]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - include/linux/tcp.h
Merge tag 'pm-5.14-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
[mirror_ubuntu-jammy-kernel.git] / include / linux / tcp.h
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
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 #ifndef _LINUX_TCP_H
14 #define _LINUX_TCP_H
15
16
17 #include <linux/skbuff.h>
18 #include <linux/win_minmax.h>
19 #include <net/sock.h>
20 #include <net/inet_connection_sock.h>
21 #include <net/inet_timewait_sock.h>
22 #include <uapi/linux/tcp.h>
23
24 static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb)
25 {
26 return (struct tcphdr *)skb_transport_header(skb);
27 }
28
29 static inline unsigned int __tcp_hdrlen(const struct tcphdr *th)
30 {
31 return th->doff * 4;
32 }
33
34 static inline unsigned int tcp_hdrlen(const struct sk_buff *skb)
35 {
36 return __tcp_hdrlen(tcp_hdr(skb));
37 }
38
39 static inline struct tcphdr *inner_tcp_hdr(const struct sk_buff *skb)
40 {
41 return (struct tcphdr *)skb_inner_transport_header(skb);
42 }
43
44 static inline unsigned int inner_tcp_hdrlen(const struct sk_buff *skb)
45 {
46 return inner_tcp_hdr(skb)->doff * 4;
47 }
48
49 static inline unsigned int tcp_optlen(const struct sk_buff *skb)
50 {
51 return (tcp_hdr(skb)->doff - 5) * 4;
52 }
53
54 /* TCP Fast Open */
55 #define TCP_FASTOPEN_COOKIE_MIN 4 /* Min Fast Open Cookie size in bytes */
56 #define TCP_FASTOPEN_COOKIE_MAX 16 /* Max Fast Open Cookie size in bytes */
57 #define TCP_FASTOPEN_COOKIE_SIZE 8 /* the size employed by this impl. */
58
59 /* TCP Fast Open Cookie as stored in memory */
60 struct tcp_fastopen_cookie {
61 __le64 val[DIV_ROUND_UP(TCP_FASTOPEN_COOKIE_MAX, sizeof(u64))];
62 s8 len;
63 bool exp; /* In RFC6994 experimental option format */
64 };
65
66 /* This defines a selective acknowledgement block. */
67 struct tcp_sack_block_wire {
68 __be32 start_seq;
69 __be32 end_seq;
70 };
71
72 struct tcp_sack_block {
73 u32 start_seq;
74 u32 end_seq;
75 };
76
77 /*These are used to set the sack_ok field in struct tcp_options_received */
78 #define TCP_SACK_SEEN (1 << 0) /*1 = peer is SACK capable, */
79 #define TCP_DSACK_SEEN (1 << 2) /*1 = DSACK was received from peer*/
80
81 struct tcp_options_received {
82 /* PAWS/RTTM data */
83 int ts_recent_stamp;/* Time we stored ts_recent (for aging) */
84 u32 ts_recent; /* Time stamp to echo next */
85 u32 rcv_tsval; /* Time stamp value */
86 u32 rcv_tsecr; /* Time stamp echo reply */
87 u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */
88 tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */
89 dsack : 1, /* D-SACK is scheduled */
90 wscale_ok : 1, /* Wscale seen on SYN packet */
91 sack_ok : 3, /* SACK seen on SYN packet */
92 smc_ok : 1, /* SMC seen on SYN packet */
93 snd_wscale : 4, /* Window scaling received from sender */
94 rcv_wscale : 4; /* Window scaling to send to receiver */
95 u8 saw_unknown:1, /* Received unknown option */
96 unused:7;
97 u8 num_sacks; /* Number of SACK blocks */
98 u16 user_mss; /* mss requested by user in ioctl */
99 u16 mss_clamp; /* Maximal mss, negotiated at connection setup */
100 };
101
102 static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
103 {
104 rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
105 rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
106 #if IS_ENABLED(CONFIG_SMC)
107 rx_opt->smc_ok = 0;
108 #endif
109 }
110
111 /* This is the max number of SACKS that we'll generate and process. It's safe
112 * to increase this, although since:
113 * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
114 * only four options will fit in a standard TCP header */
115 #define TCP_NUM_SACKS 4
116
117 struct tcp_request_sock_ops;
118
119 struct tcp_request_sock {
120 struct inet_request_sock req;
121 const struct tcp_request_sock_ops *af_specific;
122 u64 snt_synack; /* first SYNACK sent time */
123 bool tfo_listener;
124 bool is_mptcp;
125 #if IS_ENABLED(CONFIG_MPTCP)
126 bool drop_req;
127 #endif
128 u32 txhash;
129 u32 rcv_isn;
130 u32 snt_isn;
131 u32 ts_off;
132 u32 last_oow_ack_time; /* last SYNACK */
133 u32 rcv_nxt; /* the ack # by SYNACK. For
134 * FastOpen it's the seq#
135 * after data-in-SYN.
136 */
137 u8 syn_tos;
138 };
139
140 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
141 {
142 return (struct tcp_request_sock *)req;
143 }
144
145 struct tcp_sock {
146 /* inet_connection_sock has to be the first member of tcp_sock */
147 struct inet_connection_sock inet_conn;
148 u16 tcp_header_len; /* Bytes of tcp header to send */
149 u16 gso_segs; /* Max number of segs per GSO packet */
150
151 /*
152 * Header prediction flags
153 * 0x5?10 << 16 + snd_wnd in net byte order
154 */
155 __be32 pred_flags;
156
157 /*
158 * RFC793 variables by their proper names. This means you can
159 * read the code and the spec side by side (and laugh ...)
160 * See RFC793 and RFC1122. The RFC writes these in capitals.
161 */
162 u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
163 * sum(delta(rcv_nxt)), or how many bytes
164 * were acked.
165 */
166 u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
167 * total number of segments in.
168 */
169 u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
170 * total number of data segments in.
171 */
172 u32 rcv_nxt; /* What we want to receive next */
173 u32 copied_seq; /* Head of yet unread data */
174 u32 rcv_wup; /* rcv_nxt on last window update sent */
175 u32 snd_nxt; /* Next sequence we send */
176 u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
177 * The total number of segments sent.
178 */
179 u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
180 * total number of data segments sent.
181 */
182 u64 bytes_sent; /* RFC4898 tcpEStatsPerfHCDataOctetsOut
183 * total number of data bytes sent.
184 */
185 u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
186 * sum(delta(snd_una)), or how many bytes
187 * were acked.
188 */
189 u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
190 * total number of DSACK blocks received
191 */
192 u32 snd_una; /* First byte we want an ack for */
193 u32 snd_sml; /* Last byte of the most recently transmitted small packet */
194 u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */
195 u32 lsndtime; /* timestamp of last sent data packet (for restart window) */
196 u32 last_oow_ack_time; /* timestamp of last out-of-window ACK */
197 u32 compressed_ack_rcv_nxt;
198
199 u32 tsoffset; /* timestamp offset */
200
201 struct list_head tsq_node; /* anchor in tsq_tasklet.head list */
202 struct list_head tsorted_sent_queue; /* time-sorted sent but un-SACKed skbs */
203
204 u32 snd_wl1; /* Sequence for window update */
205 u32 snd_wnd; /* The window we expect to receive */
206 u32 max_window; /* Maximal window ever seen from peer */
207 u32 mss_cache; /* Cached effective mss, not including SACKS */
208
209 u32 window_clamp; /* Maximal window to advertise */
210 u32 rcv_ssthresh; /* Current window clamp */
211
212 /* Information of the most recently (s)acked skb */
213 struct tcp_rack {
214 u64 mstamp; /* (Re)sent time of the skb */
215 u32 rtt_us; /* Associated RTT */
216 u32 end_seq; /* Ending TCP sequence of the skb */
217 u32 last_delivered; /* tp->delivered at last reo_wnd adj */
218 u8 reo_wnd_steps; /* Allowed reordering window */
219 #define TCP_RACK_RECOVERY_THRESH 16
220 u8 reo_wnd_persist:5, /* No. of recovery since last adj */
221 dsack_seen:1, /* Whether DSACK seen after last adj */
222 advanced:1; /* mstamp advanced since last lost marking */
223 } rack;
224 u16 advmss; /* Advertised MSS */
225 u8 compressed_ack;
226 u8 dup_ack_counter:2,
227 tlp_retrans:1, /* TLP is a retransmission */
228 unused:5;
229 u32 chrono_start; /* Start time in jiffies of a TCP chrono */
230 u32 chrono_stat[3]; /* Time in jiffies for chrono_stat stats */
231 u8 chrono_type:2, /* current chronograph type */
232 rate_app_limited:1, /* rate_{delivered,interval_us} limited? */
233 fastopen_connect:1, /* FASTOPEN_CONNECT sockopt */
234 fastopen_no_cookie:1, /* Allow send/recv SYN+data without a cookie */
235 is_sack_reneg:1, /* in recovery from loss with SACK reneg? */
236 fastopen_client_fail:2; /* reason why fastopen failed */
237 u8 nonagle : 4,/* Disable Nagle algorithm? */
238 thin_lto : 1,/* Use linear timeouts for thin streams */
239 recvmsg_inq : 1,/* Indicate # of bytes in queue upon recvmsg */
240 repair : 1,
241 frto : 1;/* F-RTO (RFC5682) activated in CA_Loss */
242 u8 repair_queue;
243 u8 save_syn:2, /* Save headers of SYN packet */
244 syn_data:1, /* SYN includes data */
245 syn_fastopen:1, /* SYN includes Fast Open option */
246 syn_fastopen_exp:1,/* SYN includes Fast Open exp. option */
247 syn_fastopen_ch:1, /* Active TFO re-enabling probe */
248 syn_data_acked:1,/* data in SYN is acked by SYN-ACK */
249 is_cwnd_limited:1;/* forward progress limited by snd_cwnd? */
250 u32 tlp_high_seq; /* snd_nxt at the time of TLP */
251
252 u32 tcp_tx_delay; /* delay (in usec) added to TX packets */
253 u64 tcp_wstamp_ns; /* departure time for next sent data packet */
254 u64 tcp_clock_cache; /* cache last tcp_clock_ns() (see tcp_mstamp_refresh()) */
255
256 /* RTT measurement */
257 u64 tcp_mstamp; /* most recent packet received/sent */
258 u32 srtt_us; /* smoothed round trip time << 3 in usecs */
259 u32 mdev_us; /* medium deviation */
260 u32 mdev_max_us; /* maximal mdev for the last rtt period */
261 u32 rttvar_us; /* smoothed mdev_max */
262 u32 rtt_seq; /* sequence number to update rttvar */
263 struct minmax rtt_min;
264
265 u32 packets_out; /* Packets which are "in flight" */
266 u32 retrans_out; /* Retransmitted packets out */
267 u32 max_packets_out; /* max packets_out in last window */
268 u32 max_packets_seq; /* right edge of max_packets_out flight */
269
270 u16 urg_data; /* Saved octet of OOB data and control flags */
271 u8 ecn_flags; /* ECN status bits. */
272 u8 keepalive_probes; /* num of allowed keep alive probes */
273 u32 reordering; /* Packet reordering metric. */
274 u32 reord_seen; /* number of data packet reordering events */
275 u32 snd_up; /* Urgent pointer */
276
277 /*
278 * Options received (usually on last packet, some only on SYN packets).
279 */
280 struct tcp_options_received rx_opt;
281
282 /*
283 * Slow start and congestion control (see also Nagle, and Karn & Partridge)
284 */
285 u32 snd_ssthresh; /* Slow start size threshold */
286 u32 snd_cwnd; /* Sending congestion window */
287 u32 snd_cwnd_cnt; /* Linear increase counter */
288 u32 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
289 u32 snd_cwnd_used;
290 u32 snd_cwnd_stamp;
291 u32 prior_cwnd; /* cwnd right before starting loss recovery */
292 u32 prr_delivered; /* Number of newly delivered packets to
293 * receiver in Recovery. */
294 u32 prr_out; /* Total number of pkts sent during Recovery. */
295 u32 delivered; /* Total data packets delivered incl. rexmits */
296 u32 delivered_ce; /* Like the above but only ECE marked packets */
297 u32 lost; /* Total data packets lost incl. rexmits */
298 u32 app_limited; /* limited until "delivered" reaches this val */
299 u64 first_tx_mstamp; /* start of window send phase */
300 u64 delivered_mstamp; /* time we reached "delivered" */
301 u32 rate_delivered; /* saved rate sample: packets delivered */
302 u32 rate_interval_us; /* saved rate sample: time elapsed */
303
304 u32 rcv_wnd; /* Current receiver window */
305 u32 write_seq; /* Tail(+1) of data held in tcp send buffer */
306 u32 notsent_lowat; /* TCP_NOTSENT_LOWAT */
307 u32 pushed_seq; /* Last pushed seq, required to talk to windows */
308 u32 lost_out; /* Lost packets */
309 u32 sacked_out; /* SACK'd packets */
310
311 struct hrtimer pacing_timer;
312 struct hrtimer compressed_ack_timer;
313
314 /* from STCP, retrans queue hinting */
315 struct sk_buff* lost_skb_hint;
316 struct sk_buff *retransmit_skb_hint;
317
318 /* OOO segments go in this rbtree. Socket lock must be held. */
319 struct rb_root out_of_order_queue;
320 struct sk_buff *ooo_last_skb; /* cache rb_last(out_of_order_queue) */
321
322 /* SACKs data, these 2 need to be together (see tcp_options_write) */
323 struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
324 struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
325
326 struct tcp_sack_block recv_sack_cache[4];
327
328 struct sk_buff *highest_sack; /* skb just after the highest
329 * skb with SACKed bit set
330 * (validity guaranteed only if
331 * sacked_out > 0)
332 */
333
334 int lost_cnt_hint;
335
336 u32 prior_ssthresh; /* ssthresh saved at recovery start */
337 u32 high_seq; /* snd_nxt at onset of congestion */
338
339 u32 retrans_stamp; /* Timestamp of the last retransmit,
340 * also used in SYN-SENT to remember stamp of
341 * the first SYN. */
342 u32 undo_marker; /* snd_una upon a new recovery episode. */
343 int undo_retrans; /* number of undoable retransmissions. */
344 u64 bytes_retrans; /* RFC4898 tcpEStatsPerfOctetsRetrans
345 * Total data bytes retransmitted
346 */
347 u32 total_retrans; /* Total retransmits for entire connection */
348
349 u32 urg_seq; /* Seq of received urgent pointer */
350 unsigned int keepalive_time; /* time before keep alive takes place */
351 unsigned int keepalive_intvl; /* time interval between keep alive probes */
352
353 int linger2;
354
355
356 /* Sock_ops bpf program related variables */
357 #ifdef CONFIG_BPF
358 u8 bpf_sock_ops_cb_flags; /* Control calling BPF programs
359 * values defined in uapi/linux/tcp.h
360 */
361 #define BPF_SOCK_OPS_TEST_FLAG(TP, ARG) (TP->bpf_sock_ops_cb_flags & ARG)
362 #else
363 #define BPF_SOCK_OPS_TEST_FLAG(TP, ARG) 0
364 #endif
365
366 u16 timeout_rehash; /* Timeout-triggered rehash attempts */
367
368 u32 rcv_ooopack; /* Received out-of-order packets, for tcpinfo */
369
370 /* Receiver side RTT estimation */
371 u32 rcv_rtt_last_tsecr;
372 struct {
373 u32 rtt_us;
374 u32 seq;
375 u64 time;
376 } rcv_rtt_est;
377
378 /* Receiver queue space */
379 struct {
380 u32 space;
381 u32 seq;
382 u64 time;
383 } rcvq_space;
384
385 /* TCP-specific MTU probe information. */
386 struct {
387 u32 probe_seq_start;
388 u32 probe_seq_end;
389 } mtu_probe;
390 u32 mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG
391 * while socket was owned by user.
392 */
393 #if IS_ENABLED(CONFIG_MPTCP)
394 bool is_mptcp;
395 #endif
396 #if IS_ENABLED(CONFIG_SMC)
397 bool syn_smc; /* SYN includes SMC */
398 #endif
399
400 #ifdef CONFIG_TCP_MD5SIG
401 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
402 const struct tcp_sock_af_ops *af_specific;
403
404 /* TCP MD5 Signature Option information */
405 struct tcp_md5sig_info __rcu *md5sig_info;
406 #endif
407
408 /* TCP fastopen related information */
409 struct tcp_fastopen_request *fastopen_req;
410 /* fastopen_rsk points to request_sock that resulted in this big
411 * socket. Used to retransmit SYNACKs etc.
412 */
413 struct request_sock __rcu *fastopen_rsk;
414 struct saved_syn *saved_syn;
415 };
416
417 enum tsq_enum {
418 TSQ_THROTTLED,
419 TSQ_QUEUED,
420 TCP_TSQ_DEFERRED, /* tcp_tasklet_func() found socket was owned */
421 TCP_WRITE_TIMER_DEFERRED, /* tcp_write_timer() found socket was owned */
422 TCP_DELACK_TIMER_DEFERRED, /* tcp_delack_timer() found socket was owned */
423 TCP_MTU_REDUCED_DEFERRED, /* tcp_v{4|6}_err() could not call
424 * tcp_v{4|6}_mtu_reduced()
425 */
426 };
427
428 enum tsq_flags {
429 TSQF_THROTTLED = (1UL << TSQ_THROTTLED),
430 TSQF_QUEUED = (1UL << TSQ_QUEUED),
431 TCPF_TSQ_DEFERRED = (1UL << TCP_TSQ_DEFERRED),
432 TCPF_WRITE_TIMER_DEFERRED = (1UL << TCP_WRITE_TIMER_DEFERRED),
433 TCPF_DELACK_TIMER_DEFERRED = (1UL << TCP_DELACK_TIMER_DEFERRED),
434 TCPF_MTU_REDUCED_DEFERRED = (1UL << TCP_MTU_REDUCED_DEFERRED),
435 };
436
437 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
438 {
439 return (struct tcp_sock *)sk;
440 }
441
442 struct tcp_timewait_sock {
443 struct inet_timewait_sock tw_sk;
444 #define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxt
445 #define tw_snd_nxt tw_sk.__tw_common.skc_tw_snd_nxt
446 u32 tw_rcv_wnd;
447 u32 tw_ts_offset;
448 u32 tw_ts_recent;
449
450 /* The time we sent the last out-of-window ACK: */
451 u32 tw_last_oow_ack_time;
452
453 int tw_ts_recent_stamp;
454 u32 tw_tx_delay;
455 #ifdef CONFIG_TCP_MD5SIG
456 struct tcp_md5sig_key *tw_md5_key;
457 #endif
458 };
459
460 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
461 {
462 return (struct tcp_timewait_sock *)sk;
463 }
464
465 static inline bool tcp_passive_fastopen(const struct sock *sk)
466 {
467 return sk->sk_state == TCP_SYN_RECV &&
468 rcu_access_pointer(tcp_sk(sk)->fastopen_rsk) != NULL;
469 }
470
471 static inline void fastopen_queue_tune(struct sock *sk, int backlog)
472 {
473 struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue;
474 int somaxconn = READ_ONCE(sock_net(sk)->core.sysctl_somaxconn);
475
476 queue->fastopenq.max_qlen = min_t(unsigned int, backlog, somaxconn);
477 }
478
479 static inline void tcp_move_syn(struct tcp_sock *tp,
480 struct request_sock *req)
481 {
482 tp->saved_syn = req->saved_syn;
483 req->saved_syn = NULL;
484 }
485
486 static inline void tcp_saved_syn_free(struct tcp_sock *tp)
487 {
488 kfree(tp->saved_syn);
489 tp->saved_syn = NULL;
490 }
491
492 static inline u32 tcp_saved_syn_len(const struct saved_syn *saved_syn)
493 {
494 return saved_syn->mac_hdrlen + saved_syn->network_hdrlen +
495 saved_syn->tcp_hdrlen;
496 }
497
498 struct sk_buff *tcp_get_timestamping_opt_stats(const struct sock *sk,
499 const struct sk_buff *orig_skb,
500 const struct sk_buff *ack_skb);
501
502 static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss)
503 {
504 /* We use READ_ONCE() here because socket might not be locked.
505 * This happens for listeners.
506 */
507 u16 user_mss = READ_ONCE(tp->rx_opt.user_mss);
508
509 return (user_mss && user_mss < mss) ? user_mss : mss;
510 }
511
512 int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
513 int shiftlen);
514
515 void tcp_sock_set_cork(struct sock *sk, bool on);
516 int tcp_sock_set_keepcnt(struct sock *sk, int val);
517 int tcp_sock_set_keepidle_locked(struct sock *sk, int val);
518 int tcp_sock_set_keepidle(struct sock *sk, int val);
519 int tcp_sock_set_keepintvl(struct sock *sk, int val);
520 void tcp_sock_set_nodelay(struct sock *sk);
521 void tcp_sock_set_quickack(struct sock *sk, int val);
522 int tcp_sock_set_syncnt(struct sock *sk, int val);
523 void tcp_sock_set_user_timeout(struct sock *sk, u32 val);
524
525 #endif /* _LINUX_TCP_H */