~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/include/net/tcp.h

Version: ~ [ linux-5.4-rc7 ] ~ [ linux-5.3.10 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.83 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.153 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.200 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.200 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.76 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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 TCP module.
  7  *
  8  * Version:     @(#)tcp.h       1.0.5   05/23/93
  9  *
 10  * Authors:     Ross Biro
 11  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 12  *
 13  *              This program is free software; you can redistribute it and/or
 14  *              modify it under the terms of the GNU General Public License
 15  *              as published by the Free Software Foundation; either version
 16  *              2 of the License, or (at your option) any later version.
 17  */
 18 #ifndef _TCP_H
 19 #define _TCP_H
 20 
 21 #define FASTRETRANS_DEBUG 1
 22 
 23 #include <linux/list.h>
 24 #include <linux/tcp.h>
 25 #include <linux/bug.h>
 26 #include <linux/slab.h>
 27 #include <linux/cache.h>
 28 #include <linux/percpu.h>
 29 #include <linux/skbuff.h>
 30 #include <linux/dmaengine.h>
 31 #include <linux/crypto.h>
 32 #include <linux/cryptohash.h>
 33 #include <linux/kref.h>
 34 
 35 #include <net/inet_connection_sock.h>
 36 #include <net/inet_timewait_sock.h>
 37 #include <net/inet_hashtables.h>
 38 #include <net/checksum.h>
 39 #include <net/request_sock.h>
 40 #include <net/sock.h>
 41 #include <net/snmp.h>
 42 #include <net/ip.h>
 43 #include <net/tcp_states.h>
 44 #include <net/inet_ecn.h>
 45 #include <net/dst.h>
 46 
 47 #include <linux/seq_file.h>
 48 #include <linux/memcontrol.h>
 49 
 50 extern struct inet_hashinfo tcp_hashinfo;
 51 
 52 extern struct percpu_counter tcp_orphan_count;
 53 extern void tcp_time_wait(struct sock *sk, int state, int timeo);
 54 
 55 #define MAX_TCP_HEADER  (128 + MAX_HEADER)
 56 #define MAX_TCP_OPTION_SPACE 40
 57 
 58 /* 
 59  * Never offer a window over 32767 without using window scaling. Some
 60  * poor stacks do signed 16bit maths! 
 61  */
 62 #define MAX_TCP_WINDOW          32767U
 63 
 64 /* Offer an initial receive window of 10 mss. */
 65 #define TCP_DEFAULT_INIT_RCVWND 10
 66 
 67 /* Minimal accepted MSS. It is (60+60+8) - (20+20). */
 68 #define TCP_MIN_MSS             88U
 69 
 70 /* The least MTU to use for probing */
 71 #define TCP_BASE_MSS            512
 72 
 73 /* After receiving this amount of duplicate ACKs fast retransmit starts. */
 74 #define TCP_FASTRETRANS_THRESH 3
 75 
 76 /* Maximal reordering. */
 77 #define TCP_MAX_REORDERING      127
 78 
 79 /* Maximal number of ACKs sent quickly to accelerate slow-start. */
 80 #define TCP_MAX_QUICKACKS       16U
 81 
 82 /* urg_data states */
 83 #define TCP_URG_VALID   0x0100
 84 #define TCP_URG_NOTYET  0x0200
 85 #define TCP_URG_READ    0x0400
 86 
 87 #define TCP_RETR1       3       /*
 88                                  * This is how many retries it does before it
 89                                  * tries to figure out if the gateway is
 90                                  * down. Minimal RFC value is 3; it corresponds
 91                                  * to ~3sec-8min depending on RTO.
 92                                  */
 93 
 94 #define TCP_RETR2       15      /*
 95                                  * This should take at least
 96                                  * 90 minutes to time out.
 97                                  * RFC1122 says that the limit is 100 sec.
 98                                  * 15 is ~13-30min depending on RTO.
 99                                  */
100 
101 #define TCP_SYN_RETRIES  6      /* This is how many retries are done
102                                  * when active opening a connection.
103                                  * RFC1122 says the minimum retry MUST
104                                  * be at least 180secs.  Nevertheless
105                                  * this value is corresponding to
106                                  * 63secs of retransmission with the
107                                  * current initial RTO.
108                                  */
109 
110 #define TCP_SYNACK_RETRIES 5    /* This is how may retries are done
111                                  * when passive opening a connection.
112                                  * This is corresponding to 31secs of
113                                  * retransmission with the current
114                                  * initial RTO.
115                                  */
116 
117 #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
118                                   * state, about 60 seconds     */
119 #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
120                                  /* BSD style FIN_WAIT2 deadlock breaker.
121                                   * It used to be 3min, new value is 60sec,
122                                   * to combine FIN-WAIT-2 timeout with
123                                   * TIME-WAIT timer.
124                                   */
125 
126 #define TCP_DELACK_MAX  ((unsigned)(HZ/5))      /* maximal time to delay before sending an ACK */
127 #if HZ >= 100
128 #define TCP_DELACK_MIN  ((unsigned)(HZ/25))     /* minimal time to delay before sending an ACK */
129 #define TCP_ATO_MIN     ((unsigned)(HZ/25))
130 #else
131 #define TCP_DELACK_MIN  4U
132 #define TCP_ATO_MIN     4U
133 #endif
134 #define TCP_RTO_MAX     ((unsigned)(120*HZ))
135 #define TCP_RTO_MIN     ((unsigned)(HZ/5))
136 #define TCP_TIMEOUT_INIT ((unsigned)(1*HZ))     /* RFC6298 2.1 initial RTO value        */
137 #define TCP_TIMEOUT_FALLBACK ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value, now
138                                                  * used as a fallback RTO for the
139                                                  * initial data transmission if no
140                                                  * valid RTT sample has been acquired,
141                                                  * most likely due to retrans in 3WHS.
142                                                  */
143 
144 #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
145                                                          * for local resources.
146                                                          */
147 
148 #define TCP_KEEPALIVE_TIME      (120*60*HZ)     /* two hours */
149 #define TCP_KEEPALIVE_PROBES    9               /* Max of 9 keepalive probes    */
150 #define TCP_KEEPALIVE_INTVL     (75*HZ)
151 
152 #define MAX_TCP_KEEPIDLE        32767
153 #define MAX_TCP_KEEPINTVL       32767
154 #define MAX_TCP_KEEPCNT         127
155 #define MAX_TCP_SYNCNT          127
156 
157 #define TCP_SYNQ_INTERVAL       (HZ/5)  /* Period of SYNACK timer */
158 
159 #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
160 #define TCP_PAWS_MSL    60              /* Per-host timestamps are invalidated
161                                          * after this time. It should be equal
162                                          * (or greater than) TCP_TIMEWAIT_LEN
163                                          * to provide reliability equal to one
164                                          * provided by timewait state.
165                                          */
166 #define TCP_PAWS_WINDOW 1               /* Replay window for per-host
167                                          * timestamps. It must be less than
168                                          * minimal timewait lifetime.
169                                          */
170 /*
171  *      TCP option
172  */
173  
174 #define TCPOPT_NOP              1       /* Padding */
175 #define TCPOPT_EOL              0       /* End of options */
176 #define TCPOPT_MSS              2       /* Segment size negotiating */
177 #define TCPOPT_WINDOW           3       /* Window scaling */
178 #define TCPOPT_SACK_PERM        4       /* SACK Permitted */
179 #define TCPOPT_SACK             5       /* SACK Block */
180 #define TCPOPT_TIMESTAMP        8       /* Better RTT estimations/PAWS */
181 #define TCPOPT_MD5SIG           19      /* MD5 Signature (RFC2385) */
182 #define TCPOPT_EXP              254     /* Experimental */
183 /* Magic number to be after the option value for sharing TCP
184  * experimental options. See draft-ietf-tcpm-experimental-options-00.txt
185  */
186 #define TCPOPT_FASTOPEN_MAGIC   0xF989
187 
188 /*
189  *     TCP option lengths
190  */
191 
192 #define TCPOLEN_MSS            4
193 #define TCPOLEN_WINDOW         3
194 #define TCPOLEN_SACK_PERM      2
195 #define TCPOLEN_TIMESTAMP      10
196 #define TCPOLEN_MD5SIG         18
197 #define TCPOLEN_EXP_FASTOPEN_BASE  4
198 #define TCPOLEN_COOKIE_BASE    2        /* Cookie-less header extension */
199 #define TCPOLEN_COOKIE_PAIR    3        /* Cookie pair header extension */
200 #define TCPOLEN_COOKIE_MIN     (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MIN)
201 #define TCPOLEN_COOKIE_MAX     (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MAX)
202 
203 /* But this is what stacks really send out. */
204 #define TCPOLEN_TSTAMP_ALIGNED          12
205 #define TCPOLEN_WSCALE_ALIGNED          4
206 #define TCPOLEN_SACKPERM_ALIGNED        4
207 #define TCPOLEN_SACK_BASE               2
208 #define TCPOLEN_SACK_BASE_ALIGNED       4
209 #define TCPOLEN_SACK_PERBLOCK           8
210 #define TCPOLEN_MD5SIG_ALIGNED          20
211 #define TCPOLEN_MSS_ALIGNED             4
212 
213 /* Flags in tp->nonagle */
214 #define TCP_NAGLE_OFF           1       /* Nagle's algo is disabled */
215 #define TCP_NAGLE_CORK          2       /* Socket is corked         */
216 #define TCP_NAGLE_PUSH          4       /* Cork is overridden for already queued data */
217 
218 /* TCP thin-stream limits */
219 #define TCP_THIN_LINEAR_RETRIES 6       /* After 6 linear retries, do exp. backoff */
220 
221 /* TCP initial congestion window as per draft-hkchu-tcpm-initcwnd-01 */
222 #define TCP_INIT_CWND           10
223 
224 /* Bit Flags for sysctl_tcp_fastopen */
225 #define TFO_CLIENT_ENABLE       1
226 #define TFO_SERVER_ENABLE       2
227 #define TFO_CLIENT_NO_COOKIE    4       /* Data in SYN w/o cookie option */
228 
229 /* Process SYN data but skip cookie validation */
230 #define TFO_SERVER_COOKIE_NOT_CHKED     0x100
231 /* Accept SYN data w/o any cookie option */
232 #define TFO_SERVER_COOKIE_NOT_REQD      0x200
233 
234 /* Force enable TFO on all listeners, i.e., not requiring the
235  * TCP_FASTOPEN socket option. SOCKOPT1/2 determine how to set max_qlen.
236  */
237 #define TFO_SERVER_WO_SOCKOPT1  0x400
238 #define TFO_SERVER_WO_SOCKOPT2  0x800
239 /* Always create TFO child sockets on a TFO listener even when
240  * cookie/data not present. (For testing purpose!)
241  */
242 #define TFO_SERVER_ALWAYS       0x1000
243 
244 extern struct inet_timewait_death_row tcp_death_row;
245 
246 /* sysctl variables for tcp */
247 extern int sysctl_tcp_timestamps;
248 extern int sysctl_tcp_window_scaling;
249 extern int sysctl_tcp_sack;
250 extern int sysctl_tcp_fin_timeout;
251 extern int sysctl_tcp_keepalive_time;
252 extern int sysctl_tcp_keepalive_probes;
253 extern int sysctl_tcp_keepalive_intvl;
254 extern int sysctl_tcp_syn_retries;
255 extern int sysctl_tcp_synack_retries;
256 extern int sysctl_tcp_retries1;
257 extern int sysctl_tcp_retries2;
258 extern int sysctl_tcp_orphan_retries;
259 extern int sysctl_tcp_syncookies;
260 extern int sysctl_tcp_fastopen;
261 extern int sysctl_tcp_retrans_collapse;
262 extern int sysctl_tcp_stdurg;
263 extern int sysctl_tcp_rfc1337;
264 extern int sysctl_tcp_abort_on_overflow;
265 extern int sysctl_tcp_max_orphans;
266 extern int sysctl_tcp_fack;
267 extern int sysctl_tcp_reordering;
268 extern int sysctl_tcp_dsack;
269 extern int sysctl_tcp_wmem[3];
270 extern int sysctl_tcp_rmem[3];
271 extern int sysctl_tcp_app_win;
272 extern int sysctl_tcp_adv_win_scale;
273 extern int sysctl_tcp_tw_reuse;
274 extern int sysctl_tcp_frto;
275 extern int sysctl_tcp_low_latency;
276 extern int sysctl_tcp_dma_copybreak;
277 extern int sysctl_tcp_nometrics_save;
278 extern int sysctl_tcp_moderate_rcvbuf;
279 extern int sysctl_tcp_tso_win_divisor;
280 extern int sysctl_tcp_mtu_probing;
281 extern int sysctl_tcp_base_mss;
282 extern int sysctl_tcp_workaround_signed_windows;
283 extern int sysctl_tcp_slow_start_after_idle;
284 extern int sysctl_tcp_max_ssthresh;
285 extern int sysctl_tcp_thin_linear_timeouts;
286 extern int sysctl_tcp_thin_dupack;
287 extern int sysctl_tcp_early_retrans;
288 extern int sysctl_tcp_limit_output_bytes;
289 extern int sysctl_tcp_challenge_ack_limit;
290 extern int sysctl_tcp_min_tso_segs;
291 
292 extern atomic_long_t tcp_memory_allocated;
293 extern struct percpu_counter tcp_sockets_allocated;
294 extern int tcp_memory_pressure;
295 
296 /*
297  * The next routines deal with comparing 32 bit unsigned ints
298  * and worry about wraparound (automatic with unsigned arithmetic).
299  */
300 
301 static inline bool before(__u32 seq1, __u32 seq2)
302 {
303         return (__s32)(seq1-seq2) < 0;
304 }
305 #define after(seq2, seq1)       before(seq1, seq2)
306 
307 /* is s2<=s1<=s3 ? */
308 static inline bool between(__u32 seq1, __u32 seq2, __u32 seq3)
309 {
310         return seq3 - seq2 >= seq1 - seq2;
311 }
312 
313 static inline bool tcp_out_of_memory(struct sock *sk)
314 {
315         if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
316             sk_memory_allocated(sk) > sk_prot_mem_limits(sk, 2))
317                 return true;
318         return false;
319 }
320 
321 static inline bool tcp_too_many_orphans(struct sock *sk, int shift)
322 {
323         struct percpu_counter *ocp = sk->sk_prot->orphan_count;
324         int orphans = percpu_counter_read_positive(ocp);
325 
326         if (orphans << shift > sysctl_tcp_max_orphans) {
327                 orphans = percpu_counter_sum_positive(ocp);
328                 if (orphans << shift > sysctl_tcp_max_orphans)
329                         return true;
330         }
331         return false;
332 }
333 
334 extern bool tcp_check_oom(struct sock *sk, int shift);
335 
336 /* syncookies: remember time of last synqueue overflow */
337 static inline void tcp_synq_overflow(struct sock *sk)
338 {
339         tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies;
340 }
341 
342 /* syncookies: no recent synqueue overflow on this listening socket? */
343 static inline bool tcp_synq_no_recent_overflow(const struct sock *sk)
344 {
345         unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp;
346         return time_after(jiffies, last_overflow + TCP_TIMEOUT_FALLBACK);
347 }
348 
349 extern struct proto tcp_prot;
350 
351 #define TCP_INC_STATS(net, field)       SNMP_INC_STATS((net)->mib.tcp_statistics, field)
352 #define TCP_INC_STATS_BH(net, field)    SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
353 #define TCP_DEC_STATS(net, field)       SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
354 #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val)
355 #define TCP_ADD_STATS(net, field, val)  SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val)
356 
357 extern void tcp_init_mem(struct net *net);
358 
359 extern void tcp_tasklet_init(void);
360 
361 extern void tcp_v4_err(struct sk_buff *skb, u32);
362 
363 extern void tcp_shutdown (struct sock *sk, int how);
364 
365 extern void tcp_v4_early_demux(struct sk_buff *skb);
366 extern int tcp_v4_rcv(struct sk_buff *skb);
367 
368 extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
369 extern int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
370                        size_t size);
371 extern int tcp_sendpage(struct sock *sk, struct page *page, int offset,
372                         size_t size, int flags);
373 extern void tcp_release_cb(struct sock *sk);
374 extern void tcp_wfree(struct sk_buff *skb);
375 extern void tcp_write_timer_handler(struct sock *sk);
376 extern void tcp_delack_timer_handler(struct sock *sk);
377 extern int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg);
378 extern int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
379                                  const struct tcphdr *th, unsigned int len);
380 extern int tcp_rcv_established(struct sock *sk, struct sk_buff *skb,
381                                const struct tcphdr *th, unsigned int len);
382 extern void tcp_rcv_space_adjust(struct sock *sk);
383 extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
384 extern int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp);
385 extern void tcp_twsk_destructor(struct sock *sk);
386 extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
387                                struct pipe_inode_info *pipe, size_t len,
388                                unsigned int flags);
389 
390 static inline void tcp_dec_quickack_mode(struct sock *sk,
391                                          const unsigned int pkts)
392 {
393         struct inet_connection_sock *icsk = inet_csk(sk);
394 
395         if (icsk->icsk_ack.quick) {
396                 if (pkts >= icsk->icsk_ack.quick) {
397                         icsk->icsk_ack.quick = 0;
398                         /* Leaving quickack mode we deflate ATO. */
399                         icsk->icsk_ack.ato   = TCP_ATO_MIN;
400                 } else
401                         icsk->icsk_ack.quick -= pkts;
402         }
403 }
404 
405 #define TCP_ECN_OK              1
406 #define TCP_ECN_QUEUE_CWR       2
407 #define TCP_ECN_DEMAND_CWR      4
408 #define TCP_ECN_SEEN            8
409 
410 enum tcp_tw_status {
411         TCP_TW_SUCCESS = 0,
412         TCP_TW_RST = 1,
413         TCP_TW_ACK = 2,
414         TCP_TW_SYN = 3
415 };
416 
417 
418 extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
419                                                      struct sk_buff *skb,
420                                                      const struct tcphdr *th);
421 extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
422                                    struct request_sock *req,
423                                    struct request_sock **prev,
424                                    bool fastopen);
425 extern int tcp_child_process(struct sock *parent, struct sock *child,
426                              struct sk_buff *skb);
427 extern void tcp_enter_loss(struct sock *sk, int how);
428 extern void tcp_clear_retrans(struct tcp_sock *tp);
429 extern void tcp_update_metrics(struct sock *sk);
430 extern void tcp_init_metrics(struct sock *sk);
431 extern void tcp_metrics_init(void);
432 extern bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst, bool paws_check);
433 extern bool tcp_remember_stamp(struct sock *sk);
434 extern bool tcp_tw_remember_stamp(struct inet_timewait_sock *tw);
435 extern void tcp_fetch_timewait_stamp(struct sock *sk, struct dst_entry *dst);
436 extern void tcp_disable_fack(struct tcp_sock *tp);
437 extern void tcp_close(struct sock *sk, long timeout);
438 extern void tcp_init_sock(struct sock *sk);
439 extern unsigned int tcp_poll(struct file * file, struct socket *sock,
440                              struct poll_table_struct *wait);
441 extern int tcp_getsockopt(struct sock *sk, int level, int optname,
442                           char __user *optval, int __user *optlen);
443 extern int tcp_setsockopt(struct sock *sk, int level, int optname,
444                           char __user *optval, unsigned int optlen);
445 extern int compat_tcp_getsockopt(struct sock *sk, int level, int optname,
446                                  char __user *optval, int __user *optlen);
447 extern int compat_tcp_setsockopt(struct sock *sk, int level, int optname,
448                                  char __user *optval, unsigned int optlen);
449 extern void tcp_set_keepalive(struct sock *sk, int val);
450 extern void tcp_syn_ack_timeout(struct sock *sk, struct request_sock *req);
451 extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
452                        size_t len, int nonblock, int flags, int *addr_len);
453 extern void tcp_parse_options(const struct sk_buff *skb,
454                               struct tcp_options_received *opt_rx,
455                               int estab, struct tcp_fastopen_cookie *foc);
456 extern const u8 *tcp_parse_md5sig_option(const struct tcphdr *th);
457 
458 /*
459  *      TCP v4 functions exported for the inet6 API
460  */
461 
462 extern void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb);
463 void tcp_v4_mtu_reduced(struct sock *sk);
464 extern int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb);
465 extern struct sock * tcp_create_openreq_child(struct sock *sk,
466                                               struct request_sock *req,
467                                               struct sk_buff *skb);
468 extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
469                                           struct request_sock *req,
470                                           struct dst_entry *dst);
471 extern int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb);
472 extern int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr,
473                           int addr_len);
474 extern int tcp_connect(struct sock *sk);
475 extern struct sk_buff * tcp_make_synack(struct sock *sk, struct dst_entry *dst,
476                                         struct request_sock *req,
477                                         struct tcp_fastopen_cookie *foc);
478 extern int tcp_disconnect(struct sock *sk, int flags);
479 
480 void tcp_connect_init(struct sock *sk);
481 void tcp_finish_connect(struct sock *sk, struct sk_buff *skb);
482 int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size);
483 void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb);
484 
485 /* From syncookies.c */
486 extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
487 extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb, 
488                                     struct ip_options *opt);
489 #ifdef CONFIG_SYN_COOKIES
490 extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb, 
491                                      __u16 *mss);
492 #else
493 static inline __u32 cookie_v4_init_sequence(struct sock *sk,
494                                             struct sk_buff *skb,
495                                             __u16 *mss)
496 {
497         return 0;
498 }
499 #endif
500 
501 extern __u32 cookie_init_timestamp(struct request_sock *req);
502 extern bool cookie_check_timestamp(struct tcp_options_received *opt,
503                                 struct net *net, bool *ecn_ok);
504 
505 /* From net/ipv6/syncookies.c */
506 extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
507 #ifdef CONFIG_SYN_COOKIES
508 extern __u32 cookie_v6_init_sequence(struct sock *sk, const struct sk_buff *skb,
509                                      __u16 *mss);
510 #else
511 static inline __u32 cookie_v6_init_sequence(struct sock *sk,
512                                             struct sk_buff *skb,
513                                             __u16 *mss)
514 {
515         return 0;
516 }
517 #endif
518 /* tcp_output.c */
519 
520 extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
521                                       int nonagle);
522 extern bool tcp_may_send_now(struct sock *sk);
523 extern int __tcp_retransmit_skb(struct sock *, struct sk_buff *);
524 extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
525 extern void tcp_retransmit_timer(struct sock *sk);
526 extern void tcp_xmit_retransmit_queue(struct sock *);
527 extern void tcp_simple_retransmit(struct sock *);
528 extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
529 extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
530 
531 extern void tcp_send_probe0(struct sock *);
532 extern void tcp_send_partial(struct sock *);
533 extern int tcp_write_wakeup(struct sock *);
534 extern void tcp_send_fin(struct sock *sk);
535 extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
536 extern int tcp_send_synack(struct sock *);
537 extern bool tcp_syn_flood_action(struct sock *sk,
538                                  const struct sk_buff *skb,
539                                  const char *proto);
540 extern void tcp_push_one(struct sock *, unsigned int mss_now);
541 extern void tcp_send_ack(struct sock *sk);
542 extern void tcp_send_delayed_ack(struct sock *sk);
543 extern void tcp_send_loss_probe(struct sock *sk);
544 extern bool tcp_schedule_loss_probe(struct sock *sk);
545 
546 /* tcp_input.c */
547 extern void tcp_cwnd_application_limited(struct sock *sk);
548 extern void tcp_resume_early_retransmit(struct sock *sk);
549 extern void tcp_rearm_rto(struct sock *sk);
550 extern void tcp_reset(struct sock *sk);
551 
552 /* tcp_timer.c */
553 extern void tcp_init_xmit_timers(struct sock *);
554 static inline void tcp_clear_xmit_timers(struct sock *sk)
555 {
556         inet_csk_clear_xmit_timers(sk);
557 }
558 
559 extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
560 extern unsigned int tcp_current_mss(struct sock *sk);
561 
562 /* Bound MSS / TSO packet size with the half of the window */
563 static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize)
564 {
565         int cutoff;
566 
567         /* When peer uses tiny windows, there is no use in packetizing
568          * to sub-MSS pieces for the sake of SWS or making sure there
569          * are enough packets in the pipe for fast recovery.
570          *
571          * On the other hand, for extremely large MSS devices, handling
572          * smaller than MSS windows in this way does make sense.
573          */
574         if (tp->max_window >= 512)
575                 cutoff = (tp->max_window >> 1);
576         else
577                 cutoff = tp->max_window;
578 
579         if (cutoff && pktsize > cutoff)
580                 return max_t(int, cutoff, 68U - tp->tcp_header_len);
581         else
582                 return pktsize;
583 }
584 
585 /* tcp.c */
586 extern void tcp_get_info(const struct sock *, struct tcp_info *);
587 
588 /* Read 'sendfile()'-style from a TCP socket */
589 typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
590                                 unsigned int, size_t);
591 extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
592                          sk_read_actor_t recv_actor);
593 
594 extern void tcp_initialize_rcv_mss(struct sock *sk);
595 
596 extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
597 extern int tcp_mss_to_mtu(struct sock *sk, int mss);
598 extern void tcp_mtup_init(struct sock *sk);
599 extern void tcp_valid_rtt_meas(struct sock *sk, u32 seq_rtt);
600 extern void tcp_init_buffer_space(struct sock *sk);
601 
602 static inline void tcp_bound_rto(const struct sock *sk)
603 {
604         if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX)
605                 inet_csk(sk)->icsk_rto = TCP_RTO_MAX;
606 }
607 
608 static inline u32 __tcp_set_rto(const struct tcp_sock *tp)
609 {
610         return (tp->srtt >> 3) + tp->rttvar;
611 }
612 
613 extern void tcp_set_rto(struct sock *sk);
614 
615 static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
616 {
617         tp->pred_flags = htonl((tp->tcp_header_len << 26) |
618                                ntohl(TCP_FLAG_ACK) |
619                                snd_wnd);
620 }
621 
622 static inline void tcp_fast_path_on(struct tcp_sock *tp)
623 {
624         __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
625 }
626 
627 static inline void tcp_fast_path_check(struct sock *sk)
628 {
629         struct tcp_sock *tp = tcp_sk(sk);
630 
631         if (skb_queue_empty(&tp->out_of_order_queue) &&
632             tp->rcv_wnd &&
633             atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
634             !tp->urg_data)
635                 tcp_fast_path_on(tp);
636 }
637 
638 /* Compute the actual rto_min value */
639 static inline u32 tcp_rto_min(struct sock *sk)
640 {
641         const struct dst_entry *dst = __sk_dst_get(sk);
642         u32 rto_min = TCP_RTO_MIN;
643 
644         if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
645                 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
646         return rto_min;
647 }
648 
649 /* Compute the actual receive window we are currently advertising.
650  * Rcv_nxt can be after the window if our peer push more data
651  * than the offered window.
652  */
653 static inline u32 tcp_receive_window(const struct tcp_sock *tp)
654 {
655         s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
656 
657         if (win < 0)
658                 win = 0;
659         return (u32) win;
660 }
661 
662 /* Choose a new window, without checks for shrinking, and without
663  * scaling applied to the result.  The caller does these things
664  * if necessary.  This is a "raw" window selection.
665  */
666 extern u32 __tcp_select_window(struct sock *sk);
667 
668 void tcp_send_window_probe(struct sock *sk);
669 
670 /* TCP timestamps are only 32-bits, this causes a slight
671  * complication on 64-bit systems since we store a snapshot
672  * of jiffies in the buffer control blocks below.  We decided
673  * to use only the low 32-bits of jiffies and hide the ugly
674  * casts with the following macro.
675  */
676 #define tcp_time_stamp          ((__u32)(jiffies))
677 
678 #define tcp_flag_byte(th) (((u_int8_t *)th)[13])
679 
680 #define TCPHDR_FIN 0x01
681 #define TCPHDR_SYN 0x02
682 #define TCPHDR_RST 0x04
683 #define TCPHDR_PSH 0x08
684 #define TCPHDR_ACK 0x10
685 #define TCPHDR_URG 0x20
686 #define TCPHDR_ECE 0x40
687 #define TCPHDR_CWR 0x80
688 
689 /* This is what the send packet queuing engine uses to pass
690  * TCP per-packet control information to the transmission code.
691  * We also store the host-order sequence numbers in here too.
692  * This is 44 bytes if IPV6 is enabled.
693  * If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately.
694  */
695 struct tcp_skb_cb {
696         union {
697                 struct inet_skb_parm    h4;
698 #if IS_ENABLED(CONFIG_IPV6)
699                 struct inet6_skb_parm   h6;
700 #endif
701         } header;       /* For incoming frames          */
702         __u32           seq;            /* Starting sequence number     */
703         __u32           end_seq;        /* SEQ + FIN + SYN + datalen    */
704         __u32           when;           /* used to compute rtt's        */
705         __u8            tcp_flags;      /* TCP header flags. (tcp[13])  */
706 
707         __u8            sacked;         /* State flags for SACK/FACK.   */
708 #define TCPCB_SACKED_ACKED      0x01    /* SKB ACK'd by a SACK block    */
709 #define TCPCB_SACKED_RETRANS    0x02    /* SKB retransmitted            */
710 #define TCPCB_LOST              0x04    /* SKB is lost                  */
711 #define TCPCB_TAGBITS           0x07    /* All tag bits                 */
712 #define TCPCB_EVER_RETRANS      0x80    /* Ever retransmitted frame     */
713 #define TCPCB_RETRANS           (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
714 
715         __u8            ip_dsfield;     /* IPv4 tos or IPv6 dsfield     */
716         /* 1 byte hole */
717         __u32           ack_seq;        /* Sequence number ACK'd        */
718 };
719 
720 #define TCP_SKB_CB(__skb)       ((struct tcp_skb_cb *)&((__skb)->cb[0]))
721 
722 /* RFC3168 : 6.1.1 SYN packets must not have ECT/ECN bits set
723  *
724  * If we receive a SYN packet with these bits set, it means a network is
725  * playing bad games with TOS bits. In order to avoid possible false congestion
726  * notifications, we disable TCP ECN negociation.
727  */
728 static inline void
729 TCP_ECN_create_request(struct request_sock *req, const struct sk_buff *skb,
730                 struct net *net)
731 {
732         const struct tcphdr *th = tcp_hdr(skb);
733 
734         if (net->ipv4.sysctl_tcp_ecn && th->ece && th->cwr &&
735             INET_ECN_is_not_ect(TCP_SKB_CB(skb)->ip_dsfield))
736                 inet_rsk(req)->ecn_ok = 1;
737 }
738 
739 /* Due to TSO, an SKB can be composed of multiple actual
740  * packets.  To keep these tracked properly, we use this.
741  */
742 static inline int tcp_skb_pcount(const struct sk_buff *skb)
743 {
744         return skb_shinfo(skb)->gso_segs;
745 }
746 
747 /* This is valid iff tcp_skb_pcount() > 1. */
748 static inline int tcp_skb_mss(const struct sk_buff *skb)
749 {
750         return skb_shinfo(skb)->gso_size;
751 }
752 
753 /* Events passed to congestion control interface */
754 enum tcp_ca_event {
755         CA_EVENT_TX_START,      /* first transmit when no packets in flight */
756         CA_EVENT_CWND_RESTART,  /* congestion window restart */
757         CA_EVENT_COMPLETE_CWR,  /* end of congestion recovery */
758         CA_EVENT_LOSS,          /* loss timeout */
759         CA_EVENT_FAST_ACK,      /* in sequence ack */
760         CA_EVENT_SLOW_ACK,      /* other ack */
761 };
762 
763 /*
764  * Interface for adding new TCP congestion control handlers
765  */
766 #define TCP_CA_NAME_MAX 16
767 #define TCP_CA_MAX      128
768 #define TCP_CA_BUF_MAX  (TCP_CA_NAME_MAX*TCP_CA_MAX)
769 
770 #define TCP_CONG_NON_RESTRICTED 0x1
771 #define TCP_CONG_RTT_STAMP      0x2
772 
773 struct tcp_congestion_ops {
774         struct list_head        list;
775         unsigned long flags;
776 
777         /* initialize private data (optional) */
778         void (*init)(struct sock *sk);
779         /* cleanup private data  (optional) */
780         void (*release)(struct sock *sk);
781 
782         /* return slow start threshold (required) */
783         u32 (*ssthresh)(struct sock *sk);
784         /* lower bound for congestion window (optional) */
785         u32 (*min_cwnd)(const struct sock *sk);
786         /* do new cwnd calculation (required) */
787         void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
788         /* call before changing ca_state (optional) */
789         void (*set_state)(struct sock *sk, u8 new_state);
790         /* call when cwnd event occurs (optional) */
791         void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
792         /* new value of cwnd after loss (optional) */
793         u32  (*undo_cwnd)(struct sock *sk);
794         /* hook for packet ack accounting (optional) */
795         void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
796         /* get info for inet_diag (optional) */
797         void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
798 
799         char            name[TCP_CA_NAME_MAX];
800         struct module   *owner;
801 };
802 
803 extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
804 extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
805 
806 extern void tcp_init_congestion_control(struct sock *sk);
807 extern void tcp_cleanup_congestion_control(struct sock *sk);
808 extern int tcp_set_default_congestion_control(const char *name);
809 extern void tcp_get_default_congestion_control(char *name);
810 extern void tcp_get_available_congestion_control(char *buf, size_t len);
811 extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
812 extern int tcp_set_allowed_congestion_control(char *allowed);
813 extern int tcp_set_congestion_control(struct sock *sk, const char *name);
814 extern void tcp_slow_start(struct tcp_sock *tp);
815 extern void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w);
816 
817 extern struct tcp_congestion_ops tcp_init_congestion_ops;
818 extern u32 tcp_reno_ssthresh(struct sock *sk);
819 extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
820 extern u32 tcp_reno_min_cwnd(const struct sock *sk);
821 extern struct tcp_congestion_ops tcp_reno;
822 
823 static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
824 {
825         struct inet_connection_sock *icsk = inet_csk(sk);
826 
827         if (icsk->icsk_ca_ops->set_state)
828                 icsk->icsk_ca_ops->set_state(sk, ca_state);
829         icsk->icsk_ca_state = ca_state;
830 }
831 
832 static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
833 {
834         const struct inet_connection_sock *icsk = inet_csk(sk);
835 
836         if (icsk->icsk_ca_ops->cwnd_event)
837                 icsk->icsk_ca_ops->cwnd_event(sk, event);
838 }
839 
840 /* These functions determine how the current flow behaves in respect of SACK
841  * handling. SACK is negotiated with the peer, and therefore it can vary
842  * between different flows.
843  *
844  * tcp_is_sack - SACK enabled
845  * tcp_is_reno - No SACK
846  * tcp_is_fack - FACK enabled, implies SACK enabled
847  */
848 static inline int tcp_is_sack(const struct tcp_sock *tp)
849 {
850         return tp->rx_opt.sack_ok;
851 }
852 
853 static inline bool tcp_is_reno(const struct tcp_sock *tp)
854 {
855         return !tcp_is_sack(tp);
856 }
857 
858 static inline bool tcp_is_fack(const struct tcp_sock *tp)
859 {
860         return tp->rx_opt.sack_ok & TCP_FACK_ENABLED;
861 }
862 
863 static inline void tcp_enable_fack(struct tcp_sock *tp)
864 {
865         tp->rx_opt.sack_ok |= TCP_FACK_ENABLED;
866 }
867 
868 /* TCP early-retransmit (ER) is similar to but more conservative than
869  * the thin-dupack feature.  Enable ER only if thin-dupack is disabled.
870  */
871 static inline void tcp_enable_early_retrans(struct tcp_sock *tp)
872 {
873         tp->do_early_retrans = sysctl_tcp_early_retrans &&
874                 sysctl_tcp_early_retrans < 4 && !sysctl_tcp_thin_dupack &&
875                 sysctl_tcp_reordering == 3;
876 }
877 
878 static inline void tcp_disable_early_retrans(struct tcp_sock *tp)
879 {
880         tp->do_early_retrans = 0;
881 }
882 
883 static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
884 {
885         return tp->sacked_out + tp->lost_out;
886 }
887 
888 /* This determines how many packets are "in the network" to the best
889  * of our knowledge.  In many cases it is conservative, but where
890  * detailed information is available from the receiver (via SACK
891  * blocks etc.) we can make more aggressive calculations.
892  *
893  * Use this for decisions involving congestion control, use just
894  * tp->packets_out to determine if the send queue is empty or not.
895  *
896  * Read this equation as:
897  *
898  *      "Packets sent once on transmission queue" MINUS
899  *      "Packets left network, but not honestly ACKed yet" PLUS
900  *      "Packets fast retransmitted"
901  */
902 static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
903 {
904         return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
905 }
906 
907 #define TCP_INFINITE_SSTHRESH   0x7fffffff
908 
909 static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp)
910 {
911         return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH;
912 }
913 
914 static inline bool tcp_in_cwnd_reduction(const struct sock *sk)
915 {
916         return (TCPF_CA_CWR | TCPF_CA_Recovery) &
917                (1 << inet_csk(sk)->icsk_ca_state);
918 }
919 
920 /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
921  * The exception is cwnd reduction phase, when cwnd is decreasing towards
922  * ssthresh.
923  */
924 static inline __u32 tcp_current_ssthresh(const struct sock *sk)
925 {
926         const struct tcp_sock *tp = tcp_sk(sk);
927 
928         if (tcp_in_cwnd_reduction(sk))
929                 return tp->snd_ssthresh;
930         else
931                 return max(tp->snd_ssthresh,
932                            ((tp->snd_cwnd >> 1) +
933                             (tp->snd_cwnd >> 2)));
934 }
935 
936 /* Use define here intentionally to get WARN_ON location shown at the caller */
937 #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
938 
939 extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
940 extern __u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst);
941 
942 /* The maximum number of MSS of available cwnd for which TSO defers
943  * sending if not using sysctl_tcp_tso_win_divisor.
944  */
945 static inline __u32 tcp_max_tso_deferred_mss(const struct tcp_sock *tp)
946 {
947         return 3;
948 }
949 
950 /* Slow start with delack produces 3 packets of burst, so that
951  * it is safe "de facto".  This will be the default - same as
952  * the default reordering threshold - but if reordering increases,
953  * we must be able to allow cwnd to burst at least this much in order
954  * to not pull it back when holes are filled.
955  */
956 static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
957 {
958         return tp->reordering;
959 }
960 
961 /* Returns end sequence number of the receiver's advertised window */
962 static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
963 {
964         return tp->snd_una + tp->snd_wnd;
965 }
966 extern bool tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
967 
968 static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
969                                        const struct sk_buff *skb)
970 {
971         if (skb->len < mss)
972                 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
973 }
974 
975 static inline void tcp_check_probe_timer(struct sock *sk)
976 {
977         const struct tcp_sock *tp = tcp_sk(sk);
978         const struct inet_connection_sock *icsk = inet_csk(sk);
979 
980         if (!tp->packets_out && !icsk->icsk_pending)
981                 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
982                                           icsk->icsk_rto, TCP_RTO_MAX);
983 }
984 
985 static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq)
986 {
987         tp->snd_wl1 = seq;
988 }
989 
990 static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq)
991 {
992         tp->snd_wl1 = seq;
993 }
994 
995 /*
996  * Calculate(/check) TCP checksum
997  */
998 static inline __sum16 tcp_v4_check(int len, __be32 saddr,
999                                    __be32 daddr, __wsum base)
1000 {
1001         return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
1002 }
1003 
1004 static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
1005 {
1006         return __skb_checksum_complete(skb);
1007 }
1008 
1009 static inline bool tcp_checksum_complete(struct sk_buff *skb)
1010 {
1011         return !skb_csum_unnecessary(skb) &&
1012                 __tcp_checksum_complete(skb);
1013 }
1014 
1015 /* Prequeue for VJ style copy to user, combined with checksumming. */
1016 
1017 static inline void tcp_prequeue_init(struct tcp_sock *tp)
1018 {
1019         tp->ucopy.task = NULL;
1020         tp->ucopy.len = 0;
1021         tp->ucopy.memory = 0;
1022         skb_queue_head_init(&tp->ucopy.prequeue);
1023 #ifdef CONFIG_NET_DMA
1024         tp->ucopy.dma_chan = NULL;
1025         tp->ucopy.wakeup = 0;
1026         tp->ucopy.pinned_list = NULL;
1027         tp->ucopy.dma_cookie = 0;
1028 #endif
1029 }
1030 
1031 extern bool tcp_prequeue(struct sock *sk, struct sk_buff *skb);
1032 int tcp_filter(struct sock *sk, struct sk_buff *skb);
1033 
1034 #undef STATE_TRACE
1035 
1036 #ifdef STATE_TRACE
1037 static const char *statename[]={
1038         "Unused","Established","Syn Sent","Syn Recv",
1039         "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
1040         "Close Wait","Last ACK","Listen","Closing"
1041 };
1042 #endif
1043 extern void tcp_set_state(struct sock *sk, int state);
1044 
1045 extern void tcp_done(struct sock *sk);
1046 
1047 static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
1048 {
1049         rx_opt->dsack = 0;
1050         rx_opt->num_sacks = 0;
1051 }
1052 
1053 /* Determine a window scaling and initial window to offer. */
1054 extern void tcp_select_initial_window(int __space, __u32 mss,
1055                                       __u32 *rcv_wnd, __u32 *window_clamp,
1056                                       int wscale_ok, __u8 *rcv_wscale,
1057                                       __u32 init_rcv_wnd);
1058 
1059 static inline int tcp_win_from_space(int space)
1060 {
1061         return sysctl_tcp_adv_win_scale<=0 ?
1062                 (space>>(-sysctl_tcp_adv_win_scale)) :
1063                 space - (space>>sysctl_tcp_adv_win_scale);
1064 }
1065 
1066 /* Note: caller must be prepared to deal with negative returns */ 
1067 static inline int tcp_space(const struct sock *sk)
1068 {
1069         return tcp_win_from_space(sk->sk_rcvbuf -
1070                                   atomic_read(&sk->sk_rmem_alloc));
1071 } 
1072 
1073 static inline int tcp_full_space(const struct sock *sk)
1074 {
1075         return tcp_win_from_space(sk->sk_rcvbuf); 
1076 }
1077 
1078 static inline void tcp_openreq_init(struct request_sock *req,
1079                                     struct tcp_options_received *rx_opt,
1080                                     struct sk_buff *skb)
1081 {
1082         struct inet_request_sock *ireq = inet_rsk(req);
1083 
1084         req->rcv_wnd = 0;               /* So that tcp_send_synack() knows! */
1085         req->cookie_ts = 0;
1086         tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1087         tcp_rsk(req)->rcv_nxt = TCP_SKB_CB(skb)->seq + 1;
1088         tcp_rsk(req)->snt_synack = 0;
1089         req->mss = rx_opt->mss_clamp;
1090         req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
1091         ireq->tstamp_ok = rx_opt->tstamp_ok;
1092         ireq->sack_ok = rx_opt->sack_ok;
1093         ireq->snd_wscale = rx_opt->snd_wscale;
1094         ireq->wscale_ok = rx_opt->wscale_ok;
1095         ireq->acked = 0;
1096         ireq->ecn_ok = 0;
1097         ireq->rmt_port = tcp_hdr(skb)->source;
1098         ireq->loc_port = tcp_hdr(skb)->dest;
1099 }
1100 
1101 /* Compute time elapsed between SYNACK and the ACK completing 3WHS */
1102 static inline void tcp_synack_rtt_meas(struct sock *sk,
1103                                        struct request_sock *req)
1104 {
1105         if (tcp_rsk(req)->snt_synack)
1106                 tcp_valid_rtt_meas(sk,
1107                     tcp_time_stamp - tcp_rsk(req)->snt_synack);
1108 }
1109 
1110 extern void tcp_enter_memory_pressure(struct sock *sk);
1111 
1112 static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1113 {
1114         return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1115 }
1116 
1117 static inline int keepalive_time_when(const struct tcp_sock *tp)
1118 {
1119         return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1120 }
1121 
1122 static inline int keepalive_probes(const struct tcp_sock *tp)
1123 {
1124         return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes;
1125 }
1126 
1127 static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp)
1128 {
1129         const struct inet_connection_sock *icsk = &tp->inet_conn;
1130 
1131         return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime,
1132                           tcp_time_stamp - tp->rcv_tstamp);
1133 }
1134 
1135 static inline int tcp_fin_time(const struct sock *sk)
1136 {
1137         int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1138         const int rto = inet_csk(sk)->icsk_rto;
1139 
1140         if (fin_timeout < (rto << 2) - (rto >> 1))
1141                 fin_timeout = (rto << 2) - (rto >> 1);
1142 
1143         return fin_timeout;
1144 }
1145 
1146 static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt,
1147                                   int paws_win)
1148 {
1149         if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
1150                 return true;
1151         if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
1152                 return true;
1153         /*
1154          * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0,
1155          * then following tcp messages have valid values. Ignore 0 value,
1156          * or else 'negative' tsval might forbid us to accept their packets.
1157          */
1158         if (!rx_opt->ts_recent)
1159                 return true;
1160         return false;
1161 }
1162 
1163 static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt,
1164                                    int rst)
1165 {
1166         if (tcp_paws_check(rx_opt, 0))
1167                 return false;
1168 
1169         /* RST segments are not recommended to carry timestamp,
1170            and, if they do, it is recommended to ignore PAWS because
1171            "their cleanup function should take precedence over timestamps."
1172            Certainly, it is mistake. It is necessary to understand the reasons
1173            of this constraint to relax it: if peer reboots, clock may go
1174            out-of-sync and half-open connections will not be reset.
1175            Actually, the problem would be not existing if all
1176            the implementations followed draft about maintaining clock
1177            via reboots. Linux-2.2 DOES NOT!
1178 
1179            However, we can relax time bounds for RST segments to MSL.
1180          */
1181         if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1182                 return false;
1183         return true;
1184 }
1185 
1186 static inline void tcp_mib_init(struct net *net)
1187 {
1188         /* See RFC 2012 */
1189         TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1);
1190         TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1191         TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1192         TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1);
1193 }
1194 
1195 /* from STCP */
1196 static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
1197 {
1198         tp->lost_skb_hint = NULL;
1199         tp->scoreboard_skb_hint = NULL;
1200 }
1201 
1202 static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1203 {
1204         tcp_clear_retrans_hints_partial(tp);
1205         tp->retransmit_skb_hint = NULL;
1206 }
1207 
1208 /* MD5 Signature */
1209 struct crypto_hash;
1210 
1211 union tcp_md5_addr {
1212         struct in_addr  a4;
1213 #if IS_ENABLED(CONFIG_IPV6)
1214         struct in6_addr a6;
1215 #endif
1216 };
1217 
1218 /* - key database */
1219 struct tcp_md5sig_key {
1220         struct hlist_node       node;
1221         u8                      keylen;
1222         u8                      family; /* AF_INET or AF_INET6 */
1223         union tcp_md5_addr      addr;
1224         u8                      key[TCP_MD5SIG_MAXKEYLEN];
1225         struct rcu_head         rcu;
1226 };
1227 
1228 /* - sock block */
1229 struct tcp_md5sig_info {
1230         struct hlist_head       head;
1231         struct rcu_head         rcu;
1232 };
1233 
1234 /* - pseudo header */
1235 struct tcp4_pseudohdr {
1236         __be32          saddr;
1237         __be32          daddr;
1238         __u8            pad;
1239         __u8            protocol;
1240         __be16          len;
1241 };
1242 
1243 struct tcp6_pseudohdr {
1244         struct in6_addr saddr;
1245         struct in6_addr daddr;
1246         __be32          len;
1247         __be32          protocol;       /* including padding */
1248 };
1249 
1250 union tcp_md5sum_block {
1251         struct tcp4_pseudohdr ip4;
1252 #if IS_ENABLED(CONFIG_IPV6)
1253         struct tcp6_pseudohdr ip6;
1254 #endif
1255 };
1256 
1257 /* - pool: digest algorithm, hash description and scratch buffer */
1258 struct tcp_md5sig_pool {
1259         struct hash_desc        md5_desc;
1260         union tcp_md5sum_block  md5_blk;
1261 };
1262 
1263 /* - functions */
1264 extern int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
1265                                const struct sock *sk,
1266                                const struct request_sock *req,
1267                                const struct sk_buff *skb);
1268 extern int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
1269                           int family, const u8 *newkey,
1270                           u8 newkeylen, gfp_t gfp);
1271 extern int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr,
1272                           int family);
1273 extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1274                                          struct sock *addr_sk);
1275 
1276 #ifdef CONFIG_TCP_MD5SIG
1277 extern struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
1278                         const union tcp_md5_addr *addr, int family);
1279 #define tcp_twsk_md5_key(twsk)  ((twsk)->tw_md5_key)
1280 #else
1281 static inline struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
1282                                          const union tcp_md5_addr *addr,
1283                                          int family)
1284 {
1285         return NULL;
1286 }
1287 #define tcp_twsk_md5_key(twsk)  NULL
1288 #endif
1289 
1290 extern struct tcp_md5sig_pool __percpu *tcp_alloc_md5sig_pool(struct sock *);
1291 extern void tcp_free_md5sig_pool(void);
1292 
1293 extern struct tcp_md5sig_pool   *tcp_get_md5sig_pool(void);
1294 extern void tcp_put_md5sig_pool(void);
1295 
1296 extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, const struct tcphdr *);
1297 extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *,
1298                                  unsigned int header_len);
1299 extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
1300                             const struct tcp_md5sig_key *key);
1301 
1302 /* From tcp_fastopen.c */
1303 extern void tcp_fastopen_cache_get(struct sock *sk, u16 *mss,
1304                                    struct tcp_fastopen_cookie *cookie,
1305                                    int *syn_loss, unsigned long *last_syn_loss);
1306 extern void tcp_fastopen_cache_set(struct sock *sk, u16 mss,
1307                                    struct tcp_fastopen_cookie *cookie,
1308                                    bool syn_lost);
1309 struct tcp_fastopen_request {
1310         /* Fast Open cookie. Size 0 means a cookie request */
1311         struct tcp_fastopen_cookie      cookie;
1312         struct msghdr                   *data;  /* data in MSG_FASTOPEN */
1313         size_t                          size;
1314         int                             copied; /* queued in tcp_connect() */
1315 };
1316 void tcp_free_fastopen_req(struct tcp_sock *tp);
1317 
1318 extern struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
1319 int tcp_fastopen_reset_cipher(void *key, unsigned int len);
1320 void tcp_fastopen_cookie_gen(__be32 addr, struct tcp_fastopen_cookie *foc);
1321 
1322 #define TCP_FASTOPEN_KEY_LENGTH 16
1323 
1324 /* Fastopen key context */
1325 struct tcp_fastopen_context {
1326         struct crypto_cipher __rcu      *tfm;
1327         __u8                            key[TCP_FASTOPEN_KEY_LENGTH];
1328         struct rcu_head                 rcu;
1329 };
1330 
1331 /* write queue abstraction */
1332 static inline void tcp_write_queue_purge(struct sock *sk)
1333 {
1334         struct sk_buff *skb;
1335 
1336         while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
1337                 sk_wmem_free_skb(sk, skb);
1338         sk_mem_reclaim(sk);
1339         tcp_clear_all_retrans_hints(tcp_sk(sk));
1340 }
1341 
1342 static inline struct sk_buff *tcp_write_queue_head(const struct sock *sk)
1343 {
1344         return skb_peek(&sk->sk_write_queue);
1345 }
1346 
1347 static inline struct sk_buff *tcp_write_queue_tail(const struct sock *sk)
1348 {
1349         return skb_peek_tail(&sk->sk_write_queue);
1350 }
1351 
1352 static inline struct sk_buff *tcp_write_queue_next(const struct sock *sk,
1353                                                    const struct sk_buff *skb)
1354 {
1355         return skb_queue_next(&sk->sk_write_queue, skb);
1356 }
1357 
1358 static inline struct sk_buff *tcp_write_queue_prev(const struct sock *sk,
1359                                                    const struct sk_buff *skb)
1360 {
1361         return skb_queue_prev(&sk->sk_write_queue, skb);
1362 }
1363 
1364 #define tcp_for_write_queue(skb, sk)                                    \
1365         skb_queue_walk(&(sk)->sk_write_queue, skb)
1366 
1367 #define tcp_for_write_queue_from(skb, sk)                               \
1368         skb_queue_walk_from(&(sk)->sk_write_queue, skb)
1369 
1370 #define tcp_for_write_queue_from_safe(skb, tmp, sk)                     \
1371         skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
1372 
1373 static inline struct sk_buff *tcp_send_head(const struct sock *sk)
1374 {
1375         return sk->sk_send_head;
1376 }
1377 
1378 static inline bool tcp_skb_is_last(const struct sock *sk,
1379                                    const struct sk_buff *skb)
1380 {
1381         return skb_queue_is_last(&sk->sk_write_queue, skb);
1382 }
1383 
1384 static inline void tcp_advance_send_head(struct sock *sk, const struct sk_buff *skb)
1385 {
1386         if (tcp_skb_is_last(sk, skb))
1387                 sk->sk_send_head = NULL;
1388         else
1389                 sk->sk_send_head = tcp_write_queue_next(sk, skb);
1390 }
1391 
1392 static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1393 {
1394         if (sk->sk_send_head == skb_unlinked)
1395                 sk->sk_send_head = NULL;
1396         if (tcp_sk(sk)->highest_sack == skb_unlinked)
1397                 tcp_sk(sk)->highest_sack = NULL;
1398 }
1399 
1400 static inline void tcp_init_send_head(struct sock *sk)
1401 {
1402         sk->sk_send_head = NULL;
1403 }
1404 
1405 static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1406 {
1407         __skb_queue_tail(&sk->sk_write_queue, skb);
1408 }
1409 
1410 static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1411 {
1412         __tcp_add_write_queue_tail(sk, skb);
1413 
1414         /* Queue it, remembering where we must start sending. */
1415         if (sk->sk_send_head == NULL) {
1416                 sk->sk_send_head = skb;
1417 
1418                 if (tcp_sk(sk)->highest_sack == NULL)
1419                         tcp_sk(sk)->highest_sack = skb;
1420         }
1421 }
1422 
1423 static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1424 {
1425         __skb_queue_head(&sk->sk_write_queue, skb);
1426 }
1427 
1428 /* Insert buff after skb on the write queue of sk.  */
1429 static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1430                                                 struct sk_buff *buff,
1431                                                 struct sock *sk)
1432 {
1433         __skb_queue_after(&sk->sk_write_queue, skb, buff);
1434 }
1435 
1436 /* Insert new before skb on the write queue of sk.  */
1437 static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1438                                                   struct sk_buff *skb,
1439                                                   struct sock *sk)
1440 {
1441         __skb_queue_before(&sk->sk_write_queue, skb, new);
1442 
1443         if (sk->sk_send_head == skb)
1444                 sk->sk_send_head = new;
1445 }
1446 
1447 static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1448 {
1449         __skb_unlink(skb, &sk->sk_write_queue);
1450 }
1451 
1452 static inline bool tcp_write_queue_empty(struct sock *sk)
1453 {
1454         return skb_queue_empty(&sk->sk_write_queue);
1455 }
1456 
1457 static inline void tcp_push_pending_frames(struct sock *sk)
1458 {
1459         if (tcp_send_head(sk)) {
1460                 struct tcp_sock *tp = tcp_sk(sk);
1461 
1462                 __tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle);
1463         }
1464 }
1465 
1466 /* Start sequence of the skb just after the highest skb with SACKed
1467  * bit, valid only if sacked_out > 0 or when the caller has ensured
1468  * validity by itself.
1469  */
1470 static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1471 {
1472         if (!tp->sacked_out)
1473                 return tp->snd_una;
1474 
1475         if (tp->highest_sack == NULL)
1476                 return tp->snd_nxt;
1477 
1478         return TCP_SKB_CB(tp->highest_sack)->seq;
1479 }
1480 
1481 static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1482 {
1483         tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1484                                                 tcp_write_queue_next(sk, skb);
1485 }
1486 
1487 static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1488 {
1489         return tcp_sk(sk)->highest_sack;
1490 }
1491 
1492 static inline void tcp_highest_sack_reset(struct sock *sk)
1493 {
1494         tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1495 }
1496 
1497 /* Called when old skb is about to be deleted (to be combined with new skb) */
1498 static inline void tcp_highest_sack_combine(struct sock *sk,
1499                                             struct sk_buff *old,
1500                                             struct sk_buff *new)
1501 {
1502         if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1503                 tcp_sk(sk)->highest_sack = new;
1504 }
1505 
1506 /* Determines whether this is a thin stream (which may suffer from
1507  * increased latency). Used to trigger latency-reducing mechanisms.
1508  */
1509 static inline bool tcp_stream_is_thin(struct tcp_sock *tp)
1510 {
1511         return tp->packets_out < 4 && !tcp_in_initial_slowstart(tp);
1512 }
1513 
1514 /* /proc */
1515 enum tcp_seq_states {
1516         TCP_SEQ_STATE_LISTENING,
1517         TCP_SEQ_STATE_OPENREQ,
1518         TCP_SEQ_STATE_ESTABLISHED,
1519         TCP_SEQ_STATE_TIME_WAIT,
1520 };
1521 
1522 int tcp_seq_open(struct inode *inode, struct file *file);
1523 
1524 struct tcp_seq_afinfo {
1525         char                            *name;
1526         sa_family_t                     family;
1527         const struct file_operations    *seq_fops;
1528         struct seq_operations           seq_ops;
1529 };
1530 
1531 struct tcp_iter_state {
1532         struct seq_net_private  p;
1533         sa_family_t             family;
1534         enum tcp_seq_states     state;
1535         struct sock             *syn_wait_sk;
1536         int                     bucket, offset, sbucket, num;
1537         kuid_t                  uid;
1538         loff_t                  last_pos;
1539 };
1540 
1541 extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo);
1542 extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo);
1543 
1544 extern struct request_sock_ops tcp_request_sock_ops;
1545 extern struct request_sock_ops tcp6_request_sock_ops;
1546 
1547 extern void tcp_v4_destroy_sock(struct sock *sk);
1548 
1549 extern int tcp_v4_gso_send_check(struct sk_buff *skb);
1550 extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb,
1551                                        netdev_features_t features);
1552 extern struct sk_buff **tcp_gro_receive(struct sk_buff **head,
1553                                         struct sk_buff *skb);
1554 extern struct sk_buff **tcp4_gro_receive(struct sk_buff **head,
1555                                          struct sk_buff *skb);
1556 extern int tcp_gro_complete(struct sk_buff *skb);
1557 extern int tcp4_gro_complete(struct sk_buff *skb);
1558 
1559 #ifdef CONFIG_PROC_FS
1560 extern int tcp4_proc_init(void);
1561 extern void tcp4_proc_exit(void);
1562 #endif
1563 
1564 /* TCP af-specific functions */
1565 struct tcp_sock_af_ops {
1566 #ifdef CONFIG_TCP_MD5SIG
1567         struct tcp_md5sig_key   *(*md5_lookup) (struct sock *sk,
1568                                                 struct sock *addr_sk);
1569         int                     (*calc_md5_hash) (char *location,
1570                                                   struct tcp_md5sig_key *md5,
1571                                                   const struct sock *sk,
1572                                                   const struct request_sock *req,
1573                                                   const struct sk_buff *skb);
1574         int                     (*md5_parse) (struct sock *sk,
1575                                               char __user *optval,
1576                                               int optlen);
1577 #endif
1578 };
1579 
1580 struct tcp_request_sock_ops {
1581 #ifdef CONFIG_TCP_MD5SIG
1582         struct tcp_md5sig_key   *(*md5_lookup) (struct sock *sk,
1583                                                 struct request_sock *req);
1584         int                     (*calc_md5_hash) (char *location,
1585                                                   struct tcp_md5sig_key *md5,
1586                                                   const struct sock *sk,
1587                                                   const struct request_sock *req,
1588                                                   const struct sk_buff *skb);
1589 #endif
1590 };
1591 
1592 extern void tcp_v4_init(void);
1593 extern void tcp_init(void);
1594 
1595 /* At how many jiffies into the future should the RTO fire? */
1596 static inline s32 tcp_rto_delta(const struct sock *sk)
1597 {
1598         const struct sk_buff *skb = tcp_write_queue_head(sk);
1599         const u32 rto = inet_csk(sk)->icsk_rto;
1600         const u32 rto_time_stamp = TCP_SKB_CB(skb)->when + rto;
1601 
1602         return (s32)(rto_time_stamp - tcp_time_stamp);
1603 }
1604 
1605 #endif  /* _TCP_H */
1606 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp