1 /* SPDX-License-Identifier: GPL-2.0-or-later */
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.
7 * Definitions for the TCP protocol.
9 * Version: @(#)tcp.h 1.0.2 04/28/93
11 * Author: Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
17 #include <linux/skbuff.h>
18 #include <linux/win_minmax.h>
20 #include <net/inet_connection_sock.h>
21 #include <net/inet_timewait_sock.h>
22 #include <uapi/linux/tcp.h>
24 static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb)
26 return (struct tcphdr *)skb_transport_header(skb);
29 static inline unsigned int __tcp_hdrlen(const struct tcphdr *th)
34 static inline unsigned int tcp_hdrlen(const struct sk_buff *skb)
36 return __tcp_hdrlen(tcp_hdr(skb));
39 static inline struct tcphdr *inner_tcp_hdr(const struct sk_buff *skb)
41 return (struct tcphdr *)skb_inner_transport_header(skb);
44 static inline unsigned int inner_tcp_hdrlen(const struct sk_buff *skb)
46 return inner_tcp_hdr(skb)->doff * 4;
49 static inline unsigned int tcp_optlen(const struct sk_buff *skb)
51 return (tcp_hdr(skb)->doff - 5) * 4;
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. */
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))];
63 bool exp; /* In RFC6994 experimental option format */
66 /* This defines a selective acknowledgement block. */
67 struct tcp_sack_block_wire {
72 struct tcp_sack_block {
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*/
81 #if IS_ENABLED(CONFIG_MPTCP)
82 struct mptcp_options_received {
107 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
108 struct in6_addr addr6;
116 struct tcp_options_received {
118 int ts_recent_stamp;/* Time we stored ts_recent (for aging) */
119 u32 ts_recent; /* Time stamp to echo next */
120 u32 rcv_tsval; /* Time stamp value */
121 u32 rcv_tsecr; /* Time stamp echo reply */
122 u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */
123 tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */
124 dsack : 1, /* D-SACK is scheduled */
125 wscale_ok : 1, /* Wscale seen on SYN packet */
126 sack_ok : 3, /* SACK seen on SYN packet */
127 smc_ok : 1, /* SMC seen on SYN packet */
128 snd_wscale : 4, /* Window scaling received from sender */
129 rcv_wscale : 4; /* Window scaling to send to receiver */
130 u8 num_sacks; /* Number of SACK blocks */
131 u16 user_mss; /* mss requested by user in ioctl */
132 u16 mss_clamp; /* Maximal mss, negotiated at connection setup */
133 #if IS_ENABLED(CONFIG_MPTCP)
134 struct mptcp_options_received mptcp;
138 static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
140 rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
141 rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
142 #if IS_ENABLED(CONFIG_SMC)
145 #if IS_ENABLED(CONFIG_MPTCP)
146 rx_opt->mptcp.mp_capable = 0;
147 rx_opt->mptcp.mp_join = 0;
148 rx_opt->mptcp.add_addr = 0;
149 rx_opt->mptcp.rm_addr = 0;
150 rx_opt->mptcp.dss = 0;
154 /* This is the max number of SACKS that we'll generate and process. It's safe
155 * to increase this, although since:
156 * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
157 * only four options will fit in a standard TCP header */
158 #define TCP_NUM_SACKS 4
160 struct tcp_request_sock_ops;
162 struct tcp_request_sock {
163 struct inet_request_sock req;
164 const struct tcp_request_sock_ops *af_specific;
165 u64 snt_synack; /* first SYNACK sent time */
172 u32 last_oow_ack_time; /* last SYNACK */
173 u32 rcv_nxt; /* the ack # by SYNACK. For
174 * FastOpen it's the seq#
179 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
181 return (struct tcp_request_sock *)req;
185 /* inet_connection_sock has to be the first member of tcp_sock */
186 struct inet_connection_sock inet_conn;
187 u16 tcp_header_len; /* Bytes of tcp header to send */
188 u16 gso_segs; /* Max number of segs per GSO packet */
191 * Header prediction flags
192 * 0x5?10 << 16 + snd_wnd in net byte order
197 * RFC793 variables by their proper names. This means you can
198 * read the code and the spec side by side (and laugh ...)
199 * See RFC793 and RFC1122. The RFC writes these in capitals.
201 u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
202 * sum(delta(rcv_nxt)), or how many bytes
205 u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
206 * total number of segments in.
208 u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
209 * total number of data segments in.
211 u32 rcv_nxt; /* What we want to receive next */
212 u32 copied_seq; /* Head of yet unread data */
213 u32 rcv_wup; /* rcv_nxt on last window update sent */
214 u32 snd_nxt; /* Next sequence we send */
215 u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
216 * The total number of segments sent.
218 u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
219 * total number of data segments sent.
221 u64 bytes_sent; /* RFC4898 tcpEStatsPerfHCDataOctetsOut
222 * total number of data bytes sent.
224 u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
225 * sum(delta(snd_una)), or how many bytes
228 u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
229 * total number of DSACK blocks received
231 u32 snd_una; /* First byte we want an ack for */
232 u32 snd_sml; /* Last byte of the most recently transmitted small packet */
233 u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */
234 u32 lsndtime; /* timestamp of last sent data packet (for restart window) */
235 u32 last_oow_ack_time; /* timestamp of last out-of-window ACK */
236 u32 compressed_ack_rcv_nxt;
238 u32 tsoffset; /* timestamp offset */
240 struct list_head tsq_node; /* anchor in tsq_tasklet.head list */
241 struct list_head tsorted_sent_queue; /* time-sorted sent but un-SACKed skbs */
243 u32 snd_wl1; /* Sequence for window update */
244 u32 snd_wnd; /* The window we expect to receive */
245 u32 max_window; /* Maximal window ever seen from peer */
246 u32 mss_cache; /* Cached effective mss, not including SACKS */
248 u32 window_clamp; /* Maximal window to advertise */
249 u32 rcv_ssthresh; /* Current window clamp */
251 /* Information of the most recently (s)acked skb */
253 u64 mstamp; /* (Re)sent time of the skb */
254 u32 rtt_us; /* Associated RTT */
255 u32 end_seq; /* Ending TCP sequence of the skb */
256 u32 last_delivered; /* tp->delivered at last reo_wnd adj */
257 u8 reo_wnd_steps; /* Allowed reordering window */
258 #define TCP_RACK_RECOVERY_THRESH 16
259 u8 reo_wnd_persist:5, /* No. of recovery since last adj */
260 dsack_seen:1, /* Whether DSACK seen after last adj */
261 advanced:1; /* mstamp advanced since last lost marking */
263 u16 advmss; /* Advertised MSS */
265 u32 chrono_start; /* Start time in jiffies of a TCP chrono */
266 u32 chrono_stat[3]; /* Time in jiffies for chrono_stat stats */
267 u8 chrono_type:2, /* current chronograph type */
268 rate_app_limited:1, /* rate_{delivered,interval_us} limited? */
269 fastopen_connect:1, /* FASTOPEN_CONNECT sockopt */
270 fastopen_no_cookie:1, /* Allow send/recv SYN+data without a cookie */
271 is_sack_reneg:1, /* in recovery from loss with SACK reneg? */
272 fastopen_client_fail:2; /* reason why fastopen failed */
273 u8 nonagle : 4,/* Disable Nagle algorithm? */
274 thin_lto : 1,/* Use linear timeouts for thin streams */
275 recvmsg_inq : 1,/* Indicate # of bytes in queue upon recvmsg */
277 frto : 1;/* F-RTO (RFC5682) activated in CA_Loss */
279 u8 syn_data:1, /* SYN includes data */
280 syn_fastopen:1, /* SYN includes Fast Open option */
281 syn_fastopen_exp:1,/* SYN includes Fast Open exp. option */
282 syn_fastopen_ch:1, /* Active TFO re-enabling probe */
283 syn_data_acked:1,/* data in SYN is acked by SYN-ACK */
284 save_syn:1, /* Save headers of SYN packet */
285 is_cwnd_limited:1,/* forward progress limited by snd_cwnd? */
286 syn_smc:1; /* SYN includes SMC */
287 u32 tlp_high_seq; /* snd_nxt at the time of TLP retransmit. */
289 u32 tcp_tx_delay; /* delay (in usec) added to TX packets */
290 u64 tcp_wstamp_ns; /* departure time for next sent data packet */
291 u64 tcp_clock_cache; /* cache last tcp_clock_ns() (see tcp_mstamp_refresh()) */
293 /* RTT measurement */
294 u64 tcp_mstamp; /* most recent packet received/sent */
295 u32 srtt_us; /* smoothed round trip time << 3 in usecs */
296 u32 mdev_us; /* medium deviation */
297 u32 mdev_max_us; /* maximal mdev for the last rtt period */
298 u32 rttvar_us; /* smoothed mdev_max */
299 u32 rtt_seq; /* sequence number to update rttvar */
300 struct minmax rtt_min;
302 u32 packets_out; /* Packets which are "in flight" */
303 u32 retrans_out; /* Retransmitted packets out */
304 u32 max_packets_out; /* max packets_out in last window */
305 u32 max_packets_seq; /* right edge of max_packets_out flight */
307 u16 urg_data; /* Saved octet of OOB data and control flags */
308 u8 ecn_flags; /* ECN status bits. */
309 u8 keepalive_probes; /* num of allowed keep alive probes */
310 u32 reordering; /* Packet reordering metric. */
311 u32 reord_seen; /* number of data packet reordering events */
312 u32 snd_up; /* Urgent pointer */
315 * Options received (usually on last packet, some only on SYN packets).
317 struct tcp_options_received rx_opt;
320 * Slow start and congestion control (see also Nagle, and Karn & Partridge)
322 u32 snd_ssthresh; /* Slow start size threshold */
323 u32 snd_cwnd; /* Sending congestion window */
324 u32 snd_cwnd_cnt; /* Linear increase counter */
325 u32 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
328 u32 prior_cwnd; /* cwnd right before starting loss recovery */
329 u32 prr_delivered; /* Number of newly delivered packets to
330 * receiver in Recovery. */
331 u32 prr_out; /* Total number of pkts sent during Recovery. */
332 u32 delivered; /* Total data packets delivered incl. rexmits */
333 u32 delivered_ce; /* Like the above but only ECE marked packets */
334 u32 lost; /* Total data packets lost incl. rexmits */
335 u32 app_limited; /* limited until "delivered" reaches this val */
336 u64 first_tx_mstamp; /* start of window send phase */
337 u64 delivered_mstamp; /* time we reached "delivered" */
338 u32 rate_delivered; /* saved rate sample: packets delivered */
339 u32 rate_interval_us; /* saved rate sample: time elapsed */
341 u32 rcv_wnd; /* Current receiver window */
342 u32 write_seq; /* Tail(+1) of data held in tcp send buffer */
343 u32 notsent_lowat; /* TCP_NOTSENT_LOWAT */
344 u32 pushed_seq; /* Last pushed seq, required to talk to windows */
345 u32 lost_out; /* Lost packets */
346 u32 sacked_out; /* SACK'd packets */
348 struct hrtimer pacing_timer;
349 struct hrtimer compressed_ack_timer;
351 /* from STCP, retrans queue hinting */
352 struct sk_buff* lost_skb_hint;
353 struct sk_buff *retransmit_skb_hint;
355 /* OOO segments go in this rbtree. Socket lock must be held. */
356 struct rb_root out_of_order_queue;
357 struct sk_buff *ooo_last_skb; /* cache rb_last(out_of_order_queue) */
359 /* SACKs data, these 2 need to be together (see tcp_options_write) */
360 struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
361 struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
363 struct tcp_sack_block recv_sack_cache[4];
365 struct sk_buff *highest_sack; /* skb just after the highest
366 * skb with SACKed bit set
367 * (validity guaranteed only if
373 u32 prior_ssthresh; /* ssthresh saved at recovery start */
374 u32 high_seq; /* snd_nxt at onset of congestion */
376 u32 retrans_stamp; /* Timestamp of the last retransmit,
377 * also used in SYN-SENT to remember stamp of
379 u32 undo_marker; /* snd_una upon a new recovery episode. */
380 int undo_retrans; /* number of undoable retransmissions. */
381 u64 bytes_retrans; /* RFC4898 tcpEStatsPerfOctetsRetrans
382 * Total data bytes retransmitted
384 u32 total_retrans; /* Total retransmits for entire connection */
386 u32 urg_seq; /* Seq of received urgent pointer */
387 unsigned int keepalive_time; /* time before keep alive takes place */
388 unsigned int keepalive_intvl; /* time interval between keep alive probes */
393 /* Sock_ops bpf program related variables */
395 u8 bpf_sock_ops_cb_flags; /* Control calling BPF programs
396 * values defined in uapi/linux/tcp.h
398 #define BPF_SOCK_OPS_TEST_FLAG(TP, ARG) (TP->bpf_sock_ops_cb_flags & ARG)
400 #define BPF_SOCK_OPS_TEST_FLAG(TP, ARG) 0
403 u16 timeout_rehash; /* Timeout-triggered rehash attempts */
405 u32 rcv_ooopack; /* Received out-of-order packets, for tcpinfo */
407 /* Receiver side RTT estimation */
408 u32 rcv_rtt_last_tsecr;
415 /* Receiver queue space */
422 /* TCP-specific MTU probe information. */
427 u32 mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG
428 * while socket was owned by user.
430 #if IS_ENABLED(CONFIG_MPTCP)
434 #ifdef CONFIG_TCP_MD5SIG
435 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
436 const struct tcp_sock_af_ops *af_specific;
438 /* TCP MD5 Signature Option information */
439 struct tcp_md5sig_info __rcu *md5sig_info;
442 /* TCP fastopen related information */
443 struct tcp_fastopen_request *fastopen_req;
444 /* fastopen_rsk points to request_sock that resulted in this big
445 * socket. Used to retransmit SYNACKs etc.
447 struct request_sock __rcu *fastopen_rsk;
454 TCP_TSQ_DEFERRED, /* tcp_tasklet_func() found socket was owned */
455 TCP_WRITE_TIMER_DEFERRED, /* tcp_write_timer() found socket was owned */
456 TCP_DELACK_TIMER_DEFERRED, /* tcp_delack_timer() found socket was owned */
457 TCP_MTU_REDUCED_DEFERRED, /* tcp_v{4|6}_err() could not call
458 * tcp_v{4|6}_mtu_reduced()
463 TSQF_THROTTLED = (1UL << TSQ_THROTTLED),
464 TSQF_QUEUED = (1UL << TSQ_QUEUED),
465 TCPF_TSQ_DEFERRED = (1UL << TCP_TSQ_DEFERRED),
466 TCPF_WRITE_TIMER_DEFERRED = (1UL << TCP_WRITE_TIMER_DEFERRED),
467 TCPF_DELACK_TIMER_DEFERRED = (1UL << TCP_DELACK_TIMER_DEFERRED),
468 TCPF_MTU_REDUCED_DEFERRED = (1UL << TCP_MTU_REDUCED_DEFERRED),
471 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
473 return (struct tcp_sock *)sk;
476 struct tcp_timewait_sock {
477 struct inet_timewait_sock tw_sk;
478 #define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxt
479 #define tw_snd_nxt tw_sk.__tw_common.skc_tw_snd_nxt
484 /* The time we sent the last out-of-window ACK: */
485 u32 tw_last_oow_ack_time;
487 int tw_ts_recent_stamp;
489 #ifdef CONFIG_TCP_MD5SIG
490 struct tcp_md5sig_key *tw_md5_key;
494 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
496 return (struct tcp_timewait_sock *)sk;
499 static inline bool tcp_passive_fastopen(const struct sock *sk)
501 return sk->sk_state == TCP_SYN_RECV &&
502 rcu_access_pointer(tcp_sk(sk)->fastopen_rsk) != NULL;
505 static inline void fastopen_queue_tune(struct sock *sk, int backlog)
507 struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue;
508 int somaxconn = READ_ONCE(sock_net(sk)->core.sysctl_somaxconn);
510 queue->fastopenq.max_qlen = min_t(unsigned int, backlog, somaxconn);
513 static inline void tcp_move_syn(struct tcp_sock *tp,
514 struct request_sock *req)
516 tp->saved_syn = req->saved_syn;
517 req->saved_syn = NULL;
520 static inline void tcp_saved_syn_free(struct tcp_sock *tp)
522 kfree(tp->saved_syn);
523 tp->saved_syn = NULL;
526 struct sk_buff *tcp_get_timestamping_opt_stats(const struct sock *sk);
528 static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss)
530 /* We use READ_ONCE() here because socket might not be locked.
531 * This happens for listeners.
533 u16 user_mss = READ_ONCE(tp->rx_opt.user_mss);
535 return (user_mss && user_mss < mss) ? user_mss : mss;
538 int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
541 #endif /* _LINUX_TCP_H */