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
{
62 u8 val
[TCP_FASTOPEN_COOKIE_MAX
];
63 #if IS_ENABLED(CONFIG_IPV6)
68 bool exp
; /* In RFC6994 experimental option format */
71 /* This defines a selective acknowledgement block. */
72 struct tcp_sack_block_wire
{
77 struct tcp_sack_block
{
82 /*These are used to set the sack_ok field in struct tcp_options_received */
83 #define TCP_SACK_SEEN (1 << 0) /*1 = peer is SACK capable, */
84 #define TCP_DSACK_SEEN (1 << 2) /*1 = DSACK was received from peer*/
86 struct tcp_options_received
{
88 int ts_recent_stamp
;/* Time we stored ts_recent (for aging) */
89 u32 ts_recent
; /* Time stamp to echo next */
90 u32 rcv_tsval
; /* Time stamp value */
91 u32 rcv_tsecr
; /* Time stamp echo reply */
92 u16 saw_tstamp
: 1, /* Saw TIMESTAMP on last packet */
93 tstamp_ok
: 1, /* TIMESTAMP seen on SYN packet */
94 dsack
: 1, /* D-SACK is scheduled */
95 wscale_ok
: 1, /* Wscale seen on SYN packet */
96 sack_ok
: 3, /* SACK seen on SYN packet */
97 smc_ok
: 1, /* SMC seen on SYN packet */
98 snd_wscale
: 4, /* Window scaling received from sender */
99 rcv_wscale
: 4; /* Window scaling to send to receiver */
100 u8 num_sacks
; /* Number of SACK blocks */
101 u16 user_mss
; /* mss requested by user in ioctl */
102 u16 mss_clamp
; /* Maximal mss, negotiated at connection setup */
105 static inline void tcp_clear_options(struct tcp_options_received
*rx_opt
)
107 rx_opt
->tstamp_ok
= rx_opt
->sack_ok
= 0;
108 rx_opt
->wscale_ok
= rx_opt
->snd_wscale
= 0;
109 #if IS_ENABLED(CONFIG_SMC)
114 /* This is the max number of SACKS that we'll generate and process. It's safe
115 * to increase this, although since:
116 * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
117 * only four options will fit in a standard TCP header */
118 #define TCP_NUM_SACKS 4
120 struct tcp_request_sock_ops
;
122 struct tcp_request_sock
{
123 struct inet_request_sock req
;
124 const struct tcp_request_sock_ops
*af_specific
;
125 u64 snt_synack
; /* first SYNACK sent time */
131 u32 last_oow_ack_time
; /* last SYNACK */
132 u32 rcv_nxt
; /* the ack # by SYNACK. For
133 * FastOpen it's the seq#
138 static inline struct tcp_request_sock
*tcp_rsk(const struct request_sock
*req
)
140 return (struct tcp_request_sock
*)req
;
144 /* inet_connection_sock has to be the first member of tcp_sock */
145 struct inet_connection_sock inet_conn
;
146 u16 tcp_header_len
; /* Bytes of tcp header to send */
147 u16 gso_segs
; /* Max number of segs per GSO packet */
150 * Header prediction flags
151 * 0x5?10 << 16 + snd_wnd in net byte order
156 * RFC793 variables by their proper names. This means you can
157 * read the code and the spec side by side (and laugh ...)
158 * See RFC793 and RFC1122. The RFC writes these in capitals.
160 u64 bytes_received
; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
161 * sum(delta(rcv_nxt)), or how many bytes
164 u32 segs_in
; /* RFC4898 tcpEStatsPerfSegsIn
165 * total number of segments in.
167 u32 data_segs_in
; /* RFC4898 tcpEStatsPerfDataSegsIn
168 * total number of data segments in.
170 u32 rcv_nxt
; /* What we want to receive next */
171 u32 copied_seq
; /* Head of yet unread data */
172 u32 rcv_wup
; /* rcv_nxt on last window update sent */
173 u32 snd_nxt
; /* Next sequence we send */
174 u32 segs_out
; /* RFC4898 tcpEStatsPerfSegsOut
175 * The total number of segments sent.
177 u32 data_segs_out
; /* RFC4898 tcpEStatsPerfDataSegsOut
178 * total number of data segments sent.
180 u64 bytes_sent
; /* RFC4898 tcpEStatsPerfHCDataOctetsOut
181 * total number of data bytes sent.
183 u64 bytes_acked
; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
184 * sum(delta(snd_una)), or how many bytes
187 u32 dsack_dups
; /* RFC4898 tcpEStatsStackDSACKDups
188 * total number of DSACK blocks received
190 u32 snd_una
; /* First byte we want an ack for */
191 u32 snd_sml
; /* Last byte of the most recently transmitted small packet */
192 u32 rcv_tstamp
; /* timestamp of last received ACK (for keepalives) */
193 u32 lsndtime
; /* timestamp of last sent data packet (for restart window) */
194 u32 last_oow_ack_time
; /* timestamp of last out-of-window ACK */
195 u32 compressed_ack_rcv_nxt
;
197 u32 tsoffset
; /* timestamp offset */
199 struct list_head tsq_node
; /* anchor in tsq_tasklet.head list */
200 struct list_head tsorted_sent_queue
; /* time-sorted sent but un-SACKed skbs */
202 u32 snd_wl1
; /* Sequence for window update */
203 u32 snd_wnd
; /* The window we expect to receive */
204 u32 max_window
; /* Maximal window ever seen from peer */
205 u32 mss_cache
; /* Cached effective mss, not including SACKS */
207 u32 window_clamp
; /* Maximal window to advertise */
208 u32 rcv_ssthresh
; /* Current window clamp */
210 /* Information of the most recently (s)acked skb */
212 u64 mstamp
; /* (Re)sent time of the skb */
213 u32 rtt_us
; /* Associated RTT */
214 u32 end_seq
; /* Ending TCP sequence of the skb */
215 u32 last_delivered
; /* tp->delivered at last reo_wnd adj */
216 u8 reo_wnd_steps
; /* Allowed reordering window */
217 #define TCP_RACK_RECOVERY_THRESH 16
218 u8 reo_wnd_persist
:5, /* No. of recovery since last adj */
219 dsack_seen
:1, /* Whether DSACK seen after last adj */
220 advanced
:1; /* mstamp advanced since last lost marking */
222 u16 advmss
; /* Advertised MSS */
224 u32 chrono_start
; /* Start time in jiffies of a TCP chrono */
225 u32 chrono_stat
[3]; /* Time in jiffies for chrono_stat stats */
226 u8 chrono_type
:2, /* current chronograph type */
227 rate_app_limited
:1, /* rate_{delivered,interval_us} limited? */
228 fastopen_connect
:1, /* FASTOPEN_CONNECT sockopt */
229 fastopen_no_cookie
:1, /* Allow send/recv SYN+data without a cookie */
230 is_sack_reneg
:1, /* in recovery from loss with SACK reneg? */
232 u8 nonagle
: 4,/* Disable Nagle algorithm? */
233 thin_lto
: 1,/* Use linear timeouts for thin streams */
234 recvmsg_inq
: 1,/* Indicate # of bytes in queue upon recvmsg */
236 frto
: 1;/* F-RTO (RFC5682) activated in CA_Loss */
238 u8 syn_data
:1, /* SYN includes data */
239 syn_fastopen
:1, /* SYN includes Fast Open option */
240 syn_fastopen_exp
:1,/* SYN includes Fast Open exp. option */
241 syn_fastopen_ch
:1, /* Active TFO re-enabling probe */
242 syn_data_acked
:1,/* data in SYN is acked by SYN-ACK */
243 save_syn
:1, /* Save headers of SYN packet */
244 is_cwnd_limited
:1,/* forward progress limited by snd_cwnd? */
245 syn_smc
:1; /* SYN includes SMC */
246 u32 tlp_high_seq
; /* snd_nxt at the time of TLP retransmit. */
248 u64 tcp_wstamp_ns
; /* departure time for next sent data packet */
249 u64 tcp_clock_cache
; /* cache last tcp_clock_ns() (see tcp_mstamp_refresh()) */
251 /* RTT measurement */
252 u64 tcp_mstamp
; /* most recent packet received/sent */
253 u32 srtt_us
; /* smoothed round trip time << 3 in usecs */
254 u32 mdev_us
; /* medium deviation */
255 u32 mdev_max_us
; /* maximal mdev for the last rtt period */
256 u32 rttvar_us
; /* smoothed mdev_max */
257 u32 rtt_seq
; /* sequence number to update rttvar */
258 struct minmax rtt_min
;
260 u32 packets_out
; /* Packets which are "in flight" */
261 u32 retrans_out
; /* Retransmitted packets out */
262 u32 max_packets_out
; /* max packets_out in last window */
263 u32 max_packets_seq
; /* right edge of max_packets_out flight */
265 u16 urg_data
; /* Saved octet of OOB data and control flags */
266 u8 ecn_flags
; /* ECN status bits. */
267 u8 keepalive_probes
; /* num of allowed keep alive probes */
268 u32 reordering
; /* Packet reordering metric. */
269 u32 reord_seen
; /* number of data packet reordering events */
270 u32 snd_up
; /* Urgent pointer */
273 * Options received (usually on last packet, some only on SYN packets).
275 struct tcp_options_received rx_opt
;
278 * Slow start and congestion control (see also Nagle, and Karn & Partridge)
280 u32 snd_ssthresh
; /* Slow start size threshold */
281 u32 snd_cwnd
; /* Sending congestion window */
282 u32 snd_cwnd_cnt
; /* Linear increase counter */
283 u32 snd_cwnd_clamp
; /* Do not allow snd_cwnd to grow above this */
286 u32 prior_cwnd
; /* cwnd right before starting loss recovery */
287 u32 prr_delivered
; /* Number of newly delivered packets to
288 * receiver in Recovery. */
289 u32 prr_out
; /* Total number of pkts sent during Recovery. */
290 u32 delivered
; /* Total data packets delivered incl. rexmits */
291 u32 delivered_ce
; /* Like the above but only ECE marked packets */
292 u32 lost
; /* Total data packets lost incl. rexmits */
293 u32 app_limited
; /* limited until "delivered" reaches this val */
294 u64 first_tx_mstamp
; /* start of window send phase */
295 u64 delivered_mstamp
; /* time we reached "delivered" */
296 u32 rate_delivered
; /* saved rate sample: packets delivered */
297 u32 rate_interval_us
; /* saved rate sample: time elapsed */
299 u32 rcv_wnd
; /* Current receiver window */
300 u32 write_seq
; /* Tail(+1) of data held in tcp send buffer */
301 u32 notsent_lowat
; /* TCP_NOTSENT_LOWAT */
302 u32 pushed_seq
; /* Last pushed seq, required to talk to windows */
303 u32 lost_out
; /* Lost packets */
304 u32 sacked_out
; /* SACK'd packets */
306 struct hrtimer pacing_timer
;
307 struct hrtimer compressed_ack_timer
;
309 /* from STCP, retrans queue hinting */
310 struct sk_buff
* lost_skb_hint
;
311 struct sk_buff
*retransmit_skb_hint
;
313 /* OOO segments go in this rbtree. Socket lock must be held. */
314 struct rb_root out_of_order_queue
;
315 struct sk_buff
*ooo_last_skb
; /* cache rb_last(out_of_order_queue) */
317 /* SACKs data, these 2 need to be together (see tcp_options_write) */
318 struct tcp_sack_block duplicate_sack
[1]; /* D-SACK block */
319 struct tcp_sack_block selective_acks
[4]; /* The SACKS themselves*/
321 struct tcp_sack_block recv_sack_cache
[4];
323 struct sk_buff
*highest_sack
; /* skb just after the highest
324 * skb with SACKed bit set
325 * (validity guaranteed only if
331 u32 prior_ssthresh
; /* ssthresh saved at recovery start */
332 u32 high_seq
; /* snd_nxt at onset of congestion */
334 u32 retrans_stamp
; /* Timestamp of the last retransmit,
335 * also used in SYN-SENT to remember stamp of
337 u32 undo_marker
; /* snd_una upon a new recovery episode. */
338 int undo_retrans
; /* number of undoable retransmissions. */
339 u64 bytes_retrans
; /* RFC4898 tcpEStatsPerfOctetsRetrans
340 * Total data bytes retransmitted
342 u32 total_retrans
; /* Total retransmits for entire connection */
344 u32 urg_seq
; /* Seq of received urgent pointer */
345 unsigned int keepalive_time
; /* time before keep alive takes place */
346 unsigned int keepalive_intvl
; /* time interval between keep alive probes */
351 /* Sock_ops bpf program related variables */
353 u8 bpf_sock_ops_cb_flags
; /* Control calling BPF programs
354 * values defined in uapi/linux/tcp.h
356 #define BPF_SOCK_OPS_TEST_FLAG(TP, ARG) (TP->bpf_sock_ops_cb_flags & ARG)
358 #define BPF_SOCK_OPS_TEST_FLAG(TP, ARG) 0
361 /* Receiver side RTT estimation */
362 u32 rcv_rtt_last_tsecr
;
369 /* Receiver queue space */
376 /* TCP-specific MTU probe information. */
381 u32 mtu_info
; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG
382 * while socket was owned by user.
385 #ifdef CONFIG_TCP_MD5SIG
386 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
387 const struct tcp_sock_af_ops
*af_specific
;
389 /* TCP MD5 Signature Option information */
390 struct tcp_md5sig_info __rcu
*md5sig_info
;
393 /* TCP fastopen related information */
394 struct tcp_fastopen_request
*fastopen_req
;
395 /* fastopen_rsk points to request_sock that resulted in this big
396 * socket. Used to retransmit SYNACKs etc.
398 struct request_sock
*fastopen_rsk
;
405 TCP_TSQ_DEFERRED
, /* tcp_tasklet_func() found socket was owned */
406 TCP_WRITE_TIMER_DEFERRED
, /* tcp_write_timer() found socket was owned */
407 TCP_DELACK_TIMER_DEFERRED
, /* tcp_delack_timer() found socket was owned */
408 TCP_MTU_REDUCED_DEFERRED
, /* tcp_v{4|6}_err() could not call
409 * tcp_v{4|6}_mtu_reduced()
414 TSQF_THROTTLED
= (1UL << TSQ_THROTTLED
),
415 TSQF_QUEUED
= (1UL << TSQ_QUEUED
),
416 TCPF_TSQ_DEFERRED
= (1UL << TCP_TSQ_DEFERRED
),
417 TCPF_WRITE_TIMER_DEFERRED
= (1UL << TCP_WRITE_TIMER_DEFERRED
),
418 TCPF_DELACK_TIMER_DEFERRED
= (1UL << TCP_DELACK_TIMER_DEFERRED
),
419 TCPF_MTU_REDUCED_DEFERRED
= (1UL << TCP_MTU_REDUCED_DEFERRED
),
422 static inline struct tcp_sock
*tcp_sk(const struct sock
*sk
)
424 return (struct tcp_sock
*)sk
;
427 struct tcp_timewait_sock
{
428 struct inet_timewait_sock tw_sk
;
429 #define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxt
430 #define tw_snd_nxt tw_sk.__tw_common.skc_tw_snd_nxt
435 /* The time we sent the last out-of-window ACK: */
436 u32 tw_last_oow_ack_time
;
438 int tw_ts_recent_stamp
;
439 #ifdef CONFIG_TCP_MD5SIG
440 struct tcp_md5sig_key
*tw_md5_key
;
444 static inline struct tcp_timewait_sock
*tcp_twsk(const struct sock
*sk
)
446 return (struct tcp_timewait_sock
*)sk
;
449 static inline bool tcp_passive_fastopen(const struct sock
*sk
)
451 return (sk
->sk_state
== TCP_SYN_RECV
&&
452 tcp_sk(sk
)->fastopen_rsk
!= NULL
);
455 static inline void fastopen_queue_tune(struct sock
*sk
, int backlog
)
457 struct request_sock_queue
*queue
= &inet_csk(sk
)->icsk_accept_queue
;
458 int somaxconn
= READ_ONCE(sock_net(sk
)->core
.sysctl_somaxconn
);
460 queue
->fastopenq
.max_qlen
= min_t(unsigned int, backlog
, somaxconn
);
463 static inline void tcp_move_syn(struct tcp_sock
*tp
,
464 struct request_sock
*req
)
466 tp
->saved_syn
= req
->saved_syn
;
467 req
->saved_syn
= NULL
;
470 static inline void tcp_saved_syn_free(struct tcp_sock
*tp
)
472 kfree(tp
->saved_syn
);
473 tp
->saved_syn
= NULL
;
476 struct sk_buff
*tcp_get_timestamping_opt_stats(const struct sock
*sk
);
478 static inline u16
tcp_mss_clamp(const struct tcp_sock
*tp
, u16 mss
)
480 /* We use READ_ONCE() here because socket might not be locked.
481 * This happens for listeners.
483 u16 user_mss
= READ_ONCE(tp
->rx_opt
.user_mss
);
485 return (user_mss
&& user_mss
< mss
) ? user_mss
: mss
;
488 int tcp_skb_shift(struct sk_buff
*to
, struct sk_buff
*from
, int pcount
,
491 #endif /* _LINUX_TCP_H */