| 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(th: 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 | /** |
| 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 (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 (const struct sk_buff *skb) |
| 75 | { |
| 76 | return skb_inner_transport_offset(skb) + inner_tcp_hdrlen(skb); |
| 77 | } |
| 78 | |
| 79 | static inline unsigned int tcp_optlen(const struct sk_buff *skb) |
| 80 | { |
| 81 | return (tcp_hdr(skb)->doff - 5) * 4; |
| 82 | } |
| 83 | |
| 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 */ |
| 87 | #define TCP_FASTOPEN_COOKIE_SIZE 8 /* the size employed by this impl. */ |
| 88 | |
| 89 | /* TCP Fast Open Cookie as stored in memory */ |
| 90 | struct tcp_fastopen_cookie { |
| 91 | __le64 val[DIV_ROUND_UP(TCP_FASTOPEN_COOKIE_MAX, sizeof(u64))]; |
| 92 | s8 len; |
| 93 | bool exp; /* In RFC6994 experimental option format */ |
| 94 | }; |
| 95 | |
| 96 | /* This defines a selective acknowledgement block. */ |
| 97 | struct tcp_sack_block_wire { |
| 98 | __be32 start_seq; |
| 99 | __be32 end_seq; |
| 100 | }; |
| 101 | |
| 102 | struct tcp_sack_block { |
| 103 | u32 start_seq; |
| 104 | u32 end_seq; |
| 105 | }; |
| 106 | |
| 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, */ |
| 109 | #define TCP_DSACK_SEEN (1 << 2) /*1 = DSACK was received from peer*/ |
| 110 | |
| 111 | struct tcp_options_received { |
| 112 | /* PAWS/RTTM data */ |
| 113 | int ts_recent_stamp;/* Time we stored ts_recent (for aging) */ |
| 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 */ |
| 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 */ |
| 121 | sack_ok : 3, /* SACK seen on SYN packet */ |
| 122 | smc_ok : 1, /* SMC seen on SYN packet */ |
| 123 | snd_wscale : 4, /* Window scaling received from sender */ |
| 124 | rcv_wscale : 4; /* Window scaling to send to receiver */ |
| 125 | u8 saw_unknown:1, /* Received unknown option */ |
| 126 | unused:7; |
| 127 | u8 num_sacks; /* Number of SACK blocks */ |
| 128 | u16 user_mss; /* mss requested by user in ioctl */ |
| 129 | u16 mss_clamp; /* Maximal mss, negotiated at connection setup */ |
| 130 | }; |
| 131 | |
| 132 | static inline void tcp_clear_options(struct tcp_options_received *rx_opt) |
| 133 | { |
| 134 | rx_opt->tstamp_ok = rx_opt->sack_ok = 0; |
| 135 | rx_opt->wscale_ok = rx_opt->snd_wscale = 0; |
| 136 | #if IS_ENABLED(CONFIG_SMC) |
| 137 | rx_opt->smc_ok = 0; |
| 138 | #endif |
| 139 | } |
| 140 | |
| 141 | /* This is the max number of SACKS that we'll generate and process. It's safe |
| 142 | * to increase this, although since: |
| 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 | |
| 147 | struct tcp_request_sock_ops; |
| 148 | |
| 149 | struct tcp_request_sock { |
| 150 | struct inet_request_sock req; |
| 151 | const struct tcp_request_sock_ops *af_specific; |
| 152 | u64 snt_synack; /* first SYNACK sent time */ |
| 153 | bool tfo_listener; |
| 154 | bool is_mptcp; |
| 155 | bool req_usec_ts; |
| 156 | #if IS_ENABLED(CONFIG_MPTCP) |
| 157 | bool drop_req; |
| 158 | #endif |
| 159 | u32 txhash; |
| 160 | u32 rcv_isn; |
| 161 | u32 snt_isn; |
| 162 | u32 ts_off; |
| 163 | u32 snt_tsval_first; |
| 164 | u32 snt_tsval_last; |
| 165 | u32 last_oow_ack_time; /* last SYNACK */ |
| 166 | u32 rcv_nxt; /* the ack # by SYNACK. For |
| 167 | * FastOpen it's the seq# |
| 168 | * after data-in-SYN. |
| 169 | */ |
| 170 | u8 syn_tos; |
| 171 | #ifdef CONFIG_TCP_AO |
| 172 | u8 ao_keyid; |
| 173 | u8 ao_rcv_next; |
| 174 | bool used_tcp_ao; |
| 175 | #endif |
| 176 | }; |
| 177 | |
| 178 | static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req) |
| 179 | { |
| 180 | return (struct tcp_request_sock *)req; |
| 181 | } |
| 182 | |
| 183 | static inline bool tcp_rsk_used_ao(const struct request_sock *req) |
| 184 | { |
| 185 | #ifndef CONFIG_TCP_AO |
| 186 | return false; |
| 187 | #else |
| 188 | return tcp_rsk(req)->used_tcp_ao; |
| 189 | #endif |
| 190 | } |
| 191 | |
| 192 | #define TCP_RMEM_TO_WIN_SCALE 8 |
| 193 | |
| 194 | struct tcp_sock { |
| 195 | /* Cacheline organization can be found documented in |
| 196 | * Documentation/networking/net_cachelines/tcp_sock.rst. |
| 197 | * Please update the document when adding new fields. |
| 198 | */ |
| 199 | |
| 200 | /* inet_connection_sock has to be the first member of tcp_sock */ |
| 201 | struct inet_connection_sock inet_conn; |
| 202 | |
| 203 | /* TX read-mostly hotpath cache lines */ |
| 204 | __cacheline_group_begin(tcp_sock_read_tx); |
| 205 | u32 max_window; /* Maximal window ever seen from peer */ |
| 206 | u32 rcv_ssthresh; /* Current window clamp */ |
| 207 | u32 reordering; /* Packet reordering metric. */ |
| 208 | u32 notsent_lowat; /* TCP_NOTSENT_LOWAT */ |
| 209 | u16 gso_segs; /* Max number of segs per GSO packet */ |
| 210 | /* from STCP, retrans queue hinting */ |
| 211 | struct sk_buff *lost_skb_hint; |
| 212 | struct sk_buff *retransmit_skb_hint; |
| 213 | __cacheline_group_end(tcp_sock_read_tx); |
| 214 | |
| 215 | /* TXRX read-mostly hotpath cache lines */ |
| 216 | __cacheline_group_begin(tcp_sock_read_txrx); |
| 217 | u32 tsoffset; /* timestamp offset */ |
| 218 | u32 snd_wnd; /* The window we expect to receive */ |
| 219 | u32 mss_cache; /* Cached effective mss, not including SACKS */ |
| 220 | u32 snd_cwnd; /* Sending congestion window */ |
| 221 | u32 prr_out; /* Total number of pkts sent during Recovery. */ |
| 222 | u32 lost_out; /* Lost packets */ |
| 223 | u32 sacked_out; /* SACK'd packets */ |
| 224 | u16 ; /* Bytes of tcp header to send */ |
| 225 | u8 scaling_ratio; /* see tcp_win_from_space() */ |
| 226 | u8 chrono_type : 2, /* current chronograph type */ |
| 227 | repair : 1, |
| 228 | tcp_usec_ts : 1, /* TSval values in usec */ |
| 229 | is_sack_reneg:1, /* in recovery from loss with SACK reneg? */ |
| 230 | is_cwnd_limited:1;/* forward progress limited by snd_cwnd? */ |
| 231 | __cacheline_group_end(tcp_sock_read_txrx); |
| 232 | |
| 233 | /* RX read-mostly hotpath cache lines */ |
| 234 | __cacheline_group_begin(tcp_sock_read_rx); |
| 235 | u32 copied_seq; /* Head of yet unread data */ |
| 236 | u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */ |
| 237 | u32 snd_wl1; /* Sequence for window update */ |
| 238 | u32 tlp_high_seq; /* snd_nxt at the time of TLP */ |
| 239 | u32 rttvar_us; /* smoothed mdev_max */ |
| 240 | u32 retrans_out; /* Retransmitted packets out */ |
| 241 | u16 advmss; /* Advertised MSS */ |
| 242 | u16 urg_data; /* Saved octet of OOB data and control flags */ |
| 243 | u32 lost; /* Total data packets lost incl. rexmits */ |
| 244 | struct minmax rtt_min; |
| 245 | /* OOO segments go in this rbtree. Socket lock must be held. */ |
| 246 | struct rb_root out_of_order_queue; |
| 247 | #if defined(CONFIG_TLS_DEVICE) |
| 248 | void (*tcp_clean_acked)(struct sock *sk, u32 acked_seq); |
| 249 | #endif |
| 250 | u32 snd_ssthresh; /* Slow start size threshold */ |
| 251 | u8 recvmsg_inq : 1;/* Indicate # of bytes in queue upon recvmsg */ |
| 252 | __cacheline_group_end(tcp_sock_read_rx); |
| 253 | |
| 254 | /* TX read-write hotpath cache lines */ |
| 255 | __cacheline_group_begin(tcp_sock_write_tx) ____cacheline_aligned; |
| 256 | u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut |
| 257 | * The total number of segments sent. |
| 258 | */ |
| 259 | u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut |
| 260 | * total number of data segments sent. |
| 261 | */ |
| 262 | u64 bytes_sent; /* RFC4898 tcpEStatsPerfHCDataOctetsOut |
| 263 | * total number of data bytes sent. |
| 264 | */ |
| 265 | u32 snd_sml; /* Last byte of the most recently transmitted small packet */ |
| 266 | u32 chrono_start; /* Start time in jiffies of a TCP chrono */ |
| 267 | u32 chrono_stat[3]; /* Time in jiffies for chrono_stat stats */ |
| 268 | u32 write_seq; /* Tail(+1) of data held in tcp send buffer */ |
| 269 | u32 pushed_seq; /* Last pushed seq, required to talk to windows */ |
| 270 | u32 lsndtime; /* timestamp of last sent data packet (for restart window) */ |
| 271 | u32 mdev_us; /* medium deviation */ |
| 272 | u32 rtt_seq; /* sequence number to update rttvar */ |
| 273 | u64 tcp_wstamp_ns; /* departure time for next sent data packet */ |
| 274 | struct list_head tsorted_sent_queue; /* time-sorted sent but un-SACKed skbs */ |
| 275 | struct sk_buff *highest_sack; /* skb just after the highest |
| 276 | * skb with SACKed bit set |
| 277 | * (validity guaranteed only if |
| 278 | * sacked_out > 0) |
| 279 | */ |
| 280 | u8 ecn_flags; /* ECN status bits. */ |
| 281 | __cacheline_group_end(tcp_sock_write_tx); |
| 282 | |
| 283 | /* TXRX read-write hotpath cache lines */ |
| 284 | __cacheline_group_begin(tcp_sock_write_txrx); |
| 285 | /* |
| 286 | * Header prediction flags |
| 287 | * 0x5?10 << 16 + snd_wnd in net byte order |
| 288 | */ |
| 289 | __be32 pred_flags; |
| 290 | u64 tcp_clock_cache; /* cache last tcp_clock_ns() (see tcp_mstamp_refresh()) */ |
| 291 | u64 tcp_mstamp; /* most recent packet received/sent */ |
| 292 | u32 rcv_nxt; /* What we want to receive next */ |
| 293 | u32 snd_nxt; /* Next sequence we send */ |
| 294 | u32 snd_una; /* First byte we want an ack for */ |
| 295 | u32 window_clamp; /* Maximal window to advertise */ |
| 296 | u32 srtt_us; /* smoothed round trip time << 3 in usecs */ |
| 297 | u32 packets_out; /* Packets which are "in flight" */ |
| 298 | u32 snd_up; /* Urgent pointer */ |
| 299 | u32 delivered; /* Total data packets delivered incl. rexmits */ |
| 300 | u32 delivered_ce; /* Like the above but only ECE marked packets */ |
| 301 | u32 app_limited; /* limited until "delivered" reaches this val */ |
| 302 | u32 rcv_wnd; /* Current receiver window */ |
| 303 | /* |
| 304 | * Options received (usually on last packet, some only on SYN packets). |
| 305 | */ |
| 306 | struct tcp_options_received rx_opt; |
| 307 | u8 nonagle : 4,/* Disable Nagle algorithm? */ |
| 308 | rate_app_limited:1; /* rate_{delivered,interval_us} limited? */ |
| 309 | __cacheline_group_end(tcp_sock_write_txrx); |
| 310 | |
| 311 | /* RX read-write hotpath cache lines */ |
| 312 | __cacheline_group_begin(tcp_sock_write_rx) __aligned(8); |
| 313 | u64 bytes_received; |
| 314 | /* RFC4898 tcpEStatsAppHCThruOctetsReceived |
| 315 | * sum(delta(rcv_nxt)), or how many bytes |
| 316 | * were acked. |
| 317 | */ |
| 318 | u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn |
| 319 | * total number of segments in. |
| 320 | */ |
| 321 | u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn |
| 322 | * total number of data segments in. |
| 323 | */ |
| 324 | u32 rcv_wup; /* rcv_nxt on last window update sent */ |
| 325 | u32 max_packets_out; /* max packets_out in last window */ |
| 326 | u32 cwnd_usage_seq; /* right edge of cwnd usage tracking flight */ |
| 327 | u32 rate_delivered; /* saved rate sample: packets delivered */ |
| 328 | u32 rate_interval_us; /* saved rate sample: time elapsed */ |
| 329 | u32 rcv_rtt_last_tsecr; |
| 330 | u64 first_tx_mstamp; /* start of window send phase */ |
| 331 | u64 delivered_mstamp; /* time we reached "delivered" */ |
| 332 | u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked |
| 333 | * sum(delta(snd_una)), or how many bytes |
| 334 | * were acked. |
| 335 | */ |
| 336 | struct { |
| 337 | u32 rtt_us; |
| 338 | u32 seq; |
| 339 | u64 time; |
| 340 | } rcv_rtt_est; |
| 341 | /* Receiver queue space */ |
| 342 | struct { |
| 343 | int space; |
| 344 | u32 seq; |
| 345 | u64 time; |
| 346 | } rcvq_space; |
| 347 | __cacheline_group_end(tcp_sock_write_rx); |
| 348 | /* End of Hot Path */ |
| 349 | |
| 350 | /* |
| 351 | * RFC793 variables by their proper names. This means you can |
| 352 | * read the code and the spec side by side (and laugh ...) |
| 353 | * See RFC793 and RFC1122. The RFC writes these in capitals. |
| 354 | */ |
| 355 | u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups |
| 356 | * total number of DSACK blocks received |
| 357 | */ |
| 358 | u32 compressed_ack_rcv_nxt; |
| 359 | struct list_head tsq_node; /* anchor in tsq_tasklet.head list */ |
| 360 | |
| 361 | /* Information of the most recently (s)acked skb */ |
| 362 | struct tcp_rack { |
| 363 | u64 mstamp; /* (Re)sent time of the skb */ |
| 364 | u32 rtt_us; /* Associated RTT */ |
| 365 | u32 end_seq; /* Ending TCP sequence of the skb */ |
| 366 | u32 last_delivered; /* tp->delivered at last reo_wnd adj */ |
| 367 | u8 reo_wnd_steps; /* Allowed reordering window */ |
| 368 | #define TCP_RACK_RECOVERY_THRESH 16 |
| 369 | u8 reo_wnd_persist:5, /* No. of recovery since last adj */ |
| 370 | dsack_seen:1, /* Whether DSACK seen after last adj */ |
| 371 | advanced:1; /* mstamp advanced since last lost marking */ |
| 372 | } rack; |
| 373 | u8 compressed_ack; |
| 374 | u8 dup_ack_counter:2, |
| 375 | tlp_retrans:1, /* TLP is a retransmission */ |
| 376 | unused:5; |
| 377 | u8 thin_lto : 1,/* Use linear timeouts for thin streams */ |
| 378 | fastopen_connect:1, /* FASTOPEN_CONNECT sockopt */ |
| 379 | fastopen_no_cookie:1, /* Allow send/recv SYN+data without a cookie */ |
| 380 | fastopen_client_fail:2, /* reason why fastopen failed */ |
| 381 | frto : 1;/* F-RTO (RFC5682) activated in CA_Loss */ |
| 382 | u8 repair_queue; |
| 383 | u8 save_syn:2, /* Save headers of SYN packet */ |
| 384 | syn_data:1, /* SYN includes data */ |
| 385 | syn_fastopen:1, /* SYN includes Fast Open option */ |
| 386 | syn_fastopen_exp:1,/* SYN includes Fast Open exp. option */ |
| 387 | syn_fastopen_ch:1, /* Active TFO re-enabling probe */ |
| 388 | syn_data_acked:1,/* data in SYN is acked by SYN-ACK */ |
| 389 | syn_fastopen_child:1; /* created TFO passive child socket */ |
| 390 | |
| 391 | u8 keepalive_probes; /* num of allowed keep alive probes */ |
| 392 | u32 tcp_tx_delay; /* delay (in usec) added to TX packets */ |
| 393 | |
| 394 | /* RTT measurement */ |
| 395 | u32 mdev_max_us; /* maximal mdev for the last rtt period */ |
| 396 | |
| 397 | u32 reord_seen; /* number of data packet reordering events */ |
| 398 | |
| 399 | /* |
| 400 | * Slow start and congestion control (see also Nagle, and Karn & Partridge) |
| 401 | */ |
| 402 | u32 snd_cwnd_cnt; /* Linear increase counter */ |
| 403 | u32 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */ |
| 404 | u32 snd_cwnd_used; |
| 405 | u32 snd_cwnd_stamp; |
| 406 | u32 prior_cwnd; /* cwnd right before starting loss recovery */ |
| 407 | u32 prr_delivered; /* Number of newly delivered packets to |
| 408 | * receiver in Recovery. */ |
| 409 | u32 last_oow_ack_time; /* timestamp of last out-of-window ACK */ |
| 410 | |
| 411 | struct hrtimer pacing_timer; |
| 412 | struct hrtimer compressed_ack_timer; |
| 413 | |
| 414 | struct sk_buff *ooo_last_skb; /* cache rb_last(out_of_order_queue) */ |
| 415 | |
| 416 | /* SACKs data, these 2 need to be together (see tcp_options_write) */ |
| 417 | struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */ |
| 418 | struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/ |
| 419 | |
| 420 | struct tcp_sack_block recv_sack_cache[4]; |
| 421 | |
| 422 | int lost_cnt_hint; |
| 423 | |
| 424 | u32 prior_ssthresh; /* ssthresh saved at recovery start */ |
| 425 | u32 high_seq; /* snd_nxt at onset of congestion */ |
| 426 | |
| 427 | u32 retrans_stamp; /* Timestamp of the last retransmit, |
| 428 | * also used in SYN-SENT to remember stamp of |
| 429 | * the first SYN. */ |
| 430 | u32 undo_marker; /* snd_una upon a new recovery episode. */ |
| 431 | int undo_retrans; /* number of undoable retransmissions. */ |
| 432 | u64 bytes_retrans; /* RFC4898 tcpEStatsPerfOctetsRetrans |
| 433 | * Total data bytes retransmitted |
| 434 | */ |
| 435 | u32 total_retrans; /* Total retransmits for entire connection */ |
| 436 | u32 rto_stamp; /* Start time (ms) of last CA_Loss recovery */ |
| 437 | u16 total_rto; /* Total number of RTO timeouts, including |
| 438 | * SYN/SYN-ACK and recurring timeouts. |
| 439 | */ |
| 440 | u16 total_rto_recoveries; /* Total number of RTO recoveries, |
| 441 | * including any unfinished recovery. |
| 442 | */ |
| 443 | u32 total_rto_time; /* ms spent in (completed) RTO recoveries. */ |
| 444 | |
| 445 | u32 urg_seq; /* Seq of received urgent pointer */ |
| 446 | unsigned int keepalive_time; /* time before keep alive takes place */ |
| 447 | unsigned int keepalive_intvl; /* time interval between keep alive probes */ |
| 448 | |
| 449 | int linger2; |
| 450 | |
| 451 | |
| 452 | /* Sock_ops bpf program related variables */ |
| 453 | #ifdef CONFIG_BPF |
| 454 | u8 bpf_sock_ops_cb_flags; /* Control calling BPF programs |
| 455 | * values defined in uapi/linux/tcp.h |
| 456 | */ |
| 457 | u8 bpf_chg_cc_inprogress:1; /* In the middle of |
| 458 | * bpf_setsockopt(TCP_CONGESTION), |
| 459 | * it is to avoid the bpf_tcp_cc->init() |
| 460 | * to recur itself by calling |
| 461 | * bpf_setsockopt(TCP_CONGESTION, "itself"). |
| 462 | */ |
| 463 | #define BPF_SOCK_OPS_TEST_FLAG(TP, ARG) (TP->bpf_sock_ops_cb_flags & ARG) |
| 464 | #else |
| 465 | #define BPF_SOCK_OPS_TEST_FLAG(TP, ARG) 0 |
| 466 | #endif |
| 467 | |
| 468 | u16 timeout_rehash; /* Timeout-triggered rehash attempts */ |
| 469 | |
| 470 | u32 rcv_ooopack; /* Received out-of-order packets, for tcpinfo */ |
| 471 | |
| 472 | /* TCP-specific MTU probe information. */ |
| 473 | struct { |
| 474 | u32 probe_seq_start; |
| 475 | u32 probe_seq_end; |
| 476 | } mtu_probe; |
| 477 | u32 plb_rehash; /* PLB-triggered rehash attempts */ |
| 478 | u32 mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG |
| 479 | * while socket was owned by user. |
| 480 | */ |
| 481 | #if IS_ENABLED(CONFIG_MPTCP) |
| 482 | bool is_mptcp; |
| 483 | #endif |
| 484 | #if IS_ENABLED(CONFIG_SMC) |
| 485 | bool syn_smc; /* SYN includes SMC */ |
| 486 | bool (*smc_hs_congested)(const struct sock *sk); |
| 487 | #endif |
| 488 | |
| 489 | #if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO) |
| 490 | /* TCP AF-Specific parts; only used by TCP-AO/MD5 Signature support so far */ |
| 491 | const struct tcp_sock_af_ops *af_specific; |
| 492 | |
| 493 | #ifdef CONFIG_TCP_MD5SIG |
| 494 | /* TCP MD5 Signature Option information */ |
| 495 | struct tcp_md5sig_info __rcu *md5sig_info; |
| 496 | #endif |
| 497 | #ifdef CONFIG_TCP_AO |
| 498 | struct tcp_ao_info __rcu *ao_info; |
| 499 | #endif |
| 500 | #endif |
| 501 | |
| 502 | /* TCP fastopen related information */ |
| 503 | struct tcp_fastopen_request *fastopen_req; |
| 504 | /* fastopen_rsk points to request_sock that resulted in this big |
| 505 | * socket. Used to retransmit SYNACKs etc. |
| 506 | */ |
| 507 | struct request_sock __rcu *fastopen_rsk; |
| 508 | struct saved_syn *saved_syn; |
| 509 | }; |
| 510 | |
| 511 | enum tsq_enum { |
| 512 | TSQ_THROTTLED, |
| 513 | TSQ_QUEUED, |
| 514 | TCP_TSQ_DEFERRED, /* tcp_tasklet_func() found socket was owned */ |
| 515 | TCP_WRITE_TIMER_DEFERRED, /* tcp_write_timer() found socket was owned */ |
| 516 | TCP_DELACK_TIMER_DEFERRED, /* tcp_delack_timer() found socket was owned */ |
| 517 | TCP_MTU_REDUCED_DEFERRED, /* tcp_v{4|6}_err() could not call |
| 518 | * tcp_v{4|6}_mtu_reduced() |
| 519 | */ |
| 520 | TCP_ACK_DEFERRED, /* TX pure ack is deferred */ |
| 521 | }; |
| 522 | |
| 523 | enum tsq_flags { |
| 524 | TSQF_THROTTLED = BIT(TSQ_THROTTLED), |
| 525 | TSQF_QUEUED = BIT(TSQ_QUEUED), |
| 526 | TCPF_TSQ_DEFERRED = BIT(TCP_TSQ_DEFERRED), |
| 527 | TCPF_WRITE_TIMER_DEFERRED = BIT(TCP_WRITE_TIMER_DEFERRED), |
| 528 | TCPF_DELACK_TIMER_DEFERRED = BIT(TCP_DELACK_TIMER_DEFERRED), |
| 529 | TCPF_MTU_REDUCED_DEFERRED = BIT(TCP_MTU_REDUCED_DEFERRED), |
| 530 | TCPF_ACK_DEFERRED = BIT(TCP_ACK_DEFERRED), |
| 531 | }; |
| 532 | |
| 533 | #define tcp_sk(ptr) container_of_const(ptr, struct tcp_sock, inet_conn.icsk_inet.sk) |
| 534 | |
| 535 | /* Variant of tcp_sk() upgrading a const sock to a read/write tcp socket. |
| 536 | * Used in context of (lockless) tcp listeners. |
| 537 | */ |
| 538 | #define tcp_sk_rw(ptr) container_of(ptr, struct tcp_sock, inet_conn.icsk_inet.sk) |
| 539 | |
| 540 | struct tcp_timewait_sock { |
| 541 | struct inet_timewait_sock tw_sk; |
| 542 | #define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxt |
| 543 | #define tw_snd_nxt tw_sk.__tw_common.skc_tw_snd_nxt |
| 544 | u32 tw_rcv_wnd; |
| 545 | u32 tw_ts_offset; |
| 546 | u32 tw_ts_recent; |
| 547 | |
| 548 | /* The time we sent the last out-of-window ACK: */ |
| 549 | u32 tw_last_oow_ack_time; |
| 550 | |
| 551 | int tw_ts_recent_stamp; |
| 552 | u32 tw_tx_delay; |
| 553 | #ifdef CONFIG_TCP_MD5SIG |
| 554 | struct tcp_md5sig_key *tw_md5_key; |
| 555 | #endif |
| 556 | #ifdef CONFIG_TCP_AO |
| 557 | struct tcp_ao_info __rcu *ao_info; |
| 558 | #endif |
| 559 | }; |
| 560 | |
| 561 | static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk) |
| 562 | { |
| 563 | return (struct tcp_timewait_sock *)sk; |
| 564 | } |
| 565 | |
| 566 | static inline bool tcp_passive_fastopen(const struct sock *sk) |
| 567 | { |
| 568 | return sk->sk_state == TCP_SYN_RECV && |
| 569 | rcu_access_pointer(tcp_sk(sk)->fastopen_rsk) != NULL; |
| 570 | } |
| 571 | |
| 572 | static inline void fastopen_queue_tune(struct sock *sk, int backlog) |
| 573 | { |
| 574 | struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue; |
| 575 | int somaxconn = READ_ONCE(sock_net(sk)->core.sysctl_somaxconn); |
| 576 | |
| 577 | WRITE_ONCE(queue->fastopenq.max_qlen, min_t(unsigned int, backlog, somaxconn)); |
| 578 | } |
| 579 | |
| 580 | static inline void tcp_move_syn(struct tcp_sock *tp, |
| 581 | struct request_sock *req) |
| 582 | { |
| 583 | tp->saved_syn = req->saved_syn; |
| 584 | req->saved_syn = NULL; |
| 585 | } |
| 586 | |
| 587 | static inline void tcp_saved_syn_free(struct tcp_sock *tp) |
| 588 | { |
| 589 | kfree(objp: tp->saved_syn); |
| 590 | tp->saved_syn = NULL; |
| 591 | } |
| 592 | |
| 593 | static inline u32 tcp_saved_syn_len(const struct saved_syn *saved_syn) |
| 594 | { |
| 595 | return saved_syn->mac_hdrlen + saved_syn->network_hdrlen + |
| 596 | saved_syn->tcp_hdrlen; |
| 597 | } |
| 598 | |
| 599 | struct sk_buff *tcp_get_timestamping_opt_stats(const struct sock *sk, |
| 600 | const struct sk_buff *orig_skb, |
| 601 | const struct sk_buff *ack_skb); |
| 602 | |
| 603 | static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss) |
| 604 | { |
| 605 | /* We use READ_ONCE() here because socket might not be locked. |
| 606 | * This happens for listeners. |
| 607 | */ |
| 608 | u16 user_mss = READ_ONCE(tp->rx_opt.user_mss); |
| 609 | |
| 610 | return (user_mss && user_mss < mss) ? user_mss : mss; |
| 611 | } |
| 612 | |
| 613 | int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount, |
| 614 | int shiftlen); |
| 615 | |
| 616 | void __tcp_sock_set_cork(struct sock *sk, bool on); |
| 617 | void tcp_sock_set_cork(struct sock *sk, bool on); |
| 618 | int tcp_sock_set_keepcnt(struct sock *sk, int val); |
| 619 | int tcp_sock_set_keepidle_locked(struct sock *sk, int val); |
| 620 | int tcp_sock_set_keepidle(struct sock *sk, int val); |
| 621 | int tcp_sock_set_keepintvl(struct sock *sk, int val); |
| 622 | void __tcp_sock_set_nodelay(struct sock *sk, bool on); |
| 623 | void tcp_sock_set_nodelay(struct sock *sk); |
| 624 | void tcp_sock_set_quickack(struct sock *sk, int val); |
| 625 | int tcp_sock_set_syncnt(struct sock *sk, int val); |
| 626 | int tcp_sock_set_user_timeout(struct sock *sk, int val); |
| 627 | |
| 628 | static inline bool dst_tcp_usec_ts(const struct dst_entry *dst) |
| 629 | { |
| 630 | return dst_feature(dst, RTAX_FEATURE_TCP_USEC_TS); |
| 631 | } |
| 632 | |
| 633 | #endif /* _LINUX_TCP_H */ |
| 634 | |