xref: /linux-6.15/include/linux/tcp.h (revision 4bedea94)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Definitions for the TCP protocol.
7  *
8  * Version:	@(#)tcp.h	1.0.2	04/28/93
9  *
10  * Author:	Fred N. van Kempen, <[email protected]>
11  *
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.
16  */
17 #ifndef _LINUX_TCP_H
18 #define _LINUX_TCP_H
19 
20 #include <linux/types.h>
21 #include <asm/byteorder.h>
22 
23 struct tcphdr {
24 	__u16	source;
25 	__u16	dest;
26 	__u32	seq;
27 	__u32	ack_seq;
28 #if defined(__LITTLE_ENDIAN_BITFIELD)
29 	__u16	res1:4,
30 		doff:4,
31 		fin:1,
32 		syn:1,
33 		rst:1,
34 		psh:1,
35 		ack:1,
36 		urg:1,
37 		ece:1,
38 		cwr:1;
39 #elif defined(__BIG_ENDIAN_BITFIELD)
40 	__u16	doff:4,
41 		res1:4,
42 		cwr:1,
43 		ece:1,
44 		urg:1,
45 		ack:1,
46 		psh:1,
47 		rst:1,
48 		syn:1,
49 		fin:1;
50 #else
51 #error	"Adjust your <asm/byteorder.h> defines"
52 #endif
53 	__u16	window;
54 	__u16	check;
55 	__u16	urg_ptr;
56 };
57 
58 
59 enum {
60   TCP_ESTABLISHED = 1,
61   TCP_SYN_SENT,
62   TCP_SYN_RECV,
63   TCP_FIN_WAIT1,
64   TCP_FIN_WAIT2,
65   TCP_TIME_WAIT,
66   TCP_CLOSE,
67   TCP_CLOSE_WAIT,
68   TCP_LAST_ACK,
69   TCP_LISTEN,
70   TCP_CLOSING,	 /* now a valid state */
71 
72   TCP_MAX_STATES /* Leave at the end! */
73 };
74 
75 #define TCP_STATE_MASK	0xF
76 #define TCP_ACTION_FIN	(1 << 7)
77 
78 enum {
79   TCPF_ESTABLISHED = (1 << 1),
80   TCPF_SYN_SENT  = (1 << 2),
81   TCPF_SYN_RECV  = (1 << 3),
82   TCPF_FIN_WAIT1 = (1 << 4),
83   TCPF_FIN_WAIT2 = (1 << 5),
84   TCPF_TIME_WAIT = (1 << 6),
85   TCPF_CLOSE     = (1 << 7),
86   TCPF_CLOSE_WAIT = (1 << 8),
87   TCPF_LAST_ACK  = (1 << 9),
88   TCPF_LISTEN    = (1 << 10),
89   TCPF_CLOSING   = (1 << 11)
90 };
91 
92 /*
93  *	The union cast uses a gcc extension to avoid aliasing problems
94  *  (union is compatible to any of its members)
95  *  This means this part of the code is -fstrict-aliasing safe now.
96  */
97 union tcp_word_hdr {
98 	struct tcphdr hdr;
99 	__u32 		  words[5];
100 };
101 
102 #define tcp_flag_word(tp) ( ((union tcp_word_hdr *)(tp))->words [3])
103 
104 enum {
105 	TCP_FLAG_CWR = __constant_htonl(0x00800000),
106 	TCP_FLAG_ECE = __constant_htonl(0x00400000),
107 	TCP_FLAG_URG = __constant_htonl(0x00200000),
108 	TCP_FLAG_ACK = __constant_htonl(0x00100000),
109 	TCP_FLAG_PSH = __constant_htonl(0x00080000),
110 	TCP_FLAG_RST = __constant_htonl(0x00040000),
111 	TCP_FLAG_SYN = __constant_htonl(0x00020000),
112 	TCP_FLAG_FIN = __constant_htonl(0x00010000),
113 	TCP_RESERVED_BITS = __constant_htonl(0x0F000000),
114 	TCP_DATA_OFFSET = __constant_htonl(0xF0000000)
115 };
116 
117 /* TCP socket options */
118 #define TCP_NODELAY		1	/* Turn off Nagle's algorithm. */
119 #define TCP_MAXSEG		2	/* Limit MSS */
120 #define TCP_CORK		3	/* Never send partially complete segments */
121 #define TCP_KEEPIDLE		4	/* Start keeplives after this period */
122 #define TCP_KEEPINTVL		5	/* Interval between keepalives */
123 #define TCP_KEEPCNT		6	/* Number of keepalives before death */
124 #define TCP_SYNCNT		7	/* Number of SYN retransmits */
125 #define TCP_LINGER2		8	/* Life time of orphaned FIN-WAIT-2 state */
126 #define TCP_DEFER_ACCEPT	9	/* Wake up listener only when data arrive */
127 #define TCP_WINDOW_CLAMP	10	/* Bound advertised window */
128 #define TCP_INFO		11	/* Information about this connection. */
129 #define TCP_QUICKACK		12	/* Block/reenable quick acks */
130 
131 #define TCPI_OPT_TIMESTAMPS	1
132 #define TCPI_OPT_SACK		2
133 #define TCPI_OPT_WSCALE		4
134 #define TCPI_OPT_ECN		8
135 
136 enum tcp_ca_state
137 {
138 	TCP_CA_Open = 0,
139 #define TCPF_CA_Open	(1<<TCP_CA_Open)
140 	TCP_CA_Disorder = 1,
141 #define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
142 	TCP_CA_CWR = 2,
143 #define TCPF_CA_CWR	(1<<TCP_CA_CWR)
144 	TCP_CA_Recovery = 3,
145 #define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
146 	TCP_CA_Loss = 4
147 #define TCPF_CA_Loss	(1<<TCP_CA_Loss)
148 };
149 
150 struct tcp_info
151 {
152 	__u8	tcpi_state;
153 	__u8	tcpi_ca_state;
154 	__u8	tcpi_retransmits;
155 	__u8	tcpi_probes;
156 	__u8	tcpi_backoff;
157 	__u8	tcpi_options;
158 	__u8	tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
159 
160 	__u32	tcpi_rto;
161 	__u32	tcpi_ato;
162 	__u32	tcpi_snd_mss;
163 	__u32	tcpi_rcv_mss;
164 
165 	__u32	tcpi_unacked;
166 	__u32	tcpi_sacked;
167 	__u32	tcpi_lost;
168 	__u32	tcpi_retrans;
169 	__u32	tcpi_fackets;
170 
171 	/* Times. */
172 	__u32	tcpi_last_data_sent;
173 	__u32	tcpi_last_ack_sent;     /* Not remembered, sorry. */
174 	__u32	tcpi_last_data_recv;
175 	__u32	tcpi_last_ack_recv;
176 
177 	/* Metrics. */
178 	__u32	tcpi_pmtu;
179 	__u32	tcpi_rcv_ssthresh;
180 	__u32	tcpi_rtt;
181 	__u32	tcpi_rttvar;
182 	__u32	tcpi_snd_ssthresh;
183 	__u32	tcpi_snd_cwnd;
184 	__u32	tcpi_advmss;
185 	__u32	tcpi_reordering;
186 
187 	__u32	tcpi_rcv_rtt;
188 	__u32	tcpi_rcv_space;
189 
190 	__u32	tcpi_total_retrans;
191 };
192 
193 #ifdef __KERNEL__
194 
195 #include <linux/config.h>
196 #include <linux/skbuff.h>
197 #include <linux/ip.h>
198 #include <net/sock.h>
199 
200 /* This defines a selective acknowledgement block. */
201 struct tcp_sack_block {
202 	__u32	start_seq;
203 	__u32	end_seq;
204 };
205 
206 struct tcp_options_received {
207 /*	PAWS/RTTM data	*/
208 	long	ts_recent_stamp;/* Time we stored ts_recent (for aging) */
209 	__u32	ts_recent;	/* Time stamp to echo next		*/
210 	__u32	rcv_tsval;	/* Time stamp value             	*/
211 	__u32	rcv_tsecr;	/* Time stamp echo reply        	*/
212 	__u16 	saw_tstamp : 1,	/* Saw TIMESTAMP on last packet		*/
213 		tstamp_ok : 1,	/* TIMESTAMP seen on SYN packet		*/
214 		dsack : 1,	/* D-SACK is scheduled			*/
215 		wscale_ok : 1,	/* Wscale seen on SYN packet		*/
216 		sack_ok : 4,	/* SACK seen on SYN packet		*/
217 		snd_wscale : 4,	/* Window scaling received from sender	*/
218 		rcv_wscale : 4;	/* Window scaling to send to receiver	*/
219 /*	SACKs data	*/
220 	__u8	eff_sacks;	/* Size of SACK array to send with next packet */
221 	__u8	num_sacks;	/* Number of SACK blocks		*/
222 	__u16	user_mss;  	/* mss requested by user in ioctl */
223 	__u16	mss_clamp;	/* Maximal mss, negotiated at connection setup */
224 };
225 
226 struct tcp_request_sock {
227 	struct inet_request_sock req;
228 	__u32			 rcv_isn;
229 	__u32			 snt_isn;
230 };
231 
232 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
233 {
234 	return (struct tcp_request_sock *)req;
235 }
236 
237 struct tcp_sock {
238 	/* inet_sock has to be the first member of tcp_sock */
239 	struct inet_sock	inet;
240 	int	tcp_header_len;	/* Bytes of tcp header to send		*/
241 
242 /*
243  *	Header prediction flags
244  *	0x5?10 << 16 + snd_wnd in net byte order
245  */
246 	__u32	pred_flags;
247 
248 /*
249  *	RFC793 variables by their proper names. This means you can
250  *	read the code and the spec side by side (and laugh ...)
251  *	See RFC793 and RFC1122. The RFC writes these in capitals.
252  */
253  	__u32	rcv_nxt;	/* What we want to receive next 	*/
254  	__u32	snd_nxt;	/* Next sequence we send		*/
255 
256  	__u32	snd_una;	/* First byte we want an ack for	*/
257  	__u32	snd_sml;	/* Last byte of the most recently transmitted small packet */
258 	__u32	rcv_tstamp;	/* timestamp of last received ACK (for keepalives) */
259 	__u32	lsndtime;	/* timestamp of last sent data packet (for restart window) */
260 	struct tcp_bind_bucket *bind_hash;
261 	/* Delayed ACK control data */
262 	struct {
263 		__u8	pending;	/* ACK is pending */
264 		__u8	quick;		/* Scheduled number of quick acks	*/
265 		__u8	pingpong;	/* The session is interactive		*/
266 		__u8	blocked;	/* Delayed ACK was blocked by socket lock*/
267 		__u32	ato;		/* Predicted tick of soft clock		*/
268 		unsigned long timeout;	/* Currently scheduled timeout		*/
269 		__u32	lrcvtime;	/* timestamp of last received data packet*/
270 		__u16	last_seg_size;	/* Size of last incoming segment	*/
271 		__u16	rcv_mss;	/* MSS used for delayed ACK decisions	*/
272 	} ack;
273 
274 	/* Data for direct copy to user */
275 	struct {
276 		struct sk_buff_head	prequeue;
277 		struct task_struct	*task;
278 		struct iovec		*iov;
279 		int			memory;
280 		int			len;
281 	} ucopy;
282 
283 	__u32	snd_wl1;	/* Sequence for window update		*/
284 	__u32	snd_wnd;	/* The window we expect to receive	*/
285 	__u32	max_window;	/* Maximal window ever seen from peer	*/
286 	__u32	pmtu_cookie;	/* Last pmtu seen by socket		*/
287 	__u32	mss_cache;	/* Cached effective mss, not including SACKS */
288 	__u16	mss_cache_std;	/* Like mss_cache, but without TSO */
289 	__u16	ext_header_len;	/* Network protocol overhead (IP/IPv6 options) */
290 	__u8	ca_state;	/* State of fast-retransmit machine 	*/
291 	__u8	retransmits;	/* Number of unrecovered RTO timeouts.	*/
292 
293 	__u16	advmss;		/* Advertised MSS			*/
294 	__u32	window_clamp;	/* Maximal window to advertise		*/
295 	__u32	rcv_ssthresh;	/* Current window clamp			*/
296 
297 	__u32	frto_highmark;	/* snd_nxt when RTO occurred */
298 	__u8	reordering;	/* Packet reordering metric.		*/
299 	__u8	frto_counter;	/* Number of new acks after RTO */
300 
301 	__u8	unused;
302 	__u8	defer_accept;	/* User waits for some data after accept() */
303 
304 /* RTT measurement */
305 	__u32	srtt;		/* smoothed round trip time << 3	*/
306 	__u32	mdev;		/* medium deviation			*/
307 	__u32	mdev_max;	/* maximal mdev for the last rtt period	*/
308 	__u32	rttvar;		/* smoothed mdev_max			*/
309 	__u32	rtt_seq;	/* sequence number to update rttvar	*/
310 	__u32	rto;		/* retransmit timeout			*/
311 
312 	__u32	packets_out;	/* Packets which are "in flight"	*/
313 	__u32	left_out;	/* Packets which leaved network	*/
314 	__u32	retrans_out;	/* Retransmitted packets out		*/
315 	__u8	backoff;	/* backoff				*/
316 /*
317  *      Options received (usually on last packet, some only on SYN packets).
318  */
319 	__u8	nonagle;	/* Disable Nagle algorithm?             */
320 	__u8	keepalive_probes; /* num of allowed keep alive probes	*/
321 
322 	__u8	probes_out;	/* unanswered 0 window probes		*/
323 	struct tcp_options_received rx_opt;
324 
325 /*
326  *	Slow start and congestion control (see also Nagle, and Karn & Partridge)
327  */
328  	__u32	snd_ssthresh;	/* Slow start size threshold		*/
329  	__u32	snd_cwnd;	/* Sending congestion window		*/
330  	__u16	snd_cwnd_cnt;	/* Linear increase counter		*/
331 	__u16	snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
332 	__u32	snd_cwnd_used;
333 	__u32	snd_cwnd_stamp;
334 
335 	/* Two commonly used timers in both sender and receiver paths. */
336 	unsigned long		timeout;
337  	struct timer_list	retransmit_timer;	/* Resend (no ack)	*/
338  	struct timer_list	delack_timer;		/* Ack delay 		*/
339 
340 	struct sk_buff_head	out_of_order_queue; /* Out of order segments go here */
341 
342 	struct tcp_func		*af_specific;	/* Operations which are AF_INET{4,6} specific	*/
343 
344  	__u32	rcv_wnd;	/* Current receiver window		*/
345 	__u32	rcv_wup;	/* rcv_nxt on last window update sent	*/
346 	__u32	write_seq;	/* Tail(+1) of data held in tcp send buffer */
347 	__u32	pushed_seq;	/* Last pushed seq, required to talk to windows */
348 	__u32	copied_seq;	/* Head of yet unread data		*/
349 
350 /*	SACKs data	*/
351 	struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
352 	struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
353 
354 	__u8	syn_retries;	/* num of allowed syn retries */
355 	__u8	ecn_flags;	/* ECN status bits.			*/
356 	__u16	prior_ssthresh; /* ssthresh saved at recovery start	*/
357 	__u32	lost_out;	/* Lost packets			*/
358 	__u32	sacked_out;	/* SACK'd packets			*/
359 	__u32	fackets_out;	/* FACK'd packets			*/
360 	__u32	high_seq;	/* snd_nxt at onset of congestion	*/
361 
362 	__u32	retrans_stamp;	/* Timestamp of the last retransmit,
363 				 * also used in SYN-SENT to remember stamp of
364 				 * the first SYN. */
365 	__u32	undo_marker;	/* tracking retrans started here. */
366 	int	undo_retrans;	/* number of undoable retransmissions. */
367 	__u32	urg_seq;	/* Seq of received urgent pointer */
368 	__u16	urg_data;	/* Saved octet of OOB data and control flags */
369 	__u8	pending;	/* Scheduled timer event	*/
370 	__u8	urg_mode;	/* In urgent mode		*/
371 	__u32	snd_up;		/* Urgent pointer		*/
372 
373 	__u32	total_retrans;	/* Total retransmits for entire connection */
374 
375 	struct request_sock_queue accept_queue; /* FIFO of established children */
376 
377 	unsigned int		keepalive_time;	  /* time before keep alive takes place */
378 	unsigned int		keepalive_intvl;  /* time interval between keep alive probes */
379 	int			linger2;
380 
381 	unsigned long last_synq_overflow;
382 
383 /* Receiver side RTT estimation */
384 	struct {
385 		__u32	rtt;
386 		__u32	seq;
387 		__u32	time;
388 	} rcv_rtt_est;
389 
390 /* Receiver queue space */
391 	struct {
392 		int	space;
393 		__u32	seq;
394 		__u32	time;
395 	} rcvq_space;
396 
397 	/* Pluggable TCP congestion control hook */
398 	struct tcp_congestion_ops *ca_ops;
399 	u32	ca_priv[16];
400 #define TCP_CA_PRIV_SIZE	(16*sizeof(u32))
401 };
402 
403 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
404 {
405 	return (struct tcp_sock *)sk;
406 }
407 
408 static inline void *tcp_ca(const struct tcp_sock *tp)
409 {
410 	return (void *) tp->ca_priv;
411 }
412 
413 #endif
414 
415 #endif	/* _LINUX_TCP_H */
416