| 1 | /* |
| 2 | * INET An implementation of the TCP/IP protocol suite for the LINUX |
| 3 | * operating system. INET is implemented using the BSD Socket |
| 4 | * interface as the means of communication with the user level. |
| 5 | * |
| 6 | * Definitions for the TCP protocol. |
| 7 | * |
| 8 | * Version: @(#)tcp.h 1.0.2 04/28/93 |
| 9 | * |
| 10 | * Author: Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
| 11 | * |
| 12 | * This program is free software; you can redistribute it and/or |
| 13 | * modify it under the terms of the GNU General Public License |
| 14 | * as published by the Free Software Foundation; either version |
| 15 | * 2 of the License, or (at your option) any later version. |
| 16 | */ |
| 17 | #ifndef _LINUX_TCP_H |
| 18 | #define _LINUX_TCP_H |
| 19 | |
| 20 | |
| 21 | #include <linux/skbuff.h> |
| 22 | #include <linux/dmaengine.h> |
| 23 | #include <net/sock.h> |
| 24 | #include <net/inet_connection_sock.h> |
| 25 | #include <net/inet_timewait_sock.h> |
| 26 | #include <uapi/linux/tcp.h> |
| 27 | |
| 28 | static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb) |
| 29 | { |
| 30 | return (struct tcphdr *)skb_transport_header(skb); |
| 31 | } |
| 32 | |
| 33 | static inline unsigned int tcp_hdrlen(const struct sk_buff *skb) |
| 34 | { |
| 35 | return tcp_hdr(skb)->doff * 4; |
| 36 | } |
| 37 | |
| 38 | static inline struct tcphdr *inner_tcp_hdr(const struct sk_buff *skb) |
| 39 | { |
| 40 | return (struct tcphdr *)skb_inner_transport_header(skb); |
| 41 | } |
| 42 | |
| 43 | static inline unsigned int inner_tcp_hdrlen(const struct sk_buff *skb) |
| 44 | { |
| 45 | return inner_tcp_hdr(skb)->doff * 4; |
| 46 | } |
| 47 | |
| 48 | static inline unsigned int tcp_optlen(const struct sk_buff *skb) |
| 49 | { |
| 50 | return (tcp_hdr(skb)->doff - 5) * 4; |
| 51 | } |
| 52 | |
| 53 | /* TCP Fast Open */ |
| 54 | #define TCP_FASTOPEN_COOKIE_MIN 4 /* Min Fast Open Cookie size in bytes */ |
| 55 | #define TCP_FASTOPEN_COOKIE_MAX 16 /* Max Fast Open Cookie size in bytes */ |
| 56 | #define TCP_FASTOPEN_COOKIE_SIZE 8 /* the size employed by this impl. */ |
| 57 | |
| 58 | /* TCP Fast Open Cookie as stored in memory */ |
| 59 | struct tcp_fastopen_cookie { |
| 60 | s8 len; |
| 61 | u8 val[TCP_FASTOPEN_COOKIE_MAX]; |
| 62 | }; |
| 63 | |
| 64 | /* This defines a selective acknowledgement block. */ |
| 65 | struct tcp_sack_block_wire { |
| 66 | __be32 start_seq; |
| 67 | __be32 end_seq; |
| 68 | }; |
| 69 | |
| 70 | struct tcp_sack_block { |
| 71 | u32 start_seq; |
| 72 | u32 end_seq; |
| 73 | }; |
| 74 | |
| 75 | /*These are used to set the sack_ok field in struct tcp_options_received */ |
| 76 | #define TCP_SACK_SEEN (1 << 0) /*1 = peer is SACK capable, */ |
| 77 | #define TCP_FACK_ENABLED (1 << 1) /*1 = FACK is enabled locally*/ |
| 78 | #define TCP_DSACK_SEEN (1 << 2) /*1 = DSACK was received from peer*/ |
| 79 | |
| 80 | struct tcp_options_received { |
| 81 | /* PAWS/RTTM data */ |
| 82 | long ts_recent_stamp;/* Time we stored ts_recent (for aging) */ |
| 83 | u32 ts_recent; /* Time stamp to echo next */ |
| 84 | u32 rcv_tsval; /* Time stamp value */ |
| 85 | u32 rcv_tsecr; /* Time stamp echo reply */ |
| 86 | u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */ |
| 87 | tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */ |
| 88 | dsack : 1, /* D-SACK is scheduled */ |
| 89 | wscale_ok : 1, /* Wscale seen on SYN packet */ |
| 90 | sack_ok : 4, /* SACK seen on SYN packet */ |
| 91 | snd_wscale : 4, /* Window scaling received from sender */ |
| 92 | rcv_wscale : 4; /* Window scaling to send to receiver */ |
| 93 | u8 num_sacks; /* Number of SACK blocks */ |
| 94 | u16 user_mss; /* mss requested by user in ioctl */ |
| 95 | u16 mss_clamp; /* Maximal mss, negotiated at connection setup */ |
| 96 | }; |
| 97 | |
| 98 | static inline void tcp_clear_options(struct tcp_options_received *rx_opt) |
| 99 | { |
| 100 | rx_opt->tstamp_ok = rx_opt->sack_ok = 0; |
| 101 | rx_opt->wscale_ok = rx_opt->snd_wscale = 0; |
| 102 | } |
| 103 | |
| 104 | /* This is the max number of SACKS that we'll generate and process. It's safe |
| 105 | * to increase this, although since: |
| 106 | * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8) |
| 107 | * only four options will fit in a standard TCP header */ |
| 108 | #define TCP_NUM_SACKS 4 |
| 109 | |
| 110 | struct tcp_request_sock_ops; |
| 111 | |
| 112 | struct tcp_request_sock { |
| 113 | struct inet_request_sock req; |
| 114 | #ifdef CONFIG_TCP_MD5SIG |
| 115 | /* Only used by TCP MD5 Signature so far. */ |
| 116 | const struct tcp_request_sock_ops *af_specific; |
| 117 | #endif |
| 118 | struct sock *listener; /* needed for TFO */ |
| 119 | u32 rcv_isn; |
| 120 | u32 snt_isn; |
| 121 | u32 snt_synack; /* synack sent time */ |
| 122 | u32 rcv_nxt; /* the ack # by SYNACK. For |
| 123 | * FastOpen it's the seq# |
| 124 | * after data-in-SYN. |
| 125 | */ |
| 126 | }; |
| 127 | |
| 128 | static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req) |
| 129 | { |
| 130 | return (struct tcp_request_sock *)req; |
| 131 | } |
| 132 | |
| 133 | struct tcp_sock { |
| 134 | /* inet_connection_sock has to be the first member of tcp_sock */ |
| 135 | struct inet_connection_sock inet_conn; |
| 136 | u16 tcp_header_len; /* Bytes of tcp header to send */ |
| 137 | u16 xmit_size_goal_segs; /* Goal for segmenting output packets */ |
| 138 | |
| 139 | /* |
| 140 | * Header prediction flags |
| 141 | * 0x5?10 << 16 + snd_wnd in net byte order |
| 142 | */ |
| 143 | __be32 pred_flags; |
| 144 | |
| 145 | /* |
| 146 | * RFC793 variables by their proper names. This means you can |
| 147 | * read the code and the spec side by side (and laugh ...) |
| 148 | * See RFC793 and RFC1122. The RFC writes these in capitals. |
| 149 | */ |
| 150 | u32 rcv_nxt; /* What we want to receive next */ |
| 151 | u32 copied_seq; /* Head of yet unread data */ |
| 152 | u32 rcv_wup; /* rcv_nxt on last window update sent */ |
| 153 | u32 snd_nxt; /* Next sequence we send */ |
| 154 | |
| 155 | u32 snd_una; /* First byte we want an ack for */ |
| 156 | u32 snd_sml; /* Last byte of the most recently transmitted small packet */ |
| 157 | u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */ |
| 158 | u32 lsndtime; /* timestamp of last sent data packet (for restart window) */ |
| 159 | |
| 160 | u32 tsoffset; /* timestamp offset */ |
| 161 | |
| 162 | struct list_head tsq_node; /* anchor in tsq_tasklet.head list */ |
| 163 | unsigned long tsq_flags; |
| 164 | |
| 165 | /* Data for direct copy to user */ |
| 166 | struct { |
| 167 | struct sk_buff_head prequeue; |
| 168 | struct task_struct *task; |
| 169 | struct iovec *iov; |
| 170 | int memory; |
| 171 | int len; |
| 172 | #ifdef CONFIG_NET_DMA |
| 173 | /* members for async copy */ |
| 174 | struct dma_chan *dma_chan; |
| 175 | int wakeup; |
| 176 | struct dma_pinned_list *pinned_list; |
| 177 | dma_cookie_t dma_cookie; |
| 178 | #endif |
| 179 | } ucopy; |
| 180 | |
| 181 | u32 snd_wl1; /* Sequence for window update */ |
| 182 | u32 snd_wnd; /* The window we expect to receive */ |
| 183 | u32 max_window; /* Maximal window ever seen from peer */ |
| 184 | u32 mss_cache; /* Cached effective mss, not including SACKS */ |
| 185 | |
| 186 | u32 window_clamp; /* Maximal window to advertise */ |
| 187 | u32 rcv_ssthresh; /* Current window clamp */ |
| 188 | |
| 189 | u16 advmss; /* Advertised MSS */ |
| 190 | u8 unused; |
| 191 | u8 nonagle : 4,/* Disable Nagle algorithm? */ |
| 192 | thin_lto : 1,/* Use linear timeouts for thin streams */ |
| 193 | thin_dupack : 1,/* Fast retransmit on first dupack */ |
| 194 | repair : 1, |
| 195 | frto : 1;/* F-RTO (RFC5682) activated in CA_Loss */ |
| 196 | u8 repair_queue; |
| 197 | u8 do_early_retrans:1,/* Enable RFC5827 early-retransmit */ |
| 198 | syn_data:1, /* SYN includes data */ |
| 199 | syn_fastopen:1, /* SYN includes Fast Open option */ |
| 200 | syn_data_acked:1,/* data in SYN is acked by SYN-ACK */ |
| 201 | is_cwnd_limited:1;/* forward progress limited by snd_cwnd? */ |
| 202 | u32 tlp_high_seq; /* snd_nxt at the time of TLP retransmit. */ |
| 203 | |
| 204 | /* RTT measurement */ |
| 205 | u32 srtt_us; /* smoothed round trip time << 3 in usecs */ |
| 206 | u32 mdev_us; /* medium deviation */ |
| 207 | u32 mdev_max_us; /* maximal mdev for the last rtt period */ |
| 208 | u32 rttvar_us; /* smoothed mdev_max */ |
| 209 | u32 rtt_seq; /* sequence number to update rttvar */ |
| 210 | |
| 211 | u32 packets_out; /* Packets which are "in flight" */ |
| 212 | u32 retrans_out; /* Retransmitted packets out */ |
| 213 | u32 max_packets_out; /* max packets_out in last window */ |
| 214 | u32 max_packets_seq; /* right edge of max_packets_out flight */ |
| 215 | |
| 216 | u16 urg_data; /* Saved octet of OOB data and control flags */ |
| 217 | u8 ecn_flags; /* ECN status bits. */ |
| 218 | u8 reordering; /* Packet reordering metric. */ |
| 219 | u32 snd_up; /* Urgent pointer */ |
| 220 | |
| 221 | u8 keepalive_probes; /* num of allowed keep alive probes */ |
| 222 | /* |
| 223 | * Options received (usually on last packet, some only on SYN packets). |
| 224 | */ |
| 225 | struct tcp_options_received rx_opt; |
| 226 | |
| 227 | /* |
| 228 | * Slow start and congestion control (see also Nagle, and Karn & Partridge) |
| 229 | */ |
| 230 | u32 snd_ssthresh; /* Slow start size threshold */ |
| 231 | u32 snd_cwnd; /* Sending congestion window */ |
| 232 | u32 snd_cwnd_cnt; /* Linear increase counter */ |
| 233 | u32 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */ |
| 234 | u32 snd_cwnd_used; |
| 235 | u32 snd_cwnd_stamp; |
| 236 | u32 prior_cwnd; /* Congestion window at start of Recovery. */ |
| 237 | u32 prr_delivered; /* Number of newly delivered packets to |
| 238 | * receiver in Recovery. */ |
| 239 | u32 prr_out; /* Total number of pkts sent during Recovery. */ |
| 240 | |
| 241 | u32 rcv_wnd; /* Current receiver window */ |
| 242 | u32 write_seq; /* Tail(+1) of data held in tcp send buffer */ |
| 243 | u32 notsent_lowat; /* TCP_NOTSENT_LOWAT */ |
| 244 | u32 pushed_seq; /* Last pushed seq, required to talk to windows */ |
| 245 | u32 lost_out; /* Lost packets */ |
| 246 | u32 sacked_out; /* SACK'd packets */ |
| 247 | u32 fackets_out; /* FACK'd packets */ |
| 248 | u32 tso_deferred; |
| 249 | |
| 250 | /* from STCP, retrans queue hinting */ |
| 251 | struct sk_buff* lost_skb_hint; |
| 252 | struct sk_buff *retransmit_skb_hint; |
| 253 | |
| 254 | /* OOO segments go in this list. Note that socket lock must be held, |
| 255 | * as we do not use sk_buff_head lock. |
| 256 | */ |
| 257 | struct sk_buff_head out_of_order_queue; |
| 258 | |
| 259 | /* SACKs data, these 2 need to be together (see tcp_options_write) */ |
| 260 | struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */ |
| 261 | struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/ |
| 262 | |
| 263 | struct tcp_sack_block recv_sack_cache[4]; |
| 264 | |
| 265 | struct sk_buff *highest_sack; /* skb just after the highest |
| 266 | * skb with SACKed bit set |
| 267 | * (validity guaranteed only if |
| 268 | * sacked_out > 0) |
| 269 | */ |
| 270 | |
| 271 | int lost_cnt_hint; |
| 272 | u32 retransmit_high; /* L-bits may be on up to this seqno */ |
| 273 | |
| 274 | u32 lost_retrans_low; /* Sent seq after any rxmit (lowest) */ |
| 275 | |
| 276 | u32 prior_ssthresh; /* ssthresh saved at recovery start */ |
| 277 | u32 high_seq; /* snd_nxt at onset of congestion */ |
| 278 | |
| 279 | u32 retrans_stamp; /* Timestamp of the last retransmit, |
| 280 | * also used in SYN-SENT to remember stamp of |
| 281 | * the first SYN. */ |
| 282 | u32 undo_marker; /* tracking retrans started here. */ |
| 283 | int undo_retrans; /* number of undoable retransmissions. */ |
| 284 | u32 total_retrans; /* Total retransmits for entire connection */ |
| 285 | |
| 286 | u32 urg_seq; /* Seq of received urgent pointer */ |
| 287 | unsigned int keepalive_time; /* time before keep alive takes place */ |
| 288 | unsigned int keepalive_intvl; /* time interval between keep alive probes */ |
| 289 | |
| 290 | int linger2; |
| 291 | |
| 292 | /* Receiver side RTT estimation */ |
| 293 | struct { |
| 294 | u32 rtt; |
| 295 | u32 seq; |
| 296 | u32 time; |
| 297 | } rcv_rtt_est; |
| 298 | |
| 299 | /* Receiver queue space */ |
| 300 | struct { |
| 301 | int space; |
| 302 | u32 seq; |
| 303 | u32 time; |
| 304 | } rcvq_space; |
| 305 | |
| 306 | /* TCP-specific MTU probe information. */ |
| 307 | struct { |
| 308 | u32 probe_seq_start; |
| 309 | u32 probe_seq_end; |
| 310 | } mtu_probe; |
| 311 | u32 mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG |
| 312 | * while socket was owned by user. |
| 313 | */ |
| 314 | |
| 315 | #ifdef CONFIG_TCP_MD5SIG |
| 316 | /* TCP AF-Specific parts; only used by MD5 Signature support so far */ |
| 317 | const struct tcp_sock_af_ops *af_specific; |
| 318 | |
| 319 | /* TCP MD5 Signature Option information */ |
| 320 | struct tcp_md5sig_info __rcu *md5sig_info; |
| 321 | #endif |
| 322 | |
| 323 | /* TCP fastopen related information */ |
| 324 | struct tcp_fastopen_request *fastopen_req; |
| 325 | /* fastopen_rsk points to request_sock that resulted in this big |
| 326 | * socket. Used to retransmit SYNACKs etc. |
| 327 | */ |
| 328 | struct request_sock *fastopen_rsk; |
| 329 | }; |
| 330 | |
| 331 | enum tsq_flags { |
| 332 | TSQ_THROTTLED, |
| 333 | TSQ_QUEUED, |
| 334 | TCP_TSQ_DEFERRED, /* tcp_tasklet_func() found socket was owned */ |
| 335 | TCP_WRITE_TIMER_DEFERRED, /* tcp_write_timer() found socket was owned */ |
| 336 | TCP_DELACK_TIMER_DEFERRED, /* tcp_delack_timer() found socket was owned */ |
| 337 | TCP_MTU_REDUCED_DEFERRED, /* tcp_v{4|6}_err() could not call |
| 338 | * tcp_v{4|6}_mtu_reduced() |
| 339 | */ |
| 340 | }; |
| 341 | |
| 342 | static inline struct tcp_sock *tcp_sk(const struct sock *sk) |
| 343 | { |
| 344 | return (struct tcp_sock *)sk; |
| 345 | } |
| 346 | |
| 347 | struct tcp_timewait_sock { |
| 348 | struct inet_timewait_sock tw_sk; |
| 349 | u32 tw_rcv_nxt; |
| 350 | u32 tw_snd_nxt; |
| 351 | u32 tw_rcv_wnd; |
| 352 | u32 tw_ts_offset; |
| 353 | u32 tw_ts_recent; |
| 354 | long tw_ts_recent_stamp; |
| 355 | #ifdef CONFIG_TCP_MD5SIG |
| 356 | struct tcp_md5sig_key *tw_md5_key; |
| 357 | #endif |
| 358 | }; |
| 359 | |
| 360 | static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk) |
| 361 | { |
| 362 | return (struct tcp_timewait_sock *)sk; |
| 363 | } |
| 364 | |
| 365 | static inline bool tcp_passive_fastopen(const struct sock *sk) |
| 366 | { |
| 367 | return (sk->sk_state == TCP_SYN_RECV && |
| 368 | tcp_sk(sk)->fastopen_rsk != NULL); |
| 369 | } |
| 370 | |
| 371 | extern void tcp_sock_destruct(struct sock *sk); |
| 372 | |
| 373 | static inline int fastopen_init_queue(struct sock *sk, int backlog) |
| 374 | { |
| 375 | struct request_sock_queue *queue = |
| 376 | &inet_csk(sk)->icsk_accept_queue; |
| 377 | |
| 378 | if (queue->fastopenq == NULL) { |
| 379 | queue->fastopenq = kzalloc( |
| 380 | sizeof(struct fastopen_queue), |
| 381 | sk->sk_allocation); |
| 382 | if (queue->fastopenq == NULL) |
| 383 | return -ENOMEM; |
| 384 | |
| 385 | sk->sk_destruct = tcp_sock_destruct; |
| 386 | spin_lock_init(&queue->fastopenq->lock); |
| 387 | } |
| 388 | queue->fastopenq->max_qlen = backlog; |
| 389 | return 0; |
| 390 | } |
| 391 | |
| 392 | #endif /* _LINUX_TCP_H */ |