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 protocol.
8 * Version: @(#)tcp.h 1.0.2 04/28/93
10 * Author: Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
20 #include <linux/types.h>
21 #include <asm/byteorder.h>
22 #include <linux/socket.h>
29 #if defined(__LITTLE_ENDIAN_BITFIELD)
40 #elif defined(__BIG_ENDIAN_BITFIELD)
52 #error "Adjust your <asm/byteorder.h> defines"
60 * The union cast uses a gcc extension to avoid aliasing problems
61 * (union is compatible to any of its members)
62 * This means this part of the code is -fstrict-aliasing safe now.
69 #define tcp_flag_word(tp) ( ((union tcp_word_hdr *)(tp))->words [3])
72 TCP_FLAG_CWR
= __constant_cpu_to_be32(0x00800000),
73 TCP_FLAG_ECE
= __constant_cpu_to_be32(0x00400000),
74 TCP_FLAG_URG
= __constant_cpu_to_be32(0x00200000),
75 TCP_FLAG_ACK
= __constant_cpu_to_be32(0x00100000),
76 TCP_FLAG_PSH
= __constant_cpu_to_be32(0x00080000),
77 TCP_FLAG_RST
= __constant_cpu_to_be32(0x00040000),
78 TCP_FLAG_SYN
= __constant_cpu_to_be32(0x00020000),
79 TCP_FLAG_FIN
= __constant_cpu_to_be32(0x00010000),
80 TCP_RESERVED_BITS
= __constant_cpu_to_be32(0x0F000000),
81 TCP_DATA_OFFSET
= __constant_cpu_to_be32(0xF0000000)
85 * TCP general constants
87 #define TCP_MSS_DEFAULT 536U /* IPv4 (RFC1122, RFC2581) */
88 #define TCP_MSS_DESIRED 1220U /* IPv6 (tunneled), EDNS0 (RFC3226) */
90 /* TCP socket options */
91 #define TCP_NODELAY 1 /* Turn off Nagle's algorithm. */
92 #define TCP_MAXSEG 2 /* Limit MSS */
93 #define TCP_CORK 3 /* Never send partially complete segments */
94 #define TCP_KEEPIDLE 4 /* Start keeplives after this period */
95 #define TCP_KEEPINTVL 5 /* Interval between keepalives */
96 #define TCP_KEEPCNT 6 /* Number of keepalives before death */
97 #define TCP_SYNCNT 7 /* Number of SYN retransmits */
98 #define TCP_LINGER2 8 /* Life time of orphaned FIN-WAIT-2 state */
99 #define TCP_DEFER_ACCEPT 9 /* Wake up listener only when data arrive */
100 #define TCP_WINDOW_CLAMP 10 /* Bound advertised window */
101 #define TCP_INFO 11 /* Information about this connection. */
102 #define TCP_QUICKACK 12 /* Block/reenable quick acks */
103 #define TCP_CONGESTION 13 /* Congestion control algorithm */
104 #define TCP_MD5SIG 14 /* TCP MD5 Signature (RFC2385) */
105 #define TCP_COOKIE_TRANSACTIONS 15 /* TCP Cookie Transactions */
106 #define TCP_THIN_LINEAR_TIMEOUTS 16 /* Use linear timeouts for thin streams*/
107 #define TCP_THIN_DUPACK 17 /* Fast retrans. after 1 dupack */
108 #define TCP_USER_TIMEOUT 18 /* How long for loss retry before timeout */
109 #define TCP_REPAIR 19 /* TCP sock is under repair right now */
110 #define TCP_REPAIR_QUEUE 20
111 #define TCP_QUEUE_SEQ 21
112 #define TCP_REPAIR_OPTIONS 22
113 #define TCP_FASTOPEN 23 /* Enable FastOpen on listeners */
115 struct tcp_repair_opt
{
127 /* for TCP_INFO socket option */
128 #define TCPI_OPT_TIMESTAMPS 1
129 #define TCPI_OPT_SACK 2
130 #define TCPI_OPT_WSCALE 4
131 #define TCPI_OPT_ECN 8 /* ECN was negociated at TCP session init */
132 #define TCPI_OPT_ECN_SEEN 16 /* we received at least one packet with ECT */
136 #define TCPF_CA_Open (1<<TCP_CA_Open)
138 #define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
140 #define TCPF_CA_CWR (1<<TCP_CA_CWR)
142 #define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
144 #define TCPF_CA_Loss (1<<TCP_CA_Loss)
150 __u8 tcpi_retransmits
;
154 __u8 tcpi_snd_wscale
: 4, tcpi_rcv_wscale
: 4;
168 __u32 tcpi_last_data_sent
;
169 __u32 tcpi_last_ack_sent
; /* Not remembered, sorry. */
170 __u32 tcpi_last_data_recv
;
171 __u32 tcpi_last_ack_recv
;
175 __u32 tcpi_rcv_ssthresh
;
178 __u32 tcpi_snd_ssthresh
;
181 __u32 tcpi_reordering
;
184 __u32 tcpi_rcv_space
;
186 __u32 tcpi_total_retrans
;
189 /* for TCP_MD5SIG socket option */
190 #define TCP_MD5SIG_MAXKEYLEN 80
193 struct __kernel_sockaddr_storage tcpm_addr
; /* address associated */
194 __u16 __tcpm_pad1
; /* zero */
195 __u16 tcpm_keylen
; /* key length */
196 __u32 __tcpm_pad2
; /* zero */
197 __u8 tcpm_key
[TCP_MD5SIG_MAXKEYLEN
]; /* key (binary) */
200 /* for TCP_COOKIE_TRANSACTIONS (TCPCT) socket option */
201 #define TCP_COOKIE_MIN 8 /* 64-bits */
202 #define TCP_COOKIE_MAX 16 /* 128-bits */
203 #define TCP_COOKIE_PAIR_SIZE (2*TCP_COOKIE_MAX)
205 /* Flags for both getsockopt and setsockopt */
206 #define TCP_COOKIE_IN_ALWAYS (1 << 0) /* Discard SYN without cookie */
207 #define TCP_COOKIE_OUT_NEVER (1 << 1) /* Prohibit outgoing cookies,
208 * supercedes everything. */
210 /* Flags for getsockopt */
211 #define TCP_S_DATA_IN (1 << 2) /* Was data received? */
212 #define TCP_S_DATA_OUT (1 << 3) /* Was data sent? */
214 /* TCP_COOKIE_TRANSACTIONS data */
215 struct tcp_cookie_transactions
{
216 __u16 tcpct_flags
; /* see above */
217 __u8 __tcpct_pad1
; /* zero */
218 __u8 tcpct_cookie_desired
; /* bytes */
219 __u16 tcpct_s_data_desired
; /* bytes of variable data */
220 __u16 tcpct_used
; /* bytes in value */
221 __u8 tcpct_value
[TCP_MSS_DEFAULT
];
226 #include <linux/skbuff.h>
227 #include <linux/dmaengine.h>
228 #include <net/sock.h>
229 #include <net/inet_connection_sock.h>
230 #include <net/inet_timewait_sock.h>
232 static inline struct tcphdr
*tcp_hdr(const struct sk_buff
*skb
)
234 return (struct tcphdr
*)skb_transport_header(skb
);
237 static inline unsigned int tcp_hdrlen(const struct sk_buff
*skb
)
239 return tcp_hdr(skb
)->doff
* 4;
242 static inline unsigned int tcp_optlen(const struct sk_buff
*skb
)
244 return (tcp_hdr(skb
)->doff
- 5) * 4;
248 #define TCP_FASTOPEN_COOKIE_MIN 4 /* Min Fast Open Cookie size in bytes */
249 #define TCP_FASTOPEN_COOKIE_MAX 16 /* Max Fast Open Cookie size in bytes */
250 #define TCP_FASTOPEN_COOKIE_SIZE 8 /* the size employed by this impl. */
252 /* TCP Fast Open Cookie as stored in memory */
253 struct tcp_fastopen_cookie
{
255 u8 val
[TCP_FASTOPEN_COOKIE_MAX
];
258 /* This defines a selective acknowledgement block. */
259 struct tcp_sack_block_wire
{
264 struct tcp_sack_block
{
269 /*These are used to set the sack_ok field in struct tcp_options_received */
270 #define TCP_SACK_SEEN (1 << 0) /*1 = peer is SACK capable, */
271 #define TCP_FACK_ENABLED (1 << 1) /*1 = FACK is enabled locally*/
272 #define TCP_DSACK_SEEN (1 << 2) /*1 = DSACK was received from peer*/
274 struct tcp_options_received
{
276 long ts_recent_stamp
;/* Time we stored ts_recent (for aging) */
277 u32 ts_recent
; /* Time stamp to echo next */
278 u32 rcv_tsval
; /* Time stamp value */
279 u32 rcv_tsecr
; /* Time stamp echo reply */
280 u16 saw_tstamp
: 1, /* Saw TIMESTAMP on last packet */
281 tstamp_ok
: 1, /* TIMESTAMP seen on SYN packet */
282 dsack
: 1, /* D-SACK is scheduled */
283 wscale_ok
: 1, /* Wscale seen on SYN packet */
284 sack_ok
: 4, /* SACK seen on SYN packet */
285 snd_wscale
: 4, /* Window scaling received from sender */
286 rcv_wscale
: 4; /* Window scaling to send to receiver */
287 u8 cookie_plus
:6, /* bytes in authenticator/cookie option */
290 u8 num_sacks
; /* Number of SACK blocks */
291 u16 user_mss
; /* mss requested by user in ioctl */
292 u16 mss_clamp
; /* Maximal mss, negotiated at connection setup */
295 static inline void tcp_clear_options(struct tcp_options_received
*rx_opt
)
297 rx_opt
->tstamp_ok
= rx_opt
->sack_ok
= 0;
298 rx_opt
->wscale_ok
= rx_opt
->snd_wscale
= 0;
299 rx_opt
->cookie_plus
= 0;
302 /* This is the max number of SACKS that we'll generate and process. It's safe
303 * to increase this, although since:
304 * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
305 * only four options will fit in a standard TCP header */
306 #define TCP_NUM_SACKS 4
308 struct tcp_cookie_values
;
309 struct tcp_request_sock_ops
;
311 struct tcp_request_sock
{
312 struct inet_request_sock req
;
313 #ifdef CONFIG_TCP_MD5SIG
314 /* Only used by TCP MD5 Signature so far. */
315 const struct tcp_request_sock_ops
*af_specific
;
317 struct sock
*listener
; /* needed for TFO */
320 u32 snt_synack
; /* synack sent time */
321 u32 rcv_nxt
; /* the ack # by SYNACK. For
322 * FastOpen it's the seq#
327 static inline struct tcp_request_sock
*tcp_rsk(const struct request_sock
*req
)
329 return (struct tcp_request_sock
*)req
;
333 /* inet_connection_sock has to be the first member of tcp_sock */
334 struct inet_connection_sock inet_conn
;
335 u16 tcp_header_len
; /* Bytes of tcp header to send */
336 u16 xmit_size_goal_segs
; /* Goal for segmenting output packets */
339 * Header prediction flags
340 * 0x5?10 << 16 + snd_wnd in net byte order
345 * RFC793 variables by their proper names. This means you can
346 * read the code and the spec side by side (and laugh ...)
347 * See RFC793 and RFC1122. The RFC writes these in capitals.
349 u32 rcv_nxt
; /* What we want to receive next */
350 u32 copied_seq
; /* Head of yet unread data */
351 u32 rcv_wup
; /* rcv_nxt on last window update sent */
352 u32 snd_nxt
; /* Next sequence we send */
354 u32 snd_una
; /* First byte we want an ack for */
355 u32 snd_sml
; /* Last byte of the most recently transmitted small packet */
356 u32 rcv_tstamp
; /* timestamp of last received ACK (for keepalives) */
357 u32 lsndtime
; /* timestamp of last sent data packet (for restart window) */
359 struct list_head tsq_node
; /* anchor in tsq_tasklet.head list */
360 unsigned long tsq_flags
;
362 /* Data for direct copy to user */
364 struct sk_buff_head prequeue
;
365 struct task_struct
*task
;
369 #ifdef CONFIG_NET_DMA
370 /* members for async copy */
371 struct dma_chan
*dma_chan
;
373 struct dma_pinned_list
*pinned_list
;
374 dma_cookie_t dma_cookie
;
378 u32 snd_wl1
; /* Sequence for window update */
379 u32 snd_wnd
; /* The window we expect to receive */
380 u32 max_window
; /* Maximal window ever seen from peer */
381 u32 mss_cache
; /* Cached effective mss, not including SACKS */
383 u32 window_clamp
; /* Maximal window to advertise */
384 u32 rcv_ssthresh
; /* Current window clamp */
386 u32 frto_highmark
; /* snd_nxt when RTO occurred */
387 u16 advmss
; /* Advertised MSS */
388 u8 frto_counter
; /* Number of new acks after RTO */
389 u8 nonagle
: 4,/* Disable Nagle algorithm? */
390 thin_lto
: 1,/* Use linear timeouts for thin streams */
391 thin_dupack
: 1,/* Fast retransmit on first dupack */
395 u8 do_early_retrans
:1,/* Enable RFC5827 early-retransmit */
396 early_retrans_delayed
:1, /* Delayed ER timer installed */
397 syn_data
:1, /* SYN includes data */
398 syn_fastopen
:1; /* SYN includes Fast Open option */
400 /* RTT measurement */
401 u32 srtt
; /* smoothed round trip time << 3 */
402 u32 mdev
; /* medium deviation */
403 u32 mdev_max
; /* maximal mdev for the last rtt period */
404 u32 rttvar
; /* smoothed mdev_max */
405 u32 rtt_seq
; /* sequence number to update rttvar */
407 u32 packets_out
; /* Packets which are "in flight" */
408 u32 retrans_out
; /* Retransmitted packets out */
410 u16 urg_data
; /* Saved octet of OOB data and control flags */
411 u8 ecn_flags
; /* ECN status bits. */
412 u8 reordering
; /* Packet reordering metric. */
413 u32 snd_up
; /* Urgent pointer */
415 u8 keepalive_probes
; /* num of allowed keep alive probes */
417 * Options received (usually on last packet, some only on SYN packets).
419 struct tcp_options_received rx_opt
;
422 * Slow start and congestion control (see also Nagle, and Karn & Partridge)
424 u32 snd_ssthresh
; /* Slow start size threshold */
425 u32 snd_cwnd
; /* Sending congestion window */
426 u32 snd_cwnd_cnt
; /* Linear increase counter */
427 u32 snd_cwnd_clamp
; /* Do not allow snd_cwnd to grow above this */
430 u32 prior_cwnd
; /* Congestion window at start of Recovery. */
431 u32 prr_delivered
; /* Number of newly delivered packets to
432 * receiver in Recovery. */
433 u32 prr_out
; /* Total number of pkts sent during Recovery. */
435 u32 rcv_wnd
; /* Current receiver window */
436 u32 write_seq
; /* Tail(+1) of data held in tcp send buffer */
437 u32 pushed_seq
; /* Last pushed seq, required to talk to windows */
438 u32 lost_out
; /* Lost packets */
439 u32 sacked_out
; /* SACK'd packets */
440 u32 fackets_out
; /* FACK'd packets */
442 u32 bytes_acked
; /* Appropriate Byte Counting - RFC3465 */
444 /* from STCP, retrans queue hinting */
445 struct sk_buff
* lost_skb_hint
;
446 struct sk_buff
*scoreboard_skb_hint
;
447 struct sk_buff
*retransmit_skb_hint
;
449 struct sk_buff_head out_of_order_queue
; /* Out of order segments go here */
451 /* SACKs data, these 2 need to be together (see tcp_options_write) */
452 struct tcp_sack_block duplicate_sack
[1]; /* D-SACK block */
453 struct tcp_sack_block selective_acks
[4]; /* The SACKS themselves*/
455 struct tcp_sack_block recv_sack_cache
[4];
457 struct sk_buff
*highest_sack
; /* skb just after the highest
458 * skb with SACKed bit set
459 * (validity guaranteed only if
464 u32 retransmit_high
; /* L-bits may be on up to this seqno */
466 u32 lost_retrans_low
; /* Sent seq after any rxmit (lowest) */
468 u32 prior_ssthresh
; /* ssthresh saved at recovery start */
469 u32 high_seq
; /* snd_nxt at onset of congestion */
471 u32 retrans_stamp
; /* Timestamp of the last retransmit,
472 * also used in SYN-SENT to remember stamp of
474 u32 undo_marker
; /* tracking retrans started here. */
475 int undo_retrans
; /* number of undoable retransmissions. */
476 u32 total_retrans
; /* Total retransmits for entire connection */
478 u32 urg_seq
; /* Seq of received urgent pointer */
479 unsigned int keepalive_time
; /* time before keep alive takes place */
480 unsigned int keepalive_intvl
; /* time interval between keep alive probes */
484 /* Receiver side RTT estimation */
491 /* Receiver queue space */
498 /* TCP-specific MTU probe information. */
503 u32 mtu_info
; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG
504 * while socket was owned by user.
507 #ifdef CONFIG_TCP_MD5SIG
508 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
509 const struct tcp_sock_af_ops
*af_specific
;
511 /* TCP MD5 Signature Option information */
512 struct tcp_md5sig_info __rcu
*md5sig_info
;
515 /* When the cookie options are generated and exchanged, then this
516 * object holds a reference to them (cookie_values->kref). Also
517 * contains related tcp_cookie_transactions fields.
519 struct tcp_cookie_values
*cookie_values
;
521 /* TCP fastopen related information */
522 struct tcp_fastopen_request
*fastopen_req
;
523 /* fastopen_rsk points to request_sock that resulted in this big
524 * socket. Used to retransmit SYNACKs etc.
526 struct request_sock
*fastopen_rsk
;
532 TCP_TSQ_DEFERRED
, /* tcp_tasklet_func() found socket was owned */
533 TCP_WRITE_TIMER_DEFERRED
, /* tcp_write_timer() found socket was owned */
534 TCP_DELACK_TIMER_DEFERRED
, /* tcp_delack_timer() found socket was owned */
535 TCP_MTU_REDUCED_DEFERRED
, /* tcp_v{4|6}_err() could not call
536 * tcp_v{4|6}_mtu_reduced()
540 static inline struct tcp_sock
*tcp_sk(const struct sock
*sk
)
542 return (struct tcp_sock
*)sk
;
545 struct tcp_timewait_sock
{
546 struct inet_timewait_sock tw_sk
;
551 long tw_ts_recent_stamp
;
552 #ifdef CONFIG_TCP_MD5SIG
553 struct tcp_md5sig_key
*tw_md5_key
;
555 /* Few sockets in timewait have cookies; in that case, then this
556 * object holds a reference to them (tw_cookie_values->kref).
558 struct tcp_cookie_values
*tw_cookie_values
;
561 static inline struct tcp_timewait_sock
*tcp_twsk(const struct sock
*sk
)
563 return (struct tcp_timewait_sock
*)sk
;
566 static inline bool tcp_passive_fastopen(const struct sock
*sk
)
568 return (sk
->sk_state
== TCP_SYN_RECV
&&
569 tcp_sk(sk
)->fastopen_rsk
!= NULL
);
572 static inline bool fastopen_cookie_present(struct tcp_fastopen_cookie
*foc
)
574 return foc
->len
!= -1;
577 extern void tcp_sock_destruct(struct sock
*sk
);
579 static inline int fastopen_init_queue(struct sock
*sk
, int backlog
)
581 struct request_sock_queue
*queue
=
582 &inet_csk(sk
)->icsk_accept_queue
;
584 if (queue
->fastopenq
== NULL
) {
585 queue
->fastopenq
= kzalloc(
586 sizeof(struct fastopen_queue
),
588 if (queue
->fastopenq
== NULL
)
591 sk
->sk_destruct
= tcp_sock_destruct
;
592 spin_lock_init(&queue
->fastopenq
->lock
);
594 queue
->fastopenq
->max_qlen
= backlog
;
598 #endif /* __KERNEL__ */
600 #endif /* _LINUX_TCP_H */