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

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

Version: ~ [ linux-5.4-rc7 ] ~ [ linux-5.3.11 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.84 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.154 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.201 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.201 ] ~ [ 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-3.16.77 ] ~ [ 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 /* Minimal accepted MSS. It is (60+60+8) - (20+20). */
 65 #define TCP_MIN_MSS             88U
 66 
 67 /* The least MTU to use for probing */
 68 #define TCP_BASE_MSS            512
 69 
 70 /* After receiving this amount of duplicate ACKs fast retransmit starts. */
 71 #define TCP_FASTRETRANS_THRESH 3
 72 
 73 /* Maximal reordering. */
 74 #define TCP_MAX_REORDERING      127
 75 
 76 /* Maximal number of ACKs sent quickly to accelerate slow-start. */
 77 #define TCP_MAX_QUICKACKS       16U
 78 
 79 /* urg_data states */
 80 #define TCP_URG_VALID   0x0100
 81 #define TCP_URG_NOTYET  0x0200
 82 #define TCP_URG_READ    0x0400
 83 
 84 #define TCP_RETR1       3       /*
 85                                  * This is how many retries it does before it
 86                                  * tries to figure out if the gateway is
 87                                  * down. Minimal RFC value is 3; it corresponds
 88                                  * to ~3sec-8min depending on RTO.
 89                                  */
 90 
 91 #define TCP_RETR2       15      /*
 92                                  * This should take at least
 93                                  * 90 minutes to time out.
 94                                  * RFC1122 says that the limit is 100 sec.
 95                                  * 15 is ~13-30min depending on RTO.
 96                                  */
 97 
 98 #define TCP_SYN_RETRIES  6      /* This is how many retries are done
 99                                  * when active opening a connection.
100                                  * RFC1122 says the minimum retry MUST
101                                  * be at least 180secs.  Nevertheless
102                                  * this value is corresponding to
103                                  * 63secs of retransmission with the
104                                  * current initial RTO.
105                                  */
106 
107 #define TCP_SYNACK_RETRIES 5    /* This is how may retries are done
108                                  * when passive opening a connection.
109                                  * This is corresponding to 31secs of
110                                  * retransmission with the current
111                                  * initial RTO.
112                                  */
113 
114 #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
115                                   * state, about 60 seconds     */
116 #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
117                                  /* BSD style FIN_WAIT2 deadlock breaker.
118                                   * It used to be 3min, new value is 60sec,
119                                   * to combine FIN-WAIT-2 timeout with
120                                   * TIME-WAIT timer.
121                                   */
122 
123 #define TCP_DELACK_MAX  ((unsigned)(HZ/5))      /* maximal time to delay before sending an ACK */
124 #if HZ >= 100
125 #define TCP_DELACK_MIN  ((unsigned)(HZ/25))     /* minimal time to delay before sending an ACK */
126 #define TCP_ATO_MIN     ((unsigned)(HZ/25))
127 #else
128 #define TCP_DELACK_MIN  4U
129 #define TCP_ATO_MIN     4U
130 #endif
131 #define TCP_RTO_MAX     ((unsigned)(120*HZ))
132 #define TCP_RTO_MIN     ((unsigned)(HZ/5))
133 #define TCP_TIMEOUT_INIT ((unsigned)(1*HZ))     /* RFC6298 2.1 initial RTO value        */
134 #define TCP_TIMEOUT_FALLBACK ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value, now
135                                                  * used as a fallback RTO for the
136                                                  * initial data transmission if no
137                                                  * valid RTT sample has been acquired,
138                                                  * most likely due to retrans in 3WHS.
139                                                  */
140 
141 #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
142                                                          * for local resources.
143                                                          */
144 
145 #define TCP_KEEPALIVE_TIME      (120*60*HZ)     /* two hours */
146 #define TCP_KEEPALIVE_PROBES    9               /* Max of 9 keepalive probes    */
147 #define TCP_KEEPALIVE_INTVL     (75*HZ)
148 
149 #define MAX_TCP_KEEPIDLE        32767
150 #define MAX_TCP_KEEPINTVL       32767
151 #define MAX_TCP_KEEPCNT         127
152 #define MAX_TCP_SYNCNT          127
153 
154 #define TCP_SYNQ_INTERVAL       (HZ/5)  /* Period of SYNACK timer */
155 
156 #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
157 #define TCP_PAWS_MSL    60              /* Per-host timestamps are invalidated
158                                          * after this time. It should be equal
159                                          * (or greater than) TCP_TIMEWAIT_LEN
160                                          * to provide reliability equal to one
161                                          * provided by timewait state.
162                                          */
163 #define TCP_PAWS_WINDOW 1               /* Replay window for per-host
164                                          * timestamps. It must be less than
165                                          * minimal timewait lifetime.
166                                          */
167 /*
168  *      TCP option
169  */
170  
171 #define TCPOPT_NOP              1       /* Padding */
172 #define TCPOPT_EOL              0       /* End of options */
173 #define TCPOPT_MSS              2       /* Segment size negotiating */
174 #define TCPOPT_WINDOW           3       /* Window scaling */
175 #define TCPOPT_SACK_PERM        4       /* SACK Permitted */
176 #define TCPOPT_SACK             5       /* SACK Block */
177 #define TCPOPT_TIMESTAMP        8       /* Better RTT estimations/PAWS */
178 #define TCPOPT_MD5SIG           19      /* MD5 Signature (RFC2385) */
179 #define TCPOPT_EXP              254     /* Experimental */
180 /* Magic number to be after the option value for sharing TCP
181  * experimental options. See draft-ietf-tcpm-experimental-options-00.txt
182  */
183 #define TCPOPT_FASTOPEN_MAGIC   0xF989
184 
185 /*
186  *     TCP option lengths
187  */
188 
189 #define TCPOLEN_MSS            4
190 #define TCPOLEN_WINDOW         3
191 #define TCPOLEN_SACK_PERM      2
192 #define TCPOLEN_TIMESTAMP      10
193 #define TCPOLEN_MD5SIG         18
194 #define TCPOLEN_EXP_FASTOPEN_BASE  4
195 #define TCPOLEN_COOKIE_BASE    2        /* Cookie-less header extension */
196 #define TCPOLEN_COOKIE_PAIR    3        /* Cookie pair header extension */
197 #define TCPOLEN_COOKIE_MIN     (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MIN)
198 #define TCPOLEN_COOKIE_MAX     (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MAX)
199 
200 /* But this is what stacks really send out. */
201 #define TCPOLEN_TSTAMP_ALIGNED          12
202 #define TCPOLEN_WSCALE_ALIGNED          4
203 #define TCPOLEN_SACKPERM_ALIGNED        4
204 #define TCPOLEN_SACK_BASE               2
205 #define TCPOLEN_SACK_BASE_ALIGNED       4
206 #define TCPOLEN_SACK_PERBLOCK           8
207 #define TCPOLEN_MD5SIG_ALIGNED          20
208 #define TCPOLEN_MSS_ALIGNED             4
209 
210 /* Flags in tp->nonagle */
211 #define TCP_NAGLE_OFF           1       /* Nagle's algo is disabled */
212 #define TCP_NAGLE_CORK          2       /* Socket is corked         */
213 #define TCP_NAGLE_PUSH          4       /* Cork is overridden for already queued data */
214 
215 /* TCP thin-stream limits */
216 #define TCP_THIN_LINEAR_RETRIES 6       /* After 6 linear retries, do exp. backoff */
217 
218 /* TCP initial congestion window as per draft-hkchu-tcpm-initcwnd-01 */
219 #define TCP_INIT_CWND           10
220 
221 /* Bit Flags for sysctl_tcp_fastopen */
222 #define TFO_CLIENT_ENABLE       1
223 #define TFO_SERVER_ENABLE       2
224 #define TFO_CLIENT_NO_COOKIE    4       /* Data in SYN w/o cookie option */
225 
226 /* Process SYN data but skip cookie validation */
227 #define TFO_SERVER_COOKIE_NOT_CHKED     0x100
228 /* Accept SYN data w/o any cookie option */
229 #define TFO_SERVER_COOKIE_NOT_REQD      0x200
230 
231 /* Force enable TFO on all listeners, i.e., not requiring the
232  * TCP_FASTOPEN socket option. SOCKOPT1/2 determine how to set max_qlen.
233  */
234 #define TFO_SERVER_WO_SOCKOPT1  0x400
235 #define TFO_SERVER_WO_SOCKOPT2  0x800
236 /* Always create TFO child sockets on a TFO listener even when
237  * cookie/data not present. (For testing purpose!)
238  */
239 #define TFO_SERVER_ALWAYS       0x1000
240 
241 extern struct inet_timewait_death_row tcp_death_row;
242 
243 /* sysctl variables for tcp */
244 extern int sysctl_tcp_timestamps;
245 extern int sysctl_tcp_window_scaling;
246 extern int sysctl_tcp_sack;
247 extern int sysctl_tcp_fin_timeout;
248 extern int sysctl_tcp_keepalive_time;
249 extern int sysctl_tcp_keepalive_probes;
250 extern int sysctl_tcp_keepalive_intvl;
251 extern int sysctl_tcp_syn_retries;
252 extern int sysctl_tcp_synack_retries;
253 extern int sysctl_tcp_retries1;
254 extern int sysctl_tcp_retries2;
255 extern int sysctl_tcp_orphan_retries;
256 extern int sysctl_tcp_syncookies;
257 extern int sysctl_tcp_fastopen;
258 extern int sysctl_tcp_retrans_collapse;
259 extern int sysctl_tcp_stdurg;
260 extern int sysctl_tcp_rfc1337;
261 extern int sysctl_tcp_abort_on_overflow;
262 extern int sysctl_tcp_max_orphans;
263 extern int sysctl_tcp_fack;
264 extern int sysctl_tcp_reordering;
265 extern int sysctl_tcp_dsack;
266 extern int sysctl_tcp_wmem[3];
267 extern int sysctl_tcp_rmem[3];
268 extern int sysctl_tcp_app_win;
269 extern int sysctl_tcp_adv_win_scale;
270 extern int sysctl_tcp_tw_reuse;
271 extern int sysctl_tcp_frto;
272 extern int sysctl_tcp_low_latency;
273 extern int sysctl_tcp_dma_copybreak;
274 extern int sysctl_tcp_nometrics_save;
275 extern int sysctl_tcp_moderate_rcvbuf;
276 extern int sysctl_tcp_tso_win_divisor;
277 extern int sysctl_tcp_mtu_probing;
278 extern int sysctl_tcp_base_mss;
279 extern int sysctl_tcp_workaround_signed_windows;
280 extern int sysctl_tcp_slow_start_after_idle;
281 extern int sysctl_tcp_max_ssthresh;
282 extern int sysctl_tcp_thin_linear_timeouts;
283 extern int sysctl_tcp_thin_dupack;
284 extern int sysctl_tcp_early_retrans;
285 extern int sysctl_tcp_limit_output_bytes;
286 extern int sysctl_tcp_challenge_ack_limit;
287 extern int sysctl_tcp_min_tso_segs;
288 
289 extern atomic_long_t tcp_memory_allocated;
290 extern struct percpu_counter tcp_sockets_allocated;
291 extern int tcp_memory_pressure;
292 
293 /*
294  * The next routines deal with comparing 32 bit unsigned ints
295  * and worry about wraparound (automatic with unsigned arithmetic).
296  */
297 
298 static inline bool before(__u32 seq1, __u32 seq2)
299 {
300         return (__s32)(seq1-seq2) < 0;
301 }
302 #define after(seq2, seq1)       before(seq1, seq2)
303 
304 /* is s2<=s1<=s3 ? */
305 static inline bool between(__u32 seq1, __u32 seq2, __u32 seq3)
306 {
307         return seq3 - seq2 >= seq1 - seq2;
308 }
309 
310 static inline bool tcp_out_of_memory(struct sock *sk)
311 {
312         if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
313             sk_memory_allocated(sk) > sk_prot_mem_limits(sk, 2))
314                 return true;
315         return false;
316 }
317 
318 static inline bool tcp_too_many_orphans(struct sock *sk, int shift)
319 {
320         struct percpu_counter *ocp = sk->sk_prot->orphan_count;
321         int orphans = percpu_counter_read_positive(ocp);
322 
323         if (orphans << shift > sysctl_tcp_max_orphans) {
324                 orphans = percpu_counter_sum_positive(ocp);
325                 if (orphans << shift > sysctl_tcp_max_orphans)
326                         return true;
327         }
328         return false;
329 }
330 
331 extern bool tcp_check_oom(struct sock *sk, int shift);
332 
333 /* syncookies: remember time of last synqueue overflow */
334 static inline void tcp_synq_overflow(struct sock *sk)
335 {
336         tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies;
337 }
338 
339 /* syncookies: no recent synqueue overflow on this listening socket? */
340 static inline bool tcp_synq_no_recent_overflow(const struct sock *sk)
341 {
342         unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp;
343         return time_after(jiffies, last_overflow + TCP_TIMEOUT_FALLBACK);
344 }
345 
346 extern struct proto tcp_prot;
347 
348 #define TCP_INC_STATS(net, field)       SNMP_INC_STATS((net)->mib.tcp_statistics, field)
349 #define TCP_INC_STATS_BH(net, field)    SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
350 #define TCP_DEC_STATS(net, field)       SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
351 #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val)
352 #define TCP_ADD_STATS(net, field, val)  SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val)
353 
354 extern void tcp_init_mem(struct net *net);
355 
356 extern void tcp_tasklet_init(void);
357 
358 extern void tcp_v4_err(struct sk_buff *skb, u32);
359 
360 extern void tcp_shutdown (struct sock *sk, int how);
361 
362 extern void tcp_v4_early_demux(struct sk_buff *skb);
363 extern int tcp_v4_rcv(struct sk_buff *skb);
364 
365 extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
366 extern int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
367                        size_t size);
368 extern int tcp_sendpage(struct sock *sk, struct page *page, int offset,
369                         size_t size, int flags);
370 extern void tcp_release_cb(struct sock *sk);
371 extern void tcp_wfree(struct sk_buff *skb);
372 extern void tcp_write_timer_handler(struct sock *sk);
373 extern void tcp_delack_timer_handler(struct sock *sk);
374 extern int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg);
375 extern int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
376                                  const struct tcphdr *th, unsigned int len);
377 extern int tcp_rcv_established(struct sock *sk, struct sk_buff *skb,
378                                const struct tcphdr *th, unsigned int len);
379 extern void tcp_rcv_space_adjust(struct sock *sk);
380 extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
381 extern int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp);
382 extern void tcp_twsk_destructor(struct sock *sk);
383 extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
384                                struct pipe_inode_info *pipe, size_t len,
385                                unsigned int flags);
386 
387 static inline void tcp_dec_quickack_mode(struct sock *sk,
388                                          const unsigned int pkts)
389 {
390         struct inet_connection_sock *icsk = inet_csk(sk);
391 
392         if (icsk->icsk_ack.quick) {
393                 if (pkts >= icsk->icsk_ack.quick) {
394                         icsk->icsk_ack.quick = 0;
395                         /* Leaving quickack mode we deflate ATO. */
396                         icsk->icsk_ack.ato   = TCP_ATO_MIN;
397                 } else
398                         icsk->icsk_ack.quick -= pkts;
399         }
400 }
401 
402 #define TCP_ECN_OK              1
403 #define TCP_ECN_QUEUE_CWR       2
404 #define TCP_ECN_DEMAND_CWR      4
405 #define TCP_ECN_SEEN            8
406 
407 enum tcp_tw_status {
408         TCP_TW_SUCCESS = 0,
409         TCP_TW_RST = 1,
410         TCP_TW_ACK = 2,
411         TCP_TW_SYN = 3
412 };
413 
414 
415 extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
416                                                      struct sk_buff *skb,
417                                                      const struct tcphdr *th);
418 extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
419                                    struct request_sock *req,
420                                    struct request_sock **prev,
421                                    bool fastopen);
422 extern int tcp_child_process(struct sock *parent, struct sock *child,
423                              struct sk_buff *skb);
424 extern void tcp_enter_loss(struct sock *sk, int how);
425 extern void tcp_clear_retrans(struct tcp_sock *tp);
426 extern void tcp_update_metrics(struct sock *sk);
427 extern void tcp_init_metrics(struct sock *sk);
428 extern void tcp_metrics_init(void);
429 extern bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst, bool paws_check);
430 extern bool tcp_remember_stamp(struct sock *sk);
431 extern bool tcp_tw_remember_stamp(struct inet_timewait_sock *tw);
432 extern void tcp_fetch_timewait_stamp(struct sock *sk, struct dst_entry *dst);
433 extern void tcp_disable_fack(struct tcp_sock *tp);
434 extern void tcp_close(struct sock *sk, long timeout);
435 extern void tcp_init_sock(struct sock *sk);
436 extern unsigned int tcp_poll(struct file * file, struct socket *sock,
437                              struct poll_table_struct *wait);
438 extern int tcp_getsockopt(struct sock *sk, int level, int optname,
439                           char __user *optval, int __user *optlen);
440 extern int tcp_setsockopt(struct sock *sk, int level, int optname,
441                           char __user *optval, unsigned int optlen);
442 extern int compat_tcp_getsockopt(struct sock *sk, int level, int optname,
443                                  char __user *optval, int __user *optlen);
444 extern int compat_tcp_setsockopt(struct sock *sk, int level, int optname,
445                                  char __user *optval, unsigned int optlen);
446 extern void tcp_set_keepalive(struct sock *sk, int val);
447 extern void tcp_syn_ack_timeout(struct sock *sk, struct request_sock *req);
448 extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
449                        size_t len, int nonblock, int flags, int *addr_len);
450 extern void tcp_parse_options(const struct sk_buff *skb,
451                               struct tcp_options_received *opt_rx,
452                               int estab, struct tcp_fastopen_cookie *foc);
453 extern const u8 *tcp_parse_md5sig_option(const struct tcphdr *th);
454 
455 /*
456  *      TCP v4 functions exported for the inet6 API
457  */
458 
459 extern void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb);
460 extern int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb);
461 extern struct sock * tcp_create_openreq_child(struct sock *sk,
462                                               struct request_sock *req,
463                                               struct sk_buff *skb);
464 extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
465                                           struct request_sock *req,
466                                           struct dst_entry *dst);
467 extern int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb);
468 extern int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr,
469                           int addr_len);
470 extern int tcp_connect(struct sock *sk);
471 extern struct sk_buff * tcp_make_synack(struct sock *sk, struct dst_entry *dst,
472                                         struct request_sock *req,
473                                         struct tcp_fastopen_cookie *foc);
474 extern int tcp_disconnect(struct sock *sk, int flags);
475 
476 void tcp_connect_init(struct sock *sk);
477 void tcp_finish_connect(struct sock *sk, struct sk_buff *skb);
478 int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size);
479 void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb);
480 
481 /* From syncookies.c */
482 extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
483 extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb, 
484                                     struct ip_options *opt);
485 #ifdef CONFIG_SYN_COOKIES
486 extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb, 
487                                      __u16 *mss);
488 #else
489 static inline __u32 cookie_v4_init_sequence(struct sock *sk,
490                                             struct sk_buff *skb,
491                                             __u16 *mss)
492 {
493         return 0;
494 }
495 #endif
496 
497 extern __u32 cookie_init_timestamp(struct request_sock *req);
498 extern bool cookie_check_timestamp(struct tcp_options_received *opt,
499                                 struct net *net, bool *ecn_ok);
500 
501 /* From net/ipv6/syncookies.c */
502 extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
503 #ifdef CONFIG_SYN_COOKIES
504 extern __u32 cookie_v6_init_sequence(struct sock *sk, const struct sk_buff *skb,
505                                      __u16 *mss);
506 #else
507 static inline __u32 cookie_v6_init_sequence(struct sock *sk,
508                                             struct sk_buff *skb,
509                                             __u16 *mss)
510 {
511         return 0;
512 }
513 #endif
514 /* tcp_output.c */
515 
516 extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
517                                       int nonagle);
518 extern bool tcp_may_send_now(struct sock *sk);
519 extern int __tcp_retransmit_skb(struct sock *, struct sk_buff *);
520 extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
521 extern void tcp_retransmit_timer(struct sock *sk);
522 extern void tcp_xmit_retransmit_queue(struct sock *);
523 extern void tcp_simple_retransmit(struct sock *);
524 extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
525 extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
526 
527 extern void tcp_send_probe0(struct sock *);
528 extern void tcp_send_partial(struct sock *);
529 extern int tcp_write_wakeup(struct sock *);
530 extern void tcp_send_fin(struct sock *sk);
531 extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
532 extern int tcp_send_synack(struct sock *);
533 extern bool tcp_syn_flood_action(struct sock *sk,
534                                  const struct sk_buff *skb,
535                                  const char *proto);
536 extern void tcp_push_one(struct sock *, unsigned int mss_now);
537 extern void tcp_send_ack(struct sock *sk);
538 extern void tcp_send_delayed_ack(struct sock *sk);
539 extern void tcp_send_loss_probe(struct sock *sk);
540 extern bool tcp_schedule_loss_probe(struct sock *sk);
541 
542 /* tcp_input.c */
543 extern void tcp_cwnd_application_limited(struct sock *sk);
544 extern void tcp_resume_early_retransmit(struct sock *sk);
545 extern void tcp_rearm_rto(struct sock *sk);
546 extern void tcp_reset(struct sock *sk);
547 
548 /* tcp_timer.c */
549 extern void tcp_init_xmit_timers(struct sock *);
550 static inline void tcp_clear_xmit_timers(struct sock *sk)
551 {
552         inet_csk_clear_xmit_timers(sk);
553 }
554 
555 extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
556 extern unsigned int tcp_current_mss(struct sock *sk);
557 
558 /* Bound MSS / TSO packet size with the half of the window */
559 static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize)
560 {
561         int cutoff;
562 
563         /* When peer uses tiny windows, there is no use in packetizing
564          * to sub-MSS pieces for the sake of SWS or making sure there
565          * are enough packets in the pipe for fast recovery.
566          *
567          * On the other hand, for extremely large MSS devices, handling
568          * smaller than MSS windows in this way does make sense.
569          */
570         if (tp->max_window >= 512)
571                 cutoff = (tp->max_window >> 1);
572         else
573                 cutoff = tp->max_window;
574 
575         if (cutoff && pktsize > cutoff)
576                 return max_t(int, cutoff, 68U - tp->tcp_header_len);
577         else
578                 return pktsize;
579 }
580 
581 /* tcp.c */
582 extern void tcp_get_info(const struct sock *, struct tcp_info *);
583 
584 /* Read 'sendfile()'-style from a TCP socket */
585 typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
586                                 unsigned int, size_t);
587 extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
588                          sk_read_actor_t recv_actor);
589 
590 extern void tcp_initialize_rcv_mss(struct sock *sk);
591 
592 extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
593 extern int tcp_mss_to_mtu(struct sock *sk, int mss);
594 extern void tcp_mtup_init(struct sock *sk);
595 extern void tcp_valid_rtt_meas(struct sock *sk, u32 seq_rtt);
596 extern void tcp_init_buffer_space(struct sock *sk);
597 
598 static inline void tcp_bound_rto(const struct sock *sk)
599 {
600         if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX)
601                 inet_csk(sk)->icsk_rto = TCP_RTO_MAX;
602 }
603 
604 static inline u32 __tcp_set_rto(const struct tcp_sock *tp)
605 {
606         return (tp->srtt >> 3) + tp->rttvar;
607 }
608 
609 extern void tcp_set_rto(struct sock *sk);
610 
611 static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
612 {
613         tp->pred_flags = htonl((tp->tcp_header_len << 26) |
614                                ntohl(TCP_FLAG_ACK) |
615                                snd_wnd);
616 }
617 
618 static inline void tcp_fast_path_on(struct tcp_sock *tp)
619 {
620         __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
621 }
622 
623 static inline void tcp_fast_path_check(struct sock *sk)
624 {
625         struct tcp_sock *tp = tcp_sk(sk);
626 
627         if (skb_queue_empty(&tp->out_of_order_queue) &&
628             tp->rcv_wnd &&
629             atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
630             !tp->urg_data)
631                 tcp_fast_path_on(tp);
632 }
633 
634 /* Compute the actual rto_min value */
635 static inline u32 tcp_rto_min(struct sock *sk)
636 {
637         const struct dst_entry *dst = __sk_dst_get(sk);
638         u32 rto_min = TCP_RTO_MIN;
639 
640         if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
641                 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
642         return rto_min;
643 }
644 
645 /* Compute the actual receive window we are currently advertising.
646  * Rcv_nxt can be after the window if our peer push more data
647  * than the offered window.
648  */
649 static inline u32 tcp_receive_window(const struct tcp_sock *tp)
650 {
651         s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
652 
653         if (win < 0)
654                 win = 0;
655         return (u32) win;
656 }
657 
658 /* Choose a new window, without checks for shrinking, and without
659  * scaling applied to the result.  The caller does these things
660  * if necessary.  This is a "raw" window selection.
661  */
662 extern u32 __tcp_select_window(struct sock *sk);
663 
664 void tcp_send_window_probe(struct sock *sk);
665 
666 /* TCP timestamps are only 32-bits, this causes a slight
667  * complication on 64-bit systems since we store a snapshot
668  * of jiffies in the buffer control blocks below.  We decided
669  * to use only the low 32-bits of jiffies and hide the ugly
670  * casts with the following macro.
671  */
672 #define tcp_time_stamp          ((__u32)(jiffies))
673 
674 #define tcp_flag_byte(th) (((u_int8_t *)th)[13])
675 
676 #define TCPHDR_FIN 0x01
677 #define TCPHDR_SYN 0x02
678 #define TCPHDR_RST 0x04
679 #define TCPHDR_PSH 0x08
680 #define TCPHDR_ACK 0x10
681 #define TCPHDR_URG 0x20
682 #define TCPHDR_ECE 0x40
683 #define TCPHDR_CWR 0x80
684 
685 /* This is what the send packet queuing engine uses to pass
686  * TCP per-packet control information to the transmission code.
687  * We also store the host-order sequence numbers in here too.
688  * This is 44 bytes if IPV6 is enabled.
689  * If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately.
690  */
691 struct tcp_skb_cb {
692         union {
693                 struct inet_skb_parm    h4;
694 #if IS_ENABLED(CONFIG_IPV6)
695                 struct inet6_skb_parm   h6;
696 #endif
697         } header;       /* For incoming frames          */
698         __u32           seq;            /* Starting sequence number     */
699         __u32           end_seq;        /* SEQ + FIN + SYN + datalen    */
700         __u32           when;           /* used to compute rtt's        */
701         __u8            tcp_flags;      /* TCP header flags. (tcp[13])  */
702 
703         __u8            sacked;         /* State flags for SACK/FACK.   */
704 #define TCPCB_SACKED_ACKED      0x01    /* SKB ACK'd by a SACK block    */
705 #define TCPCB_SACKED_RETRANS    0x02    /* SKB retransmitted            */
706 #define TCPCB_LOST              0x04    /* SKB is lost                  */
707 #define TCPCB_TAGBITS           0x07    /* All tag bits                 */
708 #define TCPCB_EVER_RETRANS      0x80    /* Ever retransmitted frame     */
709 #define TCPCB_RETRANS           (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
710 
711         __u8            ip_dsfield;     /* IPv4 tos or IPv6 dsfield     */
712         /* 1 byte hole */
713         __u32           ack_seq;        /* Sequence number ACK'd        */
714 };
715 
716 #define TCP_SKB_CB(__skb)       ((struct tcp_skb_cb *)&((__skb)->cb[0]))
717 
718 /* RFC3168 : 6.1.1 SYN packets must not have ECT/ECN bits set
719  *
720  * If we receive a SYN packet with these bits set, it means a network is
721  * playing bad games with TOS bits. In order to avoid possible false congestion
722  * notifications, we disable TCP ECN negociation.
723  */
724 static inline void
725 TCP_ECN_create_request(struct request_sock *req, const struct sk_buff *skb,
726                 struct net *net)
727 {
728         const struct tcphdr *th = tcp_hdr(skb);
729 
730         if (net->ipv4.sysctl_tcp_ecn && th->ece && th->cwr &&
731             INET_ECN_is_not_ect(TCP_SKB_CB(skb)->ip_dsfield))
732                 inet_rsk(req)->ecn_ok = 1;
733 }
734 
735 /* Due to TSO, an SKB can be composed of multiple actual
736  * packets.  To keep these tracked properly, we use this.
737  */
738 static inline int tcp_skb_pcount(const struct sk_buff *skb)
739 {
740         return skb_shinfo(skb)->gso_segs;
741 }
742 
743 /* This is valid iff tcp_skb_pcount() > 1. */
744 static inline int tcp_skb_mss(const struct sk_buff *skb)
745 {
746         return skb_shinfo(skb)->gso_size;
747 }
748 
749 /* Events passed to congestion control interface */
750 enum tcp_ca_event {
751         CA_EVENT_TX_START,      /* first transmit when no packets in flight */
752         CA_EVENT_CWND_RESTART,  /* congestion window restart */
753         CA_EVENT_COMPLETE_CWR,  /* end of congestion recovery */
754         CA_EVENT_LOSS,          /* loss timeout */
755         CA_EVENT_FAST_ACK,      /* in sequence ack */
756         CA_EVENT_SLOW_ACK,      /* other ack */
757 };
758 
759 /*
760  * Interface for adding new TCP congestion control handlers
761  */
762 #define TCP_CA_NAME_MAX 16
763 #define TCP_CA_MAX      128
764 #define TCP_CA_BUF_MAX  (TCP_CA_NAME_MAX*TCP_CA_MAX)
765 
766 #define TCP_CONG_NON_RESTRICTED 0x1
767 #define TCP_CONG_RTT_STAMP      0x2
768 
769 struct tcp_congestion_ops {
770         struct list_head        list;
771         unsigned long flags;
772 
773         /* initialize private data (optional) */
774         void (*init)(struct sock *sk);
775         /* cleanup private data  (optional) */
776         void (*release)(struct sock *sk);
777 
778         /* return slow start threshold (required) */
779         u32 (*ssthresh)(struct sock *sk);
780         /* lower bound for congestion window (optional) */
781         u32 (*min_cwnd)(const struct sock *sk);
782         /* do new cwnd calculation (required) */
783         void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
784         /* call before changing ca_state (optional) */
785         void (*set_state)(struct sock *sk, u8 new_state);
786         /* call when cwnd event occurs (optional) */
787         void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
788         /* new value of cwnd after loss (optional) */
789         u32  (*undo_cwnd)(struct sock *sk);
790         /* hook for packet ack accounting (optional) */
791         void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
792         /* get info for inet_diag (optional) */
793         void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
794 
795         char            name[TCP_CA_NAME_MAX];
796         struct module   *owner;
797 };
798 
799 extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
800 extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
801 
802 extern void tcp_init_congestion_control(struct sock *sk);
803 extern void tcp_cleanup_congestion_control(struct sock *sk);
804 extern int tcp_set_default_congestion_control(const char *name);
805 extern void tcp_get_default_congestion_control(char *name);
806 extern void tcp_get_available_congestion_control(char *buf, size_t len);
807 extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
808 extern int tcp_set_allowed_congestion_control(char *allowed);
809 extern int tcp_set_congestion_control(struct sock *sk, const char *name);
810 extern void tcp_slow_start(struct tcp_sock *tp);
811 extern void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w);
812 
813 extern struct tcp_congestion_ops tcp_init_congestion_ops;
814 extern u32 tcp_reno_ssthresh(struct sock *sk);
815 extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
816 extern u32 tcp_reno_min_cwnd(const struct sock *sk);
817 extern struct tcp_congestion_ops tcp_reno;
818 
819 static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
820 {
821         struct inet_connection_sock *icsk = inet_csk(sk);
822 
823         if (icsk->icsk_ca_ops->set_state)
824                 icsk->icsk_ca_ops->set_state(sk, ca_state);
825         icsk->icsk_ca_state = ca_state;
826 }
827 
828 static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
829 {
830         const struct inet_connection_sock *icsk = inet_csk(sk);
831 
832         if (icsk->icsk_ca_ops->cwnd_event)
833                 icsk->icsk_ca_ops->cwnd_event(sk, event);
834 }
835 
836 /* These functions determine how the current flow behaves in respect of SACK
837  * handling. SACK is negotiated with the peer, and therefore it can vary
838  * between different flows.
839  *
840  * tcp_is_sack - SACK enabled
841  * tcp_is_reno - No SACK
842  * tcp_is_fack - FACK enabled, implies SACK enabled
843  */
844 static inline int tcp_is_sack(const struct tcp_sock *tp)
845 {
846         return tp->rx_opt.sack_ok;
847 }
848 
849 static inline bool tcp_is_reno(const struct tcp_sock *tp)
850 {
851         return !tcp_is_sack(tp);
852 }
853 
854 static inline bool tcp_is_fack(const struct tcp_sock *tp)
855 {
856         return tp->rx_opt.sack_ok & TCP_FACK_ENABLED;
857 }
858 
859 static inline void tcp_enable_fack(struct tcp_sock *tp)
860 {
861         tp->rx_opt.sack_ok |= TCP_FACK_ENABLED;
862 }
863 
864 /* TCP early-retransmit (ER) is similar to but more conservative than
865  * the thin-dupack feature.  Enable ER only if thin-dupack is disabled.
866  */
867 static inline void tcp_enable_early_retrans(struct tcp_sock *tp)
868 {
869         tp->do_early_retrans = sysctl_tcp_early_retrans &&
870                 sysctl_tcp_early_retrans < 4 && !sysctl_tcp_thin_dupack &&
871                 sysctl_tcp_reordering == 3;
872 }
873 
874 static inline void tcp_disable_early_retrans(struct tcp_sock *tp)
875 {
876         tp->do_early_retrans = 0;
877 }
878 
879 static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
880 {
881         return tp->sacked_out + tp->lost_out;
882 }
883 
884 /* This determines how many packets are "in the network" to the best
885  * of our knowledge.  In many cases it is conservative, but where
886  * detailed information is available from the receiver (via SACK
887  * blocks etc.) we can make more aggressive calculations.
888  *
889  * Use this for decisions involving congestion control, use just
890  * tp->packets_out to determine if the send queue is empty or not.
891  *
892  * Read this equation as:
893  *
894  *      "Packets sent once on transmission queue" MINUS
895  *      "Packets left network, but not honestly ACKed yet" PLUS
896  *      "Packets fast retransmitted"
897  */
898 static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
899 {
900         return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
901 }
902 
903 #define TCP_INFINITE_SSTHRESH   0x7fffffff
904 
905 static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp)
906 {
907         return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH;
908 }
909 
910 static inline bool tcp_in_cwnd_reduction(const struct sock *sk)
911 {
912         return (TCPF_CA_CWR | TCPF_CA_Recovery) &
913                (1 << inet_csk(sk)->icsk_ca_state);
914 }
915 
916 /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
917  * The exception is cwnd reduction phase, when cwnd is decreasing towards
918  * ssthresh.
919  */
920 static inline __u32 tcp_current_ssthresh(const struct sock *sk)
921 {
922         const struct tcp_sock *tp = tcp_sk(sk);
923 
924         if (tcp_in_cwnd_reduction(sk))
925                 return tp->snd_ssthresh;
926         else
927                 return max(tp->snd_ssthresh,
928                            ((tp->snd_cwnd >> 1) +
929                             (tp->snd_cwnd >> 2)));
930 }
931 
932 /* Use define here intentionally to get WARN_ON location shown at the caller */
933 #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
934 
935 extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
936 extern __u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst);
937 
938 /* The maximum number of MSS of available cwnd for which TSO defers
939  * sending if not using sysctl_tcp_tso_win_divisor.
940  */
941 static inline __u32 tcp_max_tso_deferred_mss(const struct tcp_sock *tp)
942 {
943         return 3;
944 }
945 
946 /* Slow start with delack produces 3 packets of burst, so that
947  * it is safe "de facto".  This will be the default - same as
948  * the default reordering threshold - but if reordering increases,
949  * we must be able to allow cwnd to burst at least this much in order
950  * to not pull it back when holes are filled.
951  */
952 static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
953 {
954         return tp->reordering;
955 }
956 
957 /* Returns end sequence number of the receiver's advertised window */
958 static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
959 {
960         return tp->snd_una + tp->snd_wnd;
961 }
962 extern bool tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
963 
964 static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
965                                        const struct sk_buff *skb)
966 {
967         if (skb->len < mss)
968                 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
969 }
970 
971 static inline void tcp_check_probe_timer(struct sock *sk)
972 {
973         const struct tcp_sock *tp = tcp_sk(sk);
974         const struct inet_connection_sock *icsk = inet_csk(sk);
975 
976         if (!tp->packets_out && !icsk->icsk_pending)
977                 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
978                                           icsk->icsk_rto, TCP_RTO_MAX);
979 }
980 
981 static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq)
982 {
983         tp->snd_wl1 = seq;
984 }
985 
986 static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq)
987 {
988         tp->snd_wl1 = seq;
989 }
990 
991 /*
992  * Calculate(/check) TCP checksum
993  */
994 static inline __sum16 tcp_v4_check(int len, __be32 saddr,
995                                    __be32 daddr, __wsum base)
996 {
997         return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
998 }
999 
1000 static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
1001 {
1002         return __skb_checksum_complete(skb);
1003 }
1004 
1005 static inline bool tcp_checksum_complete(struct sk_buff *skb)
1006 {
1007         return !skb_csum_unnecessary(skb) &&
1008                 __tcp_checksum_complete(skb);
1009 }
1010 
1011 /* Prequeue for VJ style copy to user, combined with checksumming. */
1012 
1013 static inline void tcp_prequeue_init(struct tcp_sock *tp)
1014 {
1015         tp->ucopy.task = NULL;
1016         tp->ucopy.len = 0;
1017         tp->ucopy.memory = 0;
1018         skb_queue_head_init(&tp->ucopy.prequeue);
1019 #ifdef CONFIG_NET_DMA
1020         tp->ucopy.dma_chan = NULL;
1021         tp->ucopy.wakeup = 0;
1022         tp->ucopy.pinned_list = NULL;
1023         tp->ucopy.dma_cookie = 0;
1024 #endif
1025 }
1026 
1027 extern bool tcp_prequeue(struct sock *sk, struct sk_buff *skb);
1028 
1029 #undef STATE_TRACE
1030 
1031 #ifdef STATE_TRACE
1032 static const char *statename[]={
1033         "Unused","Established","Syn Sent","Syn Recv",
1034         "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
1035         "Close Wait","Last ACK","Listen","Closing"
1036 };
1037 #endif
1038 extern void tcp_set_state(struct sock *sk, int state);
1039 
1040 extern void tcp_done(struct sock *sk);
1041 
1042 static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
1043 {
1044         rx_opt->dsack = 0;
1045         rx_opt->num_sacks = 0;
1046 }
1047 
1048 extern u32 tcp_default_init_rwnd(u32 mss);
1049 
1050 /* Determine a window scaling and initial window to offer. */
1051 extern void tcp_select_initial_window(int __space, __u32 mss,
1052                                       __u32 *rcv_wnd, __u32 *window_clamp,
1053                                       int wscale_ok, __u8 *rcv_wscale,
1054                                       __u32 init_rcv_wnd);
1055 
1056 static inline int tcp_win_from_space(int space)
1057 {
1058         return sysctl_tcp_adv_win_scale<=0 ?
1059                 (space>>(-sysctl_tcp_adv_win_scale)) :
1060                 space - (space>>sysctl_tcp_adv_win_scale);
1061 }
1062 
1063 /* Note: caller must be prepared to deal with negative returns */ 
1064 static inline int tcp_space(const struct sock *sk)
1065 {
1066         return tcp_win_from_space(sk->sk_rcvbuf -
1067                                   atomic_read(&sk->sk_rmem_alloc));
1068 } 
1069 
1070 static inline int tcp_full_space(const struct sock *sk)
1071 {
1072         return tcp_win_from_space(sk->sk_rcvbuf); 
1073 }
1074 
1075 static inline void tcp_openreq_init(struct request_sock *req,
1076                                     struct tcp_options_received *rx_opt,
1077                                     struct sk_buff *skb)
1078 {
1079         struct inet_request_sock *ireq = inet_rsk(req);
1080 
1081         req->rcv_wnd = 0;               /* So that tcp_send_synack() knows! */
1082         req->cookie_ts = 0;
1083         tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1084         tcp_rsk(req)->rcv_nxt = TCP_SKB_CB(skb)->seq + 1;
1085         tcp_rsk(req)->snt_synack = 0;
1086         req->mss = rx_opt->mss_clamp;
1087         req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
1088         ireq->tstamp_ok = rx_opt->tstamp_ok;
1089         ireq->sack_ok = rx_opt->sack_ok;
1090         ireq->snd_wscale = rx_opt->snd_wscale;
1091         ireq->wscale_ok = rx_opt->wscale_ok;
1092         ireq->acked = 0;
1093         ireq->ecn_ok = 0;
1094         ireq->rmt_port = tcp_hdr(skb)->source;
1095         ireq->loc_port = tcp_hdr(skb)->dest;
1096 }
1097 
1098 /* Compute time elapsed between SYNACK and the ACK completing 3WHS */
1099 static inline void tcp_synack_rtt_meas(struct sock *sk,
1100                                        struct request_sock *req)
1101 {
1102         if (tcp_rsk(req)->snt_synack)
1103                 tcp_valid_rtt_meas(sk,
1104                     tcp_time_stamp - tcp_rsk(req)->snt_synack);
1105 }
1106 
1107 extern void tcp_enter_memory_pressure(struct sock *sk);
1108 
1109 static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1110 {
1111         return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1112 }
1113 
1114 static inline int keepalive_time_when(const struct tcp_sock *tp)
1115 {
1116         return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1117 }
1118 
1119 static inline int keepalive_probes(const struct tcp_sock *tp)
1120 {
1121         return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes;
1122 }
1123 
1124 static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp)
1125 {
1126         const struct inet_connection_sock *icsk = &tp->inet_conn;
1127 
1128         return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime,
1129                           tcp_time_stamp - tp->rcv_tstamp);
1130 }
1131 
1132 static inline int tcp_fin_time(const struct sock *sk)
1133 {
1134         int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1135         const int rto = inet_csk(sk)->icsk_rto;
1136 
1137         if (fin_timeout < (rto << 2) - (rto >> 1))
1138                 fin_timeout = (rto << 2) - (rto >> 1);
1139 
1140         return fin_timeout;
1141 }
1142 
1143 static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt,
1144                                   int paws_win)
1145 {
1146         if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
1147                 return true;
1148         if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
1149                 return true;
1150         /*
1151          * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0,
1152          * then following tcp messages have valid values. Ignore 0 value,
1153          * or else 'negative' tsval might forbid us to accept their packets.
1154          */
1155         if (!rx_opt->ts_recent)
1156                 return true;
1157         return false;
1158 }
1159 
1160 static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt,
1161                                    int rst)
1162 {
1163         if (tcp_paws_check(rx_opt, 0))
1164                 return false;
1165 
1166         /* RST segments are not recommended to carry timestamp,
1167            and, if they do, it is recommended to ignore PAWS because
1168            "their cleanup function should take precedence over timestamps."
1169            Certainly, it is mistake. It is necessary to understand the reasons
1170            of this constraint to relax it: if peer reboots, clock may go
1171            out-of-sync and half-open connections will not be reset.
1172            Actually, the problem would be not existing if all
1173            the implementations followed draft about maintaining clock
1174            via reboots. Linux-2.2 DOES NOT!
1175 
1176            However, we can relax time bounds for RST segments to MSL.
1177          */
1178         if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1179                 return false;
1180         return true;
1181 }
1182 
1183 static inline void tcp_mib_init(struct net *net)
1184 {
1185         /* See RFC 2012 */
1186         TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1);
1187         TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1188         TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1189         TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1);
1190 }
1191 
1192 /* from STCP */
1193 static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
1194 {
1195         tp->lost_skb_hint = NULL;
1196 }
1197 
1198 static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1199 {
1200         tcp_clear_retrans_hints_partial(tp);
1201         tp->retransmit_skb_hint = NULL;
1202 }
1203 
1204 /* MD5 Signature */
1205 struct crypto_hash;
1206 
1207 union tcp_md5_addr {
1208         struct in_addr  a4;
1209 #if IS_ENABLED(CONFIG_IPV6)
1210         struct in6_addr a6;
1211 #endif
1212 };
1213 
1214 /* - key database */
1215 struct tcp_md5sig_key {
1216         struct hlist_node       node;
1217         u8                      keylen;
1218         u8                      family; /* AF_INET or AF_INET6 */
1219         union tcp_md5_addr      addr;
1220         u8                      key[TCP_MD5SIG_MAXKEYLEN];
1221         struct rcu_head         rcu;
1222 };
1223 
1224 /* - sock block */
1225 struct tcp_md5sig_info {
1226         struct hlist_head       head;
1227         struct rcu_head         rcu;
1228 };
1229 
1230 /* - pseudo header */
1231 struct tcp4_pseudohdr {
1232         __be32          saddr;
1233         __be32          daddr;
1234         __u8            pad;
1235         __u8            protocol;
1236         __be16          len;
1237 };
1238 
1239 struct tcp6_pseudohdr {
1240         struct in6_addr saddr;
1241         struct in6_addr daddr;
1242         __be32          len;
1243         __be32          protocol;       /* including padding */
1244 };
1245 
1246 union tcp_md5sum_block {
1247         struct tcp4_pseudohdr ip4;
1248 #if IS_ENABLED(CONFIG_IPV6)
1249         struct tcp6_pseudohdr ip6;
1250 #endif
1251 };
1252 
1253 /* - pool: digest algorithm, hash description and scratch buffer */
1254 struct tcp_md5sig_pool {
1255         struct hash_desc        md5_desc;
1256         union tcp_md5sum_block  md5_blk;
1257 };
1258 
1259 /* - functions */
1260 extern int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
1261                                const struct sock *sk,
1262                                const struct request_sock *req,
1263                                const struct sk_buff *skb);
1264 extern int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
1265                           int family, const u8 *newkey,
1266                           u8 newkeylen, gfp_t gfp);
1267 extern int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr,
1268                           int family);
1269 extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1270                                          struct sock *addr_sk);
1271 
1272 #ifdef CONFIG_TCP_MD5SIG
1273 extern struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
1274                         const union tcp_md5_addr *addr, int family);
1275 #define tcp_twsk_md5_key(twsk)  ((twsk)->tw_md5_key)
1276 #else
1277 static inline struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
1278                                          const union tcp_md5_addr *addr,
1279                                          int family)
1280 {
1281         return NULL;
1282 }
1283 #define tcp_twsk_md5_key(twsk)  NULL
1284 #endif
1285 
1286 extern bool tcp_alloc_md5sig_pool(void);
1287 
1288 extern struct tcp_md5sig_pool   *tcp_get_md5sig_pool(void);
1289 static inline void tcp_put_md5sig_pool(void)
1290 {
1291         local_bh_enable();
1292 }
1293 
1294 extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, const struct tcphdr *);
1295 extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *,
1296                                  unsigned int header_len);
1297 extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
1298                             const struct tcp_md5sig_key *key);
1299 
1300 /* From tcp_fastopen.c */
1301 extern void tcp_fastopen_cache_get(struct sock *sk, u16 *mss,
1302                                    struct tcp_fastopen_cookie *cookie,
1303                                    int *syn_loss, unsigned long *last_syn_loss);
1304 extern void tcp_fastopen_cache_set(struct sock *sk, u16 mss,
1305                                    struct tcp_fastopen_cookie *cookie,
1306                                    bool syn_lost);
1307 struct tcp_fastopen_request {
1308         /* Fast Open cookie. Size 0 means a cookie request */
1309         struct tcp_fastopen_cookie      cookie;
1310         struct msghdr                   *data;  /* data in MSG_FASTOPEN */
1311         u16                             copied; /* queued in tcp_connect() */
1312 };
1313 void tcp_free_fastopen_req(struct tcp_sock *tp);
1314 
1315 extern struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
1316 int tcp_fastopen_reset_cipher(void *key, unsigned int len);
1317 void tcp_fastopen_cookie_gen(__be32 addr, struct tcp_fastopen_cookie *foc);
1318 
1319 #define TCP_FASTOPEN_KEY_LENGTH 16
1320 
1321 /* Fastopen key context */
1322 struct tcp_fastopen_context {
1323         struct crypto_cipher    *tfm;
1324         __u8                    key[TCP_FASTOPEN_KEY_LENGTH];
1325         struct rcu_head         rcu;
1326 };
1327 
1328 /* write queue abstraction */
1329 static inline void tcp_write_queue_purge(struct sock *sk)
1330 {
1331         struct sk_buff *skb;
1332 
1333         while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
1334                 sk_wmem_free_skb(sk, skb);
1335         sk_mem_reclaim(sk);
1336         tcp_clear_all_retrans_hints(tcp_sk(sk));
1337 }
1338 
1339 static inline struct sk_buff *tcp_write_queue_head(const struct sock *sk)
1340 {
1341         return skb_peek(&sk->sk_write_queue);
1342 }
1343 
1344 static inline struct sk_buff *tcp_write_queue_tail(const struct sock *sk)
1345 {
1346         return skb_peek_tail(&sk->sk_write_queue);
1347 }
1348 
1349 static inline struct sk_buff *tcp_write_queue_next(const struct sock *sk,
1350                                                    const struct sk_buff *skb)
1351 {
1352         return skb_queue_next(&sk->sk_write_queue, skb);
1353 }
1354 
1355 static inline struct sk_buff *tcp_write_queue_prev(const struct sock *sk,
1356                                                    const struct sk_buff *skb)
1357 {
1358         return skb_queue_prev(&sk->sk_write_queue, skb);
1359 }
1360 
1361 #define tcp_for_write_queue(skb, sk)                                    \
1362         skb_queue_walk(&(sk)->sk_write_queue, skb)
1363 
1364 #define tcp_for_write_queue_from(skb, sk)                               \
1365         skb_queue_walk_from(&(sk)->sk_write_queue, skb)
1366 
1367 #define tcp_for_write_queue_from_safe(skb, tmp, sk)                     \
1368         skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
1369 
1370 static inline struct sk_buff *tcp_send_head(const struct sock *sk)
1371 {
1372         return sk->sk_send_head;
1373 }
1374 
1375 static inline bool tcp_skb_is_last(const struct sock *sk,
1376                                    const struct sk_buff *skb)
1377 {
1378         return skb_queue_is_last(&sk->sk_write_queue, skb);
1379 }
1380 
1381 static inline void tcp_advance_send_head(struct sock *sk, const struct sk_buff *skb)
1382 {
1383         if (tcp_skb_is_last(sk, skb))
1384                 sk->sk_send_head = NULL;
1385         else
1386                 sk->sk_send_head = tcp_write_queue_next(sk, skb);
1387 }
1388 
1389 static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1390 {
1391         if (sk->sk_send_head == skb_unlinked)
1392                 sk->sk_send_head = NULL;
1393 }
1394 
1395 static inline void tcp_init_send_head(struct sock *sk)
1396 {
1397         sk->sk_send_head = NULL;
1398 }
1399 
1400 static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1401 {
1402         __skb_queue_tail(&sk->sk_write_queue, skb);
1403 }
1404 
1405 static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1406 {
1407         __tcp_add_write_queue_tail(sk, skb);
1408 
1409         /* Queue it, remembering where we must start sending. */
1410         if (sk->sk_send_head == NULL) {
1411                 sk->sk_send_head = skb;
1412 
1413                 if (tcp_sk(sk)->highest_sack == NULL)
1414                         tcp_sk(sk)->highest_sack = skb;
1415         }
1416 }
1417 
1418 static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1419 {
1420         __skb_queue_head(&sk->sk_write_queue, skb);
1421 }
1422 
1423 /* Insert buff after skb on the write queue of sk.  */
1424 static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1425                                                 struct sk_buff *buff,
1426                                                 struct sock *sk)
1427 {
1428         __skb_queue_after(&sk->sk_write_queue, skb, buff);
1429 }
1430 
1431 /* Insert new before skb on the write queue of sk.  */
1432 static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1433                                                   struct sk_buff *skb,
1434                                                   struct sock *sk)
1435 {
1436         __skb_queue_before(&sk->sk_write_queue, skb, new);
1437 
1438         if (sk->sk_send_head == skb)
1439                 sk->sk_send_head = new;
1440 }
1441 
1442 static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1443 {
1444         __skb_unlink(skb, &sk->sk_write_queue);
1445 }
1446 
1447 static inline bool tcp_write_queue_empty(struct sock *sk)
1448 {
1449         return skb_queue_empty(&sk->sk_write_queue);
1450 }
1451 
1452 static inline void tcp_push_pending_frames(struct sock *sk)
1453 {
1454         if (tcp_send_head(sk)) {
1455                 struct tcp_sock *tp = tcp_sk(sk);
1456 
1457                 __tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle);
1458         }
1459 }
1460 
1461 /* Start sequence of the skb just after the highest skb with SACKed
1462  * bit, valid only if sacked_out > 0 or when the caller has ensured
1463  * validity by itself.
1464  */
1465 static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1466 {
1467         if (!tp->sacked_out)
1468                 return tp->snd_una;
1469 
1470         if (tp->highest_sack == NULL)
1471                 return tp->snd_nxt;
1472 
1473         return TCP_SKB_CB(tp->highest_sack)->seq;
1474 }
1475 
1476 static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1477 {
1478         tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1479                                                 tcp_write_queue_next(sk, skb);
1480 }
1481 
1482 static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1483 {
1484         return tcp_sk(sk)->highest_sack;
1485 }
1486 
1487 static inline void tcp_highest_sack_reset(struct sock *sk)
1488 {
1489         tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1490 }
1491 
1492 /* Called when old skb is about to be deleted (to be combined with new skb) */
1493 static inline void tcp_highest_sack_combine(struct sock *sk,
1494                                             struct sk_buff *old,
1495                                             struct sk_buff *new)
1496 {
1497         if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1498                 tcp_sk(sk)->highest_sack = new;
1499 }
1500 
1501 /* Determines whether this is a thin stream (which may suffer from
1502  * increased latency). Used to trigger latency-reducing mechanisms.
1503  */
1504 static inline bool tcp_stream_is_thin(struct tcp_sock *tp)
1505 {
1506         return tp->packets_out < 4 && !tcp_in_initial_slowstart(tp);
1507 }
1508 
1509 /* /proc */
1510 enum tcp_seq_states {
1511         TCP_SEQ_STATE_LISTENING,
1512         TCP_SEQ_STATE_OPENREQ,
1513         TCP_SEQ_STATE_ESTABLISHED,
1514         TCP_SEQ_STATE_TIME_WAIT,
1515 };
1516 
1517 int tcp_seq_open(struct inode *inode, struct file *file);
1518 
1519 struct tcp_seq_afinfo {
1520         char                            *name;
1521         sa_family_t                     family;
1522         const struct file_operations    *seq_fops;
1523         struct seq_operations           seq_ops;
1524 };
1525 
1526 struct tcp_iter_state {
1527         struct seq_net_private  p;
1528         sa_family_t             family;
1529         enum tcp_seq_states     state;
1530         struct sock             *syn_wait_sk;
1531         int                     bucket, offset, sbucket, num;
1532         kuid_t                  uid;
1533         loff_t                  last_pos;
1534 };
1535 
1536 extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo);
1537 extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo);
1538 
1539 extern struct request_sock_ops tcp_request_sock_ops;
1540 extern struct request_sock_ops tcp6_request_sock_ops;
1541 
1542 extern void tcp_v4_destroy_sock(struct sock *sk);
1543 
1544 extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb,
1545                                        netdev_features_t features);
1546 extern struct sk_buff **tcp_gro_receive(struct sk_buff **head,
1547                                         struct sk_buff *skb);
1548 extern int tcp_gro_complete(struct sk_buff *skb);
1549 
1550 extern void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr,
1551                                 __be32 daddr);
1552 
1553 #ifdef CONFIG_PROC_FS
1554 extern int tcp4_proc_init(void);
1555 extern void tcp4_proc_exit(void);
1556 #endif
1557 
1558 /* TCP af-specific functions */
1559 struct tcp_sock_af_ops {
1560 #ifdef CONFIG_TCP_MD5SIG
1561         struct tcp_md5sig_key   *(*md5_lookup) (struct sock *sk,
1562                                                 struct sock *addr_sk);
1563         int                     (*calc_md5_hash) (char *location,
1564                                                   struct tcp_md5sig_key *md5,
1565                                                   const struct sock *sk,
1566                                                   const struct request_sock *req,
1567                                                   const struct sk_buff *skb);
1568         int                     (*md5_parse) (struct sock *sk,
1569                                               char __user *optval,
1570                                               int optlen);
1571 #endif
1572 };
1573 
1574 struct tcp_request_sock_ops {
1575 #ifdef CONFIG_TCP_MD5SIG
1576         struct tcp_md5sig_key   *(*md5_lookup) (struct sock *sk,
1577                                                 struct request_sock *req);
1578         int                     (*calc_md5_hash) (char *location,
1579                                                   struct tcp_md5sig_key *md5,
1580                                                   const struct sock *sk,
1581                                                   const struct request_sock *req,
1582                                                   const struct sk_buff *skb);
1583 #endif
1584 };
1585 
1586 extern int tcpv4_offload_init(void);
1587 
1588 extern void tcp_v4_init(void);
1589 extern void tcp_init(void);
1590 
1591 #endif  /* _TCP_H */
1592 

~ [ 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