schedule rcu_read_lock(); 1522 * wq = rcu_dereference(sk->sk_wq); 1523 * if (wq && waitqueue_active(&wq->wait)) 1524 * wake_up_interruptible(&wq->wait) 1525 * ... 1526 * } 1527 * 1528 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay 1529 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 1530 * could then endup calling schedule and sleep forever if there are no more 1531 * data on the socket. 1532 * 1533 */ 1534 static inline bool wq_has_sleeper(struct socket_wq *wq) 1535 { 1536 1537 /* 1538 * We need to be sure we are in sync with the 1539 * add_wait_queue modifications to the wait queue. 1540 * 1541 * This memory barrier is paired in the sock_poll_wait. 1542 */ 1543 smp_mb(); 1544 return wq && waitqueue_active(&wq->wait); 1545 } 1546 1547 /** 1548 * sock_poll_wait - place memory barrier behind the poll_wait call. 1549 * @filp: file 1550 * @wait_address: socket wait queue 1551 * @p: poll_table 1552 * 1553 * See the comments in the wq_has_sleeper function. 1554 */ 1555 static inline void sock_poll_wait(struct file *filp, 1556 wait_queue_head_t *wait_address, poll_table *p) 1557 { 1558 if (p && wait_address) { 1559 poll_wait(filp, wait_address, p); 1560 /* 1561 * We need to be sure we are in sync with the 1562 * socket flags modification. 1563 * 1564 * This memory barrier is paired in the wq_has_sleeper. 1565 */ 1566 smp_mb(); 1567 } 1568 } 1569 1570 /* 1571 * Queue a received datagram if it will fit. Stream and sequenced 1572 * protocols can't normally use this as they need to fit buffers in 1573 * and play with them. 1574 * 1575 * Inlined as it's very short and called for pretty much every 1576 * packet ever received. 1577 */ 1578 1579 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) 1580 { 1581 skb_orphan(skb); 1582 skb->sk = sk; 1583 skb->destructor = sock_wfree; 1584 /* 1585 * We used to take a refcount on sk, but following operation 1586 * is enough to guarantee sk_free() wont free this sock until 1587 * all in-flight packets are completed 1588 */ 1589 atomic_add(skb->truesize, &sk->sk_wmem_alloc); 1590 } 1591 1592 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) 1593 { 1594 skb_orphan(skb); 1595 skb->sk = sk; 1596 skb->destructor = sock_rfree; 1597 atomic_add(skb->truesize, &sk->sk_rmem_alloc); 1598 sk_mem_charge(sk, skb->truesize); 1599 } 1600 1601 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, 1602 unsigned long expires); 1603 1604 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); 1605 1606 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); 1607 1608 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); 1609 1610 /* 1611 * Recover an error report and clear atomically 1612 */ 1613 1614 static inline int sock_error(struct sock *sk) 1615 { 1616 int err; 1617 if (likely(!sk->sk_err)) 1618 return 0; 1619 err = xchg(&sk->sk_err, 0); 1620 return -err; 1621 } 1622 1623 static inline unsigned long sock_wspace(struct sock *sk) 1624 { 1625 int amt = 0; 1626 1627 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { 1628 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); 1629 if (amt < 0) 1630 amt = 0; 1631 } 1632 return amt; 1633 } 1634 1635 static inline void sk_wake_async(struct sock *sk, int how, int band) 1636 { 1637 if (sock_flag(sk, SOCK_FASYNC)) 1638 sock_wake_async(sk->sk_socket, how, band); 1639 } 1640 1641 #define SOCK_MIN_SNDBUF 2048 1642 /* 1643 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need 1644 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak 1645 */ 1646 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff)) 1647 1648 static inline void sk_stream_moderate_sndbuf(struct sock *sk) 1649 { 1650 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { 1651 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); 1652 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); 1653 } 1654 } 1655 1656 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp); 1657 1658 static inline struct page *sk_stream_alloc_page(struct sock *sk) 1659 { 1660 struct page *page = NULL; 1661 1662 page = alloc_pages(sk->sk_allocation, 0); 1663 if (!page) { 1664 sk->sk_prot->enter_memory_pressure(sk); 1665 sk_stream_moderate_sndbuf(sk); 1666 } 1667 return page; 1668 } 1669 1670 /* 1671 * Default write policy as shown to user space via poll/select/SIGIO 1672 */ 1673 static inline int sock_writeable(const struct sock *sk) 1674 { 1675 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); 1676 } 1677 1678 static inline gfp_t gfp_any(void) 1679 { 1680 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; 1681 } 1682 1683 static inline long sock_rcvtimeo(const struct sock *sk, int noblock) 1684 { 1685 return noblock ? 0 : sk->sk_rcvtimeo; 1686 } 1687 1688 static inline long sock_sndtimeo(const struct sock *sk, int noblock) 1689 { 1690 return noblock ? 0 : sk->sk_sndtimeo; 1691 } 1692 1693 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) 1694 { 1695 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; 1696 } 1697 1698 /* Alas, with timeout socket operations are not restartable. 1699 * Compare this to poll(). 1700 */ 1701 static inline int sock_intr_errno(long timeo) 1702 { 1703 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; 1704 } 1705 1706 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 1707 struct sk_buff *skb); 1708 1709 static __inline__ void 1710 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) 1711 { 1712 ktime_t kt = skb->tstamp; 1713 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); 1714 1715 /* 1716 * generate control messages if 1717 * - receive time stamping in software requested (SOCK_RCVTSTAMP 1718 * or SOCK_TIMESTAMPING_RX_SOFTWARE) 1719 * - software time stamp available and wanted 1720 * (SOCK_TIMESTAMPING_SOFTWARE) 1721 * - hardware time stamps available and wanted 1722 * (SOCK_TIMESTAMPING_SYS_HARDWARE or 1723 * SOCK_TIMESTAMPING_RAW_HARDWARE) 1724 */ 1725 if (sock_flag(sk, SOCK_RCVTSTAMP) || 1726 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) || 1727 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) || 1728 (hwtstamps->hwtstamp.tv64 && 1729 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) || 1730 (hwtstamps->syststamp.tv64 && 1731 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))) 1732 __sock_recv_timestamp(msg, sk, skb); 1733 else 1734 sk->sk_stamp = kt; 1735 } 1736 1737 extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, 1738 struct sk_buff *skb); 1739 1740 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, 1741 struct sk_buff *skb) 1742 { 1743 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ 1744 (1UL << SOCK_RCVTSTAMP) | \ 1745 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \ 1746 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \ 1747 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \ 1748 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE)) 1749 1750 if (sk->sk_flags & FLAGS_TS_OR_DROPS) 1751 __sock_recv_ts_and_drops(msg, sk, skb); 1752 else 1753 sk->sk_stamp = skb->tstamp; 1754 } 1755 1756 /** 1757 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped 1758 * @sk: socket sending this packet 1759 * @tx_flags: filled with instructions for time stamping 1760 * 1761 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if 1762 * parameters are invalid. 1763 */ 1764 extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags); 1765 1766 /** 1767 * sk_eat_skb - Release a skb if it is no longer needed 1768 * @sk: socket to eat this skb from 1769 * @skb: socket buffer to eat 1770 * @copied_early: flag indicating whether DMA operations copied this data early 1771 * 1772 * This routine must be called with interrupts disabled or with the socket 1773 * locked so that the sk_buff queue operation is ok. 1774 */ 1775 #ifdef CONFIG_NET_DMA 1776 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) 1777 { 1778 __skb_unlink(skb, &sk->sk_receive_queue); 1779 if (!copied_early) 1780 __kfree_skb(skb); 1781 else 1782 __skb_queue_tail(&sk->sk_async_wait_queue, skb); 1783 } 1784 #else 1785 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) 1786 { 1787 __skb_unlink(skb, &sk->sk_receive_queue); 1788 __kfree_skb(skb); 1789 } 1790 #endif 1791 1792 static inline 1793 struct net *sock_net(const struct sock *sk) 1794 { 1795 return read_pnet(&sk->sk_net); 1796 } 1797 1798 static inline 1799 void sock_net_set(struct sock *sk, struct net *net) 1800 { 1801 write_pnet(&sk->sk_net, net); 1802 } 1803 1804 /* 1805 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace. 1806 * They should not hold a reference to a namespace in order to allow 1807 * to stop it. 1808 * Sockets after sk_change_net should be released using sk_release_kernel 1809 */ 1810 static inline void sk_change_net(struct sock *sk, struct net *net) 1811 { 1812 put_net(sock_net(sk)); 1813 sock_net_set(sk, hold_net(net)); 1814 } 1815 1816 static inline struct sock *skb_steal_sock(struct sk_buff *skb) 1817 { 1818 if (unlikely(skb->sk)) { 1819 struct sock *sk = skb->sk; 1820 1821 skb->destructor = NULL; 1822 skb->sk = NULL; 1823 return sk; 1824 } 1825 return NULL; 1826 } 1827 1828 extern void sock_enable_timestamp(struct sock *sk, int flag); 1829 extern int sock_get_timestamp(struct sock *, struct timeval __user *); 1830 extern int sock_get_timestampns(struct sock *, struct timespec __user *); 1831 1832 /* 1833 * Enable debug/info messages 1834 */ 1835 extern int net_msg_warn; 1836 #define NETDEBUG(fmt, args...) \ 1837 do { if (net_msg_warn) printk(fmt,##args); } while (0) 1838 1839 #define LIMIT_NETDEBUG(fmt, args...) \ 1840 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) 1841 1842 extern __u32 sysctl_wmem_max; 1843 extern __u32 sysctl_rmem_max; 1844 1845 extern void sk_init(void); 1846 1847 extern int sysctl_optmem_max; 1848 1849 extern __u32 sysctl_wmem_default; 1850 extern __u32 sysctl_rmem_default; 1851 1852 #endif /* _SOCK_H */ 1853