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/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 #include <linux/slab.h> 55 #include <linux/uaccess.h> 56 #include <linux/memcontrol.h> 57 #include <linux/res_counter.h> 58 59 #include <linux/filter.h> 60 #include <linux/rculist_nulls.h> 61 #include <linux/poll.h> 62 63 #include <linux/atomic.h> 64 #include <net/dst.h> 65 #include <net/checksum.h> 66 67 struct cgroup; 68 struct cgroup_subsys; 69 #ifdef CONFIG_NET 70 int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss); 71 void mem_cgroup_sockets_destroy(struct cgroup *cgrp, struct cgroup_subsys *ss); 72 #else 73 static inline 74 int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss) 75 { 76 return 0; 77 } 78 static inline 79 void mem_cgroup_sockets_destroy(struct cgroup *cgrp, struct cgroup_subsys *ss) 80 { 81 } 82 #endif 83 /* 84 * This structure really needs to be cleaned up. 85 * Most of it is for TCP, and not used by any of 86 * the other protocols. 87 */ 88 89 /* Define this to get the SOCK_DBG debugging facility. */ 90 #define SOCK_DEBUGGING 91 #ifdef SOCK_DEBUGGING 92 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ 93 printk(KERN_DEBUG msg); } while (0) 94 #else 95 /* Validate arguments and do nothing */ 96 static inline __printf(2, 3) 97 void SOCK_DEBUG(struct sock *sk, const char *msg, ...) 98 { 99 } 100 #endif 101 102 /* This is the per-socket lock. The spinlock provides a synchronization 103 * between user contexts and software interrupt processing, whereas the 104 * mini-semaphore synchronizes multiple users amongst themselves. 105 */ 106 typedef struct { 107 spinlock_t slock; 108 int owned; 109 wait_queue_head_t wq; 110 /* 111 * We express the mutex-alike socket_lock semantics 112 * to the lock validator by explicitly managing 113 * the slock as a lock variant (in addition to 114 * the slock itself): 115 */ 116 #ifdef CONFIG_DEBUG_LOCK_ALLOC 117 struct lockdep_map dep_map; 118 #endif 119 } socket_lock_t; 120 121 struct sock; 122 struct proto; 123 struct net; 124 125 /** 126 * struct sock_common - minimal network layer representation of sockets 127 * @skc_daddr: Foreign IPv4 addr 128 * @skc_rcv_saddr: Bound local IPv4 addr 129 * @skc_hash: hash value used with various protocol lookup tables 130 * @skc_u16hashes: two u16 hash values used by UDP lookup tables 131 * @skc_family: network address family 132 * @skc_state: Connection state 133 * @skc_reuse: %SO_REUSEADDR setting 134 * @skc_bound_dev_if: bound device index if != 0 135 * @skc_bind_node: bind hash linkage for various protocol lookup tables 136 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol 137 * @skc_prot: protocol handlers inside a network family 138 * @skc_net: reference to the network namespace of this socket 139 * @skc_node: main hash linkage for various protocol lookup tables 140 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol 141 * @skc_tx_queue_mapping: tx queue number for this connection 142 * @skc_refcnt: reference count 143 * 144 * This is the minimal network layer representation of sockets, the header 145 * for struct sock and struct inet_timewait_sock. 146 */ 147 struct sock_common { 148 /* skc_daddr and skc_rcv_saddr must be grouped : 149 * cf INET_MATCH() and INET_TW_MATCH() 150 */ 151 __be32 skc_daddr; 152 __be32 skc_rcv_saddr; 153 154 union { 155 unsigned int skc_hash; 156 __u16 skc_u16hashes[2]; 157 }; 158 unsigned short skc_family; 159 volatile unsigned char skc_state; 160 unsigned char skc_reuse; 161 int skc_bound_dev_if; 162 union { 163 struct hlist_node skc_bind_node; 164 struct hlist_nulls_node skc_portaddr_node; 165 }; 166 struct proto *skc_prot; 167 #ifdef CONFIG_NET_NS 168 struct net *skc_net; 169 #endif 170 /* 171 * fields between dontcopy_begin/dontcopy_end 172 * are not copied in sock_copy() 173 */ 174 /* private: */ 175 int skc_dontcopy_begin[0]; 176 /* public: */ 177 union { 178 struct hlist_node skc_node; 179 struct hlist_nulls_node skc_nulls_node; 180 }; 181 int skc_tx_queue_mapping; 182 atomic_t skc_refcnt; 183 /* private: */ 184 int skc_dontcopy_end[0]; 185 /* public: */ 186 }; 187 188 struct cg_proto; 189 /** 190 * struct sock - network layer representation of sockets 191 * @__sk_common: shared layout with inet_timewait_sock 192 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN 193 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings 194 * @sk_lock: synchronizer 195 * @sk_rcvbuf: size of receive buffer in bytes 196 * @sk_wq: sock wait queue and async head 197 * @sk_dst_cache: destination cache 198 * @sk_dst_lock: destination cache lock 199 * @sk_policy: flow policy 200 * @sk_receive_queue: incoming packets 201 * @sk_wmem_alloc: transmit queue bytes committed 202 * @sk_write_queue: Packet sending queue 203 * @sk_async_wait_queue: DMA copied packets 204 * @sk_omem_alloc: "o" is "option" or "other" 205 * @sk_wmem_queued: persistent queue size 206 * @sk_forward_alloc: space allocated forward 207 * @sk_allocation: allocation mode 208 * @sk_sndbuf: size of send buffer in bytes 209 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, 210 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings 211 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets 212 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) 213 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) 214 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) 215 * @sk_gso_max_size: Maximum GSO segment size to build 216 * @sk_lingertime: %SO_LINGER l_linger setting 217 * @sk_backlog: always used with the per-socket spinlock held 218 * @sk_callback_lock: used with the callbacks in the end of this struct 219 * @sk_error_queue: rarely used 220 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, 221 * IPV6_ADDRFORM for instance) 222 * @sk_err: last error 223 * @sk_err_soft: errors that don't cause failure but are the cause of a 224 * persistent failure not just 'timed out' 225 * @sk_drops: raw/udp drops counter 226 * @sk_ack_backlog: current listen backlog 227 * @sk_max_ack_backlog: listen backlog set in listen() 228 * @sk_priority: %SO_PRIORITY setting 229 * @sk_type: socket type (%SOCK_STREAM, etc) 230 * @sk_protocol: which protocol this socket belongs in this network family 231 * @sk_peer_pid: &struct pid for this socket's peer 232 * @sk_peer_cred: %SO_PEERCRED setting 233 * @sk_rcvlowat: %SO_RCVLOWAT setting 234 * @sk_rcvtimeo: %SO_RCVTIMEO setting 235 * @sk_sndtimeo: %SO_SNDTIMEO setting 236 * @sk_rxhash: flow hash received from netif layer 237 * @sk_filter: socket filtering instructions 238 * @sk_protinfo: private area, net family specific, when not using slab 239 * @sk_timer: sock cleanup timer 240 * @sk_stamp: time stamp of last packet received 241 * @sk_socket: Identd and reporting IO signals 242 * @sk_user_data: RPC layer private data 243 * @sk_sndmsg_page: cached page for sendmsg 244 * @sk_sndmsg_off: cached offset for sendmsg 245 * @sk_send_head: front of stuff to transmit 246 * @sk_security: used by security modules 247 * @sk_mark: generic packet mark 248 * @sk_classid: this socket's cgroup classid 249 * @sk_cgrp: this socket's cgroup-specific proto data 250 * @sk_write_pending: a write to stream socket waits to start 251 * @sk_state_change: callback to indicate change in the state of the sock 252 * @sk_data_ready: callback to indicate there is data to be processed 253 * @sk_write_space: callback to indicate there is bf sending space available 254 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) 255 * @sk_backlog_rcv: callback to process the backlog 256 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 257 */ 258 struct sock { 259 /* 260 * Now struct inet_timewait_sock also uses sock_common, so please just 261 * don't add nothing before this first member (__sk_common) --acme 262 */ 263 struct sock_common __sk_common; 264 #define sk_node __sk_common.skc_node 265 #define sk_nulls_node __sk_common.skc_nulls_node 266 #define sk_refcnt __sk_common.skc_refcnt 267 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping 268 269 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin 270 #define sk_dontcopy_end __sk_common.skc_dontcopy_end 271 #define sk_hash __sk_common.skc_hash 272 #define sk_family __sk_common.skc_family 273 #define sk_state __sk_common.skc_state 274 #define sk_reuse __sk_common.skc_reuse 275 #define sk_bound_dev_if __sk_common.skc_bound_dev_if 276 #define sk_bind_node __sk_common.skc_bind_node 277 #define sk_prot __sk_common.skc_prot 278 #define sk_net __sk_common.skc_net 279 socket_lock_t sk_lock; 280 struct sk_buff_head sk_receive_queue; 281 /* 282 * The backlog queue is special, it is always used with 283 * the per-socket spinlock held and requires low latency 284 * access. Therefore we special case it's implementation. 285 * Note : rmem_alloc is in this structure to fill a hole 286 * on 64bit arches, not because its logically part of 287 * backlog. 288 */ 289 struct { 290 atomic_t rmem_alloc; 291 int len; 292 struct sk_buff *head; 293 struct sk_buff *tail; 294 } sk_backlog; 295 #define sk_rmem_alloc sk_backlog.rmem_alloc 296 int sk_forward_alloc; 297 #ifdef CONFIG_RPS 298 __u32 sk_rxhash; 299 #endif 300 atomic_t sk_drops; 301 int sk_rcvbuf; 302 303 struct sk_filter __rcu *sk_filter; 304 struct socket_wq __rcu *sk_wq; 305 306 #ifdef CONFIG_NET_DMA 307 struct sk_buff_head sk_async_wait_queue; 308 #endif 309 310 #ifdef CONFIG_XFRM 311 struct xfrm_policy *sk_policy[2]; 312 #endif 313 unsigned long sk_flags; 314 struct dst_entry *sk_dst_cache; 315 spinlock_t sk_dst_lock; 316 atomic_t sk_wmem_alloc; 317 atomic_t sk_omem_alloc; 318 int sk_sndbuf; 319 struct sk_buff_head sk_write_queue; 320 kmemcheck_bitfield_begin(flags); 321 unsigned int sk_shutdown : 2, 322 sk_no_check : 2, 323 sk_userlocks : 4, 324 sk_protocol : 8, 325 sk_type : 16; 326 kmemcheck_bitfield_end(flags); 327 int sk_wmem_queued; 328 gfp_t sk_allocation; 329 netdev_features_t sk_route_caps; 330 netdev_features_t sk_route_nocaps; 331 int sk_gso_type; 332 unsigned int sk_gso_max_size; 333 int sk_rcvlowat; 334 unsigned long sk_lingertime; 335 struct sk_buff_head sk_error_queue; 336 struct proto *sk_prot_creator; 337 rwlock_t sk_callback_lock; 338 int sk_err, 339 sk_err_soft; 340 unsigned short sk_ack_backlog; 341 unsigned short sk_max_ack_backlog; 342 __u32 sk_priority; 343 #ifdef CONFIG_CGROUPS 344 __u32 sk_cgrp_prioidx; 345 #endif 346 struct pid *sk_peer_pid; 347 const struct cred *sk_peer_cred; 348 long sk_rcvtimeo; 349 long sk_sndtimeo; 350 void *sk_protinfo; 351 struct timer_list sk_timer; 352 ktime_t sk_stamp; 353 struct socket *sk_socket; 354 void *sk_user_data; 355 struct page *sk_sndmsg_page; 356 struct sk_buff *sk_send_head; 357 __u32 sk_sndmsg_off; 358 int sk_write_pending; 359 #ifdef CONFIG_SECURITY 360 void *sk_security; 361 #endif 362 __u32 sk_mark; 363 u32 sk_classid; 364 struct cg_proto *sk_cgrp; 365 void (*sk_state_change)(struct sock *sk); 366 void (*sk_data_ready)(struct sock *sk, int bytes); 367 void (*sk_write_space)(struct sock *sk); 368 void (*sk_error_report)(struct sock *sk); 369 int (*sk_backlog_rcv)(struct sock *sk, 370 struct sk_buff *skb); 371 void (*sk_destruct)(struct sock *sk); 372 }; 373 374 /* 375 * Hashed lists helper routines 376 */ 377 static inline struct sock *sk_entry(const struct hlist_node *node) 378 { 379 return hlist_entry(node, struct sock, sk_node); 380 } 381 382 static inline struct sock *__sk_head(const struct hlist_head *head) 383 { 384 return hlist_entry(head->first, struct sock, sk_node); 385 } 386 387 static inline struct sock *sk_head(const struct hlist_head *head) 388 { 389 return hlist_empty(head) ? NULL : __sk_head(head); 390 } 391 392 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) 393 { 394 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); 395 } 396 397 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) 398 { 399 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); 400 } 401 402 static inline struct sock *sk_next(const struct sock *sk) 403 { 404 return sk->sk_node.next ? 405 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; 406 } 407 408 static inline struct sock *sk_nulls_next(const struct sock *sk) 409 { 410 return (!is_a_nulls(sk->sk_nulls_node.next)) ? 411 hlist_nulls_entry(sk->sk_nulls_node.next, 412 struct sock, sk_nulls_node) : 413 NULL; 414 } 415 416 static inline int sk_unhashed(const struct sock *sk) 417 { 418 return hlist_unhashed(&sk->sk_node); 419 } 420 421 static inline int sk_hashed(const struct sock *sk) 422 { 423 return !sk_unhashed(sk); 424 } 425 426 static __inline__ void sk_node_init(struct hlist_node *node) 427 { 428 node->pprev = NULL; 429 } 430 431 static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node) 432 { 433 node->pprev = NULL; 434 } 435 436 static __inline__ void __sk_del_node(struct sock *sk) 437 { 438 __hlist_del(&sk->sk_node); 439 } 440 441 /* NB: equivalent to hlist_del_init_rcu */ 442 static __inline__ int __sk_del_node_init(struct sock *sk) 443 { 444 if (sk_hashed(sk)) { 445 __sk_del_node(sk); 446 sk_node_init(&sk->sk_node); 447 return 1; 448 } 449 return 0; 450 } 451 452 /* Grab socket reference count. This operation is valid only 453 when sk is ALREADY grabbed f.e. it is found in hash table 454 or a list and the lookup is made under lock preventing hash table 455 modifications. 456 */ 457 458 static inline void sock_hold(struct sock *sk) 459 { 460 atomic_inc(&sk->sk_refcnt); 461 } 462 463 /* Ungrab socket in the context, which assumes that socket refcnt 464 cannot hit zero, f.e. it is true in context of any socketcall. 465 */ 466 static inline void __sock_put(struct sock *sk) 467 { 468 atomic_dec(&sk->sk_refcnt); 469 } 470 471 static __inline__ int sk_del_node_init(struct sock *sk) 472 { 473 int rc = __sk_del_node_init(sk); 474 475 if (rc) { 476 /* paranoid for a while -acme */ 477 WARN_ON(atomic_read(&sk->sk_refcnt) == 1); 478 __sock_put(sk); 479 } 480 return rc; 481 } 482 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) 483 484 static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk) 485 { 486 if (sk_hashed(sk)) { 487 hlist_nulls_del_init_rcu(&sk->sk_nulls_node); 488 return 1; 489 } 490 return 0; 491 } 492 493 static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk) 494 { 495 int rc = __sk_nulls_del_node_init_rcu(sk); 496 497 if (rc) { 498 /* paranoid for a while -acme */ 499 WARN_ON(atomic_read(&sk->sk_refcnt) == 1); 500 __sock_put(sk); 501 } 502 return rc; 503 } 504 505 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) 506 { 507 hlist_add_head(&sk->sk_node, list); 508 } 509 510 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) 511 { 512 sock_hold(sk); 513 __sk_add_node(sk, list); 514 } 515 516 static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) 517 { 518 sock_hold(sk); 519 hlist_add_head_rcu(&sk->sk_node, list); 520 } 521 522 static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) 523 { 524 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); 525 } 526 527 static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) 528 { 529 sock_hold(sk); 530 __sk_nulls_add_node_rcu(sk, list); 531 } 532 533 static __inline__ void __sk_del_bind_node(struct sock *sk) 534 { 535 __hlist_del(&sk->sk_bind_node); 536 } 537 538 static __inline__ void sk_add_bind_node(struct sock *sk, 539 struct hlist_head *list) 540 { 541 hlist_add_head(&sk->sk_bind_node, list); 542 } 543 544 #define sk_for_each(__sk, node, list) \ 545 hlist_for_each_entry(__sk, node, list, sk_node) 546 #define sk_for_each_rcu(__sk, node, list) \ 547 hlist_for_each_entry_rcu(__sk, node, list, sk_node) 548 #define sk_nulls_for_each(__sk, node, list) \ 549 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) 550 #define sk_nulls_for_each_rcu(__sk, node, list) \ 551 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) 552 #define sk_for_each_from(__sk, node) \ 553 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ 554 hlist_for_each_entry_from(__sk, node, sk_node) 555 #define sk_nulls_for_each_from(__sk, node) \ 556 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ 557 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) 558 #define sk_for_each_safe(__sk, node, tmp, list) \ 559 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) 560 #define sk_for_each_bound(__sk, node, list) \ 561 hlist_for_each_entry(__sk, node, list, sk_bind_node) 562 563 /* Sock flags */ 564 enum sock_flags { 565 SOCK_DEAD, 566 SOCK_DONE, 567 SOCK_URGINLINE, 568 SOCK_KEEPOPEN, 569 SOCK_LINGER, 570 SOCK_DESTROY, 571 SOCK_BROADCAST, 572 SOCK_TIMESTAMP, 573 SOCK_ZAPPED, 574 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ 575 SOCK_DBG, /* %SO_DEBUG setting */ 576 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ 577 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ 578 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ 579 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ 580 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */ 581 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */ 582 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */ 583 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ 584 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */ 585 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */ 586 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */ 587 SOCK_FASYNC, /* fasync() active */ 588 SOCK_RXQ_OVFL, 589 SOCK_ZEROCOPY, /* buffers from userspace */ 590 SOCK_WIFI_STATUS, /* push wifi status to userspace */ 591 }; 592 593 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) 594 { 595 nsk->sk_flags = osk->sk_flags; 596 } 597 598 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) 599 { 600 __set_bit(flag, &sk->sk_flags); 601 } 602 603 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) 604 { 605 __clear_bit(flag, &sk->sk_flags); 606 } 607 608 static inline int sock_flag(struct sock *sk, enum sock_flags flag) 609 { 610 return test_bit(flag, &sk->sk_flags); 611 } 612 613 static inline void sk_acceptq_removed(struct sock *sk) 614 { 615 sk->sk_ack_backlog--; 616 } 617 618 static inline void sk_acceptq_added(struct sock *sk) 619 { 620 sk->sk_ack_backlog++; 621 } 622 623 static inline int sk_acceptq_is_full(struct sock *sk) 624 { 625 return sk->sk_ack_backlog > sk->sk_max_ack_backlog; 626 } 627 628 /* 629 * Compute minimal free write space needed to queue new packets. 630 */ 631 static inline int sk_stream_min_wspace(struct sock *sk) 632 { 633 return sk->sk_wmem_queued >> 1; 634 } 635 636 static inline int sk_stream_wspace(struct sock *sk) 637 { 638 return sk->sk_sndbuf - sk->sk_wmem_queued; 639 } 640 641 extern void sk_stream_write_space(struct sock *sk); 642 643 static inline int sk_stream_memory_free(struct sock *sk) 644 { 645 return sk->sk_wmem_queued < sk->sk_sndbuf; 646 } 647 648 /* OOB backlog add */ 649 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) 650 { 651 /* dont let skb dst not refcounted, we are going to leave rcu lock */ 652 skb_dst_force(skb); 653 654 if (!sk->sk_backlog.tail) 655 sk->sk_backlog.head = skb; 656 else 657 sk->sk_backlog.tail->next = skb; 658 659 sk->sk_backlog.tail = skb; 660 skb->next = NULL; 661 } 662 663 /* 664 * Take into account size of receive queue and backlog queue 665 * Do not take into account this skb truesize, 666 * to allow even a single big packet to come. 667 */ 668 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb) 669 { 670 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); 671 672 return qsize > sk->sk_rcvbuf; 673 } 674 675 /* The per-socket spinlock must be held here. */ 676 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb) 677 { 678 if (sk_rcvqueues_full(sk, skb)) 679 return -ENOBUFS; 680 681 __sk_add_backlog(sk, skb); 682 sk->sk_backlog.len += skb->truesize; 683 return 0; 684 } 685 686 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) 687 { 688 return sk->sk_backlog_rcv(sk, skb); 689 } 690 691 static inline void sock_rps_record_flow(const struct sock *sk) 692 { 693 #ifdef CONFIG_RPS 694 struct rps_sock_flow_table *sock_flow_table; 695 696 rcu_read_lock(); 697 sock_flow_table = rcu_dereference(rps_sock_flow_table); 698 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash); 699 rcu_read_unlock(); 700 #endif 701 } 702 703 static inline void sock_rps_reset_flow(const struct sock *sk) 704 { 705 #ifdef CONFIG_RPS 706 struct rps_sock_flow_table *sock_flow_table; 707 708 rcu_read_lock(); 709 sock_flow_table = rcu_dereference(rps_sock_flow_table); 710 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash); 711 rcu_read_unlock(); 712 #endif 713 } 714 715 static inline void sock_rps_save_rxhash(struct sock *sk, 716 const struct sk_buff *skb) 717 { 718 #ifdef CONFIG_RPS 719 if (unlikely(sk->sk_rxhash != skb->rxhash)) { 720 sock_rps_reset_flow(sk); 721 sk->sk_rxhash = skb->rxhash; 722 } 723 #endif 724 } 725 726 static inline void sock_rps_reset_rxhash(struct sock *sk) 727 { 728 #ifdef CONFIG_RPS 729 sock_rps_reset_flow(sk); 730 sk->sk_rxhash = 0; 731 #endif 732 } 733 734 #define sk_wait_event(__sk, __timeo, __condition) \ 735 ({ int __rc; \ 736 release_sock(__sk); \ 737 __rc = __condition; \ 738 if (!__rc) { \ 739 *(__timeo) = schedule_timeout(*(__timeo)); \ 740 } \ 741 lock_sock(__sk); \ 742 __rc = __condition; \ 743 __rc; \ 744 }) 745 746 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); 747 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); 748 extern void sk_stream_wait_close(struct sock *sk, long timeo_p); 749 extern int sk_stream_error(struct sock *sk, int flags, int err); 750 extern void sk_stream_kill_queues(struct sock *sk); 751 752 extern int sk_wait_data(struct sock *sk, long *timeo); 753 754 struct request_sock_ops; 755 struct timewait_sock_ops; 756 struct inet_hashinfo; 757 struct raw_hashinfo; 758 struct module; 759 760 /* Networking protocol blocks we attach to sockets. 761 * socket layer -> transport layer interface 762 * transport -> network interface is defined by struct inet_proto 763 */ 764 struct proto { 765 void (*close)(struct sock *sk, 766 long timeout); 767 int (*connect)(struct sock *sk, 768 struct sockaddr *uaddr, 769 int addr_len); 770 int (*disconnect)(struct sock *sk, int flags); 771 772 struct sock * (*accept) (struct sock *sk, int flags, int *err); 773 774 int (*ioctl)(struct sock *sk, int cmd, 775 unsigned long arg); 776 int (*init)(struct sock *sk); 777 void (*destroy)(struct sock *sk); 778 void (*shutdown)(struct sock *sk, int how); 779 int (*setsockopt)(struct sock *sk, int level, 780 int optname, char __user *optval, 781 unsigned int optlen); 782 int (*getsockopt)(struct sock *sk, int level, 783 int optname, char __user *optval, 784 int __user *option); 785 #ifdef CONFIG_COMPAT 786 int (*compat_setsockopt)(struct sock *sk, 787 int level, 788 int optname, char __user *optval, 789 unsigned int optlen); 790 int (*compat_getsockopt)(struct sock *sk, 791 int level, 792 int optname, char __user *optval, 793 int __user *option); 794 int (*compat_ioctl)(struct sock *sk, 795 unsigned int cmd, unsigned long arg); 796 #endif 797 int (*sendmsg)(struct kiocb *iocb, struct sock *sk, 798 struct msghdr *msg, size_t len); 799 int (*recvmsg)(struct kiocb *iocb, struct sock *sk, 800 struct msghdr *msg, 801 size_t len, int noblock, int flags, 802 int *addr_len); 803 int (*sendpage)(struct sock *sk, struct page *page, 804 int offset, size_t size, int flags); 805 int (*bind)(struct sock *sk, 806 struct sockaddr *uaddr, int addr_len); 807 808 int (*backlog_rcv) (struct sock *sk, 809 struct sk_buff *skb); 810 811 /* Keeping track of sk's, looking them up, and port selection methods. */ 812 void (*hash)(struct sock *sk); 813 void (*unhash)(struct sock *sk); 814 void (*rehash)(struct sock *sk); 815 int (*get_port)(struct sock *sk, unsigned short snum); 816 void (*clear_sk)(struct sock *sk, int size); 817 818 /* Keeping track of sockets in use */ 819 #ifdef CONFIG_PROC_FS 820 unsigned int inuse_idx; 821 #endif 822 823 /* Memory pressure */ 824 void (*enter_memory_pressure)(struct sock *sk); 825 atomic_long_t *memory_allocated; /* Current allocated memory. */ 826 struct percpu_counter *sockets_allocated; /* Current number of sockets. */ 827 /* 828 * Pressure flag: try to collapse. 829 * Technical note: it is used by multiple contexts non atomically. 830 * All the __sk_mem_schedule() is of this nature: accounting 831 * is strict, actions are advisory and have some latency. 832 */ 833 int *memory_pressure; 834 long *sysctl_mem; 835 int *sysctl_wmem; 836 int *sysctl_rmem; 837 int max_header; 838 bool no_autobind; 839 840 struct kmem_cache *slab; 841 unsigned int obj_size; 842 int slab_flags; 843 844 struct percpu_counter *orphan_count; 845 846 struct request_sock_ops *rsk_prot; 847 struct timewait_sock_ops *twsk_prot; 848 849 union { 850 struct inet_hashinfo *hashinfo; 851 struct udp_table *udp_table; 852 struct raw_hashinfo *raw_hash; 853 } h; 854 855 struct module *owner; 856 857 char name[32]; 858 859 struct list_head node; 860 #ifdef SOCK_REFCNT_DEBUG 861 atomic_t socks; 862 #endif 863 #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM 864 /* 865 * cgroup specific init/deinit functions. Called once for all 866 * protocols that implement it, from cgroups populate function. 867 * This function has to setup any files the protocol want to 868 * appear in the kmem cgroup filesystem. 869 */ 870 int (*init_cgroup)(struct cgroup *cgrp, 871 struct cgroup_subsys *ss); 872 void (*destroy_cgroup)(struct cgroup *cgrp, 873 struct cgroup_subsys *ss); 874 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg); 875 #endif 876 }; 877 878 struct cg_proto { 879 void (*enter_memory_pressure)(struct sock *sk); 880 struct res_counter *memory_allocated; /* Current allocated memory. */ 881 struct percpu_counter *sockets_allocated; /* Current number of sockets. */ 882 int *memory_pressure; 883 long *sysctl_mem; 884 /* 885 * memcg field is used to find which memcg we belong directly 886 * Each memcg struct can hold more than one cg_proto, so container_of 887 * won't really cut. 888 * 889 * The elegant solution would be having an inverse function to 890 * proto_cgroup in struct proto, but that means polluting the structure 891 * for everybody, instead of just for memcg users. 892 */ 893 struct mem_cgroup *memcg; 894 }; 895 896 extern int proto_register(struct proto *prot, int alloc_slab); 897 extern void proto_unregister(struct proto *prot); 898 899 #ifdef SOCK_REFCNT_DEBUG 900 static inline void sk_refcnt_debug_inc(struct sock *sk) 901 { 902 atomic_inc(&sk->sk_prot->socks); 903 } 904 905 static inline void sk_refcnt_debug_dec(struct sock *sk) 906 { 907 atomic_dec(&sk->sk_prot->socks); 908 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", 909 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); 910 } 911 912 inline void sk_refcnt_debug_release(const struct sock *sk) 913 { 914 if (atomic_read(&sk->sk_refcnt) != 1) 915 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", 916 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); 917 } 918 #else /* SOCK_REFCNT_DEBUG */ 919 #define sk_refcnt_debug_inc(sk) do { } while (0) 920 #define sk_refcnt_debug_dec(sk) do { } while (0) 921 #define sk_refcnt_debug_release(sk) do { } while (0) 922 #endif /* SOCK_REFCNT_DEBUG */ 923 924 #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM 925 extern struct jump_label_key memcg_socket_limit_enabled; 926 static inline struct cg_proto *parent_cg_proto(struct proto *proto, 927 struct cg_proto *cg_proto) 928 { 929 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg)); 930 } 931 #define mem_cgroup_sockets_enabled static_branch(&memcg_socket_limit_enabled) 932 #else 933 #define mem_cgroup_sockets_enabled 0 934 static inline struct cg_proto *parent_cg_proto(struct proto *proto, 935 struct cg_proto *cg_proto) 936 { 937 return NULL; 938 } 939 #endif 940 941 942 static inline bool sk_has_memory_pressure(const struct sock *sk) 943 { 944 return sk->sk_prot->memory_pressure != NULL; 945 } 946 947 static inline bool sk_under_memory_pressure(const struct sock *sk) 948 { 949 if (!sk->sk_prot->memory_pressure) 950 return false; 951 952 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) 953 return !!*sk->sk_cgrp->memory_pressure; 954 955 return !!*sk->sk_prot->memory_pressure; 956 } 957 958 static inline void sk_leave_memory_pressure(struct sock *sk) 959 { 960 int *memory_pressure = sk->sk_prot->memory_pressure; 961 962 if (!memory_pressure) 963 return; 964 965 if (*memory_pressure) 966 *memory_pressure = 0; 967 968 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { 969 struct cg_proto *cg_proto = sk->sk_cgrp; 970 struct proto *prot = sk->sk_prot; 971 972 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) 973 if (*cg_proto->memory_pressure) 974 *cg_proto->memory_pressure = 0; 975 } 976 977 } 978 979 static inline void sk_enter_memory_pressure(struct sock *sk) 980 { 981 if (!sk->sk_prot->enter_memory_pressure) 982 return; 983 984 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { 985 struct cg_proto *cg_proto = sk->sk_cgrp; 986 struct proto *prot = sk->sk_prot; 987 988 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) 989 cg_proto->enter_memory_pressure(sk); 990 } 991 992 sk->sk_prot->enter_memory_pressure(sk); 993 } 994 995 static inline long sk_prot_mem_limits(const struct sock *sk, int index) 996 { 997 long *prot = sk->sk_prot->sysctl_mem; 998 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) 999 prot = sk->sk_cgrp->sysctl_mem; 1000 return prot[index]; 1001 } 1002 1003 static inline void memcg_memory_allocated_add(struct cg_proto *prot, 1004 unsigned long amt, 1005 int *parent_status) 1006 { 1007 struct res_counter *fail; 1008 int ret; 1009 1010 ret = res_counter_charge(prot->memory_allocated, 1011 amt << PAGE_SHIFT, &fail); 1012 1013 if (ret < 0) 1014 *parent_status = OVER_LIMIT; 1015 } 1016 1017 static inline void memcg_memory_allocated_sub(struct cg_proto *prot, 1018 unsigned long amt) 1019 { 1020 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT); 1021 } 1022 1023 static inline u64 memcg_memory_allocated_read(struct cg_proto *prot) 1024 { 1025 u64 ret; 1026 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE); 1027 return ret >> PAGE_SHIFT; 1028 } 1029 1030 static inline long 1031 sk_memory_allocated(const struct sock *sk) 1032 { 1033 struct proto *prot = sk->sk_prot; 1034 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) 1035 return memcg_memory_allocated_read(sk->sk_cgrp); 1036 1037 return atomic_long_read(prot->memory_allocated); 1038 } 1039 1040 static inline long 1041 sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status) 1042 { 1043 struct proto *prot = sk->sk_prot; 1044 1045 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { 1046 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status); 1047 /* update the root cgroup regardless */ 1048 atomic_long_add_return(amt, prot->memory_allocated); 1049 return memcg_memory_allocated_read(sk->sk_cgrp); 1050 } 1051 1052 return atomic_long_add_return(amt, prot->memory_allocated); 1053 } 1054 1055 static inline void 1056 sk_memory_allocated_sub(struct sock *sk, int amt, int parent_status) 1057 { 1058 struct proto *prot = sk->sk_prot; 1059 1060 if (mem_cgroup_sockets_enabled && sk->sk_cgrp && 1061 parent_status != OVER_LIMIT) /* Otherwise was uncharged already */ 1062 memcg_memory_allocated_sub(sk->sk_cgrp, amt); 1063 1064 atomic_long_sub(amt, prot->memory_allocated); 1065 } 1066 1067 static inline void sk_sockets_allocated_dec(struct sock *sk) 1068 { 1069 struct proto *prot = sk->sk_prot; 1070 1071 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { 1072 struct cg_proto *cg_proto = sk->sk_cgrp; 1073 1074 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) 1075 percpu_counter_dec(cg_proto->sockets_allocated); 1076 } 1077 1078 percpu_counter_dec(prot->sockets_allocated); 1079 } 1080 1081 static inline void sk_sockets_allocated_inc(struct sock *sk) 1082 { 1083 struct proto *prot = sk->sk_prot; 1084 1085 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { 1086 struct cg_proto *cg_proto = sk->sk_cgrp; 1087 1088 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) 1089 percpu_counter_inc(cg_proto->sockets_allocated); 1090 } 1091 1092 percpu_counter_inc(prot->sockets_allocated); 1093 } 1094 1095 static inline int 1096 sk_sockets_allocated_read_positive(struct sock *sk) 1097 { 1098 struct proto *prot = sk->sk_prot; 1099 1100 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) 1101 return percpu_counter_sum_positive(sk->sk_cgrp->sockets_allocated); 1102 1103 return percpu_counter_sum_positive(prot->sockets_allocated); 1104 } 1105 1106 static inline int 1107 proto_sockets_allocated_sum_positive(struct proto *prot) 1108 { 1109 return percpu_counter_sum_positive(prot->sockets_allocated); 1110 } 1111 1112 static inline long 1113 proto_memory_allocated(struct proto *prot) 1114 { 1115 return atomic_long_read(prot->memory_allocated); 1116 } 1117 1118 static inline bool 1119 proto_memory_pressure(struct proto *prot) 1120 { 1121 if (!prot->memory_pressure) 1122 return false; 1123 return !!*prot->memory_pressure; 1124 } 1125 1126 1127 #ifdef CONFIG_PROC_FS 1128 /* Called with local bh disabled */ 1129 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); 1130 extern int sock_prot_inuse_get(struct net *net, struct proto *proto); 1131 #else 1132 static void inline sock_prot_inuse_add(struct net *net, struct proto *prot, 1133 int inc) 1134 { 1135 } 1136 #endif 1137 1138 1139 /* With per-bucket locks this operation is not-atomic, so that 1140 * this version is not worse. 1141 */ 1142 static inline void __sk_prot_rehash(struct sock *sk) 1143 { 1144 sk->sk_prot->unhash(sk); 1145 sk->sk_prot->hash(sk); 1146 } 1147 1148 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size); 1149 1150 /* About 10 seconds */ 1151 #define SOCK_DESTROY_TIME (10*HZ) 1152 1153 /* Sockets 0-1023 can't be bound to unless you are superuser */ 1154 #define PROT_SOCK 1024 1155 1156 #define SHUTDOWN_MASK 3 1157 #define RCV_SHUTDOWN 1 1158 #define SEND_SHUTDOWN 2 1159 1160 #define SOCK_SNDBUF_LOCK 1 1161 #define SOCK_RCVBUF_LOCK 2 1162 #define SOCK_BINDADDR_LOCK 4 1163 #define SOCK_BINDPORT_LOCK 8 1164 1165 /* sock_iocb: used to kick off async processing of socket ios */ 1166 struct sock_iocb { 1167 struct list_head list; 1168 1169 int flags; 1170 int size; 1171 struct socket *sock; 1172 struct sock *sk; 1173 struct scm_cookie *scm; 1174 struct msghdr *msg, async_msg; 1175 struct kiocb *kiocb; 1176 }; 1177 1178 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) 1179 { 1180 return (struct sock_iocb *)iocb->private; 1181 } 1182 1183 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) 1184 { 1185 return si->kiocb; 1186 } 1187 1188 struct socket_alloc { 1189 struct socket socket; 1190 struct inode vfs_inode; 1191 }; 1192 1193 static inline struct socket *SOCKET_I(struct inode *inode) 1194 { 1195 return &container_of(inode, struct socket_alloc, vfs_inode)->socket; 1196 } 1197 1198 static inline struct inode *SOCK_INODE(struct socket *socket) 1199 { 1200 return &container_of(socket, struct socket_alloc, socket)->vfs_inode; 1201 } 1202 1203 /* 1204 * Functions for memory accounting 1205 */ 1206 extern int __sk_mem_schedule(struct sock *sk, int size, int kind); 1207 extern void __sk_mem_reclaim(struct sock *sk); 1208 1209 #define SK_MEM_QUANTUM ((int)PAGE_SIZE) 1210 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) 1211 #define SK_MEM_SEND 0 1212 #define SK_MEM_RECV 1 1213 1214 static inline int sk_mem_pages(int amt) 1215 { 1216 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; 1217 } 1218 1219 static inline int sk_has_account(struct sock *sk) 1220 { 1221 /* return true if protocol supports memory accounting */ 1222 return !!sk->sk_prot->memory_allocated; 1223 } 1224 1225 static inline int sk_wmem_schedule(struct sock *sk, int size) 1226 { 1227 if (!sk_has_account(sk)) 1228 return 1; 1229 return size <= sk->sk_forward_alloc || 1230 __sk_mem_schedule(sk, size, SK_MEM_SEND); 1231 } 1232 1233 static inline int sk_rmem_schedule(struct sock *sk, int size) 1234 { 1235 if (!sk_has_account(sk)) 1236 return 1; 1237 return size <= sk->sk_forward_alloc || 1238 __sk_mem_schedule(sk, size, SK_MEM_RECV); 1239 } 1240 1241 static inline void sk_mem_reclaim(struct sock *sk) 1242 { 1243 if (!sk_has_account(sk)) 1244 return; 1245 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) 1246 __sk_mem_reclaim(sk); 1247 } 1248 1249 static inline void sk_mem_reclaim_partial(struct sock *sk) 1250 { 1251 if (!sk_has_account(sk)) 1252 return; 1253 if (sk->sk_forward_alloc > SK_MEM_QUANTUM) 1254 __sk_mem_reclaim(sk); 1255 } 1256 1257 static inline void sk_mem_charge(struct sock *sk, int size) 1258 { 1259 if (!sk_has_account(sk)) 1260 return; 1261 sk->sk_forward_alloc -= size; 1262 } 1263 1264 static inline void sk_mem_uncharge(struct sock *sk, int size) 1265 { 1266 if (!sk_has_account(sk)) 1267 return; 1268 sk->sk_forward_alloc += size; 1269 } 1270 1271 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) 1272 { 1273 sock_set_flag(sk, SOCK_QUEUE_SHRUNK); 1274 sk->sk_wmem_queued -= skb->truesize; 1275 sk_mem_uncharge(sk, skb->truesize); 1276 __kfree_skb(skb); 1277 } 1278 1279 /* Used by processes to "lock" a socket state, so that 1280 * interrupts and bottom half handlers won't change it 1281 * from under us. It essentially blocks any incoming 1282 * packets, so that we won't get any new data or any 1283 * packets that change the state of the socket. 1284 * 1285 * While locked, BH processing will add new packets to 1286 * the backlog queue. This queue is processed by the 1287 * owner of the socket lock right before it is released. 1288 * 1289 * Since ~2.3.5 it is also exclusive sleep lock serializing 1290 * accesses from user process context. 1291 */ 1292 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) 1293 1294 /* 1295 * Macro so as to not evaluate some arguments when 1296 * lockdep is not enabled. 1297 * 1298 * Mark both the sk_lock and the sk_lock.slock as a 1299 * per-address-family lock class. 1300 */ 1301 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ 1302 do { \ 1303 sk->sk_lock.owned = 0; \ 1304 init_waitqueue_head(&sk->sk_lock.wq); \ 1305 spin_lock_init(&(sk)->sk_lock.slock); \ 1306 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ 1307 sizeof((sk)->sk_lock)); \ 1308 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ 1309 (skey), (sname)); \ 1310 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ 1311 } while (0) 1312 1313 extern void lock_sock_nested(struct sock *sk, int subclass); 1314 1315 static inline void lock_sock(struct sock *sk) 1316 { 1317 lock_sock_nested(sk, 0); 1318 } 1319 1320 extern void release_sock(struct sock *sk); 1321 1322 /* BH context may only use the following locking interface. */ 1323 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) 1324 #define bh_lock_sock_nested(__sk) \ 1325 spin_lock_nested(&((__sk)->sk_lock.slock), \ 1326 SINGLE_DEPTH_NESTING) 1327 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) 1328 1329 extern bool lock_sock_fast(struct sock *sk); 1330 /** 1331 * unlock_sock_fast - complement of lock_sock_fast 1332 * @sk: socket 1333 * @slow: slow mode 1334 * 1335 * fast unlock socket for user context. 1336 * If slow mode is on, we call regular release_sock() 1337 */ 1338 static inline void unlock_sock_fast(struct sock *sk, bool slow) 1339 { 1340 if (slow) 1341 release_sock(sk); 1342 else 1343 spin_unlock_bh(&sk->sk_lock.slock); 1344 } 1345 1346 1347 extern struct sock *sk_alloc(struct net *net, int family, 1348 gfp_t priority, 1349 struct proto *prot); 1350 extern void sk_free(struct sock *sk); 1351 extern void sk_release_kernel(struct sock *sk); 1352 extern struct sock *sk_clone_lock(const struct sock *sk, 1353 const gfp_t priority); 1354 1355 extern struct sk_buff *sock_wmalloc(struct sock *sk, 1356 unsigned long size, int force, 1357 gfp_t priority); 1358 extern struct sk_buff *sock_rmalloc(struct sock *sk, 1359 unsigned long size, int force, 1360 gfp_t priority); 1361 extern void sock_wfree(struct sk_buff *skb); 1362 extern void sock_rfree(struct sk_buff *skb); 1363 1364 extern int sock_setsockopt(struct socket *sock, int level, 1365 int op, char __user *optval, 1366 unsigned int optlen); 1367 1368 extern int sock_getsockopt(struct socket *sock, int level, 1369 int op, char __user *optval, 1370 int __user *optlen); 1371 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, 1372 unsigned long size, 1373 int noblock, 1374 int *errcode); 1375 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk, 1376 unsigned long header_len, 1377 unsigned long data_len, 1378 int noblock, 1379 int *errcode); 1380 extern void *sock_kmalloc(struct sock *sk, int size, 1381 gfp_t priority); 1382 extern void sock_kfree_s(struct sock *sk, void *mem, int size); 1383 extern void sk_send_sigurg(struct sock *sk); 1384 1385 #ifdef CONFIG_CGROUPS 1386 extern void sock_update_classid(struct sock *sk); 1387 #else 1388 static inline void sock_update_classid(struct sock *sk) 1389 { 1390 } 1391 #endif 1392 1393 /* 1394 * Functions to fill in entries in struct proto_ops when a protocol 1395 * does not implement a particular function. 1396 */ 1397 extern int sock_no_bind(struct socket *, 1398 struct sockaddr *, int); 1399 extern int sock_no_connect(struct socket *, 1400 struct sockaddr *, int, int); 1401 extern int sock_no_socketpair(struct socket *, 1402 struct socket *); 1403 extern int sock_no_accept(struct socket *, 1404 struct socket *, int); 1405 extern int sock_no_getname(struct socket *, 1406 struct sockaddr *, int *, int); 1407 extern unsigned int sock_no_poll(struct file *, struct socket *, 1408 struct poll_table_struct *); 1409 extern int sock_no_ioctl(struct socket *, unsigned int, 1410 unsigned long); 1411 extern int sock_no_listen(struct socket *, int); 1412 extern int sock_no_shutdown(struct socket *, int); 1413 extern int sock_no_getsockopt(struct socket *, int , int, 1414 char __user *, int __user *); 1415 extern int sock_no_setsockopt(struct socket *, int, int, 1416 char __user *, unsigned int); 1417 extern int sock_no_sendmsg(struct kiocb *, struct socket *, 1418 struct msghdr *, size_t); 1419 extern int sock_no_recvmsg(struct kiocb *, struct socket *, 1420 struct msghdr *, size_t, int); 1421 extern int sock_no_mmap(struct file *file, 1422 struct socket *sock, 1423 struct vm_area_struct *vma); 1424 extern ssize_t sock_no_sendpage(struct socket *sock, 1425 struct page *page, 1426 int offset, size_t size, 1427 int flags); 1428 1429 /* 1430 * Functions to fill in entries in struct proto_ops when a protocol 1431 * uses the inet style. 1432 */ 1433 extern int sock_common_getsockopt(struct socket *sock, int level, int optname, 1434 char __user *optval, int __user *optlen); 1435 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, 1436 struct msghdr *msg, size_t size, int flags); 1437 extern int sock_common_setsockopt(struct socket *sock, int level, int optname, 1438 char __user *optval, unsigned int optlen); 1439 extern int compat_sock_common_getsockopt(struct socket *sock, int level, 1440 int optname, char __user *optval, int __user *optlen); 1441 extern int compat_sock_common_setsockopt(struct socket *sock, int level, 1442 int optname, char __user *optval, unsigned int optlen); 1443 1444 extern void sk_common_release(struct sock *sk); 1445 1446 /* 1447 * Default socket callbacks and setup code 1448 */ 1449 1450 /* Initialise core socket variables */ 1451 extern void sock_init_data(struct socket *sock, struct sock *sk); 1452 1453 extern void sk_filter_release_rcu(struct rcu_head *rcu); 1454 1455 /** 1456 * sk_filter_release - release a socket filter 1457 * @fp: filter to remove 1458 * 1459 * Remove a filter from a socket and release its resources. 1460 */ 1461 1462 static inline void sk_filter_release(struct sk_filter *fp) 1463 { 1464 if (atomic_dec_and_test(&fp->refcnt)) 1465 call_rcu(&fp->rcu, sk_filter_release_rcu); 1466 } 1467 1468 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp) 1469 { 1470 unsigned int size = sk_filter_len(fp); 1471 1472 atomic_sub(size, &sk->sk_omem_alloc); 1473 sk_filter_release(fp); 1474 } 1475 1476 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) 1477 { 1478 atomic_inc(&fp->refcnt); 1479 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); 1480 } 1481 1482 /* 1483 * Socket reference counting postulates. 1484 * 1485 * * Each user of socket SHOULD hold a reference count. 1486 * * Each access point to socket (an hash table bucket, reference from a list, 1487 * running timer, skb in flight MUST hold a reference count. 1488 * * When reference count hits 0, it means it will never increase back. 1489 * * When reference count hits 0, it means that no references from 1490 * outside exist to this socket and current process on current CPU 1491 * is last user and may/should destroy this socket. 1492 * * sk_free is called from any context: process, BH, IRQ. When 1493 * it is called, socket has no references from outside -> sk_free 1494 * may release descendant resources allocated by the socket, but 1495 * to the time when it is called, socket is NOT referenced by any 1496 * hash tables, lists etc. 1497 * * Packets, delivered from outside (from network or from another process) 1498 * and enqueued on receive/error queues SHOULD NOT grab reference count, 1499 * when they sit in queue. Otherwise, packets will leak to hole, when 1500 * socket is looked up by one cpu and unhasing is made by another CPU. 1501 * It is true for udp/raw, netlink (leak to receive and error queues), tcp 1502 * (leak to backlog). Packet socket does all the processing inside 1503 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets 1504 * use separate SMP lock, so that they are prone too. 1505 */ 1506 1507 /* Ungrab socket and destroy it, if it was the last reference. */ 1508 static inline void sock_put(struct sock *sk) 1509 { 1510 if (atomic_dec_and_test(&sk->sk_refcnt)) 1511 sk_free(sk); 1512 } 1513 1514 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, 1515 const int nested); 1516 1517 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) 1518 { 1519 sk->sk_tx_queue_mapping = tx_queue; 1520 } 1521 1522 static inline void sk_tx_queue_clear(struct sock *sk) 1523 { 1524 sk->sk_tx_queue_mapping = -1; 1525 } 1526 1527 static inline int sk_tx_queue_get(const struct sock *sk) 1528 { 1529 return sk ? sk->sk_tx_queue_mapping : -1; 1530 } 1531 1532 static inline void sk_set_socket(struct sock *sk, struct socket *sock) 1533 { 1534 sk_tx_queue_clear(sk); 1535 sk->sk_socket = sock; 1536 } 1537 1538 static inline wait_queue_head_t *sk_sleep(struct sock *sk) 1539 { 1540 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); 1541 return &rcu_dereference_raw(sk->sk_wq)->wait; 1542 } 1543 /* Detach socket from process context. 1544 * Announce socket dead, detach it from wait queue and inode. 1545 * Note that parent inode held reference count on this struct sock, 1546 * we do not release it in this function, because protocol 1547 * probably wants some additional cleanups or even continuing 1548 * to work with this socket (TCP). 1549 */ 1550 static inline void sock_orphan(struct sock *sk) 1551 { 1552 write_lock_bh(&sk->sk_callback_lock); 1553 sock_set_flag(sk, SOCK_DEAD); 1554 sk_set_socket(sk, NULL); 1555 sk->sk_wq = NULL; 1556 write_unlock_bh(&sk->sk_callback_lock); 1557 } 1558 1559 static inline void sock_graft(struct sock *sk, struct socket *parent) 1560 { 1561 write_lock_bh(&sk->sk_callback_lock); 1562 sk->sk_wq = parent->wq; 1563 parent->sk = sk; 1564 sk_set_socket(sk, parent); 1565 security_sock_graft(sk, parent); 1566 write_unlock_bh(&sk->sk_callback_lock); 1567 } 1568 1569 extern int sock_i_uid(struct sock *sk); 1570 extern unsigned long sock_i_ino(struct sock *sk); 1571 1572 static inline struct dst_entry * 1573 __sk_dst_get(struct sock *sk) 1574 { 1575 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) || 1576 lockdep_is_held(&sk->sk_lock.slock)); 1577 } 1578 1579 static inline struct dst_entry * 1580 sk_dst_get(struct sock *sk) 1581 { 1582 struct dst_entry *dst; 1583 1584 rcu_read_lock(); 1585 dst = rcu_dereference(sk->sk_dst_cache); 1586 if (dst) 1587 dst_hold(dst); 1588 rcu_read_unlock(); 1589 return dst; 1590 } 1591 1592 extern void sk_reset_txq(struct sock *sk); 1593 1594 static inline void dst_negative_advice(struct sock *sk) 1595 { 1596 struct dst_entry *ndst, *dst = __sk_dst_get(sk); 1597 1598 if (dst && dst->ops->negative_advice) { 1599 ndst = dst->ops->negative_advice(dst); 1600 1601 if (ndst != dst) { 1602 rcu_assign_pointer(sk->sk_dst_cache, ndst); 1603 sk_reset_txq(sk); 1604 } 1605 } 1606 } 1607 1608 static inline void 1609 __sk_dst_set(struct sock *sk, struct dst_entry *dst) 1610 { 1611 struct dst_entry *old_dst; 1612 1613 sk_tx_queue_clear(sk); 1614 /* 1615 * This can be called while sk is owned by the caller only, 1616 * with no state that can be checked in a rcu_dereference_check() cond 1617 */ 1618 old_dst = rcu_dereference_raw(sk->sk_dst_cache); 1619 rcu_assign_pointer(sk->sk_dst_cache, dst); 1620 dst_release(old_dst); 1621 } 1622 1623 static inline void 1624 sk_dst_set(struct sock *sk, struct dst_entry *dst) 1625 { 1626 spin_lock(&sk->sk_dst_lock); 1627 __sk_dst_set(sk, dst); 1628 spin_unlock(&sk->sk_dst_lock); 1629 } 1630 1631 static inline void 1632 __sk_dst_reset(struct sock *sk) 1633 { 1634 __sk_dst_set(sk, NULL); 1635 } 1636 1637 static inline void 1638 sk_dst_reset(struct sock *sk) 1639 { 1640 spin_lock(&sk->sk_dst_lock); 1641 __sk_dst_reset(sk); 1642 spin_unlock(&sk->sk_dst_lock); 1643 } 1644 1645 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); 1646 1647 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); 1648 1649 static inline int sk_can_gso(const struct sock *sk) 1650 { 1651 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); 1652 } 1653 1654 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst); 1655 1656 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags) 1657 { 1658 sk->sk_route_nocaps |= flags; 1659 sk->sk_route_caps &= ~flags; 1660 } 1661 1662 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, 1663 char __user *from, char *to, 1664 int copy, int offset) 1665 { 1666 if (skb->ip_summed == CHECKSUM_NONE) { 1667 int err = 0; 1668 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err); 1669 if (err) 1670 return err; 1671 skb->csum = csum_block_add(skb->csum, csum, offset); 1672 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { 1673 if (!access_ok(VERIFY_READ, from, copy) || 1674 __copy_from_user_nocache(to, from, copy)) 1675 return -EFAULT; 1676 } else if (copy_from_user(to, from, copy)) 1677 return -EFAULT; 1678 1679 return 0; 1680 } 1681 1682 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, 1683 char __user *from, int copy) 1684 { 1685 int err, offset = skb->len; 1686 1687 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), 1688 copy, offset); 1689 if (err) 1690 __skb_trim(skb, offset); 1691 1692 return err; 1693 } 1694 1695 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from, 1696 struct sk_buff *skb, 1697 struct page *page, 1698 int off, int copy) 1699 { 1700 int err; 1701 1702 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, 1703 copy, skb->len); 1704 if (err) 1705 return err; 1706 1707 skb->len += copy; 1708 skb->data_len += copy; 1709 skb->truesize += copy; 1710 sk->sk_wmem_queued += copy; 1711 sk_mem_charge(sk, copy); 1712 return 0; 1713 } 1714 1715 static inline int skb_copy_to_page(struct sock *sk, char __user *from, 1716 struct sk_buff *skb, struct page *page, 1717 int off, int copy) 1718 { 1719 if (skb->ip_summed == CHECKSUM_NONE) { 1720 int err = 0; 1721 __wsum csum = csum_and_copy_from_user(from, 1722 page_address(page) + off, 1723 copy, 0, &err); 1724 if (err) 1725 return err; 1726 skb->csum = csum_block_add(skb->csum, csum, skb->len); 1727 } else if (copy_from_user(page_address(page) + off, from, copy)) 1728 return -EFAULT; 1729 1730 skb->len += copy; 1731 skb->data_len += copy; 1732 skb->truesize += copy; 1733 sk->sk_wmem_queued += copy; 1734 sk_mem_charge(sk, copy); 1735 return 0; 1736 } 1737 1738 /** 1739 * sk_wmem_alloc_get - returns write allocations 1740 * @sk: socket 1741 * 1742 * Returns sk_wmem_alloc minus initial offset of one 1743 */ 1744 static inline int sk_wmem_alloc_get(const struct sock *sk) 1745 { 1746 return atomic_read(&sk->sk_wmem_alloc) - 1; 1747 } 1748 1749 /** 1750 * sk_rmem_alloc_get - returns read allocations 1751 * @sk: socket 1752 * 1753 * Returns sk_rmem_alloc 1754 */ 1755 static inline int sk_rmem_alloc_get(const struct sock *sk) 1756 { 1757 return atomic_read(&sk->sk_rmem_alloc); 1758 } 1759 1760 /** 1761 * sk_has_allocations - check if allocations are outstanding 1762 * @sk: socket 1763 * 1764 * Returns true if socket has write or read allocations 1765 */ 1766 static inline int sk_has_allocations(const struct sock *sk) 1767 { 1768 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); 1769 } 1770 1771 /** 1772 * wq_has_sleeper - check if there are any waiting processes 1773 * @wq: struct socket_wq 1774 * 1775 * Returns true if socket_wq has waiting processes 1776 * 1777 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory 1778 * barrier call. They were added due to the race found within the tcp code. 1779 * 1780 * Consider following tcp code paths: 1781 * 1782 * CPU1 CPU2 1783 * 1784 * sys_select receive packet 1785 * ... ... 1786 * __add_wait_queue update tp->rcv_nxt 1787 * ... ... 1788 * tp->rcv_nxt check sock_def_readable 1789 * ... { 1790 * schedule rcu_read_lock(); 1791 * wq = rcu_dereference(sk->sk_wq); 1792 * if (wq && waitqueue_active(&wq->wait)) 1793 * wake_up_interruptible(&wq->wait) 1794 * ... 1795 * } 1796 * 1797 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay 1798 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 1799 * could then endup calling schedule and sleep forever if there are no more 1800 * data on the socket. 1801 * 1802 */ 1803 static inline bool wq_has_sleeper(struct socket_wq *wq) 1804 { 1805 1806 /* 1807 * We need to be sure we are in sync with the 1808 * add_wait_queue modifications to the wait queue. 1809 * 1810 * This memory barrier is paired in the sock_poll_wait. 1811 */ 1812 smp_mb(); 1813 return wq && waitqueue_active(&wq->wait); 1814 } 1815 1816 /** 1817 * sock_poll_wait - place memory barrier behind the poll_wait call. 1818 * @filp: file 1819 * @wait_address: socket wait queue 1820 * @p: poll_table 1821 * 1822 * See the comments in the wq_has_sleeper function. 1823 */ 1824 static inline void sock_poll_wait(struct file *filp, 1825 wait_queue_head_t *wait_address, poll_table *p) 1826 { 1827 if (p && wait_address) { 1828 poll_wait(filp, wait_address, p); 1829 /* 1830 * We need to be sure we are in sync with the 1831 * socket flags modification. 1832 * 1833 * This memory barrier is paired in the wq_has_sleeper. 1834 */ 1835 smp_mb(); 1836 } 1837 } 1838 1839 /* 1840 * Queue a received datagram if it will fit. Stream and sequenced 1841 * protocols can't normally use this as they need to fit buffers in 1842 * and play with them. 1843 * 1844 * Inlined as it's very short and called for pretty much every 1845 * packet ever received. 1846 */ 1847 1848 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) 1849 { 1850 skb_orphan(skb); 1851 skb->sk = sk; 1852 skb->destructor = sock_wfree; 1853 /* 1854 * We used to take a refcount on sk, but following operation 1855 * is enough to guarantee sk_free() wont free this sock until 1856 * all in-flight packets are completed 1857 */ 1858 atomic_add(skb->truesize, &sk->sk_wmem_alloc); 1859 } 1860 1861 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) 1862 { 1863 skb_orphan(skb); 1864 skb->sk = sk; 1865 skb->destructor = sock_rfree; 1866 atomic_add(skb->truesize, &sk->sk_rmem_alloc); 1867 sk_mem_charge(sk, skb->truesize); 1868 } 1869 1870 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, 1871 unsigned long expires); 1872 1873 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); 1874 1875 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); 1876 1877 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); 1878 1879 /* 1880 * Recover an error report and clear atomically 1881 */ 1882 1883 static inline int sock_error(struct sock *sk) 1884 { 1885 int err; 1886 if (likely(!sk->sk_err)) 1887 return 0; 1888 err = xchg(&sk->sk_err, 0); 1889 return -err; 1890 } 1891 1892 static inline unsigned long sock_wspace(struct sock *sk) 1893 { 1894 int amt = 0; 1895 1896 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { 1897 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); 1898 if (amt < 0) 1899 amt = 0; 1900 } 1901 return amt; 1902 } 1903 1904 static inline void sk_wake_async(struct sock *sk, int how, int band) 1905 { 1906 if (sock_flag(sk, SOCK_FASYNC)) 1907 sock_wake_async(sk->sk_socket, how, band); 1908 } 1909 1910 #define SOCK_MIN_SNDBUF 2048 1911 /* 1912 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need 1913 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak 1914 */ 1915 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff)) 1916 1917 static inline void sk_stream_moderate_sndbuf(struct sock *sk) 1918 { 1919 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { 1920 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); 1921 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); 1922 } 1923 } 1924 1925 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp); 1926 1927 static inline struct page *sk_stream_alloc_page(struct sock *sk) 1928 { 1929 struct page *page = NULL; 1930 1931 page = alloc_pages(sk->sk_allocation, 0); 1932 if (!page) { 1933 sk_enter_memory_pressure(sk); 1934 sk_stream_moderate_sndbuf(sk); 1935 } 1936 return page; 1937 } 1938 1939 /* 1940 * Default write policy as shown to user space via poll/select/SIGIO 1941 */ 1942 static inline int sock_writeable(const struct sock *sk) 1943 { 1944 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); 1945 } 1946 1947 static inline gfp_t gfp_any(void) 1948 { 1949 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; 1950 } 1951 1952 static inline long sock_rcvtimeo(const struct sock *sk, int noblock) 1953 { 1954 return noblock ? 0 : sk->sk_rcvtimeo; 1955 } 1956 1957 static inline long sock_sndtimeo(const struct sock *sk, int noblock) 1958 { 1959 return noblock ? 0 : sk->sk_sndtimeo; 1960 } 1961 1962 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) 1963 { 1964 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; 1965 } 1966 1967 /* Alas, with timeout socket operations are not restartable. 1968 * Compare this to poll(). 1969 */ 1970 static inline int sock_intr_errno(long timeo) 1971 { 1972 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; 1973 } 1974 1975 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 1976 struct sk_buff *skb); 1977 extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, 1978 struct sk_buff *skb); 1979 1980 static __inline__ void 1981 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) 1982 { 1983 ktime_t kt = skb->tstamp; 1984 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); 1985 1986 /* 1987 * generate control messages if 1988 * - receive time stamping in software requested (SOCK_RCVTSTAMP 1989 * or SOCK_TIMESTAMPING_RX_SOFTWARE) 1990 * - software time stamp available and wanted 1991 * (SOCK_TIMESTAMPING_SOFTWARE) 1992 * - hardware time stamps available and wanted 1993 * (SOCK_TIMESTAMPING_SYS_HARDWARE or 1994 * SOCK_TIMESTAMPING_RAW_HARDWARE) 1995 */ 1996 if (sock_flag(sk, SOCK_RCVTSTAMP) || 1997 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) || 1998 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) || 1999 (hwtstamps->hwtstamp.tv64 && 2000 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) || 2001 (hwtstamps->syststamp.tv64 && 2002 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))) 2003 __sock_recv_timestamp(msg, sk, skb); 2004 else 2005 sk->sk_stamp = kt; 2006 2007 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid) 2008 __sock_recv_wifi_status(msg, sk, skb); 2009 } 2010 2011 extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, 2012 struct sk_buff *skb); 2013 2014 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, 2015 struct sk_buff *skb) 2016 { 2017 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ 2018 (1UL << SOCK_RCVTSTAMP) | \ 2019 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \ 2020 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \ 2021 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \ 2022 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE)) 2023 2024 if (sk->sk_flags & FLAGS_TS_OR_DROPS) 2025 __sock_recv_ts_and_drops(msg, sk, skb); 2026 else 2027 sk->sk_stamp = skb->tstamp; 2028 } 2029 2030 /** 2031 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped 2032 * @sk: socket sending this packet 2033 * @tx_flags: filled with instructions for time stamping 2034 * 2035 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if 2036 * parameters are invalid. 2037 */ 2038 extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags); 2039 2040 /** 2041 * sk_eat_skb - Release a skb if it is no longer needed 2042 * @sk: socket to eat this skb from 2043 * @skb: socket buffer to eat 2044 * @copied_early: flag indicating whether DMA operations copied this data early 2045 * 2046 * This routine must be called with interrupts disabled or with the socket 2047 * locked so that the sk_buff queue operation is ok. 2048 */ 2049 #ifdef CONFIG_NET_DMA 2050 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) 2051 { 2052 __skb_unlink(skb, &sk->sk_receive_queue); 2053 if (!copied_early) 2054 __kfree_skb(skb); 2055 else 2056 __skb_queue_tail(&sk->sk_async_wait_queue, skb); 2057 } 2058 #else 2059 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) 2060 { 2061 __skb_unlink(skb, &sk->sk_receive_queue); 2062 __kfree_skb(skb); 2063 } 2064 #endif 2065 2066 static inline 2067 struct net *sock_net(const struct sock *sk) 2068 { 2069 return read_pnet(&sk->sk_net); 2070 } 2071 2072 static inline 2073 void sock_net_set(struct sock *sk, struct net *net) 2074 { 2075 write_pnet(&sk->sk_net, net); 2076 } 2077 2078 /* 2079 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace. 2080 * They should not hold a reference to a namespace in order to allow 2081 * to stop it. 2082 * Sockets after sk_change_net should be released using sk_release_kernel 2083 */ 2084 static inline void sk_change_net(struct sock *sk, struct net *net) 2085 { 2086 put_net(sock_net(sk)); 2087 sock_net_set(sk, hold_net(net)); 2088 } 2089 2090 static inline struct sock *skb_steal_sock(struct sk_buff *skb) 2091 { 2092 if (unlikely(skb->sk)) { 2093 struct sock *sk = skb->sk; 2094 2095 skb->destructor = NULL; 2096 skb->sk = NULL; 2097 return sk; 2098 } 2099 return NULL; 2100 } 2101 2102 extern void sock_enable_timestamp(struct sock *sk, int flag); 2103 extern int sock_get_timestamp(struct sock *, struct timeval __user *); 2104 extern int sock_get_timestampns(struct sock *, struct timespec __user *); 2105 2106 /* 2107 * Enable debug/info messages 2108 */ 2109 extern int net_msg_warn; 2110 #define NETDEBUG(fmt, args...) \ 2111 do { if (net_msg_warn) printk(fmt,##args); } while (0) 2112 2113 #define LIMIT_NETDEBUG(fmt, args...) \ 2114 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) 2115 2116 extern __u32 sysctl_wmem_max; 2117 extern __u32 sysctl_rmem_max; 2118 2119 extern void sk_init(void); 2120 2121 extern int sysctl_optmem_max; 2122 2123 extern __u32 sysctl_wmem_default; 2124 extern __u32 sysctl_rmem_default; 2125 2126 #endif /* _SOCK_H */ 2127