linux/include/net/sock.h
<<
>>
Prefs
   1/*
   2 * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3 *              operating system.  INET is implemented using the  BSD Socket
   4 *              interface as the means of communication with the user level.
   5 *
   6 *              Definitions for the AF_INET socket handler.
   7 *
   8 * Version:     @(#)sock.h      1.0.4   05/13/93
   9 *
  10 * Authors:     Ross Biro
  11 *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  12 *              Corey Minyard <wf-rch!minyard@relay.EU.net>
  13 *              Florian La Roche <flla@stud.uni-sb.de>
  14 *
  15 * Fixes:
  16 *              Alan Cox        :       Volatiles in skbuff pointers. See
  17 *                                      skbuff comments. May be overdone,
  18 *                                      better to prove they can be removed
  19 *                                      than the reverse.
  20 *              Alan Cox        :       Added a zapped field for tcp to note
  21 *                                      a socket is reset and must stay shut up
  22 *              Alan Cox        :       New fields for options
  23 *      Pauline Middelink       :       identd support
  24 *              Alan Cox        :       Eliminate low level recv/recvfrom
  25 *              David S. Miller :       New socket lookup architecture.
  26 *              Steve Whitehouse:       Default routines for sock_ops
  27 *              Arnaldo C. Melo :       removed net_pinfo, tp_pinfo and made
  28 *                                      protinfo be just a void pointer, as the
  29 *                                      protocol specific parts were moved to
  30 *                                      respective headers and ipv4/v6, etc now
  31 *                                      use private slabcaches for its socks
  32 *              Pedro Hortas    :       New flags field for socket options
  33 *
  34 *
  35 *              This program is free software; you can redistribute it and/or
  36 *              modify it under the terms of the GNU General Public License
  37 *              as published by the Free Software Foundation; either version
  38 *              2 of the License, or (at your option) any later version.
  39 */
  40#ifndef _SOCK_H
  41#define _SOCK_H
  42
  43#include <linux/kernel.h>
  44#include <linux/list.h>
  45#include <linux/list_nulls.h>
  46#include <linux/timer.h>
  47#include <linux/cache.h>
  48#include <linux/module.h>
  49#include <linux/lockdep.h>
  50#include <linux/netdevice.h>
  51#include <linux/skbuff.h>       /* struct sk_buff */
  52#include <linux/mm.h>
  53#include <linux/security.h>
  54#include <linux/slab.h>
  55
  56#include <linux/filter.h>
  57#include <linux/rculist_nulls.h>
  58#include <linux/poll.h>
  59
  60#include <asm/atomic.h>
  61#include <net/dst.h>
  62#include <net/checksum.h>
  63
  64/*
  65 * This structure really needs to be cleaned up.
  66 * Most of it is for TCP, and not used by any of
  67 * the other protocols.
  68 */
  69
  70/* Define this to get the SOCK_DBG debugging facility. */
  71#define SOCK_DEBUGGING
  72#ifdef SOCK_DEBUGGING
  73#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
  74                                        printk(KERN_DEBUG msg); } while (0)
  75#else
  76/* Validate arguments and do nothing */
  77static inline void __attribute__ ((format (printf, 2, 3)))
  78SOCK_DEBUG(struct sock *sk, const char *msg, ...)
  79{
  80}
  81#endif
  82
  83/* This is the per-socket lock.  The spinlock provides a synchronization
  84 * between user contexts and software interrupt processing, whereas the
  85 * mini-semaphore synchronizes multiple users amongst themselves.
  86 */
  87typedef struct {
  88        spinlock_t              slock;
  89        int                     owned;
  90        wait_queue_head_t       wq;
  91        /*
  92         * We express the mutex-alike socket_lock semantics
  93         * to the lock validator by explicitly managing
  94         * the slock as a lock variant (in addition to
  95         * the slock itself):
  96         */
  97#ifdef CONFIG_DEBUG_LOCK_ALLOC
  98        struct lockdep_map dep_map;
  99#endif
 100} socket_lock_t;
 101
 102struct sock;
 103struct proto;
 104struct net;
 105
 106/**
 107 *      struct sock_common - minimal network layer representation of sockets
 108 *      @skc_node: main hash linkage for various protocol lookup tables
 109 *      @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
 110 *      @skc_refcnt: reference count
 111 *      @skc_tx_queue_mapping: tx queue number for this connection
 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 *
 123 *      This is the minimal network layer representation of sockets, the header
 124 *      for struct sock and struct inet_timewait_sock.
 125 */
 126struct sock_common {
 127        /*
 128         * first fields are not copied in sock_copy()
 129         */
 130        union {
 131                struct hlist_node       skc_node;
 132                struct hlist_nulls_node skc_nulls_node;
 133        };
 134        atomic_t                skc_refcnt;
 135        int                     skc_tx_queue_mapping;
 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
 155/**
 156  *     struct sock - network layer representation of sockets
 157  *     @__sk_common: shared layout with inet_timewait_sock
 158  *     @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
 159  *     @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
 160  *     @sk_lock:       synchronizer
 161  *     @sk_rcvbuf: size of receive buffer in bytes
 162  *     @sk_wq: sock wait queue and async head
 163  *     @sk_dst_cache: destination cache
 164  *     @sk_dst_lock: destination cache lock
 165  *     @sk_policy: flow policy
 166  *     @sk_rmem_alloc: receive queue bytes committed
 167  *     @sk_receive_queue: incoming packets
 168  *     @sk_wmem_alloc: transmit queue bytes committed
 169  *     @sk_write_queue: Packet sending queue
 170  *     @sk_async_wait_queue: DMA copied packets
 171  *     @sk_omem_alloc: "o" is "option" or "other"
 172  *     @sk_wmem_queued: persistent queue size
 173  *     @sk_forward_alloc: space allocated forward
 174  *     @sk_allocation: allocation mode
 175  *     @sk_sndbuf: size of send buffer in bytes
 176  *     @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
 177  *                %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
 178  *     @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
 179  *     @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
 180  *     @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
 181  *     @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
 182  *     @sk_gso_max_size: Maximum GSO segment size to build
 183  *     @sk_lingertime: %SO_LINGER l_linger setting
 184  *     @sk_backlog: always used with the per-socket spinlock held
 185  *     @sk_callback_lock: used with the callbacks in the end of this struct
 186  *     @sk_error_queue: rarely used
 187  *     @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
 188  *                       IPV6_ADDRFORM for instance)
 189  *     @sk_err: last error
 190  *     @sk_err_soft: errors that don't cause failure but are the cause of a
 191  *                   persistent failure not just 'timed out'
 192  *     @sk_drops: raw/udp drops counter
 193  *     @sk_ack_backlog: current listen backlog
 194  *     @sk_max_ack_backlog: listen backlog set in listen()
 195  *     @sk_priority: %SO_PRIORITY setting
 196  *     @sk_type: socket type (%SOCK_STREAM, etc)
 197  *     @sk_protocol: which protocol this socket belongs in this network family
 198  *     @sk_peer_pid: &struct pid for this socket's peer
 199  *     @sk_peer_cred: %SO_PEERCRED setting
 200  *     @sk_rcvlowat: %SO_RCVLOWAT setting
 201  *     @sk_rcvtimeo: %SO_RCVTIMEO setting
 202  *     @sk_sndtimeo: %SO_SNDTIMEO setting
 203  *     @sk_rxhash: flow hash received from netif layer
 204  *     @sk_filter: socket filtering instructions
 205  *     @sk_protinfo: private area, net family specific, when not using slab
 206  *     @sk_timer: sock cleanup timer
 207  *     @sk_stamp: time stamp of last packet received
 208  *     @sk_socket: Identd and reporting IO signals
 209  *     @sk_user_data: RPC layer private data
 210  *     @sk_sndmsg_page: cached page for sendmsg
 211  *     @sk_sndmsg_off: cached offset for sendmsg
 212  *     @sk_send_head: front of stuff to transmit
 213  *     @sk_security: used by security modules
 214  *     @sk_mark: generic packet mark
 215  *     @sk_classid: this socket's cgroup classid
 216  *     @sk_write_pending: a write to stream socket waits to start
 217  *     @sk_state_change: callback to indicate change in the state of the sock
 218  *     @sk_data_ready: callback to indicate there is data to be processed
 219  *     @sk_write_space: callback to indicate there is bf sending space available
 220  *     @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
 221  *     @sk_backlog_rcv: callback to process the backlog
 222  *     @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
 223 */
 224struct sock {
 225        /*
 226         * Now struct inet_timewait_sock also uses sock_common, so please just
 227         * don't add nothing before this first member (__sk_common) --acme
 228         */
 229        struct sock_common      __sk_common;
 230#define sk_node                 __sk_common.skc_node
 231#define sk_nulls_node           __sk_common.skc_nulls_node
 232#define sk_refcnt               __sk_common.skc_refcnt
 233#define sk_tx_queue_mapping     __sk_common.skc_tx_queue_mapping
 234
 235#define sk_copy_start           __sk_common.skc_hash
 236#define sk_hash                 __sk_common.skc_hash
 237#define sk_family               __sk_common.skc_family
 238#define sk_state                __sk_common.skc_state
 239#define sk_reuse                __sk_common.skc_reuse
 240#define sk_bound_dev_if         __sk_common.skc_bound_dev_if
 241#define sk_bind_node            __sk_common.skc_bind_node
 242#define sk_prot                 __sk_common.skc_prot
 243#define sk_net                  __sk_common.skc_net
 244        kmemcheck_bitfield_begin(flags);
 245        unsigned int            sk_shutdown  : 2,
 246                                sk_no_check  : 2,
 247                                sk_userlocks : 4,
 248                                sk_protocol  : 8,
 249                                sk_type      : 16;
 250        kmemcheck_bitfield_end(flags);
 251        int                     sk_rcvbuf;
 252        socket_lock_t           sk_lock;
 253        /*
 254         * The backlog queue is special, it is always used with
 255         * the per-socket spinlock held and requires low latency
 256         * access. Therefore we special case it's implementation.
 257         */
 258        struct {
 259                struct sk_buff *head;
 260                struct sk_buff *tail;
 261                int len;
 262        } sk_backlog;
 263        struct socket_wq        *sk_wq;
 264        struct dst_entry        *sk_dst_cache;
 265#ifdef CONFIG_XFRM
 266        struct xfrm_policy      *sk_policy[2];
 267#endif
 268        spinlock_t              sk_dst_lock;
 269        atomic_t                sk_rmem_alloc;
 270        atomic_t                sk_wmem_alloc;
 271        atomic_t                sk_omem_alloc;
 272        int                     sk_sndbuf;
 273        struct sk_buff_head     sk_receive_queue;
 274        struct sk_buff_head     sk_write_queue;
 275#ifdef CONFIG_NET_DMA
 276        struct sk_buff_head     sk_async_wait_queue;
 277#endif
 278        int                     sk_wmem_queued;
 279        int                     sk_forward_alloc;
 280        gfp_t                   sk_allocation;
 281        int                     sk_route_caps;
 282        int                     sk_route_nocaps;
 283        int                     sk_gso_type;
 284        unsigned int            sk_gso_max_size;
 285        int                     sk_rcvlowat;
 286#ifdef CONFIG_RPS
 287        __u32                   sk_rxhash;
 288#endif
 289        unsigned long           sk_flags;
 290        unsigned long           sk_lingertime;
 291        struct sk_buff_head     sk_error_queue;
 292        struct proto            *sk_prot_creator;
 293        rwlock_t                sk_callback_lock;
 294        int                     sk_err,
 295                                sk_err_soft;
 296        atomic_t                sk_drops;
 297        unsigned short          sk_ack_backlog;
 298        unsigned short          sk_max_ack_backlog;
 299        __u32                   sk_priority;
 300        struct pid              *sk_peer_pid;
 301        const struct cred       *sk_peer_cred;
 302        long                    sk_rcvtimeo;
 303        long                    sk_sndtimeo;
 304        struct sk_filter __rcu  *sk_filter;
 305        void                    *sk_protinfo;
 306        struct timer_list       sk_timer;
 307        ktime_t                 sk_stamp;
 308        struct socket           *sk_socket;
 309        void                    *sk_user_data;
 310        struct page             *sk_sndmsg_page;
 311        struct sk_buff          *sk_send_head;
 312        __u32                   sk_sndmsg_off;
 313        int                     sk_write_pending;
 314#ifdef CONFIG_SECURITY
 315        void                    *sk_security;
 316#endif
 317        __u32                   sk_mark;
 318        u32                     sk_classid;
 319        void                    (*sk_state_change)(struct sock *sk);
 320        void                    (*sk_data_ready)(struct sock *sk, int bytes);
 321        void                    (*sk_write_space)(struct sock *sk);
 322        void                    (*sk_error_report)(struct sock *sk);
 323        int                     (*sk_backlog_rcv)(struct sock *sk,
 324                                                  struct sk_buff *skb);  
 325        void                    (*sk_destruct)(struct sock *sk);
 326};
 327
 328/*
 329 * Hashed lists helper routines
 330 */
 331static inline struct sock *sk_entry(const struct hlist_node *node)
 332{
 333        return hlist_entry(node, struct sock, sk_node);
 334}
 335
 336static inline struct sock *__sk_head(const struct hlist_head *head)
 337{
 338        return hlist_entry(head->first, struct sock, sk_node);
 339}
 340
 341static inline struct sock *sk_head(const struct hlist_head *head)
 342{
 343        return hlist_empty(head) ? NULL : __sk_head(head);
 344}
 345
 346static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
 347{
 348        return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
 349}
 350
 351static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
 352{
 353        return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
 354}
 355
 356static inline struct sock *sk_next(const struct sock *sk)
 357{
 358        return sk->sk_node.next ?
 359                hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
 360}
 361
 362static inline struct sock *sk_nulls_next(const struct sock *sk)
 363{
 364        return (!is_a_nulls(sk->sk_nulls_node.next)) ?
 365                hlist_nulls_entry(sk->sk_nulls_node.next,
 366                                  struct sock, sk_nulls_node) :
 367                NULL;
 368}
 369
 370static inline int sk_unhashed(const struct sock *sk)
 371{
 372        return hlist_unhashed(&sk->sk_node);
 373}
 374
 375static inline int sk_hashed(const struct sock *sk)
 376{
 377        return !sk_unhashed(sk);
 378}
 379
 380static __inline__ void sk_node_init(struct hlist_node *node)
 381{
 382        node->pprev = NULL;
 383}
 384
 385static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
 386{
 387        node->pprev = NULL;
 388}
 389
 390static __inline__ void __sk_del_node(struct sock *sk)
 391{
 392        __hlist_del(&sk->sk_node);
 393}
 394
 395/* NB: equivalent to hlist_del_init_rcu */
 396static __inline__ int __sk_del_node_init(struct sock *sk)
 397{
 398        if (sk_hashed(sk)) {
 399                __sk_del_node(sk);
 400                sk_node_init(&sk->sk_node);
 401                return 1;
 402        }
 403        return 0;
 404}
 405
 406/* Grab socket reference count. This operation is valid only
 407   when sk is ALREADY grabbed f.e. it is found in hash table
 408   or a list and the lookup is made under lock preventing hash table
 409   modifications.
 410 */
 411
 412static inline void sock_hold(struct sock *sk)
 413{
 414        atomic_inc(&sk->sk_refcnt);
 415}
 416
 417/* Ungrab socket in the context, which assumes that socket refcnt
 418   cannot hit zero, f.e. it is true in context of any socketcall.
 419 */
 420static inline void __sock_put(struct sock *sk)
 421{
 422        atomic_dec(&sk->sk_refcnt);
 423}
 424
 425static __inline__ int sk_del_node_init(struct sock *sk)
 426{
 427        int rc = __sk_del_node_init(sk);
 428
 429        if (rc) {
 430                /* paranoid for a while -acme */
 431                WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
 432                __sock_put(sk);
 433        }
 434        return rc;
 435}
 436#define sk_del_node_init_rcu(sk)        sk_del_node_init(sk)
 437
 438static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
 439{
 440        if (sk_hashed(sk)) {
 441                hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
 442                return 1;
 443        }
 444        return 0;
 445}
 446
 447static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
 448{
 449        int rc = __sk_nulls_del_node_init_rcu(sk);
 450
 451        if (rc) {
 452                /* paranoid for a while -acme */
 453                WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
 454                __sock_put(sk);
 455        }
 456        return rc;
 457}
 458
 459static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
 460{
 461        hlist_add_head(&sk->sk_node, list);
 462}
 463
 464static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
 465{
 466        sock_hold(sk);
 467        __sk_add_node(sk, list);
 468}
 469
 470static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
 471{
 472        sock_hold(sk);
 473        hlist_add_head_rcu(&sk->sk_node, list);
 474}
 475
 476static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
 477{
 478        hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
 479}
 480
 481static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
 482{
 483        sock_hold(sk);
 484        __sk_nulls_add_node_rcu(sk, list);
 485}
 486
 487static __inline__ void __sk_del_bind_node(struct sock *sk)
 488{
 489        __hlist_del(&sk->sk_bind_node);
 490}
 491
 492static __inline__ void sk_add_bind_node(struct sock *sk,
 493                                        struct hlist_head *list)
 494{
 495        hlist_add_head(&sk->sk_bind_node, list);
 496}
 497
 498#define sk_for_each(__sk, node, list) \
 499        hlist_for_each_entry(__sk, node, list, sk_node)
 500#define sk_for_each_rcu(__sk, node, list) \
 501        hlist_for_each_entry_rcu(__sk, node, list, sk_node)
 502#define sk_nulls_for_each(__sk, node, list) \
 503        hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
 504#define sk_nulls_for_each_rcu(__sk, node, list) \
 505        hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
 506#define sk_for_each_from(__sk, node) \
 507        if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
 508                hlist_for_each_entry_from(__sk, node, sk_node)
 509#define sk_nulls_for_each_from(__sk, node) \
 510        if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
 511                hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
 512#define sk_for_each_continue(__sk, node) \
 513        if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
 514                hlist_for_each_entry_continue(__sk, node, sk_node)
 515#define sk_for_each_safe(__sk, node, tmp, list) \
 516        hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
 517#define sk_for_each_bound(__sk, node, list) \
 518        hlist_for_each_entry(__sk, node, list, sk_bind_node)
 519
 520/* Sock flags */
 521enum sock_flags {
 522        SOCK_DEAD,
 523        SOCK_DONE,
 524        SOCK_URGINLINE,
 525        SOCK_KEEPOPEN,
 526        SOCK_LINGER,
 527        SOCK_DESTROY,
 528        SOCK_BROADCAST,
 529        SOCK_TIMESTAMP,
 530        SOCK_ZAPPED,
 531        SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
 532        SOCK_DBG, /* %SO_DEBUG setting */
 533        SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
 534        SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
 535        SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
 536        SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
 537        SOCK_TIMESTAMPING_TX_HARDWARE,  /* %SOF_TIMESTAMPING_TX_HARDWARE */
 538        SOCK_TIMESTAMPING_TX_SOFTWARE,  /* %SOF_TIMESTAMPING_TX_SOFTWARE */
 539        SOCK_TIMESTAMPING_RX_HARDWARE,  /* %SOF_TIMESTAMPING_RX_HARDWARE */
 540        SOCK_TIMESTAMPING_RX_SOFTWARE,  /* %SOF_TIMESTAMPING_RX_SOFTWARE */
 541        SOCK_TIMESTAMPING_SOFTWARE,     /* %SOF_TIMESTAMPING_SOFTWARE */
 542        SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
 543        SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
 544        SOCK_FASYNC, /* fasync() active */
 545        SOCK_RXQ_OVFL,
 546};
 547
 548static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
 549{
 550        nsk->sk_flags = osk->sk_flags;
 551}
 552
 553static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
 554{
 555        __set_bit(flag, &sk->sk_flags);
 556}
 557
 558static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
 559{
 560        __clear_bit(flag, &sk->sk_flags);
 561}
 562
 563static inline int sock_flag(struct sock *sk, enum sock_flags flag)
 564{
 565        return test_bit(flag, &sk->sk_flags);
 566}
 567
 568static inline void sk_acceptq_removed(struct sock *sk)
 569{
 570        sk->sk_ack_backlog--;
 571}
 572
 573static inline void sk_acceptq_added(struct sock *sk)
 574{
 575        sk->sk_ack_backlog++;
 576}
 577
 578static inline int sk_acceptq_is_full(struct sock *sk)
 579{
 580        return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
 581}
 582
 583/*
 584 * Compute minimal free write space needed to queue new packets.
 585 */
 586static inline int sk_stream_min_wspace(struct sock *sk)
 587{
 588        return sk->sk_wmem_queued >> 1;
 589}
 590
 591static inline int sk_stream_wspace(struct sock *sk)
 592{
 593        return sk->sk_sndbuf - sk->sk_wmem_queued;
 594}
 595
 596extern void sk_stream_write_space(struct sock *sk);
 597
 598static inline int sk_stream_memory_free(struct sock *sk)
 599{
 600        return sk->sk_wmem_queued < sk->sk_sndbuf;
 601}
 602
 603/* OOB backlog add */
 604static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
 605{
 606        /* dont let skb dst not refcounted, we are going to leave rcu lock */
 607        skb_dst_force(skb);
 608
 609        if (!sk->sk_backlog.tail)
 610                sk->sk_backlog.head = skb;
 611        else
 612                sk->sk_backlog.tail->next = skb;
 613
 614        sk->sk_backlog.tail = skb;
 615        skb->next = NULL;
 616}
 617
 618/*
 619 * Take into account size of receive queue and backlog queue
 620 */
 621static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
 622{
 623        unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
 624
 625        return qsize + skb->truesize > sk->sk_rcvbuf;
 626}
 627
 628/* The per-socket spinlock must be held here. */
 629static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
 630{
 631        if (sk_rcvqueues_full(sk, skb))
 632                return -ENOBUFS;
 633
 634        __sk_add_backlog(sk, skb);
 635        sk->sk_backlog.len += skb->truesize;
 636        return 0;
 637}
 638
 639static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
 640{
 641        return sk->sk_backlog_rcv(sk, skb);
 642}
 643
 644static inline void sock_rps_record_flow(const struct sock *sk)
 645{
 646#ifdef CONFIG_RPS
 647        struct rps_sock_flow_table *sock_flow_table;
 648
 649        rcu_read_lock();
 650        sock_flow_table = rcu_dereference(rps_sock_flow_table);
 651        rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
 652        rcu_read_unlock();
 653#endif
 654}
 655
 656static inline void sock_rps_reset_flow(const struct sock *sk)
 657{
 658#ifdef CONFIG_RPS
 659        struct rps_sock_flow_table *sock_flow_table;
 660
 661        rcu_read_lock();
 662        sock_flow_table = rcu_dereference(rps_sock_flow_table);
 663        rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
 664        rcu_read_unlock();
 665#endif
 666}
 667
 668static inline void sock_rps_save_rxhash(struct sock *sk, u32 rxhash)
 669{
 670#ifdef CONFIG_RPS
 671        if (unlikely(sk->sk_rxhash != rxhash)) {
 672                sock_rps_reset_flow(sk);
 673                sk->sk_rxhash = rxhash;
 674        }
 675#endif
 676}
 677
 678#define sk_wait_event(__sk, __timeo, __condition)                       \
 679        ({      int __rc;                                               \
 680                release_sock(__sk);                                     \
 681                __rc = __condition;                                     \
 682                if (!__rc) {                                            \
 683                        *(__timeo) = schedule_timeout(*(__timeo));      \
 684                }                                                       \
 685                lock_sock(__sk);                                        \
 686                __rc = __condition;                                     \
 687                __rc;                                                   \
 688        })
 689
 690extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
 691extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
 692extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
 693extern int sk_stream_error(struct sock *sk, int flags, int err);
 694extern void sk_stream_kill_queues(struct sock *sk);
 695
 696extern int sk_wait_data(struct sock *sk, long *timeo);
 697
 698struct request_sock_ops;
 699struct timewait_sock_ops;
 700struct inet_hashinfo;
 701struct raw_hashinfo;
 702
 703/* Networking protocol blocks we attach to sockets.
 704 * socket layer -> transport layer interface
 705 * transport -> network interface is defined by struct inet_proto
 706 */
 707struct proto {
 708        void                    (*close)(struct sock *sk, 
 709                                        long timeout);
 710        int                     (*connect)(struct sock *sk,
 711                                        struct sockaddr *uaddr, 
 712                                        int addr_len);
 713        int                     (*disconnect)(struct sock *sk, int flags);
 714
 715        struct sock *           (*accept) (struct sock *sk, int flags, int *err);
 716
 717        int                     (*ioctl)(struct sock *sk, int cmd,
 718                                         unsigned long arg);
 719        int                     (*init)(struct sock *sk);
 720        void                    (*destroy)(struct sock *sk);
 721        void                    (*shutdown)(struct sock *sk, int how);
 722        int                     (*setsockopt)(struct sock *sk, int level, 
 723                                        int optname, char __user *optval,
 724                                        unsigned int optlen);
 725        int                     (*getsockopt)(struct sock *sk, int level, 
 726                                        int optname, char __user *optval, 
 727                                        int __user *option);     
 728#ifdef CONFIG_COMPAT
 729        int                     (*compat_setsockopt)(struct sock *sk,
 730                                        int level,
 731                                        int optname, char __user *optval,
 732                                        unsigned int optlen);
 733        int                     (*compat_getsockopt)(struct sock *sk,
 734                                        int level,
 735                                        int optname, char __user *optval,
 736                                        int __user *option);
 737#endif
 738        int                     (*sendmsg)(struct kiocb *iocb, struct sock *sk,
 739                                           struct msghdr *msg, size_t len);
 740        int                     (*recvmsg)(struct kiocb *iocb, struct sock *sk,
 741                                           struct msghdr *msg,
 742                                        size_t len, int noblock, int flags, 
 743                                        int *addr_len);
 744        int                     (*sendpage)(struct sock *sk, struct page *page,
 745                                        int offset, size_t size, int flags);
 746        int                     (*bind)(struct sock *sk, 
 747                                        struct sockaddr *uaddr, int addr_len);
 748
 749        int                     (*backlog_rcv) (struct sock *sk, 
 750                                                struct sk_buff *skb);
 751
 752        /* Keeping track of sk's, looking them up, and port selection methods. */
 753        void                    (*hash)(struct sock *sk);
 754        void                    (*unhash)(struct sock *sk);
 755        void                    (*rehash)(struct sock *sk);
 756        int                     (*get_port)(struct sock *sk, unsigned short snum);
 757        void                    (*clear_sk)(struct sock *sk, int size);
 758
 759        /* Keeping track of sockets in use */
 760#ifdef CONFIG_PROC_FS
 761        unsigned int            inuse_idx;
 762#endif
 763
 764        /* Memory pressure */
 765        void                    (*enter_memory_pressure)(struct sock *sk);
 766        atomic_long_t           *memory_allocated;      /* Current allocated memory. */
 767        struct percpu_counter   *sockets_allocated;     /* Current number of sockets. */
 768        /*
 769         * Pressure flag: try to collapse.
 770         * Technical note: it is used by multiple contexts non atomically.
 771         * All the __sk_mem_schedule() is of this nature: accounting
 772         * is strict, actions are advisory and have some latency.
 773         */
 774        int                     *memory_pressure;
 775        long                    *sysctl_mem;
 776        int                     *sysctl_wmem;
 777        int                     *sysctl_rmem;
 778        int                     max_header;
 779        bool                    no_autobind;
 780
 781        struct kmem_cache       *slab;
 782        unsigned int            obj_size;
 783        int                     slab_flags;
 784
 785        struct percpu_counter   *orphan_count;
 786
 787        struct request_sock_ops *rsk_prot;
 788        struct timewait_sock_ops *twsk_prot;
 789
 790        union {
 791                struct inet_hashinfo    *hashinfo;
 792                struct udp_table        *udp_table;
 793                struct raw_hashinfo     *raw_hash;
 794        } h;
 795
 796        struct module           *owner;
 797
 798        char                    name[32];
 799
 800        struct list_head        node;
 801#ifdef SOCK_REFCNT_DEBUG
 802        atomic_t                socks;
 803#endif
 804};
 805
 806extern int proto_register(struct proto *prot, int alloc_slab);
 807extern void proto_unregister(struct proto *prot);
 808
 809#ifdef SOCK_REFCNT_DEBUG
 810static inline void sk_refcnt_debug_inc(struct sock *sk)
 811{
 812        atomic_inc(&sk->sk_prot->socks);
 813}
 814
 815static inline void sk_refcnt_debug_dec(struct sock *sk)
 816{
 817        atomic_dec(&sk->sk_prot->socks);
 818        printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
 819               sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
 820}
 821
 822static inline void sk_refcnt_debug_release(const struct sock *sk)
 823{
 824        if (atomic_read(&sk->sk_refcnt) != 1)
 825                printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
 826                       sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
 827}
 828#else /* SOCK_REFCNT_DEBUG */
 829#define sk_refcnt_debug_inc(sk) do { } while (0)
 830#define sk_refcnt_debug_dec(sk) do { } while (0)
 831#define sk_refcnt_debug_release(sk) do { } while (0)
 832#endif /* SOCK_REFCNT_DEBUG */
 833
 834
 835#ifdef CONFIG_PROC_FS
 836/* Called with local bh disabled */
 837extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
 838extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
 839#else
 840static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
 841                int inc)
 842{
 843}
 844#endif
 845
 846
 847/* With per-bucket locks this operation is not-atomic, so that
 848 * this version is not worse.
 849 */
 850static inline void __sk_prot_rehash(struct sock *sk)
 851{
 852        sk->sk_prot->unhash(sk);
 853        sk->sk_prot->hash(sk);
 854}
 855
 856void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
 857
 858/* About 10 seconds */
 859#define SOCK_DESTROY_TIME (10*HZ)
 860
 861/* Sockets 0-1023 can't be bound to unless you are superuser */
 862#define PROT_SOCK       1024
 863
 864#define SHUTDOWN_MASK   3
 865#define RCV_SHUTDOWN    1
 866#define SEND_SHUTDOWN   2
 867
 868#define SOCK_SNDBUF_LOCK        1
 869#define SOCK_RCVBUF_LOCK        2
 870#define SOCK_BINDADDR_LOCK      4
 871#define SOCK_BINDPORT_LOCK      8
 872
 873/* sock_iocb: used to kick off async processing of socket ios */
 874struct sock_iocb {
 875        struct list_head        list;
 876
 877        int                     flags;
 878        int                     size;
 879        struct socket           *sock;
 880        struct sock             *sk;
 881        struct scm_cookie       *scm;
 882        struct msghdr           *msg, async_msg;
 883        struct kiocb            *kiocb;
 884};
 885
 886static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
 887{
 888        return (struct sock_iocb *)iocb->private;
 889}
 890
 891static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
 892{
 893        return si->kiocb;
 894}
 895
 896struct socket_alloc {
 897        struct socket socket;
 898        struct inode vfs_inode;
 899};
 900
 901static inline struct socket *SOCKET_I(struct inode *inode)
 902{
 903        return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
 904}
 905
 906static inline struct inode *SOCK_INODE(struct socket *socket)
 907{
 908        return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
 909}
 910
 911/*
 912 * Functions for memory accounting
 913 */
 914extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
 915extern void __sk_mem_reclaim(struct sock *sk);
 916
 917#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
 918#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
 919#define SK_MEM_SEND     0
 920#define SK_MEM_RECV     1
 921
 922static inline int sk_mem_pages(int amt)
 923{
 924        return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
 925}
 926
 927static inline int sk_has_account(struct sock *sk)
 928{
 929        /* return true if protocol supports memory accounting */
 930        return !!sk->sk_prot->memory_allocated;
 931}
 932
 933static inline int sk_wmem_schedule(struct sock *sk, int size)
 934{
 935        if (!sk_has_account(sk))
 936                return 1;
 937        return size <= sk->sk_forward_alloc ||
 938                __sk_mem_schedule(sk, size, SK_MEM_SEND);
 939}
 940
 941static inline int sk_rmem_schedule(struct sock *sk, int size)
 942{
 943        if (!sk_has_account(sk))
 944                return 1;
 945        return size <= sk->sk_forward_alloc ||
 946                __sk_mem_schedule(sk, size, SK_MEM_RECV);
 947}
 948
 949static inline void sk_mem_reclaim(struct sock *sk)
 950{
 951        if (!sk_has_account(sk))
 952                return;
 953        if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
 954                __sk_mem_reclaim(sk);
 955}
 956
 957static inline void sk_mem_reclaim_partial(struct sock *sk)
 958{
 959        if (!sk_has_account(sk))
 960                return;
 961        if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
 962                __sk_mem_reclaim(sk);
 963}
 964
 965static inline void sk_mem_charge(struct sock *sk, int size)
 966{
 967        if (!sk_has_account(sk))
 968                return;
 969        sk->sk_forward_alloc -= size;
 970}
 971
 972static inline void sk_mem_uncharge(struct sock *sk, int size)
 973{
 974        if (!sk_has_account(sk))
 975                return;
 976        sk->sk_forward_alloc += size;
 977}
 978
 979static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
 980{
 981        sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
 982        sk->sk_wmem_queued -= skb->truesize;
 983        sk_mem_uncharge(sk, skb->truesize);
 984        __kfree_skb(skb);
 985}
 986
 987/* Used by processes to "lock" a socket state, so that
 988 * interrupts and bottom half handlers won't change it
 989 * from under us. It essentially blocks any incoming
 990 * packets, so that we won't get any new data or any
 991 * packets that change the state of the socket.
 992 *
 993 * While locked, BH processing will add new packets to
 994 * the backlog queue.  This queue is processed by the
 995 * owner of the socket lock right before it is released.
 996 *
 997 * Since ~2.3.5 it is also exclusive sleep lock serializing
 998 * accesses from user process context.
 999 */
1000#define sock_owned_by_user(sk)  ((sk)->sk_lock.owned)
1001
1002/*
1003 * Macro so as to not evaluate some arguments when
1004 * lockdep is not enabled.
1005 *
1006 * Mark both the sk_lock and the sk_lock.slock as a
1007 * per-address-family lock class.
1008 */
1009#define sock_lock_init_class_and_name(sk, sname, skey, name, key)       \
1010do {                                                                    \
1011        sk->sk_lock.owned = 0;                                          \
1012        init_waitqueue_head(&sk->sk_lock.wq);                           \
1013        spin_lock_init(&(sk)->sk_lock.slock);                           \
1014        debug_check_no_locks_freed((void *)&(sk)->sk_lock,              \
1015                        sizeof((sk)->sk_lock));                         \
1016        lockdep_set_class_and_name(&(sk)->sk_lock.slock,                \
1017                        (skey), (sname));                               \
1018        lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0);     \
1019} while (0)
1020
1021extern void lock_sock_nested(struct sock *sk, int subclass);
1022
1023static inline void lock_sock(struct sock *sk)
1024{
1025        lock_sock_nested(sk, 0);
1026}
1027
1028extern void release_sock(struct sock *sk);
1029
1030/* BH context may only use the following locking interface. */
1031#define bh_lock_sock(__sk)      spin_lock(&((__sk)->sk_lock.slock))
1032#define bh_lock_sock_nested(__sk) \
1033                                spin_lock_nested(&((__sk)->sk_lock.slock), \
1034                                SINGLE_DEPTH_NESTING)
1035#define bh_unlock_sock(__sk)    spin_unlock(&((__sk)->sk_lock.slock))
1036
1037extern bool lock_sock_fast(struct sock *sk);
1038/**
1039 * unlock_sock_fast - complement of lock_sock_fast
1040 * @sk: socket
1041 * @slow: slow mode
1042 *
1043 * fast unlock socket for user context.
1044 * If slow mode is on, we call regular release_sock()
1045 */
1046static inline void unlock_sock_fast(struct sock *sk, bool slow)
1047{
1048        if (slow)
1049                release_sock(sk);
1050        else
1051                spin_unlock_bh(&sk->sk_lock.slock);
1052}
1053
1054
1055extern struct sock              *sk_alloc(struct net *net, int family,
1056                                          gfp_t priority,
1057                                          struct proto *prot);
1058extern void                     sk_free(struct sock *sk);
1059extern void                     sk_release_kernel(struct sock *sk);
1060extern struct sock              *sk_clone(const struct sock *sk,
1061                                          const gfp_t priority);
1062
1063extern struct sk_buff           *sock_wmalloc(struct sock *sk,
1064                                              unsigned long size, int force,
1065                                              gfp_t priority);
1066extern struct sk_buff           *sock_rmalloc(struct sock *sk,
1067                                              unsigned long size, int force,
1068                                              gfp_t priority);
1069extern void                     sock_wfree(struct sk_buff *skb);
1070extern void                     sock_rfree(struct sk_buff *skb);
1071
1072extern int                      sock_setsockopt(struct socket *sock, int level,
1073                                                int op, char __user *optval,
1074                                                unsigned int optlen);
1075
1076extern int                      sock_getsockopt(struct socket *sock, int level,
1077                                                int op, char __user *optval, 
1078                                                int __user *optlen);
1079extern struct sk_buff           *sock_alloc_send_skb(struct sock *sk,
1080                                                     unsigned long size,
1081                                                     int noblock,
1082                                                     int *errcode);
1083extern struct sk_buff           *sock_alloc_send_pskb(struct sock *sk,
1084                                                      unsigned long header_len,
1085                                                      unsigned long data_len,
1086                                                      int noblock,
1087                                                      int *errcode);
1088extern void *sock_kmalloc(struct sock *sk, int size,
1089                          gfp_t priority);
1090extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1091extern void sk_send_sigurg(struct sock *sk);
1092
1093#ifdef CONFIG_CGROUPS
1094extern void sock_update_classid(struct sock *sk);
1095#else
1096static inline void sock_update_classid(struct sock *sk)
1097{
1098}
1099#endif
1100
1101/*
1102 * Functions to fill in entries in struct proto_ops when a protocol
1103 * does not implement a particular function.
1104 */
1105extern int                      sock_no_bind(struct socket *, 
1106                                             struct sockaddr *, int);
1107extern int                      sock_no_connect(struct socket *,
1108                                                struct sockaddr *, int, int);
1109extern int                      sock_no_socketpair(struct socket *,
1110                                                   struct socket *);
1111extern int                      sock_no_accept(struct socket *,
1112                                               struct socket *, int);
1113extern int                      sock_no_getname(struct socket *,
1114                                                struct sockaddr *, int *, int);
1115extern unsigned int             sock_no_poll(struct file *, struct socket *,
1116                                             struct poll_table_struct *);
1117extern int                      sock_no_ioctl(struct socket *, unsigned int,
1118                                              unsigned long);
1119extern int                      sock_no_listen(struct socket *, int);
1120extern int                      sock_no_shutdown(struct socket *, int);
1121extern int                      sock_no_getsockopt(struct socket *, int , int,
1122                                                   char __user *, int __user *);
1123extern int                      sock_no_setsockopt(struct socket *, int, int,
1124                                                   char __user *, unsigned int);
1125extern int                      sock_no_sendmsg(struct kiocb *, struct socket *,
1126                                                struct msghdr *, size_t);
1127extern int                      sock_no_recvmsg(struct kiocb *, struct socket *,
1128                                                struct msghdr *, size_t, int);
1129extern int                      sock_no_mmap(struct file *file,
1130                                             struct socket *sock,
1131                                             struct vm_area_struct *vma);
1132extern ssize_t                  sock_no_sendpage(struct socket *sock,
1133                                                struct page *page,
1134                                                int offset, size_t size, 
1135                                                int flags);
1136
1137/*
1138 * Functions to fill in entries in struct proto_ops when a protocol
1139 * uses the inet style.
1140 */
1141extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1142                                  char __user *optval, int __user *optlen);
1143extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1144                               struct msghdr *msg, size_t size, int flags);
1145extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1146                                  char __user *optval, unsigned int optlen);
1147extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1148                int optname, char __user *optval, int __user *optlen);
1149extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1150                int optname, char __user *optval, unsigned int optlen);
1151
1152extern void sk_common_release(struct sock *sk);
1153
1154/*
1155 *      Default socket callbacks and setup code
1156 */
1157 
1158/* Initialise core socket variables */
1159extern void sock_init_data(struct socket *sock, struct sock *sk);
1160
1161extern void sk_filter_release_rcu(struct rcu_head *rcu);
1162
1163/**
1164 *      sk_filter_release - release a socket filter
1165 *      @fp: filter to remove
1166 *
1167 *      Remove a filter from a socket and release its resources.
1168 */
1169
1170static inline void sk_filter_release(struct sk_filter *fp)
1171{
1172        if (atomic_dec_and_test(&fp->refcnt))
1173                call_rcu_bh(&fp->rcu, sk_filter_release_rcu);
1174}
1175
1176static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1177{
1178        unsigned int size = sk_filter_len(fp);
1179
1180        atomic_sub(size, &sk->sk_omem_alloc);
1181        sk_filter_release(fp);
1182}
1183
1184static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1185{
1186        atomic_inc(&fp->refcnt);
1187        atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1188}
1189
1190/*
1191 * Socket reference counting postulates.
1192 *
1193 * * Each user of socket SHOULD hold a reference count.
1194 * * Each access point to socket (an hash table bucket, reference from a list,
1195 *   running timer, skb in flight MUST hold a reference count.
1196 * * When reference count hits 0, it means it will never increase back.
1197 * * When reference count hits 0, it means that no references from
1198 *   outside exist to this socket and current process on current CPU
1199 *   is last user and may/should destroy this socket.
1200 * * sk_free is called from any context: process, BH, IRQ. When
1201 *   it is called, socket has no references from outside -> sk_free
1202 *   may release descendant resources allocated by the socket, but
1203 *   to the time when it is called, socket is NOT referenced by any
1204 *   hash tables, lists etc.
1205 * * Packets, delivered from outside (from network or from another process)
1206 *   and enqueued on receive/error queues SHOULD NOT grab reference count,
1207 *   when they sit in queue. Otherwise, packets will leak to hole, when
1208 *   socket is looked up by one cpu and unhasing is made by another CPU.
1209 *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
1210 *   (leak to backlog). Packet socket does all the processing inside
1211 *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1212 *   use separate SMP lock, so that they are prone too.
1213 */
1214
1215/* Ungrab socket and destroy it, if it was the last reference. */
1216static inline void sock_put(struct sock *sk)
1217{
1218        if (atomic_dec_and_test(&sk->sk_refcnt))
1219                sk_free(sk);
1220}
1221
1222extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1223                          const int nested);
1224
1225static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1226{
1227        sk->sk_tx_queue_mapping = tx_queue;
1228}
1229
1230static inline void sk_tx_queue_clear(struct sock *sk)
1231{
1232        sk->sk_tx_queue_mapping = -1;
1233}
1234
1235static inline int sk_tx_queue_get(const struct sock *sk)
1236{
1237        return sk ? sk->sk_tx_queue_mapping : -1;
1238}
1239
1240static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1241{
1242        sk_tx_queue_clear(sk);
1243        sk->sk_socket = sock;
1244}
1245
1246static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1247{
1248        return &sk->sk_wq->wait;
1249}
1250/* Detach socket from process context.
1251 * Announce socket dead, detach it from wait queue and inode.
1252 * Note that parent inode held reference count on this struct sock,
1253 * we do not release it in this function, because protocol
1254 * probably wants some additional cleanups or even continuing
1255 * to work with this socket (TCP).
1256 */
1257static inline void sock_orphan(struct sock *sk)
1258{
1259        write_lock_bh(&sk->sk_callback_lock);
1260        sock_set_flag(sk, SOCK_DEAD);
1261        sk_set_socket(sk, NULL);
1262        sk->sk_wq  = NULL;
1263        write_unlock_bh(&sk->sk_callback_lock);
1264}
1265
1266static inline void sock_graft(struct sock *sk, struct socket *parent)
1267{
1268        write_lock_bh(&sk->sk_callback_lock);
1269        rcu_assign_pointer(sk->sk_wq, parent->wq);
1270        parent->sk = sk;
1271        sk_set_socket(sk, parent);
1272        security_sock_graft(sk, parent);
1273        write_unlock_bh(&sk->sk_callback_lock);
1274}
1275
1276extern int sock_i_uid(struct sock *sk);
1277extern unsigned long sock_i_ino(struct sock *sk);
1278
1279static inline struct dst_entry *
1280__sk_dst_get(struct sock *sk)
1281{
1282        return rcu_dereference_check(sk->sk_dst_cache, rcu_read_lock_held() ||
1283                                                       sock_owned_by_user(sk) ||
1284                                                       lockdep_is_held(&sk->sk_lock.slock));
1285}
1286
1287static inline struct dst_entry *
1288sk_dst_get(struct sock *sk)
1289{
1290        struct dst_entry *dst;
1291
1292        rcu_read_lock();
1293        dst = rcu_dereference(sk->sk_dst_cache);
1294        if (dst)
1295                dst_hold(dst);
1296        rcu_read_unlock();
1297        return dst;
1298}
1299
1300extern void sk_reset_txq(struct sock *sk);
1301
1302static inline void dst_negative_advice(struct sock *sk)
1303{
1304        struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1305
1306        if (dst && dst->ops->negative_advice) {
1307                ndst = dst->ops->negative_advice(dst);
1308
1309                if (ndst != dst) {
1310                        rcu_assign_pointer(sk->sk_dst_cache, ndst);
1311                        sk_reset_txq(sk);
1312                }
1313        }
1314}
1315
1316static inline void
1317__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1318{
1319        struct dst_entry *old_dst;
1320
1321        sk_tx_queue_clear(sk);
1322        /*
1323         * This can be called while sk is owned by the caller only,
1324         * with no state that can be checked in a rcu_dereference_check() cond
1325         */
1326        old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1327        rcu_assign_pointer(sk->sk_dst_cache, dst);
1328        dst_release(old_dst);
1329}
1330
1331static inline void
1332sk_dst_set(struct sock *sk, struct dst_entry *dst)
1333{
1334        spin_lock(&sk->sk_dst_lock);
1335        __sk_dst_set(sk, dst);
1336        spin_unlock(&sk->sk_dst_lock);
1337}
1338
1339static inline void
1340__sk_dst_reset(struct sock *sk)
1341{
1342        __sk_dst_set(sk, NULL);
1343}
1344
1345static inline void
1346sk_dst_reset(struct sock *sk)
1347{
1348        spin_lock(&sk->sk_dst_lock);
1349        __sk_dst_reset(sk);
1350        spin_unlock(&sk->sk_dst_lock);
1351}
1352
1353extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1354
1355extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1356
1357static inline int sk_can_gso(const struct sock *sk)
1358{
1359        return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1360}
1361
1362extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1363
1364static inline void sk_nocaps_add(struct sock *sk, int flags)
1365{
1366        sk->sk_route_nocaps |= flags;
1367        sk->sk_route_caps &= ~flags;
1368}
1369
1370static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1371                                   struct sk_buff *skb, struct page *page,
1372                                   int off, int copy)
1373{
1374        if (skb->ip_summed == CHECKSUM_NONE) {
1375                int err = 0;
1376                __wsum csum = csum_and_copy_from_user(from,
1377                                                     page_address(page) + off,
1378                                                            copy, 0, &err);
1379                if (err)
1380                        return err;
1381                skb->csum = csum_block_add(skb->csum, csum, skb->len);
1382        } else if (copy_from_user(page_address(page) + off, from, copy))
1383                return -EFAULT;
1384
1385        skb->len             += copy;
1386        skb->data_len        += copy;
1387        skb->truesize        += copy;
1388        sk->sk_wmem_queued   += copy;
1389        sk_mem_charge(sk, copy);
1390        return 0;
1391}
1392
1393/**
1394 * sk_wmem_alloc_get - returns write allocations
1395 * @sk: socket
1396 *
1397 * Returns sk_wmem_alloc minus initial offset of one
1398 */
1399static inline int sk_wmem_alloc_get(const struct sock *sk)
1400{
1401        return atomic_read(&sk->sk_wmem_alloc) - 1;
1402}
1403
1404/**
1405 * sk_rmem_alloc_get - returns read allocations
1406 * @sk: socket
1407 *
1408 * Returns sk_rmem_alloc
1409 */
1410static inline int sk_rmem_alloc_get(const struct sock *sk)
1411{
1412        return atomic_read(&sk->sk_rmem_alloc);
1413}
1414
1415/**
1416 * sk_has_allocations - check if allocations are outstanding
1417 * @sk: socket
1418 *
1419 * Returns true if socket has write or read allocations
1420 */
1421static inline int sk_has_allocations(const struct sock *sk)
1422{
1423        return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1424}
1425
1426/**
1427 * wq_has_sleeper - check if there are any waiting processes
1428 * @wq: struct socket_wq
1429 *
1430 * Returns true if socket_wq has waiting processes
1431 *
1432 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1433 * barrier call. They were added due to the race found within the tcp code.
1434 *
1435 * Consider following tcp code paths:
1436 *
1437 * CPU1                  CPU2
1438 *
1439 * sys_select            receive packet
1440 *   ...                 ...
1441 *   __add_wait_queue    update tp->rcv_nxt
1442 *   ...                 ...
1443 *   tp->rcv_nxt check   sock_def_readable
1444 *   ...                 {
1445 *   schedule               rcu_read_lock();
1446 *                          wq = rcu_dereference(sk->sk_wq);
1447 *                          if (wq && waitqueue_active(&wq->wait))
1448 *                              wake_up_interruptible(&wq->wait)
1449 *                          ...
1450 *                       }
1451 *
1452 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1453 * in its cache, and so does the tp->rcv_nxt update on CPU2 side.  The CPU1
1454 * could then endup calling schedule and sleep forever if there are no more
1455 * data on the socket.
1456 *
1457 */
1458static inline bool wq_has_sleeper(struct socket_wq *wq)
1459{
1460
1461        /*
1462         * We need to be sure we are in sync with the
1463         * add_wait_queue modifications to the wait queue.
1464         *
1465         * This memory barrier is paired in the sock_poll_wait.
1466         */
1467        smp_mb();
1468        return wq && waitqueue_active(&wq->wait);
1469}
1470
1471/**
1472 * sock_poll_wait - place memory barrier behind the poll_wait call.
1473 * @filp:           file
1474 * @wait_address:   socket wait queue
1475 * @p:              poll_table
1476 *
1477 * See the comments in the wq_has_sleeper function.
1478 */
1479static inline void sock_poll_wait(struct file *filp,
1480                wait_queue_head_t *wait_address, poll_table *p)
1481{
1482        if (p && wait_address) {
1483                poll_wait(filp, wait_address, p);
1484                /*
1485                 * We need to be sure we are in sync with the
1486                 * socket flags modification.
1487                 *
1488                 * This memory barrier is paired in the wq_has_sleeper.
1489                */
1490                smp_mb();
1491        }
1492}
1493
1494/*
1495 *      Queue a received datagram if it will fit. Stream and sequenced
1496 *      protocols can't normally use this as they need to fit buffers in
1497 *      and play with them.
1498 *
1499 *      Inlined as it's very short and called for pretty much every
1500 *      packet ever received.
1501 */
1502
1503static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1504{
1505        skb_orphan(skb);
1506        skb->sk = sk;
1507        skb->destructor = sock_wfree;
1508        /*
1509         * We used to take a refcount on sk, but following operation
1510         * is enough to guarantee sk_free() wont free this sock until
1511         * all in-flight packets are completed
1512         */
1513        atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1514}
1515
1516static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1517{
1518        skb_orphan(skb);
1519        skb->sk = sk;
1520        skb->destructor = sock_rfree;
1521        atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1522        sk_mem_charge(sk, skb->truesize);
1523}
1524
1525extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1526                           unsigned long expires);
1527
1528extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1529
1530extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1531
1532extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1533
1534/*
1535 *      Recover an error report and clear atomically
1536 */
1537 
1538static inline int sock_error(struct sock *sk)
1539{
1540        int err;
1541        if (likely(!sk->sk_err))
1542                return 0;
1543        err = xchg(&sk->sk_err, 0);
1544        return -err;
1545}
1546
1547static inline unsigned long sock_wspace(struct sock *sk)
1548{
1549        int amt = 0;
1550
1551        if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1552                amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1553                if (amt < 0) 
1554                        amt = 0;
1555        }
1556        return amt;
1557}
1558
1559static inline void sk_wake_async(struct sock *sk, int how, int band)
1560{
1561        if (sock_flag(sk, SOCK_FASYNC))
1562                sock_wake_async(sk->sk_socket, how, band);
1563}
1564
1565#define SOCK_MIN_SNDBUF 2048
1566/*
1567 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1568 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1569 */
1570#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1571
1572static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1573{
1574        if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1575                sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1576                sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1577        }
1578}
1579
1580struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1581
1582static inline struct page *sk_stream_alloc_page(struct sock *sk)
1583{
1584        struct page *page = NULL;
1585
1586        page = alloc_pages(sk->sk_allocation, 0);
1587        if (!page) {
1588                sk->sk_prot->enter_memory_pressure(sk);
1589                sk_stream_moderate_sndbuf(sk);
1590        }
1591        return page;
1592}
1593
1594/*
1595 *      Default write policy as shown to user space via poll/select/SIGIO
1596 */
1597static inline int sock_writeable(const struct sock *sk) 
1598{
1599        return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1600}
1601
1602static inline gfp_t gfp_any(void)
1603{
1604        return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1605}
1606
1607static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1608{
1609        return noblock ? 0 : sk->sk_rcvtimeo;
1610}
1611
1612static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1613{
1614        return noblock ? 0 : sk->sk_sndtimeo;
1615}
1616
1617static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1618{
1619        return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1620}
1621
1622/* Alas, with timeout socket operations are not restartable.
1623 * Compare this to poll().
1624 */
1625static inline int sock_intr_errno(long timeo)
1626{
1627        return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1628}
1629
1630extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
1631        struct sk_buff *skb);
1632
1633static __inline__ void
1634sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1635{
1636        ktime_t kt = skb->tstamp;
1637        struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
1638
1639        /*
1640         * generate control messages if
1641         * - receive time stamping in software requested (SOCK_RCVTSTAMP
1642         *   or SOCK_TIMESTAMPING_RX_SOFTWARE)
1643         * - software time stamp available and wanted
1644         *   (SOCK_TIMESTAMPING_SOFTWARE)
1645         * - hardware time stamps available and wanted
1646         *   (SOCK_TIMESTAMPING_SYS_HARDWARE or
1647         *   SOCK_TIMESTAMPING_RAW_HARDWARE)
1648         */
1649        if (sock_flag(sk, SOCK_RCVTSTAMP) ||
1650            sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
1651            (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
1652            (hwtstamps->hwtstamp.tv64 &&
1653             sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
1654            (hwtstamps->syststamp.tv64 &&
1655             sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
1656                __sock_recv_timestamp(msg, sk, skb);
1657        else
1658                sk->sk_stamp = kt;
1659}
1660
1661extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1662                                     struct sk_buff *skb);
1663
1664static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
1665                                          struct sk_buff *skb)
1666{
1667#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL)                       | \
1668                           (1UL << SOCK_RCVTSTAMP)                      | \
1669                           (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)       | \
1670                           (1UL << SOCK_TIMESTAMPING_SOFTWARE)          | \
1671                           (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE)      | \
1672                           (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
1673
1674        if (sk->sk_flags & FLAGS_TS_OR_DROPS)
1675                __sock_recv_ts_and_drops(msg, sk, skb);
1676        else
1677                sk->sk_stamp = skb->tstamp;
1678}
1679
1680/**
1681 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1682 * @sk:         socket sending this packet
1683 * @tx_flags:   filled with instructions for time stamping
1684 *
1685 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1686 * parameters are invalid.
1687 */
1688extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
1689
1690/**
1691 * sk_eat_skb - Release a skb if it is no longer needed
1692 * @sk: socket to eat this skb from
1693 * @skb: socket buffer to eat
1694 * @copied_early: flag indicating whether DMA operations copied this data early
1695 *
1696 * This routine must be called with interrupts disabled or with the socket
1697 * locked so that the sk_buff queue operation is ok.
1698*/
1699#ifdef CONFIG_NET_DMA
1700static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1701{
1702        __skb_unlink(skb, &sk->sk_receive_queue);
1703        if (!copied_early)
1704                __kfree_skb(skb);
1705        else
1706                __skb_queue_tail(&sk->sk_async_wait_queue, skb);
1707}
1708#else
1709static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1710{
1711        __skb_unlink(skb, &sk->sk_receive_queue);
1712        __kfree_skb(skb);
1713}
1714#endif
1715
1716static inline
1717struct net *sock_net(const struct sock *sk)
1718{
1719        return read_pnet(&sk->sk_net);
1720}
1721
1722static inline
1723void sock_net_set(struct sock *sk, struct net *net)
1724{
1725        write_pnet(&sk->sk_net, net);
1726}
1727
1728/*
1729 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1730 * They should not hold a referrence to a namespace in order to allow
1731 * to stop it.
1732 * Sockets after sk_change_net should be released using sk_release_kernel
1733 */
1734static inline void sk_change_net(struct sock *sk, struct net *net)
1735{
1736        put_net(sock_net(sk));
1737        sock_net_set(sk, hold_net(net));
1738}
1739
1740static inline struct sock *skb_steal_sock(struct sk_buff *skb)
1741{
1742        if (unlikely(skb->sk)) {
1743                struct sock *sk = skb->sk;
1744
1745                skb->destructor = NULL;
1746                skb->sk = NULL;
1747                return sk;
1748        }
1749        return NULL;
1750}
1751
1752extern void sock_enable_timestamp(struct sock *sk, int flag);
1753extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1754extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1755
1756/* 
1757 *      Enable debug/info messages 
1758 */
1759extern int net_msg_warn;
1760#define NETDEBUG(fmt, args...) \
1761        do { if (net_msg_warn) printk(fmt,##args); } while (0)
1762
1763#define LIMIT_NETDEBUG(fmt, args...) \
1764        do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1765
1766extern __u32 sysctl_wmem_max;
1767extern __u32 sysctl_rmem_max;
1768
1769extern void sk_init(void);
1770
1771extern int sysctl_optmem_max;
1772
1773extern __u32 sysctl_wmem_default;
1774extern __u32 sysctl_rmem_default;
1775
1776#endif  /* _SOCK_H */
1777
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.