include/net/sock.h

   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, <bir7@leland.Stanford.Edu>
  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  *
  28  *              This program is free software; you can redistribute it and/or
  29  *              modify it under the terms of the GNU General Public License
  30  *              as published by the Free Software Foundation; either version
  31  *              2 of the License, or (at your option) any later version.
  32  */
  33 #ifndef _SOCK_H
  34 #define _SOCK_H
  35
  36 #include <linux/config.h>
  37 #include <linux/timer.h>
  38 #include <linux/cache.h>
  39 #include <linux/in.h>           /* struct sockaddr_in */
  40
  41 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
  42 #include <linux/in6.h>          /* struct sockaddr_in6 */
  43 #include <linux/ipv6.h>         /* dest_cache, inet6_options */
  44 #include <linux/icmpv6.h>
  45 #include <net/if_inet6.h>       /* struct ipv6_mc_socklist */
  46 #endif
  47
  48 #if defined(CONFIG_INET) || defined (CONFIG_INET_MODULE)
  49 #include <linux/icmp.h>
  50 #endif
  51 #include <linux/tcp.h>          /* struct tcphdr */
  52
  53 #include <linux/netdevice.h>
  54 #include <linux/skbuff.h>       /* struct sk_buff */
  55 #include <net/protocol.h>               /* struct inet_protocol */
  56 #if defined(CONFIG_X25) || defined(CONFIG_X25_MODULE)
  57 #include <net/x25.h>
  58 #endif
  59 #if defined(CONFIG_AX25) || defined(CONFIG_AX25_MODULE)
  60 #include <net/ax25.h>
  61 #if defined(CONFIG_NETROM) || defined(CONFIG_NETROM_MODULE)
  62 #include <net/netrom.h>
  63 #endif
  64 #if defined(CONFIG_ROSE) || defined(CONFIG_ROSE_MODULE)
  65 #include <net/rose.h>
  66 #endif
  67 #endif
  68
  69 #if defined(CONFIG_PPPOE) || defined(CONFIG_PPPOE_MODULE)
  70 #include <linux/if_pppox.h>
  71 #include <linux/ppp_channel.h>   /* struct ppp_channel */
  72 #endif
  73
  74 #if defined(CONFIG_IPX) || defined(CONFIG_IPX_MODULE)
  75 #if defined(CONFIG_SPX) || defined(CONFIG_SPX_MODULE)
  76 #include <net/spx.h>
  77 #else
  78 #include <net/ipx.h>
  79 #endif /* CONFIG_SPX */
  80 #endif /* CONFIG_IPX */
  81
  82 #if defined(CONFIG_ATALK) || defined(CONFIG_ATALK_MODULE)
  83 #include <linux/atalk.h>
  84 #endif
  85
  86 #if defined(CONFIG_DECNET) || defined(CONFIG_DECNET_MODULE)
  87 #include <net/dn.h>
  88 #endif
  89
  90 #if defined(CONFIG_IRDA) || defined(CONFIG_IRDA_MODULE)
  91 #include <net/irda/irda.h>
  92 #endif
  93
  94 #if defined(CONFIG_ATM) || defined(CONFIG_ATM_MODULE)
  95 struct atm_vcc;
  96 #endif
  97
  98 #ifdef CONFIG_FILTER
  99 #include <linux/filter.h>
 100 #endif
 101
 102 #include <asm/atomic.h>
 103 #include <net/dst.h>
 104
 105
 106 /* The AF_UNIX specific socket options */
 107 struct unix_opt {
 108         struct unix_address     *addr;
 109         struct dentry *         dentry;
 110         struct vfsmount *       mnt;
 111         struct semaphore        readsem;
 112         struct sock *           other;
 113         struct sock **          list;
 114         struct sock *           gc_tree;
 115         atomic_t                inflight;
 116         rwlock_t                lock;
 117         wait_queue_head_t       peer_wait;
 118 };
 119
 120
 121 /* Once the IPX ncpd patches are in these are going into protinfo. */
 122 #if defined(CONFIG_IPX) || defined(CONFIG_IPX_MODULE)
 123 struct ipx_opt {
 124         ipx_address             dest_addr;
 125         ipx_interface           *intrfc;
 126         unsigned short          port;
 127 #ifdef CONFIG_IPX_INTERN
 128         unsigned char           node[IPX_NODE_LEN];
 129 #endif
 130         unsigned short          type;
 131 /*
 132  * To handle special ncp connection-handling sockets for mars_nwe,
 133  * the connection number must be stored in the socket.
 134  */
 135         unsigned short          ipx_ncp_conn;
 136 };
 137 #endif
 138
 139 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
 140 struct ipv6_pinfo {
 141         struct in6_addr         saddr;
 142         struct in6_addr         rcv_saddr;
 143         struct in6_addr         daddr;
 144         struct in6_addr         *daddr_cache;
 145
 146         __u32                   flow_label;
 147         __u32                   frag_size;
 148         int                     hop_limit;
 149         int                     mcast_hops;
 150         int                     mcast_oif;
 151
 152         /* pktoption flags */
 153         union {
 154                 struct {
 155                         __u8    srcrt:2,
 156                                 rxinfo:1,
 157                                 rxhlim:1,
 158                                 hopopts:1,
 159                                 dstopts:1,
 160                                 authhdr:1,
 161                                 rxflow:1;
 162                 } bits;
 163                 __u8            all;
 164         } rxopt;
 165
 166         /* sockopt flags */
 167         __u8                    mc_loop:1,
 168                                 recverr:1,
 169                                 sndflow:1,
 170                                 pmtudisc:2;
 171
 172         struct ipv6_mc_socklist *ipv6_mc_list;
 173         struct ipv6_fl_socklist *ipv6_fl_list;
 174         __u32                   dst_cookie;
 175
 176         struct ipv6_txoptions   *opt;
 177         struct sk_buff          *pktoptions;
 178 };
 179
 180 struct raw6_opt {
 181         __u32                   checksum;       /* perform checksum */
 182         __u32                   offset;         /* checksum offset  */
 183
 184         struct icmp6_filter     filter;
 185 };
 186
 187 #endif /* IPV6 */
 188
 189 #if defined(CONFIG_INET) || defined(CONFIG_INET_MODULE)
 190 struct raw_opt {
 191         struct icmp_filter      filter;
 192 };
 193 #endif
 194
 195 #if defined(CONFIG_INET) || defined (CONFIG_INET_MODULE)
 196 struct inet_opt
 197 {
 198         int                     ttl;                    /* TTL setting */
 199         int                     tos;                    /* TOS */
 200         unsigned                cmsg_flags;
 201         struct ip_options       *opt;
 202         unsigned char           hdrincl;                /* Include headers ? */
 203         __u8                    mc_ttl;                 /* Multicasting TTL */
 204         __u8                    mc_loop;                /* Loopback */
 205         unsigned                recverr : 1,
 206                                 freebind : 1;
 207         __u8                    pmtudisc;
 208         int                     mc_index;               /* Multicast device index */
 209         __u32                   mc_addr;
 210         struct ip_mc_socklist   *mc_list;               /* Group array */
 211 };
 212 #endif
 213
 214 #if defined(CONFIG_PPPOE) || defined (CONFIG_PPPOE_MODULE)
 215 struct pppoe_opt
 216 {
 217         struct net_device      *dev;      /* device associated with socket*/
 218         struct pppoe_addr       pa;       /* what this socket is bound to*/
 219         struct sockaddr_pppox   relay;    /* what socket data will be
 220                                              relayed to (PPPoE relaying) */
 221 };
 222
 223 struct pppox_opt
 224 {
 225         struct ppp_channel      chan;
 226         struct sock             *sk;
 227         struct pppox_opt        *next;    /* for hash table */
 228         union {
 229                 struct pppoe_opt pppoe;
 230         } proto;
 231 };
 232 #define pppoe_dev       proto.pppoe.dev
 233 #define pppoe_pa        proto.pppoe.pa
 234 #define pppoe_relay     proto.pppoe.relay
 235 #endif
 236
 237 /* This defines a selective acknowledgement block. */
 238 struct tcp_sack_block {
 239         __u32   start_seq;
 240         __u32   end_seq;
 241 };
 242
 243 struct tcp_opt {
 244         int     tcp_header_len; /* Bytes of tcp header to send          */
 245
 246 /*
 247  *      Header prediction flags
 248  *      0x5?10 << 16 + snd_wnd in net byte order
 249  */
 250         __u32   pred_flags;
 251
 252 /*
 253  *      RFC793 variables by their proper names. This means you can
 254  *      read the code and the spec side by side (and laugh ...)
 255  *      See RFC793 and RFC1122. The RFC writes these in capitals.
 256  */
 257         __u32   rcv_nxt;        /* What we want to receive next         */
 258         __u32   snd_nxt;        /* Next sequence we send                */
 259
 260         __u32   snd_una;        /* First byte we want an ack for        */
 261         __u32   snd_sml;        /* Last byte of the most recently transmitted small packet */
 262         __u32   rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
 263         __u32   lsndtime;       /* timestamp of last sent data packet (for restart window) */
 264
 265         /* Delayed ACK control data */
 266         struct {
 267                 __u8    pending;        /* ACK is pending */
 268                 __u8    quick;          /* Scheduled number of quick acks       */
 269                 __u8    pingpong;       /* The session is interactive           */
 270                 __u8    blocked;        /* Delayed ACK was blocked by socket lock*/
 271                 __u32   ato;            /* Predicted tick of soft clock         */
 272                 unsigned long timeout;  /* Currently scheduled timeout          */
 273                 __u32   lrcvtime;       /* timestamp of last received data packet*/
 274                 __u16   last_seg_size;  /* Size of last incoming segment        */
 275                 __u16   rcv_mss;        /* MSS used for delayed ACK decisions   */
 276         } ack;
 277
 278         /* Data for direct copy to user */
 279         struct {
 280                 struct sk_buff_head     prequeue;
 281                 int                     memory;
 282                 struct task_struct      *task;
 283                 struct iovec            *iov;
 284                 int                     len;
 285         } ucopy;
 286
 287         __u32   snd_wl1;        /* Sequence for window update           */
 288         __u32   snd_wnd;        /* The window we expect to receive      */
 289         __u32   max_window;     /* Maximal window ever seen from peer   */
 290         __u32   pmtu_cookie;    /* Last pmtu seen by socket             */
 291         __u16   mss_cache;      /* Cached effective mss, not including SACKS */
 292         __u16   mss_clamp;      /* Maximal mss, negotiated at connection setup */
 293         __u16   ext_header_len; /* Network protocol overhead (IP/IPv6 options) */
 294         __u8    ca_state;       /* State of fast-retransmit machine     */
 295         __u8    retransmits;    /* Number of unrecovered RTO timeouts.  */
 296
 297         __u8    reordering;     /* Packet reordering metric.            */
 298         __u8    queue_shrunk;   /* Write queue has been shrunk recently.*/
 299         __u8    defer_accept;   /* User waits for some data after accept() */
 300
 301 /* RTT measurement */
 302         __u8    backoff;        /* backoff                              */
 303         __u32   srtt;           /* smothed round trip time << 3         */
 304         __u32   mdev;           /* medium deviation                     */
 305         __u32   rto;            /* retransmit timeout                   */
 306
 307         __u32   packets_out;    /* Packets which are "in flight"        */
 308         __u32   left_out;       /* Packets which leaved network         */
 309         __u32   retrans_out;    /* Retransmitted packets out            */
 310
 311
 312 /*
 313  *      Slow start and congestion control (see also Nagle, and Karn & Partridge)
 314  */
 315         __u32   snd_ssthresh;   /* Slow start size threshold            */
 316         __u32   snd_cwnd;       /* Sending congestion window            */
 317         __u16   snd_cwnd_cnt;   /* Linear increase counter              */
 318         __u16   snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
 319         __u32   snd_cwnd_used;
 320         __u32   snd_cwnd_stamp;
 321
 322         /* Two commonly used timers in both sender and receiver paths. */
 323         unsigned long           timeout;
 324         struct timer_list       retransmit_timer;       /* Resend (no ack)      */
 325         struct timer_list       delack_timer;           /* Ack delay            */
 326
 327         struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
 328
 329         struct tcp_func         *af_specific;   /* Operations which are AF_INET{4,6} specific   */
 330         struct sk_buff          *send_head;     /* Front of stuff to transmit                   */
 331
 332         __u32   rcv_wnd;        /* Current receiver window              */
 333         __u32   rcv_wup;        /* rcv_nxt on last window update sent   */
 334         __u32   write_seq;      /* Tail(+1) of data held in tcp send buffer */
 335         __u32   pushed_seq;     /* Last pushed seq, required to talk to windows */
 336         __u32   copied_seq;     /* Head of yet unread data              */
 337 /*
 338  *      Options received (usually on last packet, some only on SYN packets).
 339  */
 340         char    tstamp_ok,      /* TIMESTAMP seen on SYN packet         */
 341                 wscale_ok,      /* Wscale seen on SYN packet            */
 342                 sack_ok;        /* SACK seen on SYN packet              */
 343         char    saw_tstamp;     /* Saw TIMESTAMP on last packet         */
 344         __u8    snd_wscale;     /* Window scaling received from sender  */
 345         __u8    rcv_wscale;     /* Window scaling to send to receiver   */
 346         __u8    nonagle;        /* Disable Nagle algorithm?             */
 347         __u8    keepalive_probes; /* num of allowed keep alive probes   */
 348
 349 /*      PAWS/RTTM data  */
 350         __u32   rcv_tsval;      /* Time stamp value                     */
 351         __u32   rcv_tsecr;      /* Time stamp echo reply                */
 352         __u32   ts_recent;      /* Time stamp to echo next              */
 353         long    ts_recent_stamp;/* Time we stored ts_recent (for aging) */
 354
 355 /*      SACKs data      */
 356         __u16   user_mss;       /* mss requested by user in ioctl */
 357         __u8    dsack;          /* D-SACK is scheduled                  */
 358         __u8    eff_sacks;      /* Size of SACK array to send with next packet */
 359         struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
 360         struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 361
 362         __u32   window_clamp;   /* Maximal window to advertise          */
 363         __u32   rcv_ssthresh;   /* Current window clamp                 */
 364         __u8    probes_out;     /* unanswered 0 window probes           */
 365         __u8    num_sacks;      /* Number of SACK blocks                */
 366         __u16   advmss;         /* Advertised MSS                       */
 367
 368         __u8    syn_retries;    /* num of allowed syn retries */
 369         __u8    ecn_flags;      /* ECN status bits.                     */
 370         __u16   prior_ssthresh; /* ssthresh saved at recovery start     */
 371         __u32   lost_out;       /* Lost packets                         */
 372         __u32   sacked_out;     /* SACK'd packets                       */
 373         __u32   fackets_out;    /* FACK'd packets                       */
 374         __u32   high_seq;       /* snd_nxt at onset of congestion       */
 375
 376         __u32   retrans_stamp;  /* Timestamp of the last retransmit,
 377                                  * also used in SYN-SENT to remember stamp of
 378                                  * the first SYN. */
 379         __u32   undo_marker;    /* tracking retrans started here. */
 380         int     undo_retrans;   /* number of undoable retransmissions. */
 381         __u32   syn_seq;        /* Seq of received SYN. */
 382         __u32   fin_seq;        /* Seq of received FIN. */
 383         __u32   urg_seq;        /* Seq of received urgent pointer */
 384         __u16   urg_data;       /* Saved octet of OOB data and control flags */
 385         __u8    pending;        /* Scheduled timer event        */
 386         __u8    urg_mode;       /* In urgent mode               */
 387         __u32   snd_up;         /* Urgent pointer               */
 388
 389         /* The syn_wait_lock is necessary only to avoid tcp_get_info having
 390          * to grab the main lock sock while browsing the listening hash
 391          * (otherwise it's deadlock prone).
 392          * This lock is acquired in read mode only from tcp_get_info() and
 393          * it's acquired in write mode _only_ from code that is actively
 394          * changing the syn_wait_queue. All readers that are holding
 395          * the master sock lock don't need to grab this lock in read mode
 396          * too as the syn_wait_queue writes are always protected from
 397          * the main sock lock.
 398          */
 399         rwlock_t                syn_wait_lock;
 400         struct tcp_listen_opt   *listen_opt;
 401
 402         /* FIFO of established children */
 403         struct open_request     *accept_queue;
 404         struct open_request     *accept_queue_tail;
 405
 406         int                     write_pending;  /* A write to socket waits to start. */
 407
 408         unsigned int            keepalive_time;   /* time before keep alive takes place */
 409         unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
 410         int                     linger2;
 411 };
 412
 413
 414 /*
 415  * This structure really needs to be cleaned up.
 416  * Most of it is for TCP, and not used by any of
 417  * the other protocols.
 418  */
 419
 420 /*
 421  * The idea is to start moving to a newer struct gradualy
 422  *
 423  * IMHO the newer struct should have the following format:
 424  *
 425  *      struct sock {
 426  *              sockmem [mem, proto, callbacks]
 427  *
 428  *              union or struct {
 429  *                      ax25;
 430  *              } ll_pinfo;
 431  *
 432  *              union {
 433  *                      ipv4;
 434  *                      ipv6;
 435  *                      ipx;
 436  *                      netrom;
 437  *                      rose;
 438  *                      x25;
 439  *              } net_pinfo;
 440  *
 441  *              union {
 442  *                      tcp;
 443  *                      udp;
 444  *                      spx;
 445  *                      netrom;
 446  *              } tp_pinfo;
 447  *
 448  *      }
 449  *
 450  * The idea failed because IPv6 transition asssumes dual IP/IPv6 sockets.
 451  * So, net_pinfo is IPv6 are really, and protinfo unifies all another
 452  * private areas.
 453  */
 454
 455 /* Define this to get the sk->debug debugging facility. */
 456 #define SOCK_DEBUGGING
 457 #ifdef SOCK_DEBUGGING
 458 #define SOCK_DEBUG(sk, msg...) do { if((sk) && ((sk)->debug)) printk(KERN_DEBUG ## msg); } while (0)
 459 #else
 460 #define SOCK_DEBUG(sk, msg...) do { } while (0)
 461 #endif
 462
 463 /* This is the per-socket lock.  The spinlock provides a synchronization
 464  * between user contexts and software interrupt processing, whereas the
 465  * mini-semaphore synchronizes multiple users amongst themselves.
 466  */
 467 typedef struct {
 468         spinlock_t              slock;
 469         unsigned int            users;
 470         wait_queue_head_t       wq;
 471 } socket_lock_t;
 472
 473 #define sock_lock_init(__sk) \
 474 do {    spin_lock_init(&((__sk)->lock.slock)); \
 475         (__sk)->lock.users = 0; \
 476         init_waitqueue_head(&((__sk)->lock.wq)); \
 477 } while(0);
 478
 479 struct sock {
 480         /* Socket demultiplex comparisons on incoming packets. */
 481         __u32                   daddr;          /* Foreign IPv4 addr                    */
 482         __u32                   rcv_saddr;      /* Bound local IPv4 addr                */
 483         __u16                   dport;          /* Destination port                     */
 484         unsigned short          num;            /* Local port                           */
 485         int                     bound_dev_if;   /* Bound device index if != 0           */
 486
 487         /* Main hash linkage for various protocol lookup tables. */
 488         struct sock             *next;
 489         struct sock             **pprev;
 490         struct sock             *bind_next;
 491         struct sock             **bind_pprev;
 492
 493         volatile unsigned char  state,          /* Connection state                     */
 494                                 zapped;         /* In ax25 & ipx means not linked       */
 495         __u16                   sport;          /* Source port                          */
 496
 497         unsigned short          family;         /* Address family                       */
 498         unsigned char           reuse;          /* SO_REUSEADDR setting                 */
 499         unsigned char           shutdown;
 500         atomic_t                refcnt;         /* Reference count                      */
 501
 502         socket_lock_t           lock;           /* Synchronizer...                      */
 503         int                     rcvbuf;         /* Size of receive buffer in bytes      */
 504
 505         wait_queue_head_t       *sleep;         /* Sock wait queue                      */
 506         struct dst_entry        *dst_cache;     /* Destination cache                    */
 507         rwlock_t                dst_lock;
 508         atomic_t                rmem_alloc;     /* Receive queue bytes committed        */
 509         struct sk_buff_head     receive_queue;  /* Incoming packets                     */
 510         atomic_t                wmem_alloc;     /* Transmit queue bytes committed       */
 511         struct sk_buff_head     write_queue;    /* Packet sending queue                 */
 512         atomic_t                omem_alloc;     /* "o" is "option" or "other" */
 513         int                     wmem_queued;    /* Persistent queue size */
 514         int                     forward_alloc;  /* Space allocated forward. */
 515         __u32                   saddr;          /* Sending source                       */
 516         unsigned int            allocation;     /* Allocation mode                      */
 517         int                     sndbuf;         /* Size of send buffer in bytes         */
 518         struct sock             *prev;
 519
 520         /* Not all are volatile, but some are, so we might as well say they all are.
 521          * XXX Make this a flag word -DaveM
 522          */
 523         volatile char           dead,
 524                                 done,
 525                                 urginline,
 526                                 keepopen,
 527                                 linger,
 528                                 destroy,
 529                                 no_check,
 530                                 broadcast,
 531                                 bsdism;
 532         unsigned char           debug;
 533         unsigned char           rcvtstamp;
 534         unsigned char           userlocks;
 535         int                     proc;
 536         unsigned long           lingertime;
 537
 538         int                     hashent;
 539         struct sock             *pair;
 540
 541         /* The backlog queue is special, it is always used with
 542          * the per-socket spinlock held and requires low latency
 543          * access.  Therefore we special case it's implementation.
 544          */
 545         struct {
 546                 struct sk_buff *head;
 547                 struct sk_buff *tail;
 548         } backlog;
 549
 550         rwlock_t                callback_lock;
 551
 552         /* Error queue, rarely used. */
 553         struct sk_buff_head     error_queue;
 554
 555         struct proto            *prot;
 556
 557 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
 558         union {
 559                 struct ipv6_pinfo       af_inet6;
 560         } net_pinfo;
 561 #endif
 562
 563         union {
 564                 struct tcp_opt          af_tcp;
 565 #if defined(CONFIG_INET) || defined (CONFIG_INET_MODULE)
 566                 struct raw_opt          tp_raw4;
 567 #endif
 568 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
 569                 struct raw6_opt         tp_raw;
 570 #endif /* CONFIG_IPV6 */
 571 #if defined(CONFIG_SPX) || defined (CONFIG_SPX_MODULE)
 572                 struct spx_opt          af_spx;
 573 #endif /* CONFIG_SPX */
 574
 575         } tp_pinfo;
 576
 577         int                     err, err_soft;  /* Soft holds errors that don't
 578                                                    cause failure but are the cause
 579                                                    of a persistent failure not just
 580                                                    'timed out' */
 581         unsigned short          ack_backlog;
 582         unsigned short          max_ack_backlog;
 583         __u32                   priority;
 584         unsigned short          type;
 585         unsigned char           localroute;     /* Route locally only */
 586         unsigned char           protocol;
 587         struct ucred            peercred;
 588         int                     rcvlowat;
 589         long                    rcvtimeo;
 590         long                    sndtimeo;
 591
 592 #ifdef CONFIG_FILTER
 593         /* Socket Filtering Instructions */
 594         struct sk_filter        *filter;
 595 #endif /* CONFIG_FILTER */
 596
 597         /* This is where all the private (optional) areas that don't
 598          * overlap will eventually live.
 599          */
 600         union {
 601                 void *destruct_hook;
 602                 struct unix_opt af_unix;
 603 #if defined(CONFIG_INET) || defined (CONFIG_INET_MODULE)
 604                 struct inet_opt af_inet;
 605 #endif
 606 #if defined(CONFIG_ATALK) || defined(CONFIG_ATALK_MODULE)
 607                 struct atalk_sock       af_at;
 608 #endif
 609 #if defined(CONFIG_IPX) || defined(CONFIG_IPX_MODULE)
 610                 struct ipx_opt          af_ipx;
 611 #endif
 612 #if defined (CONFIG_DECNET) || defined(CONFIG_DECNET_MODULE)
 613                 struct dn_scp           dn;
 614 #endif
 615 #if defined (CONFIG_PACKET) || defined(CONFIG_PACKET_MODULE)
 616                 struct packet_opt       *af_packet;
 617 #endif
 618 #if defined(CONFIG_X25) || defined(CONFIG_X25_MODULE)
 619                 x25_cb                  *x25;
 620 #endif
 621 #if defined(CONFIG_AX25) || defined(CONFIG_AX25_MODULE)
 622                 ax25_cb                 *ax25;
 623 #endif
 624 #if defined(CONFIG_NETROM) || defined(CONFIG_NETROM_MODULE)
 625                 nr_cb                   *nr;
 626 #endif
 627 #if defined(CONFIG_ROSE) || defined(CONFIG_ROSE_MODULE)
 628                 rose_cb                 *rose;
 629 #endif
 630 #if defined(CONFIG_PPPOE) || defined(CONFIG_PPPOE_MODULE)
 631                 struct pppox_opt        *pppox;
 632 #endif
 633 #ifdef CONFIG_NETLINK
 634                 struct netlink_opt      *af_netlink;
 635 #endif
 636 #if defined(CONFIG_ECONET) || defined(CONFIG_ECONET_MODULE)
 637                 struct econet_opt       *af_econet;
 638 #endif
 639 #if defined(CONFIG_ATM) || defined(CONFIG_ATM_MODULE)
 640                 struct atm_vcc          *af_atm;
 641 #endif
 642 #if defined(CONFIG_IRDA) || defined(CONFIG_IRDA_MODULE)
 643                 struct irda_sock        *irda;
 644 #endif
 645         } protinfo;
 646
 647
 648         /* This part is used for the timeout functions. */
 649         struct timer_list       timer;          /* This is the sock cleanup timer. */
 650         struct timeval          stamp;
 651
 652         /* Identd and reporting IO signals */
 653         struct socket           *socket;
 654
 655         /* RPC layer private data */
 656         void                    *user_data;
 657
 658         /* Callbacks */
 659         void                    (*state_change)(struct sock *sk);
 660         void                    (*data_ready)(struct sock *sk,int bytes);
 661         void                    (*write_space)(struct sock *sk);
 662         void                    (*error_report)(struct sock *sk);
 663
 664         int                     (*backlog_rcv) (struct sock *sk,
 665                                                 struct sk_buff *skb);
 666         void                    (*destruct)(struct sock *sk);
 667 };
 668
 669 /* The per-socket spinlock must be held here. */
 670 #define sk_add_backlog(__sk, __skb)                     \
 671 do {    if((__sk)->backlog.tail == NULL) {              \
 672                 (__sk)->backlog.head =                  \
 673                      (__sk)->backlog.tail = (__skb);    \
 674         } else {                                        \
 675                 ((__sk)->backlog.tail)->next = (__skb); \
 676                 (__sk)->backlog.tail = (__skb);         \
 677         }                                               \
 678         (__skb)->next = NULL;                           \
 679 } while(0)
 680
 681 /* IP protocol blocks we attach to sockets.
 682  * socket layer -> transport layer interface
 683  * transport -> network interface is defined by struct inet_proto
 684  */
 685 struct proto {
 686         void                    (*close)(struct sock *sk,
 687                                         long timeout);
 688         int                     (*connect)(struct sock *sk,
 689                                         struct sockaddr *uaddr,
 690                                         int addr_len);
 691         int                     (*disconnect)(struct sock *sk, int flags);
 692
 693         struct sock *           (*accept) (struct sock *sk, int flags, int *err);
 694
 695         int                     (*ioctl)(struct sock *sk, int cmd,
 696                                          unsigned long arg);
 697         int                     (*init)(struct sock *sk);
 698         int                     (*destroy)(struct sock *sk);
 699         void                    (*shutdown)(struct sock *sk, int how);
 700         int                     (*setsockopt)(struct sock *sk, int level,
 701                                         int optname, char *optval, int optlen);
 702         int                     (*getsockopt)(struct sock *sk, int level,
 703                                         int optname, char *optval,
 704                                         int *option);
 705         int                     (*sendmsg)(struct sock *sk, struct msghdr *msg,
 706                                            int len);
 707         int                     (*recvmsg)(struct sock *sk, struct msghdr *msg,
 708                                         int len, int noblock, int flags,
 709                                         int *addr_len);
 710         int                     (*bind)(struct sock *sk,
 711                                         struct sockaddr *uaddr, int addr_len);
 712
 713         int                     (*backlog_rcv) (struct sock *sk,
 714                                                 struct sk_buff *skb);
 715
 716         /* Keeping track of sk's, looking them up, and port selection methods. */
 717         void                    (*hash)(struct sock *sk);
 718         void                    (*unhash)(struct sock *sk);
 719         int                     (*get_port)(struct sock *sk, unsigned short snum);
 720
 721         char                    name[32];
 722
 723         struct {
 724                 int inuse;
 725                 u8  __pad[SMP_CACHE_BYTES - sizeof(int)];
 726         } stats[NR_CPUS];
 727 };
 728
 729 /* Called with local bh disabled */
 730 static void __inline__ sock_prot_inc_use(struct proto *prot)
 731 {
 732         prot->stats[smp_processor_id()].inuse++;
 733 }
 734
 735 static void __inline__ sock_prot_dec_use(struct proto *prot)
 736 {
 737         prot->stats[smp_processor_id()].inuse--;
 738 }
 739
 740 /* About 10 seconds */
 741 #define SOCK_DESTROY_TIME (10*HZ)
 742
 743 /* Sockets 0-1023 can't be bound to unless you are superuser */
 744 #define PROT_SOCK       1024
 745
 746 #define SHUTDOWN_MASK   3
 747 #define RCV_SHUTDOWN    1
 748 #define SEND_SHUTDOWN   2
 749
 750 #define SOCK_SNDBUF_LOCK        1
 751 #define SOCK_RCVBUF_LOCK        2
 752 #define SOCK_BINDADDR_LOCK      4
 753 #define SOCK_BINDPORT_LOCK      8
 754
 755
 756 /* Used by processes to "lock" a socket state, so that
 757  * interrupts and bottom half handlers won't change it
 758  * from under us. It essentially blocks any incoming
 759  * packets, so that we won't get any new data or any
 760  * packets that change the state of the socket.
 761  *
 762  * While locked, BH processing will add new packets to
 763  * the backlog queue.  This queue is processed by the
 764  * owner of the socket lock right before it is released.
 765  *
 766  * Since ~2.3.5 it is also exclusive sleep lock serializing
 767  * accesses from user process context.
 768  */
 769 extern void __lock_sock(struct sock *sk);
 770 extern void __release_sock(struct sock *sk);
 771 #define lock_sock(__sk) \
 772 do {    spin_lock_bh(&((__sk)->lock.slock)); \
 773         if ((__sk)->lock.users != 0) \
 774                 __lock_sock(__sk); \
 775         (__sk)->lock.users = 1; \
 776         spin_unlock_bh(&((__sk)->lock.slock)); \
 777 } while(0)
 778
 779 #define release_sock(__sk) \
 780 do {    spin_lock_bh(&((__sk)->lock.slock)); \
 781         if ((__sk)->backlog.tail != NULL) \
 782                 __release_sock(__sk); \
 783         (__sk)->lock.users = 0; \
 784         if (waitqueue_active(&((__sk)->lock.wq))) wake_up(&((__sk)->lock.wq)); \
 785         spin_unlock_bh(&((__sk)->lock.slock)); \
 786 } while(0)
 787
 788 /* BH context may only use the following locking interface. */
 789 #define bh_lock_sock(__sk)      spin_lock(&((__sk)->lock.slock))
 790 #define bh_unlock_sock(__sk)    spin_unlock(&((__sk)->lock.slock))
 791
 792 /*
 793  *      This might not be the most appropriate place for this two
 794  *      but since they are used by a lot of the net related code
 795  *      at least they get declared on a include that is common to all
 796  */
 797
 798 static __inline__ int min(unsigned int a, unsigned int b)
 799 {
 800         if (a > b)
 801                 a = b;
 802         return a;
 803 }
 804
 805 static __inline__ int max(unsigned int a, unsigned int b)
 806 {
 807         if (a < b)
 808                 a = b;
 809         return a;
 810 }
 811
 812 extern struct sock *            sk_alloc(int family, int priority, int zero_it);
 813 extern void                     sk_free(struct sock *sk);
 814
 815 extern struct sk_buff           *sock_wmalloc(struct sock *sk,
 816                                               unsigned long size, int force,
 817                                               int priority);
 818 extern struct sk_buff           *sock_rmalloc(struct sock *sk,
 819                                               unsigned long size, int force,
 820                                               int priority);
 821 extern void                     sock_wfree(struct sk_buff *skb);
 822 extern void                     sock_rfree(struct sk_buff *skb);
 823
 824 extern int                      sock_setsockopt(struct socket *sock, int level,
 825                                                 int op, char *optval,
 826                                                 int optlen);
 827
 828 extern int                      sock_getsockopt(struct socket *sock, int level,
 829                                                 int op, char *optval,
 830                                                 int *optlen);
 831 extern struct sk_buff           *sock_alloc_send_skb(struct sock *sk,
 832                                                      unsigned long size,
 833                                                      unsigned long fallback,
 834                                                      int noblock,
 835                                                      int *errcode);
 836 extern void *sock_kmalloc(struct sock *sk, int size, int priority);
 837 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
 838
 839 extern int copy_and_csum_toiovec(struct iovec *iov, struct sk_buff *skb, int hlen);
 840
 841 /*
 842  * Functions to fill in entries in struct proto_ops when a protocol
 843  * does not implement a particular function.
 844  */
 845 extern int                      sock_no_release(struct socket *);
 846 extern int                      sock_no_bind(struct socket *,
 847                                              struct sockaddr *, int);
 848 extern int                      sock_no_connect(struct socket *,
 849                                                 struct sockaddr *, int, int);
 850 extern int                      sock_no_socketpair(struct socket *,
 851                                                    struct socket *);
 852 extern int                      sock_no_accept(struct socket *,
 853                                                struct socket *, int);
 854 extern int                      sock_no_getname(struct socket *,
 855                                                 struct sockaddr *, int *, int);
 856 extern unsigned int             sock_no_poll(struct file *, struct socket *,
 857                                              struct poll_table_struct *);
 858 extern int                      sock_no_ioctl(struct socket *, unsigned int,
 859                                               unsigned long);
 860 extern int                      sock_no_listen(struct socket *, int);
 861 extern int                      sock_no_shutdown(struct socket *, int);
 862 extern int                      sock_no_getsockopt(struct socket *, int , int,
 863                                                    char *, int *);
 864 extern int                      sock_no_setsockopt(struct socket *, int, int,
 865                                                    char *, int);
 866 extern int                      sock_no_fcntl(struct socket *,
 867                                               unsigned int, unsigned long);
 868 extern int                      sock_no_sendmsg(struct socket *,
 869                                                 struct msghdr *, int,
 870                                                 struct scm_cookie *);
 871 extern int                      sock_no_recvmsg(struct socket *,
 872                                                 struct msghdr *, int, int,
 873                                                 struct scm_cookie *);
 874 extern int                      sock_no_mmap(struct file *file,
 875                                              struct socket *sock,
 876                                              struct vm_area_struct *vma);
 877
 878 /*
 879  *      Default socket callbacks and setup code
 880  */
 881
 882 extern void sock_def_destruct(struct sock *);
 883
 884 /* Initialise core socket variables */
 885 extern void sock_init_data(struct socket *sock, struct sock *sk);
 886
 887 extern void sklist_remove_socket(struct sock **list, struct sock *sk);
 888 extern void sklist_insert_socket(struct sock **list, struct sock *sk);
 889 extern void sklist_destroy_socket(struct sock **list, struct sock *sk);
 890
 891 #ifdef CONFIG_FILTER
 892
 893 /**
 894  *      sk_filter - run a packet through a socket filter
 895  *      @skb: buffer to filter
 896  *      @filter: filter to apply
 897  *
 898  * Run the filter code and then cut skb->data to correct size returned by
 899  * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
 900  * than pkt_len we keep whole skb->data. This is the socket level
 901  * wrapper to sk_run_filter. It returns 0 if the packet should
 902  * be accepted or 1 if the packet should be tossed.
 903  */
 904
 905 static inline int sk_filter(struct sk_buff *skb, struct sk_filter *filter)
 906 {
 907         int pkt_len;
 908
 909         pkt_len = sk_run_filter(skb, filter->insns, filter->len);
 910         if(!pkt_len)
 911                 return 1;       /* Toss Packet */
 912         else
 913                 skb_trim(skb, pkt_len);
 914
 915         return 0;
 916 }
 917
 918 /**
 919  *      sk_filter_release: Release a socket filter
 920  *      @sk: socket
 921  *      @fp: filter to remove
 922  *
 923  *      Remove a filter from a socket and release its resources.
 924  */
 925
 926 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
 927 {
 928         unsigned int size = sk_filter_len(fp);
 929
 930         atomic_sub(size, &sk->omem_alloc);
 931
 932         if (atomic_dec_and_test(&fp->refcnt))
 933                 kfree(fp);
 934 }
 935
 936 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
 937 {
 938         atomic_inc(&fp->refcnt);
 939         atomic_add(sk_filter_len(fp), &sk->omem_alloc);
 940 }
 941
 942 #endif /* CONFIG_FILTER */
 943
 944 /*
 945  * Socket reference counting postulates.
 946  *
 947  * * Each user of socket SHOULD hold a reference count.
 948  * * Each access point to socket (an hash table bucket, reference from a list,
 949  *   running timer, skb in flight MUST hold a reference count.
 950  * * When reference count hits 0, it means it will never increase back.
 951  * * When reference count hits 0, it means that no references from
 952  *   outside exist to this socket and current process on current CPU
 953  *   is last user and may/should destroy this socket.
 954  * * sk_free is called from any context: process, BH, IRQ. When
 955  *   it is called, socket has no references from outside -> sk_free
 956  *   may release descendant resources allocated by the socket, but
 957  *   to the time when it is called, socket is NOT referenced by any
 958  *   hash tables, lists etc.
 959  * * Packets, delivered from outside (from network or from another process)
 960  *   and enqueued on receive/error queues SHOULD NOT grab reference count,
 961  *   when they sit in queue. Otherwise, packets will leak to hole, when
 962  *   socket is looked up by one cpu and unhasing is made by another CPU.
 963  *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
 964  *   (leak to backlog). Packet socket does all the processing inside
 965  *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
 966  *   use separate SMP lock, so that they are prone too.
 967  */
 968
 969 /* Grab socket reference count. This operation is valid only
 970    when sk is ALREADY grabbed f.e. it is found in hash table
 971    or a list and the lookup is made under lock preventing hash table
 972    modifications.
 973  */
 974
 975 static inline void sock_hold(struct sock *sk)
 976 {
 977         atomic_inc(&sk->refcnt);
 978 }
 979
 980 /* Ungrab socket in the context, which assumes that socket refcnt
 981    cannot hit zero, f.e. it is true in context of any socketcall.
 982  */
 983 static inline void __sock_put(struct sock *sk)
 984 {
 985         atomic_dec(&sk->refcnt);
 986 }
 987
 988 /* Ungrab socket and destroy it, if it was the last reference. */
 989 static inline void sock_put(struct sock *sk)
 990 {
 991         if (atomic_dec_and_test(&sk->refcnt))
 992                 sk_free(sk);
 993 }
 994
 995 /* Detach socket from process context.
 996  * Announce socket dead, detach it from wait queue and inode.
 997  * Note that parent inode held reference count on this struct sock,
 998  * we do not release it in this function, because protocol
 999  * probably wants some additional cleanups or even continuing
1000  * to work with this socket (TCP).
1001  */
1002 static inline void sock_orphan(struct sock *sk)
1003 {
1004         write_lock_bh(&sk->callback_lock);
1005         sk->dead = 1;
1006         sk->socket = NULL;
1007         sk->sleep = NULL;
1008         write_unlock_bh(&sk->callback_lock);
1009 }
1010
1011 static inline void sock_graft(struct sock *sk, struct socket *parent)
1012 {
1013         write_lock_bh(&sk->callback_lock);
1014         sk->sleep = &parent->wait;
1015         parent->sk = sk;
1016         sk->socket = parent;
1017         write_unlock_bh(&sk->callback_lock);
1018 }
1019
1020 static inline int sock_i_uid(struct sock *sk)
1021 {
1022         int uid;
1023
1024         read_lock(&sk->callback_lock);
1025         uid = sk->socket ? sk->socket->inode->i_uid : 0;
1026         read_unlock(&sk->callback_lock);
1027         return uid;
1028 }
1029
1030 static inline unsigned long sock_i_ino(struct sock *sk)
1031 {
1032         unsigned long ino;
1033
1034         read_lock(&sk->callback_lock);
1035         ino = sk->socket ? sk->socket->inode->i_ino : 0;
1036         read_unlock(&sk->callback_lock);
1037         return ino;
1038 }
1039
1040 static inline struct dst_entry *
1041 __sk_dst_get(struct sock *sk)
1042 {
1043         return sk->dst_cache;
1044 }
1045
1046 static inline struct dst_entry *
1047 sk_dst_get(struct sock *sk)
1048 {
1049         struct dst_entry *dst;
1050
1051         read_lock(&sk->dst_lock);
1052         dst = sk->dst_cache;
1053         if (dst)
1054                 dst_hold(dst);
1055         read_unlock(&sk->dst_lock);
1056         return dst;
1057 }
1058
1059 static inline void
1060 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1061 {
1062         struct dst_entry *old_dst;
1063
1064         old_dst = sk->dst_cache;
1065         sk->dst_cache = dst;
1066         dst_release(old_dst);
1067 }
1068
1069 static inline void
1070 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1071 {
1072         write_lock(&sk->dst_lock);
1073         __sk_dst_set(sk, dst);
1074         write_unlock(&sk->dst_lock);
1075 }
1076
1077 static inline void
1078 __sk_dst_reset(struct sock *sk)
1079 {
1080         struct dst_entry *old_dst;
1081
1082         old_dst = sk->dst_cache;
1083         sk->dst_cache = NULL;
1084         dst_release(old_dst);
1085 }
1086
1087 static inline void
1088 sk_dst_reset(struct sock *sk)
1089 {
1090         write_lock(&sk->dst_lock);
1091         __sk_dst_reset(sk);
1092         write_unlock(&sk->dst_lock);
1093 }
1094
1095 static inline struct dst_entry *
1096 __sk_dst_check(struct sock *sk, u32 cookie)
1097 {
1098         struct dst_entry *dst = sk->dst_cache;
1099
1100         if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1101                 sk->dst_cache = NULL;
1102                 return NULL;
1103         }
1104
1105         return dst;
1106 }
1107
1108 static inline struct dst_entry *
1109 sk_dst_check(struct sock *sk, u32 cookie)
1110 {
1111         struct dst_entry *dst = sk_dst_get(sk);
1112
1113         if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1114                 sk_dst_reset(sk);
1115                 return NULL;
1116         }
1117
1118         return dst;
1119 }
1120
1121
1122 /*
1123  *      Queue a received datagram if it will fit. Stream and sequenced
1124  *      protocols can't normally use this as they need to fit buffers in
1125  *      and play with them.
1126  *
1127  *      Inlined as it's very short and called for pretty much every
1128  *      packet ever received.
1129  */
1130
1131 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1132 {
1133         sock_hold(sk);
1134         skb->sk = sk;
1135         skb->destructor = sock_wfree;
1136         atomic_add(skb->truesize, &sk->wmem_alloc);
1137 }
1138
1139 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1140 {
1141         skb->sk = sk;
1142         skb->destructor = sock_rfree;
1143         atomic_add(skb->truesize, &sk->rmem_alloc);
1144 }
1145
1146 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1147 {
1148         /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1149            number of warnings when compiling with -W --ANK
1150          */
1151         if (atomic_read(&sk->rmem_alloc) + skb->truesize >= (unsigned)sk->rcvbuf)
1152                 return -ENOMEM;
1153
1154 #ifdef CONFIG_FILTER
1155         if (sk->filter) {
1156                 int err = 0;
1157                 struct sk_filter *filter;
1158
1159                 /* It would be deadlock, if sock_queue_rcv_skb is used
1160                    with socket lock! We assume that users of this
1161                    function are lock free.
1162                  */
1163                 bh_lock_sock(sk);
1164                 if ((filter = sk->filter) != NULL && sk_filter(skb, filter))
1165                         err = -EPERM;
1166                 bh_unlock_sock(sk);
1167                 if (err)
1168                         return err;     /* Toss packet */
1169         }
1170 #endif /* CONFIG_FILTER */
1171
1172         skb_set_owner_r(skb, sk);
1173         skb_queue_tail(&sk->receive_queue, skb);
1174         if (!sk->dead)
1175                 sk->data_ready(sk,skb->len);
1176         return 0;
1177 }
1178
1179 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1180 {
1181         /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1182            number of warnings when compiling with -W --ANK
1183          */
1184         if (atomic_read(&sk->rmem_alloc) + skb->truesize >= (unsigned)sk->rcvbuf)
1185                 return -ENOMEM;
1186         skb_set_owner_r(skb, sk);
1187         skb_queue_tail(&sk->error_queue,skb);
1188         if (!sk->dead)
1189                 sk->data_ready(sk,skb->len);
1190         return 0;
1191 }
1192
1193 /*
1194  *      Recover an error report and clear atomically
1195  */
1196
1197 static inline int sock_error(struct sock *sk)
1198 {
1199         int err=xchg(&sk->err,0);
1200         return -err;
1201 }
1202
1203 static inline unsigned long sock_wspace(struct sock *sk)
1204 {
1205         int amt = 0;
1206
1207         if (!(sk->shutdown & SEND_SHUTDOWN)) {
1208                 amt = sk->sndbuf - atomic_read(&sk->wmem_alloc);
1209                 if (amt < 0)
1210                         amt = 0;
1211         }
1212         return amt;
1213 }
1214
1215 static inline void sk_wake_async(struct sock *sk, int how, int band)
1216 {
1217         if (sk->socket && sk->socket->fasync_list)
1218                 sock_wake_async(sk->socket, how, band);
1219 }
1220
1221 #define SOCK_MIN_SNDBUF 2048
1222 #define SOCK_MIN_RCVBUF 256
1223 /* Must be less or equal SOCK_MIN_SNDBUF */
1224 #define SOCK_MIN_WRITE_SPACE    SOCK_MIN_SNDBUF
1225
1226 /*
1227  *      Default write policy as shown to user space via poll/select/SIGIO
1228  *      Kernel internally doesn't use the MIN_WRITE_SPACE threshold.
1229  */
1230 static inline int sock_writeable(struct sock *sk)
1231 {
1232         return sock_wspace(sk) >= SOCK_MIN_WRITE_SPACE;
1233 }
1234
1235 static inline int gfp_any(void)
1236 {
1237         return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1238 }
1239
1240 static inline long sock_rcvtimeo(struct sock *sk, int noblock)
1241 {
1242         return noblock ? 0 : sk->rcvtimeo;
1243 }
1244
1245 static inline long sock_sndtimeo(struct sock *sk, int noblock)
1246 {
1247         return noblock ? 0 : sk->sndtimeo;
1248 }
1249
1250 static inline int sock_rcvlowat(struct sock *sk, int waitall, int len)
1251 {
1252         return (waitall ? len : min(sk->rcvlowat, len)) ? : 1;
1253 }
1254
1255 /* Alas, with timeout socket operations are not restartable.
1256  * Compare this to poll().
1257  */
1258 static inline int sock_intr_errno(long timeo)
1259 {
1260         return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1261 }
1262
1263 static __inline__ void
1264 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1265 {
1266         if (sk->rcvtstamp)
1267                 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(skb->stamp), &skb->stamp);
1268         else
1269                 sk->stamp = skb->stamp;
1270 }
1271
1272 /*
1273  *      Enable debug/info messages
1274  */
1275
1276 #if 0
1277 #define NETDEBUG(x)     do { } while (0)
1278 #else
1279 #define NETDEBUG(x)     do { x; } while (0)
1280 #endif
1281
1282 /*
1283  * Macros for sleeping on a socket. Use them like this:
1284  *
1285  * SOCK_SLEEP_PRE(sk)
1286  * if (condition)
1287  *      schedule();
1288  * SOCK_SLEEP_POST(sk)
1289  *
1290  */
1291
1292 #define SOCK_SLEEP_PRE(sk)      { struct task_struct *tsk = current; \
1293                                 DECLARE_WAITQUEUE(wait, tsk); \
1294                                 tsk->state = TASK_INTERRUPTIBLE; \
1295                                 add_wait_queue((sk)->sleep, &wait); \
1296                                 release_sock(sk);
1297
1298 #define SOCK_SLEEP_POST(sk)     tsk->state = TASK_RUNNING; \
1299                                 remove_wait_queue((sk)->sleep, &wait); \
1300                                 lock_sock(sk); \
1301                                 }
1302
1303 extern __u32 sysctl_wmem_max;
1304 extern __u32 sysctl_rmem_max;
1305
1306 #endif  /* _SOCK_H */