master
1/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the TCP protocol.
8 *
9 * Version: @(#)tcp.h 1.0.2 04/28/93
10 *
11 * Author: Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 *
13 * This program is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU General Public License
15 * as published by the Free Software Foundation; either version
16 * 2 of the License, or (at your option) any later version.
17 */
18#ifndef _LINUX_TCP_H
19#define _LINUX_TCP_H
20
21#include <linux/types.h>
22#include <asm/byteorder.h>
23#include <linux/socket.h>
24
25struct tcphdr {
26 __be16 source;
27 __be16 dest;
28 __be32 seq;
29 __be32 ack_seq;
30#if defined(__LITTLE_ENDIAN_BITFIELD)
31 __u16 ae:1,
32 res1:3,
33 doff:4,
34 fin:1,
35 syn:1,
36 rst:1,
37 psh:1,
38 ack:1,
39 urg:1,
40 ece:1,
41 cwr:1;
42#elif defined(__BIG_ENDIAN_BITFIELD)
43 __u16 doff:4,
44 res1:3,
45 ae:1,
46 cwr:1,
47 ece:1,
48 urg:1,
49 ack:1,
50 psh:1,
51 rst:1,
52 syn:1,
53 fin:1;
54#else
55#error "Adjust your <asm/byteorder.h> defines"
56#endif
57 __be16 window;
58 __sum16 check;
59 __be16 urg_ptr;
60};
61
62/*
63 * The union cast uses a gcc extension to avoid aliasing problems
64 * (union is compatible to any of its members)
65 * This means this part of the code is -fstrict-aliasing safe now.
66 */
67union tcp_word_hdr {
68 struct tcphdr hdr;
69 __be32 words[5];
70};
71
72#define tcp_flag_word(tp) (((union tcp_word_hdr *)(tp))->words[3])
73
74enum {
75 TCP_FLAG_AE = __constant_cpu_to_be32(0x01000000),
76 TCP_FLAG_CWR = __constant_cpu_to_be32(0x00800000),
77 TCP_FLAG_ECE = __constant_cpu_to_be32(0x00400000),
78 TCP_FLAG_URG = __constant_cpu_to_be32(0x00200000),
79 TCP_FLAG_ACK = __constant_cpu_to_be32(0x00100000),
80 TCP_FLAG_PSH = __constant_cpu_to_be32(0x00080000),
81 TCP_FLAG_RST = __constant_cpu_to_be32(0x00040000),
82 TCP_FLAG_SYN = __constant_cpu_to_be32(0x00020000),
83 TCP_FLAG_FIN = __constant_cpu_to_be32(0x00010000),
84 TCP_RESERVED_BITS = __constant_cpu_to_be32(0x0E000000),
85 TCP_DATA_OFFSET = __constant_cpu_to_be32(0xF0000000)
86};
87
88/*
89 * TCP general constants
90 */
91#define TCP_MSS_DEFAULT 536U /* IPv4 (RFC1122, RFC2581) */
92#define TCP_MSS_DESIRED 1220U /* IPv6 (tunneled), EDNS0 (RFC3226) */
93
94/* TCP socket options */
95#define TCP_NODELAY 1 /* Turn off Nagle's algorithm. */
96#define TCP_MAXSEG 2 /* Limit MSS */
97#define TCP_CORK 3 /* Never send partially complete segments */
98#define TCP_KEEPIDLE 4 /* Start keeplives after this period */
99#define TCP_KEEPINTVL 5 /* Interval between keepalives */
100#define TCP_KEEPCNT 6 /* Number of keepalives before death */
101#define TCP_SYNCNT 7 /* Number of SYN retransmits */
102#define TCP_LINGER2 8 /* Life time of orphaned FIN-WAIT-2 state */
103#define TCP_DEFER_ACCEPT 9 /* Wake up listener only when data arrive */
104#define TCP_WINDOW_CLAMP 10 /* Bound advertised window */
105#define TCP_INFO 11 /* Information about this connection. */
106#define TCP_QUICKACK 12 /* Block/reenable quick acks */
107#define TCP_CONGESTION 13 /* Congestion control algorithm */
108#define TCP_MD5SIG 14 /* TCP MD5 Signature (RFC2385) */
109#define TCP_THIN_LINEAR_TIMEOUTS 16 /* Use linear timeouts for thin streams*/
110#define TCP_THIN_DUPACK 17 /* Fast retrans. after 1 dupack */
111#define TCP_USER_TIMEOUT 18 /* How long for loss retry before timeout */
112#define TCP_REPAIR 19 /* TCP sock is under repair right now */
113#define TCP_REPAIR_QUEUE 20
114#define TCP_QUEUE_SEQ 21
115#define TCP_REPAIR_OPTIONS 22
116#define TCP_FASTOPEN 23 /* Enable FastOpen on listeners */
117#define TCP_TIMESTAMP 24
118#define TCP_NOTSENT_LOWAT 25 /* limit number of unsent bytes in write queue */
119#define TCP_CC_INFO 26 /* Get Congestion Control (optional) info */
120#define TCP_SAVE_SYN 27 /* Record SYN headers for new connections */
121#define TCP_SAVED_SYN 28 /* Get SYN headers recorded for connection */
122#define TCP_REPAIR_WINDOW 29 /* Get/set window parameters */
123#define TCP_FASTOPEN_CONNECT 30 /* Attempt FastOpen with connect */
124#define TCP_ULP 31 /* Attach a ULP to a TCP connection */
125#define TCP_MD5SIG_EXT 32 /* TCP MD5 Signature with extensions */
126#define TCP_FASTOPEN_KEY 33 /* Set the key for Fast Open (cookie) */
127#define TCP_FASTOPEN_NO_COOKIE 34 /* Enable TFO without a TFO cookie */
128#define TCP_ZEROCOPY_RECEIVE 35
129#define TCP_INQ 36 /* Notify bytes available to read as a cmsg on read */
130
131#define TCP_CM_INQ TCP_INQ
132
133#define TCP_TX_DELAY 37 /* delay outgoing packets by XX usec */
134
135#define TCP_AO_ADD_KEY 38 /* Add/Set MKT */
136#define TCP_AO_DEL_KEY 39 /* Delete MKT */
137#define TCP_AO_INFO 40 /* Set/list TCP-AO per-socket options */
138#define TCP_AO_GET_KEYS 41 /* List MKT(s) */
139#define TCP_AO_REPAIR 42 /* Get/Set SNEs and ISNs */
140
141#define TCP_IS_MPTCP 43 /* Is MPTCP being used? */
142#define TCP_RTO_MAX_MS 44 /* max rto time in ms */
143#define TCP_RTO_MIN_US 45 /* min rto time in us */
144#define TCP_DELACK_MAX_US 46 /* max delayed ack time in us */
145
146#define TCP_REPAIR_ON 1
147#define TCP_REPAIR_OFF 0
148#define TCP_REPAIR_OFF_NO_WP -1 /* Turn off without window probes */
149
150struct tcp_repair_opt {
151 __u32 opt_code;
152 __u32 opt_val;
153};
154
155struct tcp_repair_window {
156 __u32 snd_wl1;
157 __u32 snd_wnd;
158 __u32 max_window;
159
160 __u32 rcv_wnd;
161 __u32 rcv_wup;
162};
163
164enum {
165 TCP_NO_QUEUE,
166 TCP_RECV_QUEUE,
167 TCP_SEND_QUEUE,
168 TCP_QUEUES_NR,
169};
170
171/* why fastopen failed from client perspective */
172enum tcp_fastopen_client_fail {
173 TFO_STATUS_UNSPEC, /* catch-all */
174 TFO_COOKIE_UNAVAILABLE, /* if not in TFO_CLIENT_NO_COOKIE mode */
175 TFO_DATA_NOT_ACKED, /* SYN-ACK did not ack SYN data */
176 TFO_SYN_RETRANSMITTED, /* SYN-ACK did not ack SYN data after timeout */
177};
178
179/* for TCP_INFO socket option */
180#define TCPI_OPT_TIMESTAMPS 1
181#define TCPI_OPT_SACK 2
182#define TCPI_OPT_WSCALE 4
183#define TCPI_OPT_ECN 8 /* ECN was negociated at TCP session init */
184#define TCPI_OPT_ECN_SEEN 16 /* we received at least one packet with ECT */
185#define TCPI_OPT_SYN_DATA 32 /* SYN-ACK acked data in SYN sent or rcvd */
186#define TCPI_OPT_USEC_TS 64 /* usec timestamps */
187#define TCPI_OPT_TFO_CHILD 128 /* child from a Fast Open option on SYN */
188
189/*
190 * Sender's congestion state indicating normal or abnormal situations
191 * in the last round of packets sent. The state is driven by the ACK
192 * information and timer events.
193 */
194enum tcp_ca_state {
195 /*
196 * Nothing bad has been observed recently.
197 * No apparent reordering, packet loss, or ECN marks.
198 */
199 TCP_CA_Open = 0,
200#define TCPF_CA_Open (1<<TCP_CA_Open)
201 /*
202 * The sender enters disordered state when it has received DUPACKs or
203 * SACKs in the last round of packets sent. This could be due to packet
204 * loss or reordering but needs further information to confirm packets
205 * have been lost.
206 */
207 TCP_CA_Disorder = 1,
208#define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
209 /*
210 * The sender enters Congestion Window Reduction (CWR) state when it
211 * has received ACKs with ECN-ECE marks, or has experienced congestion
212 * or packet discard on the sender host (e.g. qdisc).
213 */
214 TCP_CA_CWR = 2,
215#define TCPF_CA_CWR (1<<TCP_CA_CWR)
216 /*
217 * The sender is in fast recovery and retransmitting lost packets,
218 * typically triggered by ACK events.
219 */
220 TCP_CA_Recovery = 3,
221#define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
222 /*
223 * The sender is in loss recovery triggered by retransmission timeout.
224 */
225 TCP_CA_Loss = 4
226#define TCPF_CA_Loss (1<<TCP_CA_Loss)
227};
228
229struct tcp_info {
230 __u8 tcpi_state;
231 __u8 tcpi_ca_state;
232 __u8 tcpi_retransmits;
233 __u8 tcpi_probes;
234 __u8 tcpi_backoff;
235 __u8 tcpi_options;
236 __u8 tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
237 __u8 tcpi_delivery_rate_app_limited:1, tcpi_fastopen_client_fail:2;
238
239 __u32 tcpi_rto;
240 __u32 tcpi_ato;
241 __u32 tcpi_snd_mss;
242 __u32 tcpi_rcv_mss;
243
244 __u32 tcpi_unacked;
245 __u32 tcpi_sacked;
246 __u32 tcpi_lost;
247 __u32 tcpi_retrans;
248 __u32 tcpi_fackets;
249
250 /* Times. */
251 __u32 tcpi_last_data_sent;
252 __u32 tcpi_last_ack_sent; /* Not remembered, sorry. */
253 __u32 tcpi_last_data_recv;
254 __u32 tcpi_last_ack_recv;
255
256 /* Metrics. */
257 __u32 tcpi_pmtu;
258 __u32 tcpi_rcv_ssthresh;
259 __u32 tcpi_rtt;
260 __u32 tcpi_rttvar;
261 __u32 tcpi_snd_ssthresh;
262 __u32 tcpi_snd_cwnd;
263 __u32 tcpi_advmss;
264 __u32 tcpi_reordering;
265
266 __u32 tcpi_rcv_rtt;
267 __u32 tcpi_rcv_space;
268
269 __u32 tcpi_total_retrans;
270
271 __u64 tcpi_pacing_rate;
272 __u64 tcpi_max_pacing_rate;
273 __u64 tcpi_bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked */
274 __u64 tcpi_bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived */
275 __u32 tcpi_segs_out; /* RFC4898 tcpEStatsPerfSegsOut */
276 __u32 tcpi_segs_in; /* RFC4898 tcpEStatsPerfSegsIn */
277
278 __u32 tcpi_notsent_bytes;
279 __u32 tcpi_min_rtt;
280 __u32 tcpi_data_segs_in; /* RFC4898 tcpEStatsDataSegsIn */
281 __u32 tcpi_data_segs_out; /* RFC4898 tcpEStatsDataSegsOut */
282
283 __u64 tcpi_delivery_rate;
284
285 __u64 tcpi_busy_time; /* Time (usec) busy sending data */
286 __u64 tcpi_rwnd_limited; /* Time (usec) limited by receive window */
287 __u64 tcpi_sndbuf_limited; /* Time (usec) limited by send buffer */
288
289 __u32 tcpi_delivered;
290 __u32 tcpi_delivered_ce;
291
292 __u64 tcpi_bytes_sent; /* RFC4898 tcpEStatsPerfHCDataOctetsOut */
293 __u64 tcpi_bytes_retrans; /* RFC4898 tcpEStatsPerfOctetsRetrans */
294 __u32 tcpi_dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups */
295 __u32 tcpi_reord_seen; /* reordering events seen */
296
297 __u32 tcpi_rcv_ooopack; /* Out-of-order packets received */
298
299 __u32 tcpi_snd_wnd; /* peer's advertised receive window after
300 * scaling (bytes)
301 */
302 __u32 tcpi_rcv_wnd; /* local advertised receive window after
303 * scaling (bytes)
304 */
305
306 __u32 tcpi_rehash; /* PLB or timeout triggered rehash attempts */
307
308 __u16 tcpi_total_rto; /* Total number of RTO timeouts, including
309 * SYN/SYN-ACK and recurring timeouts.
310 */
311 __u16 tcpi_total_rto_recoveries; /* Total number of RTO
312 * recoveries, including any
313 * unfinished recovery.
314 */
315 __u32 tcpi_total_rto_time; /* Total time spent in RTO recoveries
316 * in milliseconds, including any
317 * unfinished recovery.
318 */
319};
320
321/* netlink attributes types for SCM_TIMESTAMPING_OPT_STATS */
322enum {
323 TCP_NLA_PAD,
324 TCP_NLA_BUSY, /* Time (usec) busy sending data */
325 TCP_NLA_RWND_LIMITED, /* Time (usec) limited by receive window */
326 TCP_NLA_SNDBUF_LIMITED, /* Time (usec) limited by send buffer */
327 TCP_NLA_DATA_SEGS_OUT, /* Data pkts sent including retransmission */
328 TCP_NLA_TOTAL_RETRANS, /* Data pkts retransmitted */
329 TCP_NLA_PACING_RATE, /* Pacing rate in bytes per second */
330 TCP_NLA_DELIVERY_RATE, /* Delivery rate in bytes per second */
331 TCP_NLA_SND_CWND, /* Sending congestion window */
332 TCP_NLA_REORDERING, /* Reordering metric */
333 TCP_NLA_MIN_RTT, /* minimum RTT */
334 TCP_NLA_RECUR_RETRANS, /* Recurring retransmits for the current pkt */
335 TCP_NLA_DELIVERY_RATE_APP_LMT, /* delivery rate application limited ? */
336 TCP_NLA_SNDQ_SIZE, /* Data (bytes) pending in send queue */
337 TCP_NLA_CA_STATE, /* ca_state of socket */
338 TCP_NLA_SND_SSTHRESH, /* Slow start size threshold */
339 TCP_NLA_DELIVERED, /* Data pkts delivered incl. out-of-order */
340 TCP_NLA_DELIVERED_CE, /* Like above but only ones w/ CE marks */
341 TCP_NLA_BYTES_SENT, /* Data bytes sent including retransmission */
342 TCP_NLA_BYTES_RETRANS, /* Data bytes retransmitted */
343 TCP_NLA_DSACK_DUPS, /* DSACK blocks received */
344 TCP_NLA_REORD_SEEN, /* reordering events seen */
345 TCP_NLA_SRTT, /* smoothed RTT in usecs */
346 TCP_NLA_TIMEOUT_REHASH, /* Timeout-triggered rehash attempts */
347 TCP_NLA_BYTES_NOTSENT, /* Bytes in write queue not yet sent */
348 TCP_NLA_EDT, /* Earliest departure time (CLOCK_MONOTONIC) */
349 TCP_NLA_TTL, /* TTL or hop limit of a packet received */
350 TCP_NLA_REHASH, /* PLB and timeout triggered rehash attempts */
351};
352
353/* for TCP_MD5SIG socket option */
354#define TCP_MD5SIG_MAXKEYLEN 80
355
356/* tcp_md5sig extension flags for TCP_MD5SIG_EXT */
357#define TCP_MD5SIG_FLAG_PREFIX 0x1 /* address prefix length */
358#define TCP_MD5SIG_FLAG_IFINDEX 0x2 /* ifindex set */
359
360struct tcp_md5sig {
361 struct __kernel_sockaddr_storage tcpm_addr; /* address associated */
362 __u8 tcpm_flags; /* extension flags */
363 __u8 tcpm_prefixlen; /* address prefix */
364 __u16 tcpm_keylen; /* key length */
365 int tcpm_ifindex; /* device index for scope */
366 __u8 tcpm_key[TCP_MD5SIG_MAXKEYLEN]; /* key (binary) */
367};
368
369/* INET_DIAG_MD5SIG */
370struct tcp_diag_md5sig {
371 __u8 tcpm_family;
372 __u8 tcpm_prefixlen;
373 __u16 tcpm_keylen;
374 __be32 tcpm_addr[4];
375 __u8 tcpm_key[TCP_MD5SIG_MAXKEYLEN];
376};
377
378#define TCP_AO_MAXKEYLEN 80
379
380#define TCP_AO_KEYF_IFINDEX (1 << 0) /* L3 ifindex for VRF */
381#define TCP_AO_KEYF_EXCLUDE_OPT (1 << 1) /* "Indicates whether TCP
382 * options other than TCP-AO
383 * are included in the MAC
384 * calculation"
385 */
386
387struct tcp_ao_add { /* setsockopt(TCP_AO_ADD_KEY) */
388 struct __kernel_sockaddr_storage addr; /* peer's address for the key */
389 char alg_name[64]; /* crypto hash algorithm to use */
390 __s32 ifindex; /* L3 dev index for VRF */
391 __u32 set_current :1, /* set key as Current_key at once */
392 set_rnext :1, /* request it from peer with RNext_key */
393 reserved :30; /* must be 0 */
394 __u16 reserved2; /* padding, must be 0 */
395 __u8 prefix; /* peer's address prefix */
396 __u8 sndid; /* SendID for outgoing segments */
397 __u8 rcvid; /* RecvID to match for incoming seg */
398 __u8 maclen; /* length of authentication code (hash) */
399 __u8 keyflags; /* see TCP_AO_KEYF_ */
400 __u8 keylen; /* length of ::key */
401 __u8 key[TCP_AO_MAXKEYLEN];
402} __attribute__((aligned(8)));
403
404struct tcp_ao_del { /* setsockopt(TCP_AO_DEL_KEY) */
405 struct __kernel_sockaddr_storage addr; /* peer's address for the key */
406 __s32 ifindex; /* L3 dev index for VRF */
407 __u32 set_current :1, /* corresponding ::current_key */
408 set_rnext :1, /* corresponding ::rnext */
409 del_async :1, /* only valid for listen sockets */
410 reserved :29; /* must be 0 */
411 __u16 reserved2; /* padding, must be 0 */
412 __u8 prefix; /* peer's address prefix */
413 __u8 sndid; /* SendID for outgoing segments */
414 __u8 rcvid; /* RecvID to match for incoming seg */
415 __u8 current_key; /* KeyID to set as Current_key */
416 __u8 rnext; /* KeyID to set as Rnext_key */
417 __u8 keyflags; /* see TCP_AO_KEYF_ */
418} __attribute__((aligned(8)));
419
420struct tcp_ao_info_opt { /* setsockopt(TCP_AO_INFO), getsockopt(TCP_AO_INFO) */
421 /* Here 'in' is for setsockopt(), 'out' is for getsockopt() */
422 __u32 set_current :1, /* in/out: corresponding ::current_key */
423 set_rnext :1, /* in/out: corresponding ::rnext */
424 ao_required :1, /* in/out: don't accept non-AO connects */
425 set_counters :1, /* in: set/clear ::pkt_* counters */
426 accept_icmps :1, /* in/out: accept incoming ICMPs */
427 reserved :27; /* must be 0 */
428 __u16 reserved2; /* padding, must be 0 */
429 __u8 current_key; /* in/out: KeyID of Current_key */
430 __u8 rnext; /* in/out: keyid of RNext_key */
431 __u64 pkt_good; /* in/out: verified segments */
432 __u64 pkt_bad; /* in/out: failed verification */
433 __u64 pkt_key_not_found; /* in/out: could not find a key to verify */
434 __u64 pkt_ao_required; /* in/out: segments missing TCP-AO sign */
435 __u64 pkt_dropped_icmp; /* in/out: ICMPs that were ignored */
436} __attribute__((aligned(8)));
437
438struct tcp_ao_getsockopt { /* getsockopt(TCP_AO_GET_KEYS) */
439 struct __kernel_sockaddr_storage addr; /* in/out: dump keys for peer
440 * with this address/prefix
441 */
442 char alg_name[64]; /* out: crypto hash algorithm */
443 __u8 key[TCP_AO_MAXKEYLEN];
444 __u32 nkeys; /* in: size of the userspace buffer
445 * @optval, measured in @optlen - the
446 * sizeof(struct tcp_ao_getsockopt)
447 * out: number of keys that matched
448 */
449 __u16 is_current :1, /* in: match and dump Current_key,
450 * out: the dumped key is Current_key
451 */
452
453 is_rnext :1, /* in: match and dump RNext_key,
454 * out: the dumped key is RNext_key
455 */
456 get_all :1, /* in: dump all keys */
457 reserved :13; /* padding, must be 0 */
458 __u8 sndid; /* in/out: dump keys with SendID */
459 __u8 rcvid; /* in/out: dump keys with RecvID */
460 __u8 prefix; /* in/out: dump keys with address/prefix */
461 __u8 maclen; /* out: key's length of authentication
462 * code (hash)
463 */
464 __u8 keyflags; /* in/out: see TCP_AO_KEYF_ */
465 __u8 keylen; /* out: length of ::key */
466 __s32 ifindex; /* in/out: L3 dev index for VRF */
467 __u64 pkt_good; /* out: verified segments */
468 __u64 pkt_bad; /* out: segments that failed verification */
469} __attribute__((aligned(8)));
470
471struct tcp_ao_repair { /* {s,g}etsockopt(TCP_AO_REPAIR) */
472 __be32 snt_isn;
473 __be32 rcv_isn;
474 __u32 snd_sne;
475 __u32 rcv_sne;
476} __attribute__((aligned(8)));
477
478/* setsockopt(fd, IPPROTO_TCP, TCP_ZEROCOPY_RECEIVE, ...) */
479
480#define TCP_RECEIVE_ZEROCOPY_FLAG_TLB_CLEAN_HINT 0x1
481struct tcp_zerocopy_receive {
482 __u64 address; /* in: address of mapping */
483 __u32 length; /* in/out: number of bytes to map/mapped */
484 __u32 recv_skip_hint; /* out: amount of bytes to skip */
485 __u32 inq; /* out: amount of bytes in read queue */
486 __s32 err; /* out: socket error */
487 __u64 copybuf_address; /* in: copybuf address (small reads) */
488 __s32 copybuf_len; /* in/out: copybuf bytes avail/used or error */
489 __u32 flags; /* in: flags */
490 __u64 msg_control; /* ancillary data */
491 __u64 msg_controllen;
492 __u32 msg_flags;
493 __u32 reserved; /* set to 0 for now */
494};
495#endif /* _LINUX_TCP_H */