Commit | Line | Data |
---|---|---|
2874c5fd | 1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
1da177e4 LT |
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> | |
1da177e4 LT |
12 | */ |
13 | #ifndef _LINUX_TCP_H | |
14 | #define _LINUX_TCP_H | |
15 | ||
1da177e4 | 16 | |
1da177e4 | 17 | #include <linux/skbuff.h> |
64033892 | 18 | #include <linux/win_minmax.h> |
1da177e4 | 19 | #include <net/sock.h> |
463c84b9 | 20 | #include <net/inet_connection_sock.h> |
8feaf0c0 | 21 | #include <net/inet_timewait_sock.h> |
607ca46e | 22 | #include <uapi/linux/tcp.h> |
1da177e4 | 23 | |
aa8223c7 ACM |
24 | static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb) |
25 | { | |
9c70220b | 26 | return (struct tcphdr *)skb_transport_header(skb); |
aa8223c7 ACM |
27 | } |
28 | ||
d9b3fca2 CG |
29 | static inline unsigned int __tcp_hdrlen(const struct tcphdr *th) |
30 | { | |
31 | return th->doff * 4; | |
32 | } | |
33 | ||
ab6a5bb6 ACM |
34 | static inline unsigned int tcp_hdrlen(const struct sk_buff *skb) |
35 | { | |
d9b3fca2 | 36 | return __tcp_hdrlen(tcp_hdr(skb)); |
ab6a5bb6 ACM |
37 | } |
38 | ||
6a674e9c JG |
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 | ||
504148fe ED |
49 | /** |
50 | * skb_tcp_all_headers - Returns size of all headers for a TCP packet | |
51 | * @skb: buffer | |
52 | * | |
53 | * Used in TX path, for a packet known to be a TCP one. | |
54 | * | |
55 | * if (skb_is_gso(skb)) { | |
56 | * int hlen = skb_tcp_all_headers(skb); | |
57 | * ... | |
58 | */ | |
59 | static inline int skb_tcp_all_headers(const struct sk_buff *skb) | |
60 | { | |
61 | return skb_transport_offset(skb) + tcp_hdrlen(skb); | |
62 | } | |
63 | ||
64 | /** | |
65 | * skb_inner_tcp_all_headers - Returns size of all headers for an encap TCP packet | |
66 | * @skb: buffer | |
67 | * | |
68 | * Used in TX path, for a packet known to be a TCP one. | |
69 | * | |
70 | * if (skb_is_gso(skb) && skb->encapsulation) { | |
71 | * int hlen = skb_inner_tcp_all_headers(skb); | |
72 | * ... | |
73 | */ | |
74 | static inline int skb_inner_tcp_all_headers(const struct sk_buff *skb) | |
75 | { | |
76 | return skb_inner_transport_offset(skb) + inner_tcp_hdrlen(skb); | |
77 | } | |
78 | ||
ab6a5bb6 ACM |
79 | static inline unsigned int tcp_optlen(const struct sk_buff *skb) |
80 | { | |
aa8223c7 | 81 | return (tcp_hdr(skb)->doff - 5) * 4; |
ab6a5bb6 ACM |
82 | } |
83 | ||
2100c8d2 YC |
84 | /* TCP Fast Open */ |
85 | #define TCP_FASTOPEN_COOKIE_MIN 4 /* Min Fast Open Cookie size in bytes */ | |
86 | #define TCP_FASTOPEN_COOKIE_MAX 16 /* Max Fast Open Cookie size in bytes */ | |
10467163 | 87 | #define TCP_FASTOPEN_COOKIE_SIZE 8 /* the size employed by this impl. */ |
2100c8d2 YC |
88 | |
89 | /* TCP Fast Open Cookie as stored in memory */ | |
90 | struct tcp_fastopen_cookie { | |
438ac880 | 91 | __le64 val[DIV_ROUND_UP(TCP_FASTOPEN_COOKIE_MAX, sizeof(u64))]; |
2100c8d2 | 92 | s8 len; |
7f9b838b | 93 | bool exp; /* In RFC6994 experimental option format */ |
2100c8d2 YC |
94 | }; |
95 | ||
1da177e4 | 96 | /* This defines a selective acknowledgement block. */ |
269bd27e AV |
97 | struct tcp_sack_block_wire { |
98 | __be32 start_seq; | |
99 | __be32 end_seq; | |
100 | }; | |
101 | ||
1da177e4 | 102 | struct tcp_sack_block { |
3a137d20 ACM |
103 | u32 start_seq; |
104 | u32 end_seq; | |
1da177e4 LT |
105 | }; |
106 | ||
ab56222a VS |
107 | /*These are used to set the sack_ok field in struct tcp_options_received */ |
108 | #define TCP_SACK_SEEN (1 << 0) /*1 = peer is SACK capable, */ | |
ab56222a VS |
109 | #define TCP_DSACK_SEEN (1 << 2) /*1 = DSACK was received from peer*/ |
110 | ||
1da177e4 LT |
111 | struct tcp_options_received { |
112 | /* PAWS/RTTM data */ | |
cca9bab1 | 113 | int ts_recent_stamp;/* Time we stored ts_recent (for aging) */ |
3a137d20 ACM |
114 | u32 ts_recent; /* Time stamp to echo next */ |
115 | u32 rcv_tsval; /* Time stamp value */ | |
116 | u32 rcv_tsecr; /* Time stamp echo reply */ | |
117 | u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */ | |
1da177e4 LT |
118 | tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */ |
119 | dsack : 1, /* D-SACK is scheduled */ | |
120 | wscale_ok : 1, /* Wscale seen on SYN packet */ | |
60e2a778 UB |
121 | sack_ok : 3, /* SACK seen on SYN packet */ |
122 | smc_ok : 1, /* SMC seen on SYN packet */ | |
1da177e4 LT |
123 | snd_wscale : 4, /* Window scaling received from sender */ |
124 | rcv_wscale : 4; /* Window scaling to send to receiver */ | |
7656d684 MKL |
125 | u8 saw_unknown:1, /* Received unknown option */ |
126 | unused:7; | |
3a137d20 | 127 | u8 num_sacks; /* Number of SACK blocks */ |
435cf559 | 128 | u16 user_mss; /* mss requested by user in ioctl */ |
3a137d20 | 129 | u16 mss_clamp; /* Maximal mss, negotiated at connection setup */ |
1da177e4 LT |
130 | }; |
131 | ||
519855c5 WAS |
132 | static inline void tcp_clear_options(struct tcp_options_received *rx_opt) |
133 | { | |
435cf559 WAS |
134 | rx_opt->tstamp_ok = rx_opt->sack_ok = 0; |
135 | rx_opt->wscale_ok = rx_opt->snd_wscale = 0; | |
60e2a778 UB |
136 | #if IS_ENABLED(CONFIG_SMC) |
137 | rx_opt->smc_ok = 0; | |
138 | #endif | |
519855c5 WAS |
139 | } |
140 | ||
4389dded | 141 | /* This is the max number of SACKS that we'll generate and process. It's safe |
435cf559 | 142 | * to increase this, although since: |
4389dded AL |
143 | * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8) |
144 | * only four options will fit in a standard TCP header */ | |
145 | #define TCP_NUM_SACKS 4 | |
146 | ||
435cf559 WAS |
147 | struct tcp_request_sock_ops; |
148 | ||
2e6599cb | 149 | struct tcp_request_sock { |
cfb6eeb4 | 150 | struct inet_request_sock req; |
b2e4b3de | 151 | const struct tcp_request_sock_ops *af_specific; |
9a568de4 | 152 | u64 snt_synack; /* first SYNACK sent time */ |
9439ce00 | 153 | bool tfo_listener; |
cec37a6e | 154 | bool is_mptcp; |
cdbab623 | 155 | bool req_usec_ts; |
90bf4513 PA |
156 | #if IS_ENABLED(CONFIG_MPTCP) |
157 | bool drop_req; | |
158 | #endif | |
58d607d3 | 159 | u32 txhash; |
435cf559 WAS |
160 | u32 rcv_isn; |
161 | u32 snt_isn; | |
95a22cae | 162 | u32 ts_off; |
a9b2c06d | 163 | u32 last_oow_ack_time; /* last SYNACK */ |
10467163 JC |
164 | u32 rcv_nxt; /* the ack # by SYNACK. For |
165 | * FastOpen it's the seq# | |
166 | * after data-in-SYN. | |
167 | */ | |
e9b12edc | 168 | u8 syn_tos; |
06b22ef2 DS |
169 | #ifdef CONFIG_TCP_AO |
170 | u8 ao_keyid; | |
171 | u8 ao_rcv_next; | |
9396c4ee | 172 | bool used_tcp_ao; |
06b22ef2 | 173 | #endif |
2e6599cb ACM |
174 | }; |
175 | ||
60236fdd | 176 | static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req) |
2e6599cb ACM |
177 | { |
178 | return (struct tcp_request_sock *)req; | |
179 | } | |
180 | ||
06b22ef2 DS |
181 | static inline bool tcp_rsk_used_ao(const struct request_sock *req) |
182 | { | |
06b22ef2 DS |
183 | #ifndef CONFIG_TCP_AO |
184 | return false; | |
185 | #else | |
9396c4ee | 186 | return tcp_rsk(req)->used_tcp_ao; |
06b22ef2 DS |
187 | #endif |
188 | } | |
189 | ||
dfa2f048 ED |
190 | #define TCP_RMEM_TO_WIN_SCALE 8 |
191 | ||
1da177e4 | 192 | struct tcp_sock { |
d5fed5ad CL |
193 | /* Cacheline organization can be found documented in |
194 | * Documentation/networking/net_cachelines/tcp_sock.rst. | |
195 | * Please update the document when adding new fields. | |
196 | */ | |
197 | ||
463c84b9 ACM |
198 | /* inet_connection_sock has to be the first member of tcp_sock */ |
199 | struct inet_connection_sock inet_conn; | |
d5fed5ad CL |
200 | |
201 | /* TX read-mostly hotpath cache lines */ | |
202 | __cacheline_group_begin(tcp_sock_read_tx); | |
203 | /* timestamp of last sent data packet (for restart window) */ | |
204 | u32 max_window; /* Maximal window ever seen from peer */ | |
205 | u32 rcv_ssthresh; /* Current window clamp */ | |
206 | u32 reordering; /* Packet reordering metric. */ | |
207 | u32 notsent_lowat; /* TCP_NOTSENT_LOWAT */ | |
605ad7f1 | 208 | u16 gso_segs; /* Max number of segs per GSO packet */ |
d5fed5ad CL |
209 | /* from STCP, retrans queue hinting */ |
210 | struct sk_buff *lost_skb_hint; | |
211 | struct sk_buff *retransmit_skb_hint; | |
212 | __cacheline_group_end(tcp_sock_read_tx); | |
213 | ||
214 | /* TXRX read-mostly hotpath cache lines */ | |
215 | __cacheline_group_begin(tcp_sock_read_txrx); | |
216 | u32 tsoffset; /* timestamp offset */ | |
217 | u32 snd_wnd; /* The window we expect to receive */ | |
218 | u32 mss_cache; /* Cached effective mss, not including SACKS */ | |
219 | u32 snd_cwnd; /* Sending congestion window */ | |
220 | u32 prr_out; /* Total number of pkts sent during Recovery. */ | |
221 | u32 lost_out; /* Lost packets */ | |
222 | u32 sacked_out; /* SACK'd packets */ | |
223 | u16 tcp_header_len; /* Bytes of tcp header to send */ | |
119ff048 | 224 | u8 scaling_ratio; /* see tcp_win_from_space() */ |
d5fed5ad CL |
225 | u8 chrono_type : 2, /* current chronograph type */ |
226 | repair : 1, | |
666a877d | 227 | tcp_usec_ts : 1, /* TSval values in usec */ |
d5fed5ad CL |
228 | is_sack_reneg:1, /* in recovery from loss with SACK reneg? */ |
229 | is_cwnd_limited:1;/* forward progress limited by snd_cwnd? */ | |
230 | __cacheline_group_end(tcp_sock_read_txrx); | |
231 | ||
232 | /* RX read-mostly hotpath cache lines */ | |
233 | __cacheline_group_begin(tcp_sock_read_rx); | |
234 | u32 copied_seq; /* Head of yet unread data */ | |
235 | u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */ | |
236 | u32 snd_wl1; /* Sequence for window update */ | |
237 | u32 tlp_high_seq; /* snd_nxt at the time of TLP */ | |
238 | u32 rttvar_us; /* smoothed mdev_max */ | |
239 | u32 retrans_out; /* Retransmitted packets out */ | |
240 | u16 advmss; /* Advertised MSS */ | |
241 | u16 urg_data; /* Saved octet of OOB data and control flags */ | |
242 | u32 lost; /* Total data packets lost incl. rexmits */ | |
243 | struct minmax rtt_min; | |
244 | /* OOO segments go in this rbtree. Socket lock must be held. */ | |
245 | struct rb_root out_of_order_queue; | |
246 | u32 snd_ssthresh; /* Slow start size threshold */ | |
247 | __cacheline_group_end(tcp_sock_read_rx); | |
1da177e4 | 248 | |
d5fed5ad CL |
249 | /* TX read-write hotpath cache lines */ |
250 | __cacheline_group_begin(tcp_sock_write_tx) ____cacheline_aligned; | |
251 | u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut | |
252 | * The total number of segments sent. | |
253 | */ | |
254 | u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut | |
255 | * total number of data segments sent. | |
256 | */ | |
257 | u64 bytes_sent; /* RFC4898 tcpEStatsPerfHCDataOctetsOut | |
258 | * total number of data bytes sent. | |
259 | */ | |
260 | u32 snd_sml; /* Last byte of the most recently transmitted small packet */ | |
261 | u32 chrono_start; /* Start time in jiffies of a TCP chrono */ | |
262 | u32 chrono_stat[3]; /* Time in jiffies for chrono_stat stats */ | |
263 | u32 write_seq; /* Tail(+1) of data held in tcp send buffer */ | |
264 | u32 pushed_seq; /* Last pushed seq, required to talk to windows */ | |
265 | u32 lsndtime; | |
266 | u32 mdev_us; /* medium deviation */ | |
99123622 | 267 | u32 rtt_seq; /* sequence number to update rttvar */ |
d5fed5ad CL |
268 | u64 tcp_wstamp_ns; /* departure time for next sent data packet */ |
269 | u64 tcp_clock_cache; /* cache last tcp_clock_ns() (see tcp_mstamp_refresh()) */ | |
270 | u64 tcp_mstamp; /* most recent packet received/sent */ | |
d5fed5ad CL |
271 | struct list_head tsorted_sent_queue; /* time-sorted sent but un-SACKed skbs */ |
272 | struct sk_buff *highest_sack; /* skb just after the highest | |
273 | * skb with SACKed bit set | |
274 | * (validity guaranteed only if | |
275 | * sacked_out > 0) | |
276 | */ | |
277 | u8 ecn_flags; /* ECN status bits. */ | |
278 | __cacheline_group_end(tcp_sock_write_tx); | |
279 | ||
280 | /* TXRX read-write hotpath cache lines */ | |
281 | __cacheline_group_begin(tcp_sock_write_txrx); | |
31770e34 FW |
282 | /* |
283 | * Header prediction flags | |
284 | * 0x5?10 << 16 + snd_wnd in net byte order | |
285 | */ | |
286 | __be32 pred_flags; | |
d5fed5ad CL |
287 | u32 rcv_nxt; /* What we want to receive next */ |
288 | u32 snd_nxt; /* Next sequence we send */ | |
289 | u32 snd_una; /* First byte we want an ack for */ | |
290 | u32 window_clamp; /* Maximal window to advertise */ | |
291 | u32 srtt_us; /* smoothed round trip time << 3 in usecs */ | |
292 | u32 packets_out; /* Packets which are "in flight" */ | |
293 | u32 snd_up; /* Urgent pointer */ | |
294 | u32 delivered; /* Total data packets delivered incl. rexmits */ | |
295 | u32 delivered_ce; /* Like the above but only ECE marked packets */ | |
296 | u32 app_limited; /* limited until "delivered" reaches this val */ | |
297 | u32 rcv_wnd; /* Current receiver window */ | |
1da177e4 | 298 | /* |
d5fed5ad | 299 | * Options received (usually on last packet, some only on SYN packets). |
1da177e4 | 300 | */ |
d5fed5ad CL |
301 | struct tcp_options_received rx_opt; |
302 | u8 nonagle : 4,/* Disable Nagle algorithm? */ | |
303 | rate_app_limited:1; /* rate_{delivered,interval_us} limited? */ | |
304 | __cacheline_group_end(tcp_sock_write_txrx); | |
305 | ||
306 | /* RX read-write hotpath cache lines */ | |
345a6e26 | 307 | __cacheline_group_begin(tcp_sock_write_rx) __aligned(8); |
d5fed5ad CL |
308 | u64 bytes_received; |
309 | /* RFC4898 tcpEStatsAppHCThruOctetsReceived | |
bdd1f9ed ED |
310 | * sum(delta(rcv_nxt)), or how many bytes |
311 | * were acked. | |
312 | */ | |
2efd055c MRL |
313 | u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn |
314 | * total number of segments in. | |
315 | */ | |
a44d6eac MKL |
316 | u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn |
317 | * total number of data segments in. | |
318 | */ | |
54287cc1 | 319 | u32 rcv_wup; /* rcv_nxt on last window update sent */ |
d5fed5ad CL |
320 | u32 max_packets_out; /* max packets_out in last window */ |
321 | u32 cwnd_usage_seq; /* right edge of cwnd usage tracking flight */ | |
322 | u32 rate_delivered; /* saved rate sample: packets delivered */ | |
323 | u32 rate_interval_us; /* saved rate sample: time elapsed */ | |
324 | u32 rcv_rtt_last_tsecr; | |
325 | u64 first_tx_mstamp; /* start of window send phase */ | |
326 | u64 delivered_mstamp; /* time we reached "delivered" */ | |
0df48c26 ED |
327 | u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked |
328 | * sum(delta(snd_una)), or how many bytes | |
329 | * were acked. | |
330 | */ | |
d5fed5ad CL |
331 | struct { |
332 | u32 rtt_us; | |
333 | u32 seq; | |
334 | u64 time; | |
335 | } rcv_rtt_est; | |
336 | /* Receiver queue space */ | |
337 | struct { | |
338 | u32 space; | |
339 | u32 seq; | |
340 | u64 time; | |
341 | } rcvq_space; | |
342 | __cacheline_group_end(tcp_sock_write_rx); | |
343 | /* End of Hot Path */ | |
344 | ||
345 | /* | |
346 | * RFC793 variables by their proper names. This means you can | |
347 | * read the code and the spec side by side (and laugh ...) | |
348 | * See RFC793 and RFC1122. The RFC writes these in capitals. | |
349 | */ | |
7e10b655 WW |
350 | u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups |
351 | * total number of DSACK blocks received | |
352 | */ | |
86de5921 | 353 | u32 compressed_ack_rcv_nxt; |
46d3ceab | 354 | struct list_head tsq_node; /* anchor in tsq_tasklet.head list */ |
1da177e4 | 355 | |
659a8ad5 YC |
356 | /* Information of the most recently (s)acked skb */ |
357 | struct tcp_rack { | |
9a568de4 | 358 | u64 mstamp; /* (Re)sent time of the skb */ |
deed7be7 | 359 | u32 rtt_us; /* Associated RTT */ |
1d0833df | 360 | u32 end_seq; /* Ending TCP sequence of the skb */ |
1f255691 PJ |
361 | u32 last_delivered; /* tp->delivered at last reo_wnd adj */ |
362 | u8 reo_wnd_steps; /* Allowed reordering window */ | |
363 | #define TCP_RACK_RECOVERY_THRESH 16 | |
364 | u8 reo_wnd_persist:5, /* No. of recovery since last adj */ | |
365 | dsack_seen:1, /* Whether DSACK seen after last adj */ | |
7ec65372 | 366 | advanced:1; /* mstamp advanced since last lost marking */ |
659a8ad5 | 367 | } rack; |
5d9f4262 | 368 | u8 compressed_ack; |
76be93fc YC |
369 | u8 dup_ack_counter:2, |
370 | tlp_retrans:1, /* TLP is a retransmission */ | |
666a877d | 371 | unused:5; |
d5fed5ad CL |
372 | u8 thin_lto : 1,/* Use linear timeouts for thin streams */ |
373 | recvmsg_inq : 1,/* Indicate # of bytes in queue upon recvmsg */ | |
19f6d3f3 | 374 | fastopen_connect:1, /* FASTOPEN_CONNECT sockopt */ |
71c02379 | 375 | fastopen_no_cookie:1, /* Allow send/recv SYN+data without a cookie */ |
d5fed5ad | 376 | fastopen_client_fail:2, /* reason why fastopen failed */ |
e33099f9 | 377 | frto : 1;/* F-RTO (RFC5682) activated in CA_Loss */ |
ee995283 | 378 | u8 repair_queue; |
267cf9fa MKL |
379 | u8 save_syn:2, /* Save headers of SYN packet */ |
380 | syn_data:1, /* SYN includes data */ | |
6f73601e | 381 | syn_fastopen:1, /* SYN includes Fast Open option */ |
2646c831 | 382 | syn_fastopen_exp:1,/* SYN includes Fast Open exp. option */ |
cf1ef3f0 | 383 | syn_fastopen_ch:1, /* Active TFO re-enabling probe */ |
d5fed5ad | 384 | syn_data_acked:1;/* data in SYN is acked by SYN-ACK */ |
1da177e4 | 385 | |
99123622 | 386 | u8 keepalive_probes; /* num of allowed keep alive probes */ |
a842fe14 | 387 | u32 tcp_tx_delay; /* delay (in usec) added to TX packets */ |
9799ccb0 | 388 | |
1da177e4 | 389 | /* RTT measurement */ |
740b0f18 | 390 | u32 mdev_max_us; /* maximal mdev for the last rtt period */ |
3a137d20 | 391 | |
7ec65372 | 392 | u32 reord_seen; /* number of data packet reordering events */ |
1da177e4 LT |
393 | |
394 | /* | |
395 | * Slow start and congestion control (see also Nagle, and Karn & Partridge) | |
396 | */ | |
f78a1b38 | 397 | u32 snd_cwnd_cnt; /* Linear increase counter */ |
e0ef57cc | 398 | u32 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */ |
3a137d20 ACM |
399 | u32 snd_cwnd_used; |
400 | u32 snd_cwnd_stamp; | |
4faf7839 | 401 | u32 prior_cwnd; /* cwnd right before starting loss recovery */ |
a262f0cd ND |
402 | u32 prr_delivered; /* Number of newly delivered packets to |
403 | * receiver in Recovery. */ | |
99123622 | 404 | u32 last_oow_ack_time; /* timestamp of last out-of-window ACK */ |
1da177e4 | 405 | |
218af599 | 406 | struct hrtimer pacing_timer; |
5d9f4262 | 407 | struct hrtimer compressed_ack_timer; |
218af599 | 408 | |
9f5afeae | 409 | struct sk_buff *ooo_last_skb; /* cache rb_last(out_of_order_queue) */ |
b79eeeb9 | 410 | |
c0a788c4 | 411 | /* SACKs data, these 2 need to be together (see tcp_options_write) */ |
1da177e4 LT |
412 | struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */ |
413 | struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/ | |
414 | ||
fd6dad61 | 415 | struct tcp_sack_block recv_sack_cache[4]; |
6a438bbe | 416 | |
6a438bbe | 417 | int lost_cnt_hint; |
6a438bbe | 418 | |
4b749440 | 419 | u32 prior_ssthresh; /* ssthresh saved at recovery start */ |
3a137d20 | 420 | u32 high_seq; /* snd_nxt at onset of congestion */ |
1da177e4 | 421 | |
3a137d20 | 422 | u32 retrans_stamp; /* Timestamp of the last retransmit, |
1da177e4 LT |
423 | * also used in SYN-SENT to remember stamp of |
424 | * the first SYN. */ | |
989e04c5 | 425 | u32 undo_marker; /* snd_una upon a new recovery episode. */ |
1da177e4 | 426 | int undo_retrans; /* number of undoable retransmissions. */ |
fb31c9b9 WW |
427 | u64 bytes_retrans; /* RFC4898 tcpEStatsPerfOctetsRetrans |
428 | * Total data bytes retransmitted | |
429 | */ | |
b79eeeb9 | 430 | u32 total_retrans; /* Total retransmits for entire connection */ |
3868ab0f A |
431 | u32 rto_stamp; /* Start time (ms) of last CA_Loss recovery */ |
432 | u16 total_rto; /* Total number of RTO timeouts, including | |
433 | * SYN/SYN-ACK and recurring timeouts. | |
434 | */ | |
435 | u16 total_rto_recoveries; /* Total number of RTO recoveries, | |
436 | * including any unfinished recovery. | |
437 | */ | |
438 | u32 total_rto_time; /* ms spent in (completed) RTO recoveries. */ | |
b79eeeb9 | 439 | |
3a137d20 | 440 | u32 urg_seq; /* Seq of received urgent pointer */ |
1da177e4 LT |
441 | unsigned int keepalive_time; /* time before keep alive takes place */ |
442 | unsigned int keepalive_intvl; /* time interval between keep alive probes */ | |
1da177e4 | 443 | |
a0f82f64 | 444 | int linger2; |
1da177e4 | 445 | |
b13d8807 LB |
446 | |
447 | /* Sock_ops bpf program related variables */ | |
448 | #ifdef CONFIG_BPF | |
449 | u8 bpf_sock_ops_cb_flags; /* Control calling BPF programs | |
450 | * values defined in uapi/linux/tcp.h | |
451 | */ | |
061ff040 MKL |
452 | u8 bpf_chg_cc_inprogress:1; /* In the middle of |
453 | * bpf_setsockopt(TCP_CONGESTION), | |
454 | * it is to avoid the bpf_tcp_cc->init() | |
455 | * to recur itself by calling | |
456 | * bpf_setsockopt(TCP_CONGESTION, "itself"). | |
457 | */ | |
b13d8807 LB |
458 | #define BPF_SOCK_OPS_TEST_FLAG(TP, ARG) (TP->bpf_sock_ops_cb_flags & ARG) |
459 | #else | |
460 | #define BPF_SOCK_OPS_TEST_FLAG(TP, ARG) 0 | |
461 | #endif | |
462 | ||
32efcc06 AK |
463 | u16 timeout_rehash; /* Timeout-triggered rehash attempts */ |
464 | ||
f9af2dbb TH |
465 | u32 rcv_ooopack; /* Received out-of-order packets, for tcpinfo */ |
466 | ||
0e7b1368 JH |
467 | /* TCP-specific MTU probe information. */ |
468 | struct { | |
3a137d20 ACM |
469 | u32 probe_seq_start; |
470 | u32 probe_seq_end; | |
0e7b1368 | 471 | } mtu_probe; |
29c1c446 | 472 | u32 plb_rehash; /* PLB-triggered rehash attempts */ |
563d34d0 ED |
473 | u32 mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG |
474 | * while socket was owned by user. | |
475 | */ | |
2303f994 PK |
476 | #if IS_ENABLED(CONFIG_MPTCP) |
477 | bool is_mptcp; | |
478 | #endif | |
267cf9fa MKL |
479 | #if IS_ENABLED(CONFIG_SMC) |
480 | bool syn_smc; /* SYN includes SMC */ | |
99123622 | 481 | bool (*smc_hs_congested)(const struct sock *sk); |
267cf9fa | 482 | #endif |
cfb6eeb4 | 483 | |
c845f5f3 DS |
484 | #if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO) |
485 | /* TCP AF-Specific parts; only used by TCP-AO/MD5 Signature support so far */ | |
b2e4b3de | 486 | const struct tcp_sock_af_ops *af_specific; |
cfb6eeb4 | 487 | |
c845f5f3 | 488 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 489 | /* TCP MD5 Signature Option information */ |
a8afca03 | 490 | struct tcp_md5sig_info __rcu *md5sig_info; |
cfb6eeb4 | 491 | #endif |
c845f5f3 DS |
492 | #ifdef CONFIG_TCP_AO |
493 | struct tcp_ao_info __rcu *ao_info; | |
494 | #endif | |
495 | #endif | |
435cf559 | 496 | |
10467163 JC |
497 | /* TCP fastopen related information */ |
498 | struct tcp_fastopen_request *fastopen_req; | |
499 | /* fastopen_rsk points to request_sock that resulted in this big | |
500 | * socket. Used to retransmit SYNACKs etc. | |
501 | */ | |
d983ea6f | 502 | struct request_sock __rcu *fastopen_rsk; |
70a217f1 | 503 | struct saved_syn *saved_syn; |
1da177e4 LT |
504 | }; |
505 | ||
40fc3423 | 506 | enum tsq_enum { |
46d3ceab ED |
507 | TSQ_THROTTLED, |
508 | TSQ_QUEUED, | |
6f458dfb ED |
509 | TCP_TSQ_DEFERRED, /* tcp_tasklet_func() found socket was owned */ |
510 | TCP_WRITE_TIMER_DEFERRED, /* tcp_write_timer() found socket was owned */ | |
511 | TCP_DELACK_TIMER_DEFERRED, /* tcp_delack_timer() found socket was owned */ | |
563d34d0 ED |
512 | TCP_MTU_REDUCED_DEFERRED, /* tcp_v{4|6}_err() could not call |
513 | * tcp_v{4|6}_mtu_reduced() | |
514 | */ | |
133c4c0d | 515 | TCP_ACK_DEFERRED, /* TX pure ack is deferred */ |
46d3ceab ED |
516 | }; |
517 | ||
40fc3423 | 518 | enum tsq_flags { |
133c4c0d ED |
519 | TSQF_THROTTLED = BIT(TSQ_THROTTLED), |
520 | TSQF_QUEUED = BIT(TSQ_QUEUED), | |
521 | TCPF_TSQ_DEFERRED = BIT(TCP_TSQ_DEFERRED), | |
522 | TCPF_WRITE_TIMER_DEFERRED = BIT(TCP_WRITE_TIMER_DEFERRED), | |
523 | TCPF_DELACK_TIMER_DEFERRED = BIT(TCP_DELACK_TIMER_DEFERRED), | |
524 | TCPF_MTU_REDUCED_DEFERRED = BIT(TCP_MTU_REDUCED_DEFERRED), | |
525 | TCPF_ACK_DEFERRED = BIT(TCP_ACK_DEFERRED), | |
40fc3423 ED |
526 | }; |
527 | ||
e9d9da91 ED |
528 | #define tcp_sk(ptr) container_of_const(ptr, struct tcp_sock, inet_conn.icsk_inet.sk) |
529 | ||
530 | /* Variant of tcp_sk() upgrading a const sock to a read/write tcp socket. | |
531 | * Used in context of (lockless) tcp listeners. | |
532 | */ | |
533 | #define tcp_sk_rw(ptr) container_of(ptr, struct tcp_sock, inet_conn.icsk_inet.sk) | |
1da177e4 | 534 | |
8feaf0c0 ACM |
535 | struct tcp_timewait_sock { |
536 | struct inet_timewait_sock tw_sk; | |
d475f090 ED |
537 | #define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxt |
538 | #define tw_snd_nxt tw_sk.__tw_common.skc_tw_snd_nxt | |
3a137d20 | 539 | u32 tw_rcv_wnd; |
ceaa1fef | 540 | u32 tw_ts_offset; |
3a137d20 | 541 | u32 tw_ts_recent; |
4fb17a60 NC |
542 | |
543 | /* The time we sent the last out-of-window ACK: */ | |
544 | u32 tw_last_oow_ack_time; | |
545 | ||
cca9bab1 | 546 | int tw_ts_recent_stamp; |
a842fe14 | 547 | u32 tw_tx_delay; |
cfb6eeb4 | 548 | #ifdef CONFIG_TCP_MD5SIG |
2397849b | 549 | struct tcp_md5sig_key *tw_md5_key; |
cfb6eeb4 | 550 | #endif |
decde258 DS |
551 | #ifdef CONFIG_TCP_AO |
552 | struct tcp_ao_info __rcu *ao_info; | |
553 | #endif | |
8feaf0c0 ACM |
554 | }; |
555 | ||
556 | static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk) | |
557 | { | |
558 | return (struct tcp_timewait_sock *)sk; | |
559 | } | |
560 | ||
10467163 JC |
561 | static inline bool tcp_passive_fastopen(const struct sock *sk) |
562 | { | |
d983ea6f ED |
563 | return sk->sk_state == TCP_SYN_RECV && |
564 | rcu_access_pointer(tcp_sk(sk)->fastopen_rsk) != NULL; | |
10467163 JC |
565 | } |
566 | ||
0536fcc0 | 567 | static inline void fastopen_queue_tune(struct sock *sk, int backlog) |
10467163 | 568 | { |
0536fcc0 | 569 | struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue; |
dbf650b6 | 570 | int somaxconn = READ_ONCE(sock_net(sk)->core.sysctl_somaxconn); |
0536fcc0 | 571 | |
70f360dd | 572 | WRITE_ONCE(queue->fastopenq.max_qlen, min_t(unsigned int, backlog, somaxconn)); |
10467163 JC |
573 | } |
574 | ||
805c4bc0 ED |
575 | static inline void tcp_move_syn(struct tcp_sock *tp, |
576 | struct request_sock *req) | |
577 | { | |
578 | tp->saved_syn = req->saved_syn; | |
579 | req->saved_syn = NULL; | |
580 | } | |
581 | ||
cd8ae852 ED |
582 | static inline void tcp_saved_syn_free(struct tcp_sock *tp) |
583 | { | |
584 | kfree(tp->saved_syn); | |
585 | tp->saved_syn = NULL; | |
586 | } | |
587 | ||
70a217f1 MKL |
588 | static inline u32 tcp_saved_syn_len(const struct saved_syn *saved_syn) |
589 | { | |
267cf9fa MKL |
590 | return saved_syn->mac_hdrlen + saved_syn->network_hdrlen + |
591 | saved_syn->tcp_hdrlen; | |
70a217f1 MKL |
592 | } |
593 | ||
48040793 | 594 | struct sk_buff *tcp_get_timestamping_opt_stats(const struct sock *sk, |
e7ed11ee YS |
595 | const struct sk_buff *orig_skb, |
596 | const struct sk_buff *ack_skb); | |
1c885808 | 597 | |
3541f9e8 ED |
598 | static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss) |
599 | { | |
600 | /* We use READ_ONCE() here because socket might not be locked. | |
601 | * This happens for listeners. | |
602 | */ | |
603 | u16 user_mss = READ_ONCE(tp->rx_opt.user_mss); | |
604 | ||
605 | return (user_mss && user_mss < mss) ? user_mss : mss; | |
606 | } | |
3b4929f6 ED |
607 | |
608 | int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount, | |
609 | int shiftlen); | |
610 | ||
6fadaa56 | 611 | void __tcp_sock_set_cork(struct sock *sk, bool on); |
db10538a | 612 | void tcp_sock_set_cork(struct sock *sk, bool on); |
480aeb96 | 613 | int tcp_sock_set_keepcnt(struct sock *sk, int val); |
aad4a0a9 | 614 | int tcp_sock_set_keepidle_locked(struct sock *sk, int val); |
71c48eb8 | 615 | int tcp_sock_set_keepidle(struct sock *sk, int val); |
d41ecaac | 616 | int tcp_sock_set_keepintvl(struct sock *sk, int val); |
6fadaa56 | 617 | void __tcp_sock_set_nodelay(struct sock *sk, bool on); |
12abc5ee | 618 | void tcp_sock_set_nodelay(struct sock *sk); |
ddd061b8 | 619 | void tcp_sock_set_quickack(struct sock *sk, int val); |
557eadfc | 620 | int tcp_sock_set_syncnt(struct sock *sk, int val); |
d58f2e15 | 621 | int tcp_sock_set_user_timeout(struct sock *sk, int val); |
db10538a | 622 | |
3d44de9a ED |
623 | static inline bool dst_tcp_usec_ts(const struct dst_entry *dst) |
624 | { | |
625 | return dst_feature(dst, RTAX_FEATURE_TCP_USEC_TS); | |
626 | } | |
627 | ||
1da177e4 | 628 | #endif /* _LINUX_TCP_H */ |