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.
6 * Definitions for the TCP protocol.
8 * Version: @(#)tcp.h 1.0.2 04/28/93
10 * Author: Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
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.
20 #include <linux/types.h>
21 #include <asm/byteorder.h>
22 #include <linux/socket.h>
29 #if defined(__LITTLE_ENDIAN_BITFIELD)
40 #elif defined(__BIG_ENDIAN_BITFIELD)
52 #error "Adjust your <asm/byteorder.h> defines"
60 * The union cast uses a gcc extension to avoid aliasing problems
61 * (union is compatible to any of its members)
62 * This means this part of the code is -fstrict-aliasing safe now.
69 #define tcp_flag_word(tp) ( ((union tcp_word_hdr *)(tp))->words [3])
72 TCP_FLAG_CWR = __cpu_to_be32(0x00800000),
73 TCP_FLAG_ECE = __cpu_to_be32(0x00400000),
74 TCP_FLAG_URG = __cpu_to_be32(0x00200000),
75 TCP_FLAG_ACK = __cpu_to_be32(0x00100000),
76 TCP_FLAG_PSH = __cpu_to_be32(0x00080000),
77 TCP_FLAG_RST = __cpu_to_be32(0x00040000),
78 TCP_FLAG_SYN = __cpu_to_be32(0x00020000),
79 TCP_FLAG_FIN = __cpu_to_be32(0x00010000),
80 TCP_RESERVED_BITS = __cpu_to_be32(0x0F000000),
81 TCP_DATA_OFFSET = __cpu_to_be32(0xF0000000)
85 * TCP general constants
87 #define TCP_MSS_DEFAULT 536U /* IPv4 (RFC1122, RFC2581) */
88 #define TCP_MSS_DESIRED 1220U /* IPv6 (tunneled), EDNS0 (RFC3226) */
90 /* TCP socket options */
91 #define TCP_NODELAY 1 /* Turn off Nagle's algorithm. */
92 #define TCP_MAXSEG 2 /* Limit MSS */
93 #define TCP_CORK 3 /* Never send partially complete segments */
94 #define TCP_KEEPIDLE 4 /* Start keeplives after this period */
95 #define TCP_KEEPINTVL 5 /* Interval between keepalives */
96 #define TCP_KEEPCNT 6 /* Number of keepalives before death */
97 #define TCP_SYNCNT 7 /* Number of SYN retransmits */
98 #define TCP_LINGER2 8 /* Life time of orphaned FIN-WAIT-2 state */
99 #define TCP_DEFER_ACCEPT 9 /* Wake up listener only when data arrive */
100 #define TCP_WINDOW_CLAMP 10 /* Bound advertised window */
101 #define TCP_INFO 11 /* Information about this connection. */
102 #define TCP_QUICKACK 12 /* Block/reenable quick acks */
103 #define TCP_CONGESTION 13 /* Congestion control algorithm */
104 #define TCP_MD5SIG 14 /* TCP MD5 Signature (RFC2385) */
106 #define TCPI_OPT_TIMESTAMPS 1
107 #define TCPI_OPT_SACK 2
108 #define TCPI_OPT_WSCALE 4
109 #define TCPI_OPT_ECN 8
113 #define TCPF_CA_Open (1<<TCP_CA_Open)
115 #define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
117 #define TCPF_CA_CWR (1<<TCP_CA_CWR)
119 #define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
121 #define TCPF_CA_Loss (1<<TCP_CA_Loss)
127 __u8 tcpi_retransmits;
131 __u8 tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
145 __u32 tcpi_last_data_sent;
146 __u32 tcpi_last_ack_sent; /* Not remembered, sorry. */
147 __u32 tcpi_last_data_recv;
148 __u32 tcpi_last_ack_recv;
152 __u32 tcpi_rcv_ssthresh;
155 __u32 tcpi_snd_ssthresh;
158 __u32 tcpi_reordering;
161 __u32 tcpi_rcv_space;
163 __u32 tcpi_total_retrans;
166 /* for TCP_MD5SIG socket option */
167 #define TCP_MD5SIG_MAXKEYLEN 80
170 struct __kernel_sockaddr_storage tcpm_addr; /* address associated */
171 __u16 __tcpm_pad1; /* zero */
172 __u16 tcpm_keylen; /* key length */
173 __u32 __tcpm_pad2; /* zero */
174 __u8 tcpm_key[TCP_MD5SIG_MAXKEYLEN]; /* key (binary) */
179 #include <linux/skbuff.h>
180 #include <linux/dmaengine.h>
181 #include <net/sock.h>
182 #include <net/inet_connection_sock.h>
183 #include <net/inet_timewait_sock.h>
185 static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb)
187 return (struct tcphdr *)skb_transport_header(skb);
190 static inline unsigned int tcp_hdrlen(const struct sk_buff *skb)
192 return tcp_hdr(skb)->doff * 4;
195 static inline unsigned int tcp_optlen(const struct sk_buff *skb)
197 return (tcp_hdr(skb)->doff - 5) * 4;
200 /* This defines a selective acknowledgement block. */
201 struct tcp_sack_block_wire {
206 struct tcp_sack_block {
211 struct tcp_options_received {
213 long ts_recent_stamp;/* Time we stored ts_recent (for aging) */
214 u32 ts_recent; /* Time stamp to echo next */
215 u32 rcv_tsval; /* Time stamp value */
216 u32 rcv_tsecr; /* Time stamp echo reply */
217 u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */
218 tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */
219 dsack : 1, /* D-SACK is scheduled */
220 wscale_ok : 1, /* Wscale seen on SYN packet */
221 sack_ok : 4, /* SACK seen on SYN packet */
222 snd_wscale : 4, /* Window scaling received from sender */
223 rcv_wscale : 4; /* Window scaling to send to receiver */
225 u8 num_sacks; /* Number of SACK blocks */
226 u16 user_mss; /* mss requested by user in ioctl */
227 u16 mss_clamp; /* Maximal mss, negotiated at connection setup */
230 /* This is the max number of SACKS that we'll generate and process. It's safe
231 * to increse this, although since:
232 * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
233 * only four options will fit in a standard TCP header */
234 #define TCP_NUM_SACKS 4
236 struct tcp_request_sock {
237 struct inet_request_sock req;
238 #ifdef CONFIG_TCP_MD5SIG
239 /* Only used by TCP MD5 Signature so far. */
240 const struct tcp_request_sock_ops *af_specific;
246 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
248 return (struct tcp_request_sock *)req;
252 /* inet_connection_sock has to be the first member of tcp_sock */
253 struct inet_connection_sock inet_conn;
254 u16 tcp_header_len; /* Bytes of tcp header to send */
255 u16 xmit_size_goal_segs; /* Goal for segmenting output packets */
258 * Header prediction flags
259 * 0x5?10 << 16 + snd_wnd in net byte order
264 * RFC793 variables by their proper names. This means you can
265 * read the code and the spec side by side (and laugh ...)
266 * See RFC793 and RFC1122. The RFC writes these in capitals.
268 u32 rcv_nxt; /* What we want to receive next */
269 u32 copied_seq; /* Head of yet unread data */
270 u32 rcv_wup; /* rcv_nxt on last window update sent */
271 u32 snd_nxt; /* Next sequence we send */
273 u32 snd_una; /* First byte we want an ack for */
274 u32 snd_sml; /* Last byte of the most recently transmitted small packet */
275 u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */
276 u32 lsndtime; /* timestamp of last sent data packet (for restart window) */
278 /* Data for direct copy to user */
280 struct sk_buff_head prequeue;
281 struct task_struct *task;
285 #ifdef CONFIG_NET_DMA
286 /* members for async copy */
287 struct dma_chan *dma_chan;
289 struct dma_pinned_list *pinned_list;
290 dma_cookie_t dma_cookie;
294 u32 snd_wl1; /* Sequence for window update */
295 u32 snd_wnd; /* The window we expect to receive */
296 u32 max_window; /* Maximal window ever seen from peer */
297 u32 mss_cache; /* Cached effective mss, not including SACKS */
299 u32 window_clamp; /* Maximal window to advertise */
300 u32 rcv_ssthresh; /* Current window clamp */
302 u32 frto_highmark; /* snd_nxt when RTO occurred */
303 u16 advmss; /* Advertised MSS */
304 u8 frto_counter; /* Number of new acks after RTO */
305 u8 nonagle; /* Disable Nagle algorithm? */
307 /* RTT measurement */
308 u32 srtt; /* smoothed round trip time << 3 */
309 u32 mdev; /* medium deviation */
310 u32 mdev_max; /* maximal mdev for the last rtt period */
311 u32 rttvar; /* smoothed mdev_max */
312 u32 rtt_seq; /* sequence number to update rttvar */
314 u32 packets_out; /* Packets which are "in flight" */
315 u32 retrans_out; /* Retransmitted packets out */
317 u16 urg_data; /* Saved octet of OOB data and control flags */
318 u8 ecn_flags; /* ECN status bits. */
319 u8 reordering; /* Packet reordering metric. */
320 u32 snd_up; /* Urgent pointer */
322 u8 keepalive_probes; /* num of allowed keep alive probes */
324 * Options received (usually on last packet, some only on SYN packets).
326 struct tcp_options_received rx_opt;
329 * Slow start and congestion control (see also Nagle, and Karn & Partridge)
331 u32 snd_ssthresh; /* Slow start size threshold */
332 u32 snd_cwnd; /* Sending congestion window */
333 u32 snd_cwnd_cnt; /* Linear increase counter */
334 u32 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
338 u32 rcv_wnd; /* Current receiver window */
339 u32 write_seq; /* Tail(+1) of data held in tcp send buffer */
340 u32 pushed_seq; /* Last pushed seq, required to talk to windows */
341 u32 lost_out; /* Lost packets */
342 u32 sacked_out; /* SACK'd packets */
343 u32 fackets_out; /* FACK'd packets */
345 u32 bytes_acked; /* Appropriate Byte Counting - RFC3465 */
347 /* from STCP, retrans queue hinting */
348 struct sk_buff* lost_skb_hint;
349 struct sk_buff *scoreboard_skb_hint;
350 struct sk_buff *retransmit_skb_hint;
352 struct sk_buff_head out_of_order_queue; /* Out of order segments go here */
354 /* SACKs data, these 2 need to be together (see tcp_build_and_update_options) */
355 struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
356 struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
358 struct tcp_sack_block recv_sack_cache[4];
360 struct sk_buff *highest_sack; /* highest skb with SACK received
361 * (validity guaranteed only if
366 u32 retransmit_high; /* L-bits may be on up to this seqno */
368 u32 lost_retrans_low; /* Sent seq after any rxmit (lowest) */
370 u32 prior_ssthresh; /* ssthresh saved at recovery start */
371 u32 high_seq; /* snd_nxt at onset of congestion */
373 u32 retrans_stamp; /* Timestamp of the last retransmit,
374 * also used in SYN-SENT to remember stamp of
376 u32 undo_marker; /* tracking retrans started here. */
377 int undo_retrans; /* number of undoable retransmissions. */
378 u32 total_retrans; /* Total retransmits for entire connection */
380 u32 urg_seq; /* Seq of received urgent pointer */
381 unsigned int keepalive_time; /* time before keep alive takes place */
382 unsigned int keepalive_intvl; /* time interval between keep alive probes */
386 /* Receiver side RTT estimation */
393 /* Receiver queue space */
400 /* TCP-specific MTU probe information. */
406 #ifdef CONFIG_TCP_MD5SIG
407 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
408 const struct tcp_sock_af_ops *af_specific;
410 /* TCP MD5 Signature Option information */
411 struct tcp_md5sig_info *md5sig_info;
415 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
417 return (struct tcp_sock *)sk;
420 struct tcp_timewait_sock {
421 struct inet_timewait_sock tw_sk;
426 long tw_ts_recent_stamp;
427 #ifdef CONFIG_TCP_MD5SIG
429 u8 tw_md5_key[TCP_MD5SIG_MAXKEYLEN];
433 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
435 return (struct tcp_timewait_sock *)sk;
440 #endif /* _LINUX_TCP_H */