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 20 mss. */
65 #define TCP_DEFAULT_INIT_RCVWND 20
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 10 /*
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 9 /* 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)(60*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_low_latency
;
276 extern int sysctl_tcp_dma_copybreak
;
277 extern int sysctl_tcp_nometrics_save
;
278 extern int sysctl_tcp_moderate_rcvbuf
;
279 extern int sysctl_tcp_tso_win_divisor
;
280 extern int sysctl_tcp_mtu_probing
;
281 extern int sysctl_tcp_base_mss
;
282 extern int sysctl_tcp_workaround_signed_windows
;
283 extern int sysctl_tcp_slow_start_after_idle
;
284 extern int sysctl_tcp_max_ssthresh
;
285 extern int sysctl_tcp_thin_linear_timeouts
;
286 extern int sysctl_tcp_thin_dupack
;
287 extern int sysctl_tcp_early_retrans
;
288 extern int sysctl_tcp_limit_output_bytes
;
289 extern int sysctl_tcp_challenge_ack_limit
;
290 extern int sysctl_tcp_default_init_rwnd
;
291 extern int sysctl_tcp_min_tso_segs
;
292 extern int sysctl_tcp_default_init_rwnd
;
293 extern int sysctl_tcp_rto_min
;
294 extern int sysctl_tcp_rto_max
;
295 extern atomic_long_t tcp_memory_allocated
;
296 extern struct percpu_counter tcp_sockets_allocated
;
297 extern int tcp_memory_pressure
;
300 * The next routines deal with comparing 32 bit unsigned ints
301 * and worry about wraparound (automatic with unsigned arithmetic).
304 static inline bool before(__u32 seq1
, __u32 seq2
)
306 return (__s32
)(seq1
-seq2
) < 0;
308 #define after(seq2, seq1) before(seq1, seq2)
310 /* is s2<=s1<=s3 ? */
311 static inline bool between(__u32 seq1
, __u32 seq2
, __u32 seq3
)
313 return seq3
- seq2
>= seq1
- seq2
;
316 static inline bool tcp_out_of_memory(struct sock
*sk
)
318 if (sk
->sk_wmem_queued
> SOCK_MIN_SNDBUF
&&
319 sk_memory_allocated(sk
) > sk_prot_mem_limits(sk
, 2))
324 static inline bool tcp_too_many_orphans(struct sock
*sk
, int shift
)
326 struct percpu_counter
*ocp
= sk
->sk_prot
->orphan_count
;
327 int orphans
= percpu_counter_read_positive(ocp
);
329 if (orphans
<< shift
> sysctl_tcp_max_orphans
) {
330 orphans
= percpu_counter_sum_positive(ocp
);
331 if (orphans
<< shift
> sysctl_tcp_max_orphans
)
337 extern bool tcp_check_oom(struct sock
*sk
, int shift
);
339 /* syncookies: remember time of last synqueue overflow */
340 static inline void tcp_synq_overflow(struct sock
*sk
)
342 tcp_sk(sk
)->rx_opt
.ts_recent_stamp
= jiffies
;
345 /* syncookies: no recent synqueue overflow on this listening socket? */
346 static inline bool tcp_synq_no_recent_overflow(const struct sock
*sk
)
348 unsigned long last_overflow
= tcp_sk(sk
)->rx_opt
.ts_recent_stamp
;
349 return time_after(jiffies
, last_overflow
+ TCP_TIMEOUT_FALLBACK
);
352 extern struct proto tcp_prot
;
354 #define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field)
355 #define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
356 #define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
357 #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val)
358 #define TCP_ADD_STATS(net, field, val) SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val)
360 extern void tcp_init_mem(struct net
*net
);
362 extern void tcp_tasklet_init(void);
364 extern void tcp_v4_err(struct sk_buff
*skb
, u32
);
366 extern void tcp_shutdown (struct sock
*sk
, int how
);
368 extern void tcp_v4_early_demux(struct sk_buff
*skb
);
369 extern int tcp_v4_rcv(struct sk_buff
*skb
);
371 extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock
*tw
);
372 extern int tcp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
374 extern int tcp_sendpage(struct sock
*sk
, struct page
*page
, int offset
,
375 size_t size
, int flags
);
376 extern void tcp_release_cb(struct sock
*sk
);
377 extern void tcp_wfree(struct sk_buff
*skb
);
378 extern void tcp_write_timer_handler(struct sock
*sk
);
379 extern void tcp_delack_timer_handler(struct sock
*sk
);
380 extern int tcp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
);
381 extern int tcp_rcv_state_process(struct sock
*sk
, struct sk_buff
*skb
,
382 const struct tcphdr
*th
, unsigned int len
);
383 extern int tcp_rcv_established(struct sock
*sk
, struct sk_buff
*skb
,
384 const struct tcphdr
*th
, unsigned int len
);
385 extern void tcp_rcv_space_adjust(struct sock
*sk
);
386 extern void tcp_cleanup_rbuf(struct sock
*sk
, int copied
);
387 extern int tcp_twsk_unique(struct sock
*sk
, struct sock
*sktw
, void *twp
);
388 extern void tcp_twsk_destructor(struct sock
*sk
);
389 extern ssize_t
tcp_splice_read(struct socket
*sk
, loff_t
*ppos
,
390 struct pipe_inode_info
*pipe
, size_t len
,
393 static inline void tcp_dec_quickack_mode(struct sock
*sk
,
394 const unsigned int pkts
)
396 struct inet_connection_sock
*icsk
= inet_csk(sk
);
398 if (icsk
->icsk_ack
.quick
) {
399 if (pkts
>= icsk
->icsk_ack
.quick
) {
400 icsk
->icsk_ack
.quick
= 0;
401 /* Leaving quickack mode we deflate ATO. */
402 icsk
->icsk_ack
.ato
= TCP_ATO_MIN
;
404 icsk
->icsk_ack
.quick
-= pkts
;
409 #define TCP_ECN_QUEUE_CWR 2
410 #define TCP_ECN_DEMAND_CWR 4
411 #define TCP_ECN_SEEN 8
421 extern enum tcp_tw_status
tcp_timewait_state_process(struct inet_timewait_sock
*tw
,
423 const struct tcphdr
*th
);
424 extern struct sock
* tcp_check_req(struct sock
*sk
,struct sk_buff
*skb
,
425 struct request_sock
*req
,
426 struct request_sock
**prev
,
428 extern int tcp_child_process(struct sock
*parent
, struct sock
*child
,
429 struct sk_buff
*skb
);
430 extern void tcp_enter_loss(struct sock
*sk
, int how
);
431 extern void tcp_clear_retrans(struct tcp_sock
*tp
);
432 extern void tcp_update_metrics(struct sock
*sk
);
433 extern void tcp_init_metrics(struct sock
*sk
);
434 extern void tcp_metrics_init(void);
435 extern bool tcp_peer_is_proven(struct request_sock
*req
, struct dst_entry
*dst
, bool paws_check
);
436 extern bool tcp_remember_stamp(struct sock
*sk
);
437 extern bool tcp_tw_remember_stamp(struct inet_timewait_sock
*tw
);
438 extern void tcp_fetch_timewait_stamp(struct sock
*sk
, struct dst_entry
*dst
);
439 extern void tcp_disable_fack(struct tcp_sock
*tp
);
440 extern void tcp_close(struct sock
*sk
, long timeout
);
441 extern void tcp_init_sock(struct sock
*sk
);
442 extern unsigned int tcp_poll(struct file
* file
, struct socket
*sock
,
443 struct poll_table_struct
*wait
);
444 extern int tcp_getsockopt(struct sock
*sk
, int level
, int optname
,
445 char __user
*optval
, int __user
*optlen
);
446 extern int tcp_setsockopt(struct sock
*sk
, int level
, int optname
,
447 char __user
*optval
, unsigned int optlen
);
448 extern int compat_tcp_getsockopt(struct sock
*sk
, int level
, int optname
,
449 char __user
*optval
, int __user
*optlen
);
450 extern int compat_tcp_setsockopt(struct sock
*sk
, int level
, int optname
,
451 char __user
*optval
, unsigned int optlen
);
452 extern void tcp_set_keepalive(struct sock
*sk
, int val
);
453 extern void tcp_syn_ack_timeout(struct sock
*sk
, struct request_sock
*req
);
454 extern int tcp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
455 size_t len
, int nonblock
, int flags
, int *addr_len
);
456 extern void tcp_parse_options(const struct sk_buff
*skb
,
457 struct tcp_options_received
*opt_rx
,
458 int estab
, struct tcp_fastopen_cookie
*foc
);
459 extern const u8
*tcp_parse_md5sig_option(const struct tcphdr
*th
);
462 * TCP v4 functions exported for the inet6 API
465 extern void tcp_v4_send_check(struct sock
*sk
, struct sk_buff
*skb
);
466 void tcp_v4_mtu_reduced(struct sock
*sk
);
467 extern int tcp_v4_conn_request(struct sock
*sk
, struct sk_buff
*skb
);
468 extern struct sock
* tcp_create_openreq_child(struct sock
*sk
,
469 struct request_sock
*req
,
470 struct sk_buff
*skb
);
471 extern struct sock
* tcp_v4_syn_recv_sock(struct sock
*sk
, struct sk_buff
*skb
,
472 struct request_sock
*req
,
473 struct dst_entry
*dst
);
474 extern int tcp_v4_do_rcv(struct sock
*sk
, struct sk_buff
*skb
);
475 extern int tcp_v4_connect(struct sock
*sk
, struct sockaddr
*uaddr
,
477 extern int tcp_connect(struct sock
*sk
);
478 extern struct sk_buff
* tcp_make_synack(struct sock
*sk
, struct dst_entry
*dst
,
479 struct request_sock
*req
,
480 struct tcp_fastopen_cookie
*foc
);
481 extern int tcp_disconnect(struct sock
*sk
, int flags
);
483 void tcp_connect_init(struct sock
*sk
);
484 void tcp_finish_connect(struct sock
*sk
, struct sk_buff
*skb
);
485 int tcp_send_rcvq(struct sock
*sk
, struct msghdr
*msg
, size_t size
);
486 void inet_sk_rx_dst_set(struct sock
*sk
, const struct sk_buff
*skb
);
488 /* From syncookies.c */
489 extern __u32 syncookie_secret
[2][16-4+SHA_DIGEST_WORDS
];
490 extern struct sock
*cookie_v4_check(struct sock
*sk
, struct sk_buff
*skb
,
491 struct ip_options
*opt
);
492 #ifdef CONFIG_SYN_COOKIES
493 extern __u32
cookie_v4_init_sequence(struct sock
*sk
, struct sk_buff
*skb
,
496 static inline __u32
cookie_v4_init_sequence(struct sock
*sk
,
504 extern __u32
cookie_init_timestamp(struct request_sock
*req
);
505 extern bool cookie_check_timestamp(struct tcp_options_received
*opt
,
506 struct net
*net
, bool *ecn_ok
);
508 /* From net/ipv6/syncookies.c */
509 extern struct sock
*cookie_v6_check(struct sock
*sk
, struct sk_buff
*skb
);
510 #ifdef CONFIG_SYN_COOKIES
511 extern __u32
cookie_v6_init_sequence(struct sock
*sk
, const struct sk_buff
*skb
,
514 static inline __u32
cookie_v6_init_sequence(struct sock
*sk
,
523 extern void __tcp_push_pending_frames(struct sock
*sk
, unsigned int cur_mss
,
525 extern bool tcp_may_send_now(struct sock
*sk
);
526 extern int __tcp_retransmit_skb(struct sock
*, struct sk_buff
*);
527 extern int tcp_retransmit_skb(struct sock
*, struct sk_buff
*);
528 extern void tcp_retransmit_timer(struct sock
*sk
);
529 extern void tcp_xmit_retransmit_queue(struct sock
*);
530 extern void tcp_simple_retransmit(struct sock
*);
531 extern int tcp_trim_head(struct sock
*, struct sk_buff
*, u32
);
532 extern int tcp_fragment(struct sock
*, struct sk_buff
*, u32
, unsigned int);
534 extern void tcp_send_probe0(struct sock
*);
535 extern void tcp_send_partial(struct sock
*);
536 extern int tcp_write_wakeup(struct sock
*);
537 extern void tcp_send_fin(struct sock
*sk
);
538 extern void tcp_send_active_reset(struct sock
*sk
, gfp_t priority
);
539 extern int tcp_send_synack(struct sock
*);
540 extern bool tcp_syn_flood_action(struct sock
*sk
,
541 const struct sk_buff
*skb
,
543 extern void tcp_push_one(struct sock
*, unsigned int mss_now
);
544 extern void tcp_send_ack(struct sock
*sk
);
545 extern void tcp_send_delayed_ack(struct sock
*sk
);
546 extern void tcp_send_loss_probe(struct sock
*sk
);
547 extern bool tcp_schedule_loss_probe(struct sock
*sk
);
550 extern void tcp_cwnd_application_limited(struct sock
*sk
);
551 extern void tcp_resume_early_retransmit(struct sock
*sk
);
552 extern void tcp_rearm_rto(struct sock
*sk
);
553 extern void tcp_reset(struct sock
*sk
);
556 extern void tcp_init_xmit_timers(struct sock
*);
557 static inline void tcp_clear_xmit_timers(struct sock
*sk
)
559 inet_csk_clear_xmit_timers(sk
);
562 extern unsigned int tcp_sync_mss(struct sock
*sk
, u32 pmtu
);
563 extern unsigned int tcp_current_mss(struct sock
*sk
);
565 /* Bound MSS / TSO packet size with the half of the window */
566 static inline int tcp_bound_to_half_wnd(struct tcp_sock
*tp
, int pktsize
)
570 /* When peer uses tiny windows, there is no use in packetizing
571 * to sub-MSS pieces for the sake of SWS or making sure there
572 * are enough packets in the pipe for fast recovery.
574 * On the other hand, for extremely large MSS devices, handling
575 * smaller than MSS windows in this way does make sense.
577 if (tp
->max_window
>= 512)
578 cutoff
= (tp
->max_window
>> 1);
580 cutoff
= tp
->max_window
;
582 if (cutoff
&& pktsize
> cutoff
)
583 return max_t(int, cutoff
, 68U - tp
->tcp_header_len
);
589 extern void tcp_get_info(const struct sock
*, struct tcp_info
*);
591 /* Read 'sendfile()'-style from a TCP socket */
592 typedef int (*sk_read_actor_t
)(read_descriptor_t
*, struct sk_buff
*,
593 unsigned int, size_t);
594 extern int tcp_read_sock(struct sock
*sk
, read_descriptor_t
*desc
,
595 sk_read_actor_t recv_actor
);
597 extern void tcp_initialize_rcv_mss(struct sock
*sk
);
599 extern int tcp_mtu_to_mss(struct sock
*sk
, int pmtu
);
600 extern int tcp_mss_to_mtu(struct sock
*sk
, int mss
);
601 extern void tcp_mtup_init(struct sock
*sk
);
602 extern void tcp_valid_rtt_meas(struct sock
*sk
, u32 seq_rtt
);
603 extern void tcp_init_buffer_space(struct sock
*sk
);
605 static inline void tcp_bound_rto(const struct sock
*sk
)
607 if (inet_csk(sk
)->icsk_rto
> sysctl_tcp_rto_max
)
608 inet_csk(sk
)->icsk_rto
= sysctl_tcp_rto_max
;
611 static inline u32
__tcp_set_rto(const struct tcp_sock
*tp
)
613 return (tp
->srtt
>> 3) + tp
->rttvar
;
616 extern void tcp_set_rto(struct sock
*sk
);
618 static inline void __tcp_fast_path_on(struct tcp_sock
*tp
, u32 snd_wnd
)
620 tp
->pred_flags
= htonl((tp
->tcp_header_len
<< 26) |
621 ntohl(TCP_FLAG_ACK
) |
625 static inline void tcp_fast_path_on(struct tcp_sock
*tp
)
627 __tcp_fast_path_on(tp
, tp
->snd_wnd
>> tp
->rx_opt
.snd_wscale
);
630 static inline void tcp_fast_path_check(struct sock
*sk
)
632 struct tcp_sock
*tp
= tcp_sk(sk
);
634 if (skb_queue_empty(&tp
->out_of_order_queue
) &&
636 atomic_read(&sk
->sk_rmem_alloc
) < sk
->sk_rcvbuf
&&
638 tcp_fast_path_on(tp
);
641 /* Compute the actual rto_min value */
642 static inline u32
tcp_rto_min(struct sock
*sk
)
644 const struct dst_entry
*dst
= __sk_dst_get(sk
);
645 u32 rto_min
= sysctl_tcp_rto_min
;
647 if (dst
&& dst_metric_locked(dst
, RTAX_RTO_MIN
))
648 rto_min
= dst_metric_rtt(dst
, RTAX_RTO_MIN
);
652 /* Compute the actual receive window we are currently advertising.
653 * Rcv_nxt can be after the window if our peer push more data
654 * than the offered window.
656 static inline u32
tcp_receive_window(const struct tcp_sock
*tp
)
658 s32 win
= tp
->rcv_wup
+ tp
->rcv_wnd
- tp
->rcv_nxt
;
665 /* Choose a new window, without checks for shrinking, and without
666 * scaling applied to the result. The caller does these things
667 * if necessary. This is a "raw" window selection.
669 extern u32
__tcp_select_window(struct sock
*sk
);
671 void tcp_send_window_probe(struct sock
*sk
);
673 /* TCP timestamps are only 32-bits, this causes a slight
674 * complication on 64-bit systems since we store a snapshot
675 * of jiffies in the buffer control blocks below. We decided
676 * to use only the low 32-bits of jiffies and hide the ugly
677 * casts with the following macro.
679 #define tcp_time_stamp ((__u32)(jiffies))
681 #define tcp_flag_byte(th) (((u_int8_t *)th)[13])
683 #define TCPHDR_FIN 0x01
684 #define TCPHDR_SYN 0x02
685 #define TCPHDR_RST 0x04
686 #define TCPHDR_PSH 0x08
687 #define TCPHDR_ACK 0x10
688 #define TCPHDR_URG 0x20
689 #define TCPHDR_ECE 0x40
690 #define TCPHDR_CWR 0x80
692 /* This is what the send packet queuing engine uses to pass
693 * TCP per-packet control information to the transmission code.
694 * We also store the host-order sequence numbers in here too.
695 * This is 44 bytes if IPV6 is enabled.
696 * If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately.
700 struct inet_skb_parm h4
;
701 #if IS_ENABLED(CONFIG_IPV6)
702 struct inet6_skb_parm h6
;
704 } header
; /* For incoming frames */
705 __u32 seq
; /* Starting sequence number */
706 __u32 end_seq
; /* SEQ + FIN + SYN + datalen */
707 __u32 when
; /* used to compute rtt's */
708 __u8 tcp_flags
; /* TCP header flags. (tcp[13]) */
710 __u8 sacked
; /* State flags for SACK/FACK. */
711 #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
712 #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
713 #define TCPCB_LOST 0x04 /* SKB is lost */
714 #define TCPCB_TAGBITS 0x07 /* All tag bits */
715 #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
716 #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
718 __u8 ip_dsfield
; /* IPv4 tos or IPv6 dsfield */
720 __u32 ack_seq
; /* Sequence number ACK'd */
723 #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
725 /* RFC3168 : 6.1.1 SYN packets must not have ECT/ECN bits set
727 * If we receive a SYN packet with these bits set, it means a network is
728 * playing bad games with TOS bits. In order to avoid possible false congestion
729 * notifications, we disable TCP ECN negociation.
732 TCP_ECN_create_request(struct request_sock
*req
, const struct sk_buff
*skb
,
735 const struct tcphdr
*th
= tcp_hdr(skb
);
737 if (net
->ipv4
.sysctl_tcp_ecn
&& th
->ece
&& th
->cwr
&&
738 INET_ECN_is_not_ect(TCP_SKB_CB(skb
)->ip_dsfield
))
739 inet_rsk(req
)->ecn_ok
= 1;
742 /* Due to TSO, an SKB can be composed of multiple actual
743 * packets. To keep these tracked properly, we use this.
745 static inline int tcp_skb_pcount(const struct sk_buff
*skb
)
747 return skb_shinfo(skb
)->gso_segs
;
750 /* This is valid iff tcp_skb_pcount() > 1. */
751 static inline int tcp_skb_mss(const struct sk_buff
*skb
)
753 return skb_shinfo(skb
)->gso_size
;
756 /* Events passed to congestion control interface */
758 CA_EVENT_TX_START
, /* first transmit when no packets in flight */
759 CA_EVENT_CWND_RESTART
, /* congestion window restart */
760 CA_EVENT_COMPLETE_CWR
, /* end of congestion recovery */
761 CA_EVENT_LOSS
, /* loss timeout */
762 CA_EVENT_FAST_ACK
, /* in sequence ack */
763 CA_EVENT_SLOW_ACK
, /* other ack */
767 * Interface for adding new TCP congestion control handlers
769 #define TCP_CA_NAME_MAX 16
770 #define TCP_CA_MAX 128
771 #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
773 #define TCP_CONG_NON_RESTRICTED 0x1
774 #define TCP_CONG_RTT_STAMP 0x2
776 struct tcp_congestion_ops
{
777 struct list_head list
;
780 /* initialize private data (optional) */
781 void (*init
)(struct sock
*sk
);
782 /* cleanup private data (optional) */
783 void (*release
)(struct sock
*sk
);
785 /* return slow start threshold (required) */
786 u32 (*ssthresh
)(struct sock
*sk
);
787 /* lower bound for congestion window (optional) */
788 u32 (*min_cwnd
)(const struct sock
*sk
);
789 /* do new cwnd calculation (required) */
790 void (*cong_avoid
)(struct sock
*sk
, u32 ack
, u32 in_flight
);
791 /* call before changing ca_state (optional) */
792 void (*set_state
)(struct sock
*sk
, u8 new_state
);
793 /* call when cwnd event occurs (optional) */
794 void (*cwnd_event
)(struct sock
*sk
, enum tcp_ca_event ev
);
795 /* new value of cwnd after loss (optional) */
796 u32 (*undo_cwnd
)(struct sock
*sk
);
797 /* hook for packet ack accounting (optional) */
798 void (*pkts_acked
)(struct sock
*sk
, u32 num_acked
, s32 rtt_us
);
799 /* get info for inet_diag (optional) */
800 void (*get_info
)(struct sock
*sk
, u32 ext
, struct sk_buff
*skb
);
802 char name
[TCP_CA_NAME_MAX
];
803 struct module
*owner
;
806 extern int tcp_register_congestion_control(struct tcp_congestion_ops
*type
);
807 extern void tcp_unregister_congestion_control(struct tcp_congestion_ops
*type
);
809 extern void tcp_init_congestion_control(struct sock
*sk
);
810 extern void tcp_cleanup_congestion_control(struct sock
*sk
);
811 extern int tcp_set_default_congestion_control(const char *name
);
812 extern void tcp_get_default_congestion_control(char *name
);
813 extern void tcp_get_available_congestion_control(char *buf
, size_t len
);
814 extern void tcp_get_allowed_congestion_control(char *buf
, size_t len
);
815 extern int tcp_set_allowed_congestion_control(char *allowed
);
816 extern int tcp_set_congestion_control(struct sock
*sk
, const char *name
);
817 extern void tcp_slow_start(struct tcp_sock
*tp
);
818 extern void tcp_cong_avoid_ai(struct tcp_sock
*tp
, u32 w
);
820 extern struct tcp_congestion_ops tcp_init_congestion_ops
;
821 extern u32
tcp_reno_ssthresh(struct sock
*sk
);
822 extern void tcp_reno_cong_avoid(struct sock
*sk
, u32 ack
, u32 in_flight
);
823 extern u32
tcp_reno_min_cwnd(const struct sock
*sk
);
824 extern struct tcp_congestion_ops tcp_reno
;
826 static inline void tcp_set_ca_state(struct sock
*sk
, const u8 ca_state
)
828 struct inet_connection_sock
*icsk
= inet_csk(sk
);
830 if (icsk
->icsk_ca_ops
->set_state
)
831 icsk
->icsk_ca_ops
->set_state(sk
, ca_state
);
832 icsk
->icsk_ca_state
= ca_state
;
835 static inline void tcp_ca_event(struct sock
*sk
, const enum tcp_ca_event event
)
837 const struct inet_connection_sock
*icsk
= inet_csk(sk
);
839 if (icsk
->icsk_ca_ops
->cwnd_event
)
840 icsk
->icsk_ca_ops
->cwnd_event(sk
, event
);
843 /* These functions determine how the current flow behaves in respect of SACK
844 * handling. SACK is negotiated with the peer, and therefore it can vary
845 * between different flows.
847 * tcp_is_sack - SACK enabled
848 * tcp_is_reno - No SACK
849 * tcp_is_fack - FACK enabled, implies SACK enabled
851 static inline int tcp_is_sack(const struct tcp_sock
*tp
)
853 return tp
->rx_opt
.sack_ok
;
856 static inline bool tcp_is_reno(const struct tcp_sock
*tp
)
858 return !tcp_is_sack(tp
);
861 static inline bool tcp_is_fack(const struct tcp_sock
*tp
)
863 return tp
->rx_opt
.sack_ok
& TCP_FACK_ENABLED
;
866 static inline void tcp_enable_fack(struct tcp_sock
*tp
)
868 tp
->rx_opt
.sack_ok
|= TCP_FACK_ENABLED
;
871 /* TCP early-retransmit (ER) is similar to but more conservative than
872 * the thin-dupack feature. Enable ER only if thin-dupack is disabled.
874 static inline void tcp_enable_early_retrans(struct tcp_sock
*tp
)
876 tp
->do_early_retrans
= sysctl_tcp_early_retrans
&&
877 sysctl_tcp_early_retrans
< 4 && !sysctl_tcp_thin_dupack
&&
878 sysctl_tcp_reordering
== 3;
881 static inline void tcp_disable_early_retrans(struct tcp_sock
*tp
)
883 tp
->do_early_retrans
= 0;
886 static inline unsigned int tcp_left_out(const struct tcp_sock
*tp
)
888 return tp
->sacked_out
+ tp
->lost_out
;
891 /* This determines how many packets are "in the network" to the best
892 * of our knowledge. In many cases it is conservative, but where
893 * detailed information is available from the receiver (via SACK
894 * blocks etc.) we can make more aggressive calculations.
896 * Use this for decisions involving congestion control, use just
897 * tp->packets_out to determine if the send queue is empty or not.
899 * Read this equation as:
901 * "Packets sent once on transmission queue" MINUS
902 * "Packets left network, but not honestly ACKed yet" PLUS
903 * "Packets fast retransmitted"
905 static inline unsigned int tcp_packets_in_flight(const struct tcp_sock
*tp
)
907 return tp
->packets_out
- tcp_left_out(tp
) + tp
->retrans_out
;
910 #define TCP_INFINITE_SSTHRESH 0x7fffffff
912 static inline bool tcp_in_initial_slowstart(const struct tcp_sock
*tp
)
914 return tp
->snd_ssthresh
>= TCP_INFINITE_SSTHRESH
;
917 static inline bool tcp_in_cwnd_reduction(const struct sock
*sk
)
919 return (TCPF_CA_CWR
| TCPF_CA_Recovery
) &
920 (1 << inet_csk(sk
)->icsk_ca_state
);
923 /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
924 * The exception is cwnd reduction phase, when cwnd is decreasing towards
927 static inline __u32
tcp_current_ssthresh(const struct sock
*sk
)
929 const struct tcp_sock
*tp
= tcp_sk(sk
);
931 if (tcp_in_cwnd_reduction(sk
))
932 return tp
->snd_ssthresh
;
934 return max(tp
->snd_ssthresh
,
935 ((tp
->snd_cwnd
>> 1) +
936 (tp
->snd_cwnd
>> 2)));
939 /* Use define here intentionally to get WARN_ON location shown at the caller */
940 #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
942 extern void tcp_enter_cwr(struct sock
*sk
, const int set_ssthresh
);
943 extern __u32
tcp_init_cwnd(const struct tcp_sock
*tp
, const struct dst_entry
*dst
);
945 /* The maximum number of MSS of available cwnd for which TSO defers
946 * sending if not using sysctl_tcp_tso_win_divisor.
948 static inline __u32
tcp_max_tso_deferred_mss(const struct tcp_sock
*tp
)
953 /* Slow start with delack produces 3 packets of burst, so that
954 * it is safe "de facto". This will be the default - same as
955 * the default reordering threshold - but if reordering increases,
956 * we must be able to allow cwnd to burst at least this much in order
957 * to not pull it back when holes are filled.
959 static __inline__ __u32
tcp_max_burst(const struct tcp_sock
*tp
)
961 return tp
->reordering
;
964 /* Returns end sequence number of the receiver's advertised window */
965 static inline u32
tcp_wnd_end(const struct tcp_sock
*tp
)
967 return tp
->snd_una
+ tp
->snd_wnd
;
969 extern bool tcp_is_cwnd_limited(const struct sock
*sk
, u32 in_flight
);
971 static inline void tcp_minshall_update(struct tcp_sock
*tp
, unsigned int mss
,
972 const struct sk_buff
*skb
)
975 tp
->snd_sml
= TCP_SKB_CB(skb
)->end_seq
;
978 static inline void tcp_check_probe_timer(struct sock
*sk
)
980 const struct tcp_sock
*tp
= tcp_sk(sk
);
981 const struct inet_connection_sock
*icsk
= inet_csk(sk
);
983 if (!tp
->packets_out
&& !icsk
->icsk_pending
)
984 inet_csk_reset_xmit_timer(sk
, ICSK_TIME_PROBE0
,
985 icsk
->icsk_rto
, sysctl_tcp_rto_max
);
988 static inline void tcp_init_wl(struct tcp_sock
*tp
, u32 seq
)
993 static inline void tcp_update_wl(struct tcp_sock
*tp
, u32 seq
)
999 * Calculate(/check) TCP checksum
1001 static inline __sum16
tcp_v4_check(int len
, __be32 saddr
,
1002 __be32 daddr
, __wsum base
)
1004 return csum_tcpudp_magic(saddr
,daddr
,len
,IPPROTO_TCP
,base
);
1007 static inline __sum16
__tcp_checksum_complete(struct sk_buff
*skb
)
1009 return __skb_checksum_complete(skb
);
1012 static inline bool tcp_checksum_complete(struct sk_buff
*skb
)
1014 return !skb_csum_unnecessary(skb
) &&
1015 __tcp_checksum_complete(skb
);
1018 /* Prequeue for VJ style copy to user, combined with checksumming. */
1020 static inline void tcp_prequeue_init(struct tcp_sock
*tp
)
1022 tp
->ucopy
.task
= NULL
;
1024 tp
->ucopy
.memory
= 0;
1025 skb_queue_head_init(&tp
->ucopy
.prequeue
);
1026 #ifdef CONFIG_NET_DMA
1027 tp
->ucopy
.dma_chan
= NULL
;
1028 tp
->ucopy
.wakeup
= 0;
1029 tp
->ucopy
.pinned_list
= NULL
;
1030 tp
->ucopy
.dma_cookie
= 0;
1034 extern bool tcp_prequeue(struct sock
*sk
, struct sk_buff
*skb
);
1039 static const char *statename
[]={
1040 "Unused","Established","Syn Sent","Syn Recv",
1041 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
1042 "Close Wait","Last ACK","Listen","Closing"
1045 extern void tcp_set_state(struct sock
*sk
, int state
);
1047 extern void tcp_done(struct sock
*sk
);
1049 static inline void tcp_sack_reset(struct tcp_options_received
*rx_opt
)
1052 rx_opt
->num_sacks
= 0;
1055 /* Determine a window scaling and initial window to offer. */
1056 extern void tcp_select_initial_window(int __space
, __u32 mss
,
1057 __u32
*rcv_wnd
, __u32
*window_clamp
,
1058 int wscale_ok
, __u8
*rcv_wscale
,
1059 __u32 init_rcv_wnd
);
1061 static inline int tcp_win_from_space(int space
)
1063 return sysctl_tcp_adv_win_scale
<=0 ?
1064 (space
>>(-sysctl_tcp_adv_win_scale
)) :
1065 space
- (space
>>sysctl_tcp_adv_win_scale
);
1068 /* Note: caller must be prepared to deal with negative returns */
1069 static inline int tcp_space(const struct sock
*sk
)
1071 return tcp_win_from_space(sk
->sk_rcvbuf
-
1072 atomic_read(&sk
->sk_rmem_alloc
));
1075 static inline int tcp_full_space(const struct sock
*sk
)
1077 return tcp_win_from_space(sk
->sk_rcvbuf
);
1080 static inline void tcp_openreq_init(struct request_sock
*req
,
1081 struct tcp_options_received
*rx_opt
,
1082 struct sk_buff
*skb
)
1084 struct inet_request_sock
*ireq
= inet_rsk(req
);
1086 req
->rcv_wnd
= 0; /* So that tcp_send_synack() knows! */
1088 tcp_rsk(req
)->rcv_isn
= TCP_SKB_CB(skb
)->seq
;
1089 tcp_rsk(req
)->rcv_nxt
= TCP_SKB_CB(skb
)->seq
+ 1;
1090 tcp_rsk(req
)->snt_synack
= 0;
1091 req
->mss
= rx_opt
->mss_clamp
;
1092 req
->ts_recent
= rx_opt
->saw_tstamp
? rx_opt
->rcv_tsval
: 0;
1093 ireq
->tstamp_ok
= rx_opt
->tstamp_ok
;
1094 ireq
->sack_ok
= rx_opt
->sack_ok
;
1095 ireq
->snd_wscale
= rx_opt
->snd_wscale
;
1096 ireq
->wscale_ok
= rx_opt
->wscale_ok
;
1099 ireq
->rmt_port
= tcp_hdr(skb
)->source
;
1100 ireq
->loc_port
= tcp_hdr(skb
)->dest
;
1103 /* Compute time elapsed between SYNACK and the ACK completing 3WHS */
1104 static inline void tcp_synack_rtt_meas(struct sock
*sk
,
1105 struct request_sock
*req
)
1107 if (tcp_rsk(req
)->snt_synack
)
1108 tcp_valid_rtt_meas(sk
,
1109 tcp_time_stamp
- tcp_rsk(req
)->snt_synack
);
1112 extern void tcp_enter_memory_pressure(struct sock
*sk
);
1114 static inline int keepalive_intvl_when(const struct tcp_sock
*tp
)
1116 return tp
->keepalive_intvl
? : sysctl_tcp_keepalive_intvl
;
1119 static inline int keepalive_time_when(const struct tcp_sock
*tp
)
1121 return tp
->keepalive_time
? : sysctl_tcp_keepalive_time
;
1124 static inline int keepalive_probes(const struct tcp_sock
*tp
)
1126 return tp
->keepalive_probes
? : sysctl_tcp_keepalive_probes
;
1129 static inline u32
keepalive_time_elapsed(const struct tcp_sock
*tp
)
1131 const struct inet_connection_sock
*icsk
= &tp
->inet_conn
;
1133 return min_t(u32
, tcp_time_stamp
- icsk
->icsk_ack
.lrcvtime
,
1134 tcp_time_stamp
- tp
->rcv_tstamp
);
1137 static inline int tcp_fin_time(const struct sock
*sk
)
1139 int fin_timeout
= tcp_sk(sk
)->linger2
? : sysctl_tcp_fin_timeout
;
1140 const int rto
= inet_csk(sk
)->icsk_rto
;
1142 if (fin_timeout
< (rto
<< 2) - (rto
>> 1))
1143 fin_timeout
= (rto
<< 2) - (rto
>> 1);
1148 static inline bool tcp_paws_check(const struct tcp_options_received
*rx_opt
,
1151 if ((s32
)(rx_opt
->ts_recent
- rx_opt
->rcv_tsval
) <= paws_win
)
1153 if (unlikely(get_seconds() >= rx_opt
->ts_recent_stamp
+ TCP_PAWS_24DAYS
))
1156 * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0,
1157 * then following tcp messages have valid values. Ignore 0 value,
1158 * or else 'negative' tsval might forbid us to accept their packets.
1160 if (!rx_opt
->ts_recent
)
1165 static inline bool tcp_paws_reject(const struct tcp_options_received
*rx_opt
,
1168 if (tcp_paws_check(rx_opt
, 0))
1171 /* RST segments are not recommended to carry timestamp,
1172 and, if they do, it is recommended to ignore PAWS because
1173 "their cleanup function should take precedence over timestamps."
1174 Certainly, it is mistake. It is necessary to understand the reasons
1175 of this constraint to relax it: if peer reboots, clock may go
1176 out-of-sync and half-open connections will not be reset.
1177 Actually, the problem would be not existing if all
1178 the implementations followed draft about maintaining clock
1179 via reboots. Linux-2.2 DOES NOT!
1181 However, we can relax time bounds for RST segments to MSL.
1183 if (rst
&& get_seconds() >= rx_opt
->ts_recent_stamp
+ TCP_PAWS_MSL
)
1188 static inline void tcp_mib_init(struct net
*net
)
1191 TCP_ADD_STATS_USER(net
, TCP_MIB_RTOALGORITHM
, 1);
1192 TCP_ADD_STATS_USER(net
, TCP_MIB_RTOMIN
, sysctl_tcp_rto_min
*1000/HZ
);
1193 TCP_ADD_STATS_USER(net
, TCP_MIB_RTOMAX
, sysctl_tcp_rto_max
*1000/HZ
);
1194 TCP_ADD_STATS_USER(net
, TCP_MIB_MAXCONN
, -1);
1198 static inline void tcp_clear_retrans_hints_partial(struct tcp_sock
*tp
)
1200 tp
->lost_skb_hint
= NULL
;
1201 tp
->scoreboard_skb_hint
= NULL
;
1204 static inline void tcp_clear_all_retrans_hints(struct tcp_sock
*tp
)
1206 tcp_clear_retrans_hints_partial(tp
);
1207 tp
->retransmit_skb_hint
= NULL
;
1213 union tcp_md5_addr
{
1215 #if IS_ENABLED(CONFIG_IPV6)
1220 /* - key database */
1221 struct tcp_md5sig_key
{
1222 struct hlist_node node
;
1224 u8 family
; /* AF_INET or AF_INET6 */
1225 union tcp_md5_addr addr
;
1226 u8 key
[TCP_MD5SIG_MAXKEYLEN
];
1227 struct rcu_head rcu
;
1231 struct tcp_md5sig_info
{
1232 struct hlist_head head
;
1233 struct rcu_head rcu
;
1236 /* - pseudo header */
1237 struct tcp4_pseudohdr
{
1245 struct tcp6_pseudohdr
{
1246 struct in6_addr saddr
;
1247 struct in6_addr daddr
;
1249 __be32 protocol
; /* including padding */
1252 union tcp_md5sum_block
{
1253 struct tcp4_pseudohdr ip4
;
1254 #if IS_ENABLED(CONFIG_IPV6)
1255 struct tcp6_pseudohdr ip6
;
1259 /* - pool: digest algorithm, hash description and scratch buffer */
1260 struct tcp_md5sig_pool
{
1261 struct hash_desc md5_desc
;
1262 union tcp_md5sum_block md5_blk
;
1266 extern int tcp_v4_md5_hash_skb(char *md5_hash
, struct tcp_md5sig_key
*key
,
1267 const struct sock
*sk
,
1268 const struct request_sock
*req
,
1269 const struct sk_buff
*skb
);
1270 extern int tcp_md5_do_add(struct sock
*sk
, const union tcp_md5_addr
*addr
,
1271 int family
, const u8
*newkey
,
1272 u8 newkeylen
, gfp_t gfp
);
1273 extern int tcp_md5_do_del(struct sock
*sk
, const union tcp_md5_addr
*addr
,
1275 extern struct tcp_md5sig_key
*tcp_v4_md5_lookup(struct sock
*sk
,
1276 struct sock
*addr_sk
);
1278 #ifdef CONFIG_TCP_MD5SIG
1279 extern struct tcp_md5sig_key
*tcp_md5_do_lookup(struct sock
*sk
,
1280 const union tcp_md5_addr
*addr
, int family
);
1281 #define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_key)
1283 static inline struct tcp_md5sig_key
*tcp_md5_do_lookup(struct sock
*sk
,
1284 const union tcp_md5_addr
*addr
,
1289 #define tcp_twsk_md5_key(twsk) NULL
1292 extern struct tcp_md5sig_pool __percpu
*tcp_alloc_md5sig_pool(struct sock
*);
1293 extern void tcp_free_md5sig_pool(void);
1295 extern struct tcp_md5sig_pool
*tcp_get_md5sig_pool(void);
1296 extern void tcp_put_md5sig_pool(void);
1298 extern int tcp_md5_hash_header(struct tcp_md5sig_pool
*, const struct tcphdr
*);
1299 extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool
*, const struct sk_buff
*,
1300 unsigned int header_len
);
1301 extern int tcp_md5_hash_key(struct tcp_md5sig_pool
*hp
,
1302 const struct tcp_md5sig_key
*key
);
1304 /* From tcp_fastopen.c */
1305 extern void tcp_fastopen_cache_get(struct sock
*sk
, u16
*mss
,
1306 struct tcp_fastopen_cookie
*cookie
,
1307 int *syn_loss
, unsigned long *last_syn_loss
);
1308 extern void tcp_fastopen_cache_set(struct sock
*sk
, u16 mss
,
1309 struct tcp_fastopen_cookie
*cookie
,
1311 struct tcp_fastopen_request
{
1312 /* Fast Open cookie. Size 0 means a cookie request */
1313 struct tcp_fastopen_cookie cookie
;
1314 struct msghdr
*data
; /* data in MSG_FASTOPEN */
1316 int copied
; /* queued in tcp_connect() */
1318 void tcp_free_fastopen_req(struct tcp_sock
*tp
);
1320 extern struct tcp_fastopen_context __rcu
*tcp_fastopen_ctx
;
1321 int tcp_fastopen_reset_cipher(void *key
, unsigned int len
);
1322 void tcp_fastopen_cookie_gen(__be32 addr
, struct tcp_fastopen_cookie
*foc
);
1324 #define TCP_FASTOPEN_KEY_LENGTH 16
1326 /* Fastopen key context */
1327 struct tcp_fastopen_context
{
1328 struct crypto_cipher __rcu
*tfm
;
1329 __u8 key
[TCP_FASTOPEN_KEY_LENGTH
];
1330 struct rcu_head rcu
;
1333 /* write queue abstraction */
1334 static inline void tcp_write_queue_purge(struct sock
*sk
)
1336 struct sk_buff
*skb
;
1338 while ((skb
= __skb_dequeue(&sk
->sk_write_queue
)) != NULL
)
1339 sk_wmem_free_skb(sk
, skb
);
1341 tcp_clear_all_retrans_hints(tcp_sk(sk
));
1344 static inline struct sk_buff
*tcp_write_queue_head(const struct sock
*sk
)
1346 return skb_peek(&sk
->sk_write_queue
);
1349 static inline struct sk_buff
*tcp_write_queue_tail(const struct sock
*sk
)
1351 return skb_peek_tail(&sk
->sk_write_queue
);
1354 static inline struct sk_buff
*tcp_write_queue_next(const struct sock
*sk
,
1355 const struct sk_buff
*skb
)
1357 return skb_queue_next(&sk
->sk_write_queue
, skb
);
1360 static inline struct sk_buff
*tcp_write_queue_prev(const struct sock
*sk
,
1361 const struct sk_buff
*skb
)
1363 return skb_queue_prev(&sk
->sk_write_queue
, skb
);
1366 #define tcp_for_write_queue(skb, sk) \
1367 skb_queue_walk(&(sk)->sk_write_queue, skb)
1369 #define tcp_for_write_queue_from(skb, sk) \
1370 skb_queue_walk_from(&(sk)->sk_write_queue, skb)
1372 #define tcp_for_write_queue_from_safe(skb, tmp, sk) \
1373 skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
1375 static inline struct sk_buff
*tcp_send_head(const struct sock
*sk
)
1377 return sk
->sk_send_head
;
1380 static inline bool tcp_skb_is_last(const struct sock
*sk
,
1381 const struct sk_buff
*skb
)
1383 return skb_queue_is_last(&sk
->sk_write_queue
, skb
);
1386 static inline void tcp_advance_send_head(struct sock
*sk
, const struct sk_buff
*skb
)
1388 if (tcp_skb_is_last(sk
, skb
))
1389 sk
->sk_send_head
= NULL
;
1391 sk
->sk_send_head
= tcp_write_queue_next(sk
, skb
);
1394 static inline void tcp_check_send_head(struct sock
*sk
, struct sk_buff
*skb_unlinked
)
1396 if (sk
->sk_send_head
== skb_unlinked
)
1397 sk
->sk_send_head
= NULL
;
1398 if (tcp_sk(sk
)->highest_sack
== skb_unlinked
)
1399 tcp_sk(sk
)->highest_sack
= NULL
;
1402 static inline void tcp_init_send_head(struct sock
*sk
)
1404 sk
->sk_send_head
= NULL
;
1407 static inline void __tcp_add_write_queue_tail(struct sock
*sk
, struct sk_buff
*skb
)
1409 __skb_queue_tail(&sk
->sk_write_queue
, skb
);
1412 static inline void tcp_add_write_queue_tail(struct sock
*sk
, struct sk_buff
*skb
)
1414 __tcp_add_write_queue_tail(sk
, skb
);
1416 /* Queue it, remembering where we must start sending. */
1417 if (sk
->sk_send_head
== NULL
) {
1418 sk
->sk_send_head
= skb
;
1420 if (tcp_sk(sk
)->highest_sack
== NULL
)
1421 tcp_sk(sk
)->highest_sack
= skb
;
1425 static inline void __tcp_add_write_queue_head(struct sock
*sk
, struct sk_buff
*skb
)
1427 __skb_queue_head(&sk
->sk_write_queue
, skb
);
1430 /* Insert buff after skb on the write queue of sk. */
1431 static inline void tcp_insert_write_queue_after(struct sk_buff
*skb
,
1432 struct sk_buff
*buff
,
1435 __skb_queue_after(&sk
->sk_write_queue
, skb
, buff
);
1438 /* Insert new before skb on the write queue of sk. */
1439 static inline void tcp_insert_write_queue_before(struct sk_buff
*new,
1440 struct sk_buff
*skb
,
1443 __skb_queue_before(&sk
->sk_write_queue
, skb
, new);
1445 if (sk
->sk_send_head
== skb
)
1446 sk
->sk_send_head
= new;
1449 static inline void tcp_unlink_write_queue(struct sk_buff
*skb
, struct sock
*sk
)
1451 __skb_unlink(skb
, &sk
->sk_write_queue
);
1454 static inline bool tcp_write_queue_empty(struct sock
*sk
)
1456 return skb_queue_empty(&sk
->sk_write_queue
);
1459 static inline void tcp_push_pending_frames(struct sock
*sk
)
1461 if (tcp_send_head(sk
)) {
1462 struct tcp_sock
*tp
= tcp_sk(sk
);
1464 __tcp_push_pending_frames(sk
, tcp_current_mss(sk
), tp
->nonagle
);
1468 /* Start sequence of the skb just after the highest skb with SACKed
1469 * bit, valid only if sacked_out > 0 or when the caller has ensured
1470 * validity by itself.
1472 static inline u32
tcp_highest_sack_seq(struct tcp_sock
*tp
)
1474 if (!tp
->sacked_out
)
1477 if (tp
->highest_sack
== NULL
)
1480 return TCP_SKB_CB(tp
->highest_sack
)->seq
;
1483 static inline void tcp_advance_highest_sack(struct sock
*sk
, struct sk_buff
*skb
)
1485 tcp_sk(sk
)->highest_sack
= tcp_skb_is_last(sk
, skb
) ? NULL
:
1486 tcp_write_queue_next(sk
, skb
);
1489 static inline struct sk_buff
*tcp_highest_sack(struct sock
*sk
)
1491 return tcp_sk(sk
)->highest_sack
;
1494 static inline void tcp_highest_sack_reset(struct sock
*sk
)
1496 tcp_sk(sk
)->highest_sack
= tcp_write_queue_head(sk
);
1499 /* Called when old skb is about to be deleted (to be combined with new skb) */
1500 static inline void tcp_highest_sack_combine(struct sock
*sk
,
1501 struct sk_buff
*old
,
1502 struct sk_buff
*new)
1504 if (tcp_sk(sk
)->sacked_out
&& (old
== tcp_sk(sk
)->highest_sack
))
1505 tcp_sk(sk
)->highest_sack
= new;
1508 /* Determines whether this is a thin stream (which may suffer from
1509 * increased latency). Used to trigger latency-reducing mechanisms.
1511 static inline bool tcp_stream_is_thin(struct tcp_sock
*tp
)
1513 return tp
->packets_out
< 4 && !tcp_in_initial_slowstart(tp
);
1517 enum tcp_seq_states
{
1518 TCP_SEQ_STATE_LISTENING
,
1519 TCP_SEQ_STATE_OPENREQ
,
1520 TCP_SEQ_STATE_ESTABLISHED
,
1521 TCP_SEQ_STATE_TIME_WAIT
,
1524 int tcp_seq_open(struct inode
*inode
, struct file
*file
);
1526 struct tcp_seq_afinfo
{
1529 const struct file_operations
*seq_fops
;
1530 struct seq_operations seq_ops
;
1533 struct tcp_iter_state
{
1534 struct seq_net_private p
;
1536 enum tcp_seq_states state
;
1537 struct sock
*syn_wait_sk
;
1538 int bucket
, offset
, sbucket
, num
;
1543 /* MTK_NET_CHANGES */
1545 * reset tcp connection by uid
1552 extern int tcp_proc_register(struct net
*net
, struct tcp_seq_afinfo
*afinfo
);
1553 extern void tcp_proc_unregister(struct net
*net
, struct tcp_seq_afinfo
*afinfo
);
1555 extern struct request_sock_ops tcp_request_sock_ops
;
1556 extern struct request_sock_ops tcp6_request_sock_ops
;
1558 extern void tcp_v4_destroy_sock(struct sock
*sk
);
1560 extern int tcp_v4_gso_send_check(struct sk_buff
*skb
);
1561 extern struct sk_buff
*tcp_tso_segment(struct sk_buff
*skb
,
1562 netdev_features_t features
);
1563 extern struct sk_buff
**tcp_gro_receive(struct sk_buff
**head
,
1564 struct sk_buff
*skb
);
1565 extern struct sk_buff
**tcp4_gro_receive(struct sk_buff
**head
,
1566 struct sk_buff
*skb
);
1567 extern int tcp_gro_complete(struct sk_buff
*skb
);
1568 extern int tcp4_gro_complete(struct sk_buff
*skb
);
1570 extern int tcp_nuke_addr(struct net
*net
, struct sockaddr
*addr
);
1571 /* MTK_NET_CHANGES */
1572 extern void tcp_v4_reset_connections_by_uid(struct uid_err uid_e
);
1573 extern void tcp_v4_handle_retrans_time_by_uid(struct uid_err uid_e
);
1575 #ifdef CONFIG_PROC_FS
1576 extern int tcp4_proc_init(void);
1577 extern void tcp4_proc_exit(void);
1580 /* TCP af-specific functions */
1581 struct tcp_sock_af_ops
{
1582 #ifdef CONFIG_TCP_MD5SIG
1583 struct tcp_md5sig_key
*(*md5_lookup
) (struct sock
*sk
,
1584 struct sock
*addr_sk
);
1585 int (*calc_md5_hash
) (char *location
,
1586 struct tcp_md5sig_key
*md5
,
1587 const struct sock
*sk
,
1588 const struct request_sock
*req
,
1589 const struct sk_buff
*skb
);
1590 int (*md5_parse
) (struct sock
*sk
,
1591 char __user
*optval
,
1596 struct tcp_request_sock_ops
{
1597 #ifdef CONFIG_TCP_MD5SIG
1598 struct tcp_md5sig_key
*(*md5_lookup
) (struct sock
*sk
,
1599 struct request_sock
*req
);
1600 int (*calc_md5_hash
) (char *location
,
1601 struct tcp_md5sig_key
*md5
,
1602 const struct sock
*sk
,
1603 const struct request_sock
*req
,
1604 const struct sk_buff
*skb
);
1608 extern void tcp_v4_init(void);
1609 extern void tcp_init(void);