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.
6 * Definitions for the TCP module.
8 * Version: @(#)tcp.h 1.0.5 05/23/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
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.
21 #define FASTRETRANS_DEBUG 1
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>
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>
43 #include <net/tcp_states.h>
44 #include <net/inet_ecn.h>
47 #include <linux/seq_file.h>
48 #include <linux/memcontrol.h>
50 extern struct inet_hashinfo tcp_hashinfo;
52 extern struct percpu_counter tcp_orphan_count;
53 extern void tcp_time_wait(struct sock *sk, int state, int timeo);
55 #define MAX_TCP_HEADER (128 + MAX_HEADER)
56 #define MAX_TCP_OPTION_SPACE 40
59 * Never offer a window over 32767 without using window scaling. Some
60 * poor stacks do signed 16bit maths!
62 #define MAX_TCP_WINDOW 32767U
64 /* Offer an initial receive window of 10 mss. */
65 #define TCP_DEFAULT_INIT_RCVWND 10
67 /* Minimal accepted MSS. It is (60+60+8) - (20+20). */
68 #define TCP_MIN_MSS 88U
70 /* The least MTU to use for probing */
71 #define TCP_BASE_MSS 512
73 /* After receiving this amount of duplicate ACKs fast retransmit starts. */
74 #define TCP_FASTRETRANS_THRESH 3
76 /* Maximal reordering. */
77 #define TCP_MAX_REORDERING 127
79 /* Maximal number of ACKs sent quickly to accelerate slow-start. */
80 #define TCP_MAX_QUICKACKS 16U
83 #define TCP_URG_VALID 0x0100
84 #define TCP_URG_NOTYET 0x0200
85 #define TCP_URG_READ 0x0400
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.
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.
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.
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
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
126 #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
128 #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
129 #define TCP_ATO_MIN ((unsigned)(HZ/25))
131 #define TCP_DELACK_MIN 4U
132 #define TCP_ATO_MIN 4U
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.
144 #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
145 * for local resources.
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)
152 #define MAX_TCP_KEEPIDLE 32767
153 #define MAX_TCP_KEEPINTVL 32767
154 #define MAX_TCP_KEEPCNT 127
155 #define MAX_TCP_SYNCNT 127
157 #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
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.
166 #define TCP_PAWS_WINDOW 1 /* Replay window for per-host
167 * timestamps. It must be less than
168 * minimal timewait lifetime.
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
186 #define TCPOPT_FASTOPEN_MAGIC 0xF989
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)
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
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 */
218 /* TCP thin-stream limits */
219 #define TCP_THIN_LINEAR_RETRIES 6 /* After 6 linear retries, do exp. backoff */
221 /* TCP initial congestion window as per draft-hkchu-tcpm-initcwnd-01 */
222 #define TCP_INIT_CWND 10
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 */
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
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.
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!)
242 #define TFO_SERVER_ALWAYS 0x1000
244 extern struct inet_timewait_death_row tcp_death_row;
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_frto_response;
276 extern int sysctl_tcp_low_latency;
277 extern int sysctl_tcp_dma_copybreak;
278 extern int sysctl_tcp_nometrics_save;
279 extern int sysctl_tcp_moderate_rcvbuf;
280 extern int sysctl_tcp_tso_win_divisor;
281 extern int sysctl_tcp_mtu_probing;
282 extern int sysctl_tcp_base_mss;
283 extern int sysctl_tcp_workaround_signed_windows;
284 extern int sysctl_tcp_slow_start_after_idle;
285 extern int sysctl_tcp_max_ssthresh;
286 extern int sysctl_tcp_cookie_size;
287 extern int sysctl_tcp_thin_linear_timeouts;
288 extern int sysctl_tcp_thin_dupack;
289 extern int sysctl_tcp_early_retrans;
290 extern int sysctl_tcp_limit_output_bytes;
291 extern int sysctl_tcp_challenge_ack_limit;
293 extern atomic_long_t tcp_memory_allocated;
294 extern struct percpu_counter tcp_sockets_allocated;
295 extern int tcp_memory_pressure;
298 * The next routines deal with comparing 32 bit unsigned ints
299 * and worry about wraparound (automatic with unsigned arithmetic).
302 static inline bool before(__u32 seq1, __u32 seq2)
304 return (__s32)(seq1-seq2) < 0;
306 #define after(seq2, seq1) before(seq1, seq2)
308 /* is s2<=s1<=s3 ? */
309 static inline bool between(__u32 seq1, __u32 seq2, __u32 seq3)
311 return seq3 - seq2 >= seq1 - seq2;
314 static inline bool tcp_out_of_memory(struct sock *sk)
316 if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
317 sk_memory_allocated(sk) > sk_prot_mem_limits(sk, 2))
322 static inline bool tcp_too_many_orphans(struct sock *sk, int shift)
324 struct percpu_counter *ocp = sk->sk_prot->orphan_count;
325 int orphans = percpu_counter_read_positive(ocp);
327 if (orphans << shift > sysctl_tcp_max_orphans) {
328 orphans = percpu_counter_sum_positive(ocp);
329 if (orphans << shift > sysctl_tcp_max_orphans)
335 extern bool tcp_check_oom(struct sock *sk, int shift);
337 /* syncookies: remember time of last synqueue overflow */
338 static inline void tcp_synq_overflow(struct sock *sk)
340 tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies;
343 /* syncookies: no recent synqueue overflow on this listening socket? */
344 static inline bool tcp_synq_no_recent_overflow(const struct sock *sk)
346 unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp;
347 return time_after(jiffies, last_overflow + TCP_TIMEOUT_FALLBACK);
350 extern struct proto tcp_prot;
352 #define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field)
353 #define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
354 #define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
355 #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val)
356 #define TCP_ADD_STATS(net, field, val) SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val)
358 extern void tcp_init_mem(struct net *net);
360 extern void tcp_tasklet_init(void);
362 extern void tcp_v4_err(struct sk_buff *skb, u32);
364 extern void tcp_shutdown (struct sock *sk, int how);
366 extern void tcp_v4_early_demux(struct sk_buff *skb);
367 extern int tcp_v4_rcv(struct sk_buff *skb);
369 extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
370 extern int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
372 extern int tcp_sendpage(struct sock *sk, struct page *page, int offset,
373 size_t size, int flags);
374 extern void tcp_release_cb(struct sock *sk);
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,
390 static inline void tcp_dec_quickack_mode(struct sock *sk,
391 const unsigned int pkts)
393 struct inet_connection_sock *icsk = inet_csk(sk);
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;
401 icsk->icsk_ack.quick -= pkts;
406 #define TCP_ECN_QUEUE_CWR 2
407 #define TCP_ECN_DEMAND_CWR 4
408 #define TCP_ECN_SEEN 8
418 extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
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,
425 extern int tcp_child_process(struct sock *parent, struct sock *child,
426 struct sk_buff *skb);
427 extern bool tcp_use_frto(struct sock *sk);
428 extern void tcp_enter_frto(struct sock *sk);
429 extern void tcp_enter_loss(struct sock *sk, int how);
430 extern void tcp_clear_retrans(struct tcp_sock *tp);
431 extern void tcp_update_metrics(struct sock *sk);
432 extern void tcp_init_metrics(struct sock *sk);
433 extern void tcp_metrics_init(void);
434 extern bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst, bool paws_check);
435 extern bool tcp_remember_stamp(struct sock *sk);
436 extern bool tcp_tw_remember_stamp(struct inet_timewait_sock *tw);
437 extern void tcp_fetch_timewait_stamp(struct sock *sk, struct dst_entry *dst);
438 extern void tcp_disable_fack(struct tcp_sock *tp);
439 extern void tcp_close(struct sock *sk, long timeout);
440 extern void tcp_init_sock(struct sock *sk);
441 extern unsigned int tcp_poll(struct file * file, struct socket *sock,
442 struct poll_table_struct *wait);
443 extern int tcp_getsockopt(struct sock *sk, int level, int optname,
444 char __user *optval, int __user *optlen);
445 extern int tcp_setsockopt(struct sock *sk, int level, int optname,
446 char __user *optval, unsigned int optlen);
447 extern int compat_tcp_getsockopt(struct sock *sk, int level, int optname,
448 char __user *optval, int __user *optlen);
449 extern int compat_tcp_setsockopt(struct sock *sk, int level, int optname,
450 char __user *optval, unsigned int optlen);
451 extern void tcp_set_keepalive(struct sock *sk, int val);
452 extern void tcp_syn_ack_timeout(struct sock *sk, struct request_sock *req);
453 extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
454 size_t len, int nonblock, int flags, int *addr_len);
455 extern void tcp_parse_options(const struct sk_buff *skb,
456 struct tcp_options_received *opt_rx,
457 int estab, struct tcp_fastopen_cookie *foc);
458 extern const u8 *tcp_parse_md5sig_option(const struct tcphdr *th);
461 * TCP v4 functions exported for the inet6 API
464 extern void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb);
465 extern int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb);
466 extern struct sock * tcp_create_openreq_child(struct sock *sk,
467 struct request_sock *req,
468 struct sk_buff *skb);
469 extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
470 struct request_sock *req,
471 struct dst_entry *dst);
472 extern int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb);
473 extern int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr,
475 extern int tcp_connect(struct sock *sk);
476 extern struct sk_buff * tcp_make_synack(struct sock *sk, struct dst_entry *dst,
477 struct request_sock *req,
478 struct tcp_fastopen_cookie *foc);
479 extern int tcp_disconnect(struct sock *sk, int flags);
481 void tcp_connect_init(struct sock *sk);
482 void tcp_finish_connect(struct sock *sk, struct sk_buff *skb);
483 int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size);
484 void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb);
486 /* From syncookies.c */
487 extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
488 extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
489 struct ip_options *opt);
490 #ifdef CONFIG_SYN_COOKIES
491 extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
494 static inline __u32 cookie_v4_init_sequence(struct sock *sk,
502 extern __u32 cookie_init_timestamp(struct request_sock *req);
503 extern bool cookie_check_timestamp(struct tcp_options_received *opt,
504 struct net *net, bool *ecn_ok);
506 /* From net/ipv6/syncookies.c */
507 extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
508 #ifdef CONFIG_SYN_COOKIES
509 extern __u32 cookie_v6_init_sequence(struct sock *sk, const struct sk_buff *skb,
512 static inline __u32 cookie_v6_init_sequence(struct sock *sk,
521 extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
523 extern bool tcp_may_send_now(struct sock *sk);
524 extern int __tcp_retransmit_skb(struct sock *, struct sk_buff *);
525 extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
526 extern void tcp_retransmit_timer(struct sock *sk);
527 extern void tcp_xmit_retransmit_queue(struct sock *);
528 extern void tcp_simple_retransmit(struct sock *);
529 extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
530 extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
532 extern void tcp_send_probe0(struct sock *);
533 extern void tcp_send_partial(struct sock *);
534 extern int tcp_write_wakeup(struct sock *);
535 extern void tcp_send_fin(struct sock *sk);
536 extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
537 extern int tcp_send_synack(struct sock *);
538 extern bool tcp_syn_flood_action(struct sock *sk,
539 const struct sk_buff *skb,
541 extern void tcp_push_one(struct sock *, unsigned int mss_now);
542 extern void tcp_send_ack(struct sock *sk);
543 extern void tcp_send_delayed_ack(struct sock *sk);
544 extern void tcp_send_loss_probe(struct sock *sk);
545 extern bool tcp_schedule_loss_probe(struct sock *sk);
548 extern void tcp_cwnd_application_limited(struct sock *sk);
549 extern void tcp_resume_early_retransmit(struct sock *sk);
550 extern void tcp_rearm_rto(struct sock *sk);
551 extern void tcp_reset(struct sock *sk);
554 extern void tcp_init_xmit_timers(struct sock *);
555 static inline void tcp_clear_xmit_timers(struct sock *sk)
557 inet_csk_clear_xmit_timers(sk);
560 extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
561 extern unsigned int tcp_current_mss(struct sock *sk);
563 /* Bound MSS / TSO packet size with the half of the window */
564 static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize)
568 /* When peer uses tiny windows, there is no use in packetizing
569 * to sub-MSS pieces for the sake of SWS or making sure there
570 * are enough packets in the pipe for fast recovery.
572 * On the other hand, for extremely large MSS devices, handling
573 * smaller than MSS windows in this way does make sense.
575 if (tp->max_window >= 512)
576 cutoff = (tp->max_window >> 1);
578 cutoff = tp->max_window;
580 if (cutoff && pktsize > cutoff)
581 return max_t(int, cutoff, 68U - tp->tcp_header_len);
587 extern void tcp_get_info(const struct sock *, struct tcp_info *);
589 /* Read 'sendfile()'-style from a TCP socket */
590 typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
591 unsigned int, size_t);
592 extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
593 sk_read_actor_t recv_actor);
595 extern void tcp_initialize_rcv_mss(struct sock *sk);
597 extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
598 extern int tcp_mss_to_mtu(struct sock *sk, int mss);
599 extern void tcp_mtup_init(struct sock *sk);
600 extern void tcp_valid_rtt_meas(struct sock *sk, u32 seq_rtt);
601 extern void tcp_init_buffer_space(struct sock *sk);
603 static inline void tcp_bound_rto(const struct sock *sk)
605 if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX)
606 inet_csk(sk)->icsk_rto = TCP_RTO_MAX;
609 static inline u32 __tcp_set_rto(const struct tcp_sock *tp)
611 return (tp->srtt >> 3) + tp->rttvar;
614 extern void tcp_set_rto(struct sock *sk);
616 static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
618 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
619 ntohl(TCP_FLAG_ACK) |
623 static inline void tcp_fast_path_on(struct tcp_sock *tp)
625 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
628 static inline void tcp_fast_path_check(struct sock *sk)
630 struct tcp_sock *tp = tcp_sk(sk);
632 if (skb_queue_empty(&tp->out_of_order_queue) &&
634 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
636 tcp_fast_path_on(tp);
639 /* Compute the actual rto_min value */
640 static inline u32 tcp_rto_min(struct sock *sk)
642 const struct dst_entry *dst = __sk_dst_get(sk);
643 u32 rto_min = TCP_RTO_MIN;
645 if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
646 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
650 /* Compute the actual receive window we are currently advertising.
651 * Rcv_nxt can be after the window if our peer push more data
652 * than the offered window.
654 static inline u32 tcp_receive_window(const struct tcp_sock *tp)
656 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
663 /* Choose a new window, without checks for shrinking, and without
664 * scaling applied to the result. The caller does these things
665 * if necessary. This is a "raw" window selection.
667 extern u32 __tcp_select_window(struct sock *sk);
669 void tcp_send_window_probe(struct sock *sk);
671 /* TCP timestamps are only 32-bits, this causes a slight
672 * complication on 64-bit systems since we store a snapshot
673 * of jiffies in the buffer control blocks below. We decided
674 * to use only the low 32-bits of jiffies and hide the ugly
675 * casts with the following macro.
677 #define tcp_time_stamp ((__u32)(jiffies))
679 #define tcp_flag_byte(th) (((u_int8_t *)th)[13])
681 #define TCPHDR_FIN 0x01
682 #define TCPHDR_SYN 0x02
683 #define TCPHDR_RST 0x04
684 #define TCPHDR_PSH 0x08
685 #define TCPHDR_ACK 0x10
686 #define TCPHDR_URG 0x20
687 #define TCPHDR_ECE 0x40
688 #define TCPHDR_CWR 0x80
690 /* This is what the send packet queuing engine uses to pass
691 * TCP per-packet control information to the transmission code.
692 * We also store the host-order sequence numbers in here too.
693 * This is 44 bytes if IPV6 is enabled.
694 * If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately.
698 struct inet_skb_parm h4;
699 #if IS_ENABLED(CONFIG_IPV6)
700 struct inet6_skb_parm h6;
702 } header; /* For incoming frames */
703 __u32 seq; /* Starting sequence number */
704 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
705 __u32 when; /* used to compute rtt's */
706 __u8 tcp_flags; /* TCP header flags. (tcp[13]) */
708 __u8 sacked; /* State flags for SACK/FACK. */
709 #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
710 #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
711 #define TCPCB_LOST 0x04 /* SKB is lost */
712 #define TCPCB_TAGBITS 0x07 /* All tag bits */
713 #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
714 #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
716 __u8 ip_dsfield; /* IPv4 tos or IPv6 dsfield */
718 __u32 ack_seq; /* Sequence number ACK'd */
721 #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
723 /* RFC3168 : 6.1.1 SYN packets must not have ECT/ECN bits set
725 * If we receive a SYN packet with these bits set, it means a network is
726 * playing bad games with TOS bits. In order to avoid possible false congestion
727 * notifications, we disable TCP ECN negociation.
730 TCP_ECN_create_request(struct request_sock *req, const struct sk_buff *skb,
733 const struct tcphdr *th = tcp_hdr(skb);
735 if (net->ipv4.sysctl_tcp_ecn && th->ece && th->cwr &&
736 INET_ECN_is_not_ect(TCP_SKB_CB(skb)->ip_dsfield))
737 inet_rsk(req)->ecn_ok = 1;
740 /* Due to TSO, an SKB can be composed of multiple actual
741 * packets. To keep these tracked properly, we use this.
743 static inline int tcp_skb_pcount(const struct sk_buff *skb)
745 return skb_shinfo(skb)->gso_segs;
748 /* This is valid iff tcp_skb_pcount() > 1. */
749 static inline int tcp_skb_mss(const struct sk_buff *skb)
751 return skb_shinfo(skb)->gso_size;
754 /* Events passed to congestion control interface */
756 CA_EVENT_TX_START, /* first transmit when no packets in flight */
757 CA_EVENT_CWND_RESTART, /* congestion window restart */
758 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
759 CA_EVENT_FRTO, /* fast recovery timeout */
760 CA_EVENT_LOSS, /* loss timeout */
761 CA_EVENT_FAST_ACK, /* in sequence ack */
762 CA_EVENT_SLOW_ACK, /* other ack */
766 * Interface for adding new TCP congestion control handlers
768 #define TCP_CA_NAME_MAX 16
769 #define TCP_CA_MAX 128
770 #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
772 #define TCP_CONG_NON_RESTRICTED 0x1
773 #define TCP_CONG_RTT_STAMP 0x2
775 struct tcp_congestion_ops {
776 struct list_head list;
779 /* initialize private data (optional) */
780 void (*init)(struct sock *sk);
781 /* cleanup private data (optional) */
782 void (*release)(struct sock *sk);
784 /* return slow start threshold (required) */
785 u32 (*ssthresh)(struct sock *sk);
786 /* lower bound for congestion window (optional) */
787 u32 (*min_cwnd)(const struct sock *sk);
788 /* do new cwnd calculation (required) */
789 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
790 /* call before changing ca_state (optional) */
791 void (*set_state)(struct sock *sk, u8 new_state);
792 /* call when cwnd event occurs (optional) */
793 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
794 /* new value of cwnd after loss (optional) */
795 u32 (*undo_cwnd)(struct sock *sk);
796 /* hook for packet ack accounting (optional) */
797 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
798 /* get info for inet_diag (optional) */
799 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
801 char name[TCP_CA_NAME_MAX];
802 struct module *owner;
805 extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
806 extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
808 extern void tcp_init_congestion_control(struct sock *sk);
809 extern void tcp_cleanup_congestion_control(struct sock *sk);
810 extern int tcp_set_default_congestion_control(const char *name);
811 extern void tcp_get_default_congestion_control(char *name);
812 extern void tcp_get_available_congestion_control(char *buf, size_t len);
813 extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
814 extern int tcp_set_allowed_congestion_control(char *allowed);
815 extern int tcp_set_congestion_control(struct sock *sk, const char *name);
816 extern void tcp_slow_start(struct tcp_sock *tp);
817 extern void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w);
819 extern struct tcp_congestion_ops tcp_init_congestion_ops;
820 extern u32 tcp_reno_ssthresh(struct sock *sk);
821 extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
822 extern u32 tcp_reno_min_cwnd(const struct sock *sk);
823 extern struct tcp_congestion_ops tcp_reno;
825 static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
827 struct inet_connection_sock *icsk = inet_csk(sk);
829 if (icsk->icsk_ca_ops->set_state)
830 icsk->icsk_ca_ops->set_state(sk, ca_state);
831 icsk->icsk_ca_state = ca_state;
834 static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
836 const struct inet_connection_sock *icsk = inet_csk(sk);
838 if (icsk->icsk_ca_ops->cwnd_event)
839 icsk->icsk_ca_ops->cwnd_event(sk, event);
842 /* These functions determine how the current flow behaves in respect of SACK
843 * handling. SACK is negotiated with the peer, and therefore it can vary
844 * between different flows.
846 * tcp_is_sack - SACK enabled
847 * tcp_is_reno - No SACK
848 * tcp_is_fack - FACK enabled, implies SACK enabled
850 static inline int tcp_is_sack(const struct tcp_sock *tp)
852 return tp->rx_opt.sack_ok;
855 static inline bool tcp_is_reno(const struct tcp_sock *tp)
857 return !tcp_is_sack(tp);
860 static inline bool tcp_is_fack(const struct tcp_sock *tp)
862 return tp->rx_opt.sack_ok & TCP_FACK_ENABLED;
865 static inline void tcp_enable_fack(struct tcp_sock *tp)
867 tp->rx_opt.sack_ok |= TCP_FACK_ENABLED;
870 /* TCP early-retransmit (ER) is similar to but more conservative than
871 * the thin-dupack feature. Enable ER only if thin-dupack is disabled.
873 static inline void tcp_enable_early_retrans(struct tcp_sock *tp)
875 tp->do_early_retrans = sysctl_tcp_early_retrans &&
876 sysctl_tcp_early_retrans < 4 && !sysctl_tcp_thin_dupack &&
877 sysctl_tcp_reordering == 3;
880 static inline void tcp_disable_early_retrans(struct tcp_sock *tp)
882 tp->do_early_retrans = 0;
885 static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
887 return tp->sacked_out + tp->lost_out;
890 /* This determines how many packets are "in the network" to the best
891 * of our knowledge. In many cases it is conservative, but where
892 * detailed information is available from the receiver (via SACK
893 * blocks etc.) we can make more aggressive calculations.
895 * Use this for decisions involving congestion control, use just
896 * tp->packets_out to determine if the send queue is empty or not.
898 * Read this equation as:
900 * "Packets sent once on transmission queue" MINUS
901 * "Packets left network, but not honestly ACKed yet" PLUS
902 * "Packets fast retransmitted"
904 static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
906 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
909 #define TCP_INFINITE_SSTHRESH 0x7fffffff
911 static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp)
913 return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH;
916 static inline bool tcp_in_cwnd_reduction(const struct sock *sk)
918 return (TCPF_CA_CWR | TCPF_CA_Recovery) &
919 (1 << inet_csk(sk)->icsk_ca_state);
922 /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
923 * The exception is cwnd reduction phase, when cwnd is decreasing towards
926 static inline __u32 tcp_current_ssthresh(const struct sock *sk)
928 const struct tcp_sock *tp = tcp_sk(sk);
930 if (tcp_in_cwnd_reduction(sk))
931 return tp->snd_ssthresh;
933 return max(tp->snd_ssthresh,
934 ((tp->snd_cwnd >> 1) +
935 (tp->snd_cwnd >> 2)));
938 /* Use define here intentionally to get WARN_ON location shown at the caller */
939 #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
941 extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
942 extern __u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst);
944 /* The maximum number of MSS of available cwnd for which TSO defers
945 * sending if not using sysctl_tcp_tso_win_divisor.
947 static inline __u32 tcp_max_tso_deferred_mss(const struct tcp_sock *tp)
952 /* Slow start with delack produces 3 packets of burst, so that
953 * it is safe "de facto". This will be the default - same as
954 * the default reordering threshold - but if reordering increases,
955 * we must be able to allow cwnd to burst at least this much in order
956 * to not pull it back when holes are filled.
958 static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
960 return tp->reordering;
963 /* Returns end sequence number of the receiver's advertised window */
964 static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
966 return tp->snd_una + tp->snd_wnd;
968 extern bool tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
970 static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
971 const struct sk_buff *skb)
974 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
977 static inline void tcp_check_probe_timer(struct sock *sk)
979 const struct tcp_sock *tp = tcp_sk(sk);
980 const struct inet_connection_sock *icsk = inet_csk(sk);
982 if (!tp->packets_out && !icsk->icsk_pending)
983 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
984 icsk->icsk_rto, TCP_RTO_MAX);
987 static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq)
992 static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq)
998 * Calculate(/check) TCP checksum
1000 static inline __sum16 tcp_v4_check(int len, __be32 saddr,
1001 __be32 daddr, __wsum base)
1003 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
1006 static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
1008 return __skb_checksum_complete(skb);
1011 static inline bool tcp_checksum_complete(struct sk_buff *skb)
1013 return !skb_csum_unnecessary(skb) &&
1014 __tcp_checksum_complete(skb);
1017 /* Prequeue for VJ style copy to user, combined with checksumming. */
1019 static inline void tcp_prequeue_init(struct tcp_sock *tp)
1021 tp->ucopy.task = NULL;
1023 tp->ucopy.memory = 0;
1024 skb_queue_head_init(&tp->ucopy.prequeue);
1025 #ifdef CONFIG_NET_DMA
1026 tp->ucopy.dma_chan = NULL;
1027 tp->ucopy.wakeup = 0;
1028 tp->ucopy.pinned_list = NULL;
1029 tp->ucopy.dma_cookie = 0;
1033 extern bool tcp_prequeue(struct sock *sk, struct sk_buff *skb);
1038 static const char *statename[]={
1039 "Unused","Established","Syn Sent","Syn Recv",
1040 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
1041 "Close Wait","Last ACK","Listen","Closing"
1044 extern void tcp_set_state(struct sock *sk, int state);
1046 extern void tcp_done(struct sock *sk);
1048 static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
1051 rx_opt->num_sacks = 0;
1054 /* Determine a window scaling and initial window to offer. */
1055 extern void tcp_select_initial_window(int __space, __u32 mss,
1056 __u32 *rcv_wnd, __u32 *window_clamp,
1057 int wscale_ok, __u8 *rcv_wscale,
1058 __u32 init_rcv_wnd);
1060 static inline int tcp_win_from_space(int space)
1062 return sysctl_tcp_adv_win_scale<=0 ?
1063 (space>>(-sysctl_tcp_adv_win_scale)) :
1064 space - (space>>sysctl_tcp_adv_win_scale);
1067 /* Note: caller must be prepared to deal with negative returns */
1068 static inline int tcp_space(const struct sock *sk)
1070 return tcp_win_from_space(sk->sk_rcvbuf -
1071 atomic_read(&sk->sk_rmem_alloc));
1074 static inline int tcp_full_space(const struct sock *sk)
1076 return tcp_win_from_space(sk->sk_rcvbuf);
1079 static inline void tcp_openreq_init(struct request_sock *req,
1080 struct tcp_options_received *rx_opt,
1081 struct sk_buff *skb)
1083 struct inet_request_sock *ireq = inet_rsk(req);
1085 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
1087 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1088 tcp_rsk(req)->rcv_nxt = TCP_SKB_CB(skb)->seq + 1;
1089 tcp_rsk(req)->snt_synack = 0;
1090 req->mss = rx_opt->mss_clamp;
1091 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
1092 ireq->tstamp_ok = rx_opt->tstamp_ok;
1093 ireq->sack_ok = rx_opt->sack_ok;
1094 ireq->snd_wscale = rx_opt->snd_wscale;
1095 ireq->wscale_ok = rx_opt->wscale_ok;
1098 ireq->rmt_port = tcp_hdr(skb)->source;
1099 ireq->loc_port = tcp_hdr(skb)->dest;
1102 /* Compute time elapsed between SYNACK and the ACK completing 3WHS */
1103 static inline void tcp_synack_rtt_meas(struct sock *sk,
1104 struct request_sock *req)
1106 if (tcp_rsk(req)->snt_synack)
1107 tcp_valid_rtt_meas(sk,
1108 tcp_time_stamp - tcp_rsk(req)->snt_synack);
1111 extern void tcp_enter_memory_pressure(struct sock *sk);
1113 static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1115 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1118 static inline int keepalive_time_when(const struct tcp_sock *tp)
1120 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1123 static inline int keepalive_probes(const struct tcp_sock *tp)
1125 return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes;
1128 static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp)
1130 const struct inet_connection_sock *icsk = &tp->inet_conn;
1132 return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime,
1133 tcp_time_stamp - tp->rcv_tstamp);
1136 static inline int tcp_fin_time(const struct sock *sk)
1138 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1139 const int rto = inet_csk(sk)->icsk_rto;
1141 if (fin_timeout < (rto << 2) - (rto >> 1))
1142 fin_timeout = (rto << 2) - (rto >> 1);
1147 static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt,
1150 if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
1152 if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
1155 * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0,
1156 * then following tcp messages have valid values. Ignore 0 value,
1157 * or else 'negative' tsval might forbid us to accept their packets.
1159 if (!rx_opt->ts_recent)
1164 static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt,
1167 if (tcp_paws_check(rx_opt, 0))
1170 /* RST segments are not recommended to carry timestamp,
1171 and, if they do, it is recommended to ignore PAWS because
1172 "their cleanup function should take precedence over timestamps."
1173 Certainly, it is mistake. It is necessary to understand the reasons
1174 of this constraint to relax it: if peer reboots, clock may go
1175 out-of-sync and half-open connections will not be reset.
1176 Actually, the problem would be not existing if all
1177 the implementations followed draft about maintaining clock
1178 via reboots. Linux-2.2 DOES NOT!
1180 However, we can relax time bounds for RST segments to MSL.
1182 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1187 static inline void tcp_mib_init(struct net *net)
1190 TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1);
1191 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1192 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1193 TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1);
1197 static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
1199 tp->lost_skb_hint = NULL;
1200 tp->scoreboard_skb_hint = NULL;
1203 static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1205 tcp_clear_retrans_hints_partial(tp);
1206 tp->retransmit_skb_hint = NULL;
1212 union tcp_md5_addr {
1214 #if IS_ENABLED(CONFIG_IPV6)
1219 /* - key database */
1220 struct tcp_md5sig_key {
1221 struct hlist_node node;
1223 u8 family; /* AF_INET or AF_INET6 */
1224 union tcp_md5_addr addr;
1225 u8 key[TCP_MD5SIG_MAXKEYLEN];
1226 struct rcu_head rcu;
1230 struct tcp_md5sig_info {
1231 struct hlist_head head;
1232 struct rcu_head rcu;
1235 /* - pseudo header */
1236 struct tcp4_pseudohdr {
1244 struct tcp6_pseudohdr {
1245 struct in6_addr saddr;
1246 struct in6_addr daddr;
1248 __be32 protocol; /* including padding */
1251 union tcp_md5sum_block {
1252 struct tcp4_pseudohdr ip4;
1253 #if IS_ENABLED(CONFIG_IPV6)
1254 struct tcp6_pseudohdr ip6;
1258 /* - pool: digest algorithm, hash description and scratch buffer */
1259 struct tcp_md5sig_pool {
1260 struct hash_desc md5_desc;
1261 union tcp_md5sum_block md5_blk;
1265 extern int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
1266 const struct sock *sk,
1267 const struct request_sock *req,
1268 const struct sk_buff *skb);
1269 extern int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
1270 int family, const u8 *newkey,
1271 u8 newkeylen, gfp_t gfp);
1272 extern int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr,
1274 extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1275 struct sock *addr_sk);
1277 #ifdef CONFIG_TCP_MD5SIG
1278 extern struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
1279 const union tcp_md5_addr *addr, int family);
1280 #define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_key)
1282 static inline struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
1283 const union tcp_md5_addr *addr,
1288 #define tcp_twsk_md5_key(twsk) NULL
1291 extern struct tcp_md5sig_pool __percpu *tcp_alloc_md5sig_pool(struct sock *);
1292 extern void tcp_free_md5sig_pool(void);
1294 extern struct tcp_md5sig_pool *tcp_get_md5sig_pool(void);
1295 extern void tcp_put_md5sig_pool(void);
1297 extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, const struct tcphdr *);
1298 extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *,
1299 unsigned int header_len);
1300 extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
1301 const struct tcp_md5sig_key *key);
1303 /* From tcp_fastopen.c */
1304 extern void tcp_fastopen_cache_get(struct sock *sk, u16 *mss,
1305 struct tcp_fastopen_cookie *cookie,
1306 int *syn_loss, unsigned long *last_syn_loss);
1307 extern void tcp_fastopen_cache_set(struct sock *sk, u16 mss,
1308 struct tcp_fastopen_cookie *cookie,
1310 struct tcp_fastopen_request {
1311 /* Fast Open cookie. Size 0 means a cookie request */
1312 struct tcp_fastopen_cookie cookie;
1313 struct msghdr *data; /* data in MSG_FASTOPEN */
1314 u16 copied; /* queued in tcp_connect() */
1316 void tcp_free_fastopen_req(struct tcp_sock *tp);
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);
1322 #define TCP_FASTOPEN_KEY_LENGTH 16
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;
1331 /* write queue abstraction */
1332 static inline void tcp_write_queue_purge(struct sock *sk)
1334 struct sk_buff *skb;
1336 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
1337 sk_wmem_free_skb(sk, skb);
1339 tcp_clear_all_retrans_hints(tcp_sk(sk));
1342 static inline struct sk_buff *tcp_write_queue_head(const struct sock *sk)
1344 return skb_peek(&sk->sk_write_queue);
1347 static inline struct sk_buff *tcp_write_queue_tail(const struct sock *sk)
1349 return skb_peek_tail(&sk->sk_write_queue);
1352 static inline struct sk_buff *tcp_write_queue_next(const struct sock *sk,
1353 const struct sk_buff *skb)
1355 return skb_queue_next(&sk->sk_write_queue, skb);
1358 static inline struct sk_buff *tcp_write_queue_prev(const struct sock *sk,
1359 const struct sk_buff *skb)
1361 return skb_queue_prev(&sk->sk_write_queue, skb);
1364 #define tcp_for_write_queue(skb, sk) \
1365 skb_queue_walk(&(sk)->sk_write_queue, skb)
1367 #define tcp_for_write_queue_from(skb, sk) \
1368 skb_queue_walk_from(&(sk)->sk_write_queue, skb)
1370 #define tcp_for_write_queue_from_safe(skb, tmp, sk) \
1371 skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
1373 static inline struct sk_buff *tcp_send_head(const struct sock *sk)
1375 return sk->sk_send_head;
1378 static inline bool tcp_skb_is_last(const struct sock *sk,
1379 const struct sk_buff *skb)
1381 return skb_queue_is_last(&sk->sk_write_queue, skb);
1384 static inline void tcp_advance_send_head(struct sock *sk, const struct sk_buff *skb)
1386 if (tcp_skb_is_last(sk, skb))
1387 sk->sk_send_head = NULL;
1389 sk->sk_send_head = tcp_write_queue_next(sk, skb);
1392 static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1394 if (sk->sk_send_head == skb_unlinked)
1395 sk->sk_send_head = NULL;
1398 static inline void tcp_init_send_head(struct sock *sk)
1400 sk->sk_send_head = NULL;
1403 static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1405 __skb_queue_tail(&sk->sk_write_queue, skb);
1408 static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1410 __tcp_add_write_queue_tail(sk, skb);
1412 /* Queue it, remembering where we must start sending. */
1413 if (sk->sk_send_head == NULL) {
1414 sk->sk_send_head = skb;
1416 if (tcp_sk(sk)->highest_sack == NULL)
1417 tcp_sk(sk)->highest_sack = skb;
1421 static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1423 __skb_queue_head(&sk->sk_write_queue, skb);
1426 /* Insert buff after skb on the write queue of sk. */
1427 static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1428 struct sk_buff *buff,
1431 __skb_queue_after(&sk->sk_write_queue, skb, buff);
1434 /* Insert new before skb on the write queue of sk. */
1435 static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1436 struct sk_buff *skb,
1439 __skb_queue_before(&sk->sk_write_queue, skb, new);
1441 if (sk->sk_send_head == skb)
1442 sk->sk_send_head = new;
1445 static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1447 __skb_unlink(skb, &sk->sk_write_queue);
1450 static inline bool tcp_write_queue_empty(struct sock *sk)
1452 return skb_queue_empty(&sk->sk_write_queue);
1455 static inline void tcp_push_pending_frames(struct sock *sk)
1457 if (tcp_send_head(sk)) {
1458 struct tcp_sock *tp = tcp_sk(sk);
1460 __tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle);
1464 /* Start sequence of the skb just after the highest skb with SACKed
1465 * bit, valid only if sacked_out > 0 or when the caller has ensured
1466 * validity by itself.
1468 static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1470 if (!tp->sacked_out)
1473 if (tp->highest_sack == NULL)
1476 return TCP_SKB_CB(tp->highest_sack)->seq;
1479 static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1481 tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1482 tcp_write_queue_next(sk, skb);
1485 static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1487 return tcp_sk(sk)->highest_sack;
1490 static inline void tcp_highest_sack_reset(struct sock *sk)
1492 tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1495 /* Called when old skb is about to be deleted (to be combined with new skb) */
1496 static inline void tcp_highest_sack_combine(struct sock *sk,
1497 struct sk_buff *old,
1498 struct sk_buff *new)
1500 if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1501 tcp_sk(sk)->highest_sack = new;
1504 /* Determines whether this is a thin stream (which may suffer from
1505 * increased latency). Used to trigger latency-reducing mechanisms.
1507 static inline bool tcp_stream_is_thin(struct tcp_sock *tp)
1509 return tp->packets_out < 4 && !tcp_in_initial_slowstart(tp);
1513 enum tcp_seq_states {
1514 TCP_SEQ_STATE_LISTENING,
1515 TCP_SEQ_STATE_OPENREQ,
1516 TCP_SEQ_STATE_ESTABLISHED,
1517 TCP_SEQ_STATE_TIME_WAIT,
1520 int tcp_seq_open(struct inode *inode, struct file *file);
1522 struct tcp_seq_afinfo {
1525 const struct file_operations *seq_fops;
1526 struct seq_operations seq_ops;
1529 struct tcp_iter_state {
1530 struct seq_net_private p;
1532 enum tcp_seq_states state;
1533 struct sock *syn_wait_sk;
1534 int bucket, offset, sbucket, num;
1539 extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo);
1540 extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo);
1542 extern struct request_sock_ops tcp_request_sock_ops;
1543 extern struct request_sock_ops tcp6_request_sock_ops;
1545 extern void tcp_v4_destroy_sock(struct sock *sk);
1547 extern int tcp_v4_gso_send_check(struct sk_buff *skb);
1548 extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb,
1549 netdev_features_t features);
1550 extern struct sk_buff **tcp_gro_receive(struct sk_buff **head,
1551 struct sk_buff *skb);
1552 extern struct sk_buff **tcp4_gro_receive(struct sk_buff **head,
1553 struct sk_buff *skb);
1554 extern int tcp_gro_complete(struct sk_buff *skb);
1555 extern int tcp4_gro_complete(struct sk_buff *skb);
1557 #ifdef CONFIG_PROC_FS
1558 extern int tcp4_proc_init(void);
1559 extern void tcp4_proc_exit(void);
1562 /* TCP af-specific functions */
1563 struct tcp_sock_af_ops {
1564 #ifdef CONFIG_TCP_MD5SIG
1565 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1566 struct sock *addr_sk);
1567 int (*calc_md5_hash) (char *location,
1568 struct tcp_md5sig_key *md5,
1569 const struct sock *sk,
1570 const struct request_sock *req,
1571 const struct sk_buff *skb);
1572 int (*md5_parse) (struct sock *sk,
1573 char __user *optval,
1578 struct tcp_request_sock_ops {
1579 #ifdef CONFIG_TCP_MD5SIG
1580 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1581 struct request_sock *req);
1582 int (*calc_md5_hash) (char *location,
1583 struct tcp_md5sig_key *md5,
1584 const struct sock *sk,
1585 const struct request_sock *req,
1586 const struct sk_buff *skb);
1590 extern void tcp_v4_init(void);
1591 extern void tcp_init(void);