LXR linux/net/dccp/dccp.h

   1/* SPDX-License-Identifier: GPL-2.0-only */
   2#ifndef _DCCP_H
   3#define _DCCP_H
   4/*
   5 *  net/dccp/dccp.h
   6 *
   7 *  An implementation of the DCCP protocol
   8 *  Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@conectiva.com.br>
   9 *  Copyright (c) 2005-6 Ian McDonald <ian.mcdonald@jandi.co.nz>
  10 */
  11
  12#include <linux/dccp.h>
  13#include <linux/ktime.h>
  14#include <net/snmp.h>
  15#include <net/sock.h>
  16#include <net/tcp.h>
  17#include "ackvec.h"
  18
  19/*
  20 *      DCCP - specific warning and debugging macros.
  21 */
  22#define DCCP_WARN(fmt, ...)                                             \
  23        net_warn_ratelimited("%s: " fmt, __func__, ##__VA_ARGS__)
  24#define DCCP_CRIT(fmt, a...) printk(KERN_CRIT fmt " at %s:%d/%s()\n", ##a, \
  25                                         __FILE__, __LINE__, __func__)
  26#define DCCP_BUG(a...)       do { DCCP_CRIT("BUG: " a); dump_stack(); } while(0)
  27#define DCCP_BUG_ON(cond)    do { if (unlikely((cond) != 0))               \
  28                                     DCCP_BUG("\"%s\" holds (exception!)", \
  29                                              __stringify(cond));          \
  30                             } while (0)
  31
  32#define DCCP_PRINTK(enable, fmt, args...)       do { if (enable)             \
  33                                                        printk(fmt, ##args); \
  34                                                } while(0)
  35#define DCCP_PR_DEBUG(enable, fmt, a...)        DCCP_PRINTK(enable, KERN_DEBUG \
  36                                                  "%s: " fmt, __func__, ##a)
  37
  38#ifdef CONFIG_IP_DCCP_DEBUG
  39extern bool dccp_debug;
  40#define dccp_pr_debug(format, a...)       DCCP_PR_DEBUG(dccp_debug, format, ##a)
  41#define dccp_pr_debug_cat(format, a...)   DCCP_PRINTK(dccp_debug, format, ##a)
  42#define dccp_debug(fmt, a...)             dccp_pr_debug_cat(KERN_DEBUG fmt, ##a)
  43#else
  44#define dccp_pr_debug(format, a...)
  45#define dccp_pr_debug_cat(format, a...)
  46#define dccp_debug(format, a...)
  47#endif
  48
  49extern struct inet_hashinfo dccp_hashinfo;
  50
  51extern struct percpu_counter dccp_orphan_count;
  52
  53void dccp_time_wait(struct sock *sk, int state, int timeo);
  54
  55/*
  56 *  Set safe upper bounds for header and option length. Since Data Offset is 8
  57 *  bits (RFC 4340, sec. 5.1), the total header length can never be more than
  58 *  4 * 255 = 1020 bytes. The largest possible header length is 28 bytes (X=1):
  59 *    - DCCP-Response with ACK Subheader and 4 bytes of Service code      OR
  60 *    - DCCP-Reset    with ACK Subheader and 4 bytes of Reset Code fields
  61 *  Hence a safe upper bound for the maximum option length is 1020-28 = 992
  62 */
  63#define MAX_DCCP_SPECIFIC_HEADER (255 * sizeof(uint32_t))
  64#define DCCP_MAX_PACKET_HDR 28
  65#define DCCP_MAX_OPT_LEN (MAX_DCCP_SPECIFIC_HEADER - DCCP_MAX_PACKET_HDR)
  66#define MAX_DCCP_HEADER (MAX_DCCP_SPECIFIC_HEADER + MAX_HEADER)
  67
  68/* Upper bound for initial feature-negotiation overhead (padded to 32 bits) */
  69#define DCCP_FEATNEG_OVERHEAD    (32 * sizeof(uint32_t))
  70
  71#define DCCP_TIMEWAIT_LEN (60 * HZ) /* how long to wait to destroy TIME-WAIT
  72                                     * state, about 60 seconds */
  73
  74/* RFC 1122, 4.2.3.1 initial RTO value */
  75#define DCCP_TIMEOUT_INIT ((unsigned int)(3 * HZ))
  76
  77/*
  78 * The maximum back-off value for retransmissions. This is needed for
  79 *  - retransmitting client-Requests (sec. 8.1.1),
  80 *  - retransmitting Close/CloseReq when closing (sec. 8.3),
  81 *  - feature-negotiation retransmission (sec. 6.6.3),
  82 *  - Acks in client-PARTOPEN state (sec. 8.1.5).
  83 */
  84#define DCCP_RTO_MAX ((unsigned int)(64 * HZ))
  85
  86/*
  87 * RTT sampling: sanity bounds and fallback RTT value from RFC 4340, section 3.4
  88 */
  89#define DCCP_SANE_RTT_MIN       100
  90#define DCCP_FALLBACK_RTT       (USEC_PER_SEC / 5)
  91#define DCCP_SANE_RTT_MAX       (3 * USEC_PER_SEC)
  92
  93/* sysctl variables for DCCP */
  94extern int  sysctl_dccp_request_retries;
  95extern int  sysctl_dccp_retries1;
  96extern int  sysctl_dccp_retries2;
  97extern int  sysctl_dccp_tx_qlen;
  98extern int  sysctl_dccp_sync_ratelimit;
  99
 100/*
 101 *      48-bit sequence number arithmetic (signed and unsigned)
 102 */
 103#define INT48_MIN         0x800000000000LL              /* 2^47     */
 104#define UINT48_MAX        0xFFFFFFFFFFFFLL              /* 2^48 - 1 */
 105#define COMPLEMENT48(x)  (0x1000000000000LL - (x))      /* 2^48 - x */
 106#define TO_SIGNED48(x)   (((x) < INT48_MIN)? (x) : -COMPLEMENT48( (x)))
 107#define TO_UNSIGNED48(x) (((x) >= 0)?        (x) :  COMPLEMENT48(-(x)))
 108#define ADD48(a, b)      (((a) + (b)) & UINT48_MAX)
 109#define SUB48(a, b)      ADD48((a), COMPLEMENT48(b))
 110
 111static inline void dccp_set_seqno(u64 *seqno, u64 value)
 112{
 113        *seqno = value & UINT48_MAX;
 114}
 115
 116static inline void dccp_inc_seqno(u64 *seqno)
 117{
 118        *seqno = ADD48(*seqno, 1);
 119}
 120
 121/* signed mod-2^48 distance: pos. if seqno1 < seqno2, neg. if seqno1 > seqno2 */
 122static inline s64 dccp_delta_seqno(const u64 seqno1, const u64 seqno2)
 123{
 124        u64 delta = SUB48(seqno2, seqno1);
 125
 126        return TO_SIGNED48(delta);
 127}
 128
 129/* is seq1 < seq2 ? */
 130static inline int before48(const u64 seq1, const u64 seq2)
 131{
 132        return (s64)((seq2 << 16) - (seq1 << 16)) > 0;
 133}
 134
 135/* is seq1 > seq2 ? */
 136#define after48(seq1, seq2)     before48(seq2, seq1)
 137
 138/* is seq2 <= seq1 <= seq3 ? */
 139static inline int between48(const u64 seq1, const u64 seq2, const u64 seq3)
 140{
 141        return (seq3 << 16) - (seq2 << 16) >= (seq1 << 16) - (seq2 << 16);
 142}
 143
 144static inline u64 max48(const u64 seq1, const u64 seq2)
 145{
 146        return after48(seq1, seq2) ? seq1 : seq2;
 147}
 148
 149/**
 150 * dccp_loss_count - Approximate the number of lost data packets in a burst loss
 151 * @s1:  last known sequence number before the loss ('hole')
 152 * @s2:  first sequence number seen after the 'hole'
 153 * @ndp: NDP count on packet with sequence number @s2
 154 */
 155static inline u64 dccp_loss_count(const u64 s1, const u64 s2, const u64 ndp)
 156{
 157        s64 delta = dccp_delta_seqno(s1, s2);
 158
 159        WARN_ON(delta < 0);
 160        delta -= ndp + 1;
 161
 162        return delta > 0 ? delta : 0;
 163}
 164
 165/**
 166 * dccp_loss_free - Evaluate condition for data loss from RFC 4340, 7.7.1
 167 */
 168static inline bool dccp_loss_free(const u64 s1, const u64 s2, const u64 ndp)
 169{
 170        return dccp_loss_count(s1, s2, ndp) == 0;
 171}
 172
 173enum {
 174        DCCP_MIB_NUM = 0,
 175        DCCP_MIB_ACTIVEOPENS,                   /* ActiveOpens */
 176        DCCP_MIB_ESTABRESETS,                   /* EstabResets */
 177        DCCP_MIB_CURRESTAB,                     /* CurrEstab */
 178        DCCP_MIB_OUTSEGS,                       /* OutSegs */
 179        DCCP_MIB_OUTRSTS,
 180        DCCP_MIB_ABORTONTIMEOUT,
 181        DCCP_MIB_TIMEOUTS,
 182        DCCP_MIB_ABORTFAILED,
 183        DCCP_MIB_PASSIVEOPENS,
 184        DCCP_MIB_ATTEMPTFAILS,
 185        DCCP_MIB_OUTDATAGRAMS,
 186        DCCP_MIB_INERRS,
 187        DCCP_MIB_OPTMANDATORYERROR,
 188        DCCP_MIB_INVALIDOPT,
 189        __DCCP_MIB_MAX
 190};
 191
 192#define DCCP_MIB_MAX    __DCCP_MIB_MAX
 193struct dccp_mib {
 194        unsigned long   mibs[DCCP_MIB_MAX];
 195};
 196
 197DECLARE_SNMP_STAT(struct dccp_mib, dccp_statistics);
 198#define DCCP_INC_STATS(field)   SNMP_INC_STATS(dccp_statistics, field)
 199#define __DCCP_INC_STATS(field) __SNMP_INC_STATS(dccp_statistics, field)
 200#define DCCP_DEC_STATS(field)   SNMP_DEC_STATS(dccp_statistics, field)
 201
 202/*
 203 *      Checksumming routines
 204 */
 205static inline unsigned int dccp_csum_coverage(const struct sk_buff *skb)
 206{
 207        const struct dccp_hdr* dh = dccp_hdr(skb);
 208
 209        if (dh->dccph_cscov == 0)
 210                return skb->len;
 211        return (dh->dccph_doff + dh->dccph_cscov - 1) * sizeof(u32);
 212}
 213
 214static inline void dccp_csum_outgoing(struct sk_buff *skb)
 215{
 216        unsigned int cov = dccp_csum_coverage(skb);
 217
 218        if (cov >= skb->len)
 219                dccp_hdr(skb)->dccph_cscov = 0;
 220
 221        skb->csum = skb_checksum(skb, 0, (cov > skb->len)? skb->len : cov, 0);
 222}
 223
 224void dccp_v4_send_check(struct sock *sk, struct sk_buff *skb);
 225
 226int dccp_retransmit_skb(struct sock *sk);
 227
 228void dccp_send_ack(struct sock *sk);
 229void dccp_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
 230                         struct request_sock *rsk);
 231
 232void dccp_send_sync(struct sock *sk, const u64 seq,
 233                    const enum dccp_pkt_type pkt_type);
 234
 235/*
 236 * TX Packet Dequeueing Interface
 237 */
 238void dccp_qpolicy_push(struct sock *sk, struct sk_buff *skb);
 239bool dccp_qpolicy_full(struct sock *sk);
 240void dccp_qpolicy_drop(struct sock *sk, struct sk_buff *skb);
 241struct sk_buff *dccp_qpolicy_top(struct sock *sk);
 242struct sk_buff *dccp_qpolicy_pop(struct sock *sk);
 243bool dccp_qpolicy_param_ok(struct sock *sk, __be32 param);
 244
 245/*
 246 * TX Packet Output and TX Timers
 247 */
 248void dccp_write_xmit(struct sock *sk);
 249void dccp_write_space(struct sock *sk);
 250void dccp_flush_write_queue(struct sock *sk, long *time_budget);
 251
 252void dccp_init_xmit_timers(struct sock *sk);
 253static inline void dccp_clear_xmit_timers(struct sock *sk)
 254{
 255        inet_csk_clear_xmit_timers(sk);
 256}
 257
 258unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu);
 259
 260const char *dccp_packet_name(const int type);
 261
 262void dccp_set_state(struct sock *sk, const int state);
 263void dccp_done(struct sock *sk);
 264
 265int dccp_reqsk_init(struct request_sock *rq, struct dccp_sock const *dp,
 266                    struct sk_buff const *skb);
 267
 268int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb);
 269
 270struct sock *dccp_create_openreq_child(const struct sock *sk,
 271                                       const struct request_sock *req,
 272                                       const struct sk_buff *skb);
 273
 274int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb);
 275
 276struct sock *dccp_v4_request_recv_sock(const struct sock *sk, struct sk_buff *skb,
 277                                       struct request_sock *req,
 278                                       struct dst_entry *dst,
 279                                       struct request_sock *req_unhash,
 280                                       bool *own_req);
 281struct sock *dccp_check_req(struct sock *sk, struct sk_buff *skb,
 282                            struct request_sock *req);
 283
 284int dccp_child_process(struct sock *parent, struct sock *child,
 285                       struct sk_buff *skb);
 286int dccp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
 287                           struct dccp_hdr *dh, unsigned int len);
 288int dccp_rcv_established(struct sock *sk, struct sk_buff *skb,
 289                         const struct dccp_hdr *dh, const unsigned int len);
 290
 291int dccp_init_sock(struct sock *sk, const __u8 ctl_sock_initialized);
 292void dccp_destroy_sock(struct sock *sk);
 293
 294void dccp_close(struct sock *sk, long timeout);
 295struct sk_buff *dccp_make_response(const struct sock *sk, struct dst_entry *dst,
 296                                   struct request_sock *req);
 297
 298int dccp_connect(struct sock *sk);
 299int dccp_disconnect(struct sock *sk, int flags);
 300int dccp_getsockopt(struct sock *sk, int level, int optname,
 301                    char __user *optval, int __user *optlen);
 302int dccp_setsockopt(struct sock *sk, int level, int optname,
 303                    char __user *optval, unsigned int optlen);
 304#ifdef CONFIG_COMPAT
 305int compat_dccp_getsockopt(struct sock *sk, int level, int optname,
 306                           char __user *optval, int __user *optlen);
 307int compat_dccp_setsockopt(struct sock *sk, int level, int optname,
 308                           char __user *optval, unsigned int optlen);
 309#endif
 310int dccp_ioctl(struct sock *sk, int cmd, unsigned long arg);
 311int dccp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size);
 312int dccp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
 313                 int flags, int *addr_len);
 314void dccp_shutdown(struct sock *sk, int how);
 315int inet_dccp_listen(struct socket *sock, int backlog);
 316__poll_t dccp_poll(struct file *file, struct socket *sock,
 317                       poll_table *wait);
 318int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
 319void dccp_req_err(struct sock *sk, u64 seq);
 320
 321struct sk_buff *dccp_ctl_make_reset(struct sock *sk, struct sk_buff *skb);
 322int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code);
 323void dccp_send_close(struct sock *sk, const int active);
 324int dccp_invalid_packet(struct sk_buff *skb);
 325u32 dccp_sample_rtt(struct sock *sk, long delta);
 326
 327static inline bool dccp_bad_service_code(const struct sock *sk,
 328                                        const __be32 service)
 329{
 330        const struct dccp_sock *dp = dccp_sk(sk);
 331
 332        if (dp->dccps_service == service)
 333                return false;
 334        return !dccp_list_has_service(dp->dccps_service_list, service);
 335}
 336
 337/**
 338 * dccp_skb_cb  -  DCCP per-packet control information
 339 * @dccpd_type: one of %dccp_pkt_type (or unknown)
 340 * @dccpd_ccval: CCVal field (5.1), see e.g. RFC 4342, 8.1
 341 * @dccpd_reset_code: one of %dccp_reset_codes
 342 * @dccpd_reset_data: Data1..3 fields (depend on @dccpd_reset_code)
 343 * @dccpd_opt_len: total length of all options (5.8) in the packet
 344 * @dccpd_seq: sequence number
 345 * @dccpd_ack_seq: acknowledgment number subheader field value
 346 *
 347 * This is used for transmission as well as for reception.
 348 */
 349struct dccp_skb_cb {
 350        union {
 351                struct inet_skb_parm    h4;
 352#if IS_ENABLED(CONFIG_IPV6)
 353                struct inet6_skb_parm   h6;
 354#endif
 355        } header;
 356        __u8  dccpd_type:4;
 357        __u8  dccpd_ccval:4;
 358        __u8  dccpd_reset_code,
 359              dccpd_reset_data[3];
 360        __u16 dccpd_opt_len;
 361        __u64 dccpd_seq;
 362        __u64 dccpd_ack_seq;
 363};
 364
 365#define DCCP_SKB_CB(__skb) ((struct dccp_skb_cb *)&((__skb)->cb[0]))
 366
 367/* RFC 4340, sec. 7.7 */
 368static inline int dccp_non_data_packet(const struct sk_buff *skb)
 369{
 370        const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
 371
 372        return type == DCCP_PKT_ACK      ||
 373               type == DCCP_PKT_CLOSE    ||
 374               type == DCCP_PKT_CLOSEREQ ||
 375               type == DCCP_PKT_RESET    ||
 376               type == DCCP_PKT_SYNC     ||
 377               type == DCCP_PKT_SYNCACK;
 378}
 379
 380/* RFC 4340, sec. 7.7 */
 381static inline int dccp_data_packet(const struct sk_buff *skb)
 382{
 383        const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
 384
 385        return type == DCCP_PKT_DATA     ||
 386               type == DCCP_PKT_DATAACK  ||
 387               type == DCCP_PKT_REQUEST  ||
 388               type == DCCP_PKT_RESPONSE;
 389}
 390
 391static inline int dccp_packet_without_ack(const struct sk_buff *skb)
 392{
 393        const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
 394
 395        return type == DCCP_PKT_DATA || type == DCCP_PKT_REQUEST;
 396}
 397
 398#define DCCP_PKT_WITHOUT_ACK_SEQ (UINT48_MAX << 2)
 399
 400static inline void dccp_hdr_set_seq(struct dccp_hdr *dh, const u64 gss)
 401{
 402        struct dccp_hdr_ext *dhx = (struct dccp_hdr_ext *)((void *)dh +
 403                                                           sizeof(*dh));
 404        dh->dccph_seq2 = 0;
 405        dh->dccph_seq = htons((gss >> 32) & 0xfffff);
 406        dhx->dccph_seq_low = htonl(gss & 0xffffffff);
 407}
 408
 409static inline void dccp_hdr_set_ack(struct dccp_hdr_ack_bits *dhack,
 410                                    const u64 gsr)
 411{
 412        dhack->dccph_reserved1 = 0;
 413        dhack->dccph_ack_nr_high = htons(gsr >> 32);
 414        dhack->dccph_ack_nr_low  = htonl(gsr & 0xffffffff);
 415}
 416
 417static inline void dccp_update_gsr(struct sock *sk, u64 seq)
 418{
 419        struct dccp_sock *dp = dccp_sk(sk);
 420
 421        if (after48(seq, dp->dccps_gsr))
 422                dp->dccps_gsr = seq;
 423        /* Sequence validity window depends on remote Sequence Window (7.5.1) */
 424        dp->dccps_swl = SUB48(ADD48(dp->dccps_gsr, 1), dp->dccps_r_seq_win / 4);
 425        /*
 426         * Adjust SWL so that it is not below ISR. In contrast to RFC 4340,
 427         * 7.5.1 we perform this check beyond the initial handshake: W/W' are
 428         * always > 32, so for the first W/W' packets in the lifetime of a
 429         * connection we always have to adjust SWL.
 430         * A second reason why we are doing this is that the window depends on
 431         * the feature-remote value of Sequence Window: nothing stops the peer
 432         * from updating this value while we are busy adjusting SWL for the
 433         * first W packets (we would have to count from scratch again then).
 434         * Therefore it is safer to always make sure that the Sequence Window
 435         * is not artificially extended by a peer who grows SWL downwards by
 436         * continually updating the feature-remote Sequence-Window.
 437         * If sequence numbers wrap it is bad luck. But that will take a while
 438         * (48 bit), and this measure prevents Sequence-number attacks.
 439         */
 440        if (before48(dp->dccps_swl, dp->dccps_isr))
 441                dp->dccps_swl = dp->dccps_isr;
 442        dp->dccps_swh = ADD48(dp->dccps_gsr, (3 * dp->dccps_r_seq_win) / 4);
 443}
 444
 445static inline void dccp_update_gss(struct sock *sk, u64 seq)
 446{
 447        struct dccp_sock *dp = dccp_sk(sk);
 448
 449        dp->dccps_gss = seq;
 450        /* Ack validity window depends on local Sequence Window value (7.5.1) */
 451        dp->dccps_awl = SUB48(ADD48(dp->dccps_gss, 1), dp->dccps_l_seq_win);
 452        /* Adjust AWL so that it is not below ISS - see comment above for SWL */
 453        if (before48(dp->dccps_awl, dp->dccps_iss))
 454                dp->dccps_awl = dp->dccps_iss;
 455        dp->dccps_awh = dp->dccps_gss;
 456}
 457
 458static inline int dccp_ackvec_pending(const struct sock *sk)
 459{
 460        return dccp_sk(sk)->dccps_hc_rx_ackvec != NULL &&
 461               !dccp_ackvec_is_empty(dccp_sk(sk)->dccps_hc_rx_ackvec);
 462}
 463
 464static inline int dccp_ack_pending(const struct sock *sk)
 465{
 466        return dccp_ackvec_pending(sk) || inet_csk_ack_scheduled(sk);
 467}
 468
 469int dccp_feat_signal_nn_change(struct sock *sk, u8 feat, u64 nn_val);
 470int dccp_feat_finalise_settings(struct dccp_sock *dp);
 471int dccp_feat_server_ccid_dependencies(struct dccp_request_sock *dreq);
 472int dccp_feat_insert_opts(struct dccp_sock*, struct dccp_request_sock*,
 473                          struct sk_buff *skb);
 474int dccp_feat_activate_values(struct sock *sk, struct list_head *fn);
 475void dccp_feat_list_purge(struct list_head *fn_list);
 476
 477int dccp_insert_options(struct sock *sk, struct sk_buff *skb);
 478int dccp_insert_options_rsk(struct dccp_request_sock *, struct sk_buff *);
 479u32 dccp_timestamp(void);
 480void dccp_timestamping_init(void);
 481int dccp_insert_option(struct sk_buff *skb, unsigned char option,
 482                       const void *value, unsigned char len);
 483
 484#ifdef CONFIG_SYSCTL
 485int dccp_sysctl_init(void);
 486void dccp_sysctl_exit(void);
 487#else
 488static inline int dccp_sysctl_init(void)
 489{
 490        return 0;
 491}
 492
 493static inline void dccp_sysctl_exit(void)
 494{
 495}
 496#endif
 497
 498#endif /* _DCCP_H */
 499