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 #include <linux/types.h>
21 #include <asm/byteorder.h>
22 #include <linux/socket.h>
23
24 struct tcphdr {
25 __be16 source;
26 __be16 dest;
27 __be32 seq;
28 __be32 ack_seq;
29 #if defined(__LITTLE_ENDIAN_BITFIELD)
30 __u16 res1:4,
31 doff:4,
32 fin:1,
33 syn:1,
34 rst:1,
35 psh:1,
36 ack:1,
37 urg:1,
38 ece:1,
39 cwr:1;
40 #elif defined(__BIG_ENDIAN_BITFIELD)
41 __u16 doff:4,
42 res1:4,
43 cwr:1,
44 ece:1,
45 urg:1,
46 ack:1,
47 psh:1,
48 rst:1,
49 syn:1,
50 fin:1;
51 #else
52 #error "Adjust your <asm/byteorder.h> defines"
53 #endif
54 __be16 window;
55 __sum16 check;
56 __be16 urg_ptr;
57 };
58
59 /*
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.
63 */
64 union tcp_word_hdr {
65 struct tcphdr hdr;
66 __be32 words[5];
67 };
68
69 #define tcp_flag_word(tp) ( ((union tcp_word_hdr *)(tp))->words [3])
70
71 enum {
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)
82 };
83
84 /*
85 * TCP general constants
86 */
87 #define TCP_MSS_DEFAULT 536U /* IPv4 (RFC1122, RFC2581) */
88 #define TCP_MSS_DESIRED 1220U /* IPv6 (tunneled), EDNS0 (RFC3226) */
89
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) */
105 #define TCP_COOKIE_TRANSACTIONS 15 /* TCP Cookie Transactions */
106 #define TCP_THIN_LINEAR_TIMEOUTS 16 /* Use linear timeouts for thin streams*/
107 #define TCP_THIN_DUPACK 17 /* Fast retrans. after 1 dupack */
108 #define TCP_USER_TIMEOUT 18 /* How long for loss retry before timeout */
109
110 /* for TCP_INFO socket option */
111 #define TCPI_OPT_TIMESTAMPS 1
112 #define TCPI_OPT_SACK 2
113 #define TCPI_OPT_WSCALE 4
114 #define TCPI_OPT_ECN 8 /* ECN was negociated at TCP session init */
115 #define TCPI_OPT_ECN_SEEN 16 /* we received at least one packet with ECT */
116
117 enum tcp_ca_state {
118 TCP_CA_Open = 0,
119 #define TCPF_CA_Open (1<<TCP_CA_Open)
120 TCP_CA_Disorder = 1,
121 #define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
122 TCP_CA_CWR = 2,
123 #define TCPF_CA_CWR (1<<TCP_CA_CWR)
124 TCP_CA_Recovery = 3,
125 #define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
126 TCP_CA_Loss = 4
127 #define TCPF_CA_Loss (1<<TCP_CA_Loss)
128 };
129
130 struct tcp_info {
131 __u8 tcpi_state;
132 __u8 tcpi_ca_state;
133 __u8 tcpi_retransmits;
134 __u8 tcpi_probes;
135 __u8 tcpi_backoff;
136 __u8 tcpi_options;
137 __u8 tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
138
139 __u32 tcpi_rto;
140 __u32 tcpi_ato;
141 __u32 tcpi_snd_mss;
142 __u32 tcpi_rcv_mss;
143
144 __u32 tcpi_unacked;
145 __u32 tcpi_sacked;
146 __u32 tcpi_lost;
147 __u32 tcpi_retrans;
148 __u32 tcpi_fackets;
149
150 /* Times. */
151 __u32 tcpi_last_data_sent;
152 __u32 tcpi_last_ack_sent; /* Not remembered, sorry. */
153 __u32 tcpi_last_data_recv;
154 __u32 tcpi_last_ack_recv;
155
156 /* Metrics. */
157 __u32 tcpi_pmtu;
158 __u32 tcpi_rcv_ssthresh;
159 __u32 tcpi_rtt;
160 __u32 tcpi_rttvar;
161 __u32 tcpi_snd_ssthresh;
162 __u32 tcpi_snd_cwnd;
163 __u32 tcpi_advmss;
164 __u32 tcpi_reordering;
165
166 __u32 tcpi_rcv_rtt;
167 __u32 tcpi_rcv_space;
168
169 __u32 tcpi_total_retrans;
170 };
171
172 /* for TCP_MD5SIG socket option */
173 #define TCP_MD5SIG_MAXKEYLEN 80
174
175 struct tcp_md5sig {
176 struct __kernel_sockaddr_storage tcpm_addr; /* address associated */
177 __u16 __tcpm_pad1; /* zero */
178 __u16 tcpm_keylen; /* key length */
179 __u32 __tcpm_pad2; /* zero */
180 __u8 tcpm_key[TCP_MD5SIG_MAXKEYLEN]; /* key (binary) */
181 };
182
183 /* for TCP_COOKIE_TRANSACTIONS (TCPCT) socket option */
184 #define TCP_COOKIE_MIN 8 /* 64-bits */
185 #define TCP_COOKIE_MAX 16 /* 128-bits */
186 #define TCP_COOKIE_PAIR_SIZE (2*TCP_COOKIE_MAX)
187
188 /* Flags for both getsockopt and setsockopt */
189 #define TCP_COOKIE_IN_ALWAYS (1 << 0) /* Discard SYN without cookie */
190 #define TCP_COOKIE_OUT_NEVER (1 << 1) /* Prohibit outgoing cookies,
191 * supercedes everything. */
192
193 /* Flags for getsockopt */
194 #define TCP_S_DATA_IN (1 << 2) /* Was data received? */
195 #define TCP_S_DATA_OUT (1 << 3) /* Was data sent? */
196
197 /* TCP_COOKIE_TRANSACTIONS data */
198 struct tcp_cookie_transactions {
199 __u16 tcpct_flags; /* see above */
200 __u8 __tcpct_pad1; /* zero */
201 __u8 tcpct_cookie_desired; /* bytes */
202 __u16 tcpct_s_data_desired; /* bytes of variable data */
203 __u16 tcpct_used; /* bytes in value */
204 __u8 tcpct_value[TCP_MSS_DEFAULT];
205 };
206
207 #ifdef __KERNEL__
208
209 #include <linux/skbuff.h>
210 #include <linux/dmaengine.h>
211 #include <net/sock.h>
212 #include <net/inet_connection_sock.h>
213 #include <net/inet_timewait_sock.h>
214
tcp_hdr(const struct sk_buff * skb)215 static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb)
216 {
217 return (struct tcphdr *)skb_transport_header(skb);
218 }
219
tcp_hdrlen(const struct sk_buff * skb)220 static inline unsigned int tcp_hdrlen(const struct sk_buff *skb)
221 {
222 return tcp_hdr(skb)->doff * 4;
223 }
224
tcp_optlen(const struct sk_buff * skb)225 static inline unsigned int tcp_optlen(const struct sk_buff *skb)
226 {
227 return (tcp_hdr(skb)->doff - 5) * 4;
228 }
229
230 /* This defines a selective acknowledgement block. */
231 struct tcp_sack_block_wire {
232 __be32 start_seq;
233 __be32 end_seq;
234 };
235
236 struct tcp_sack_block {
237 u32 start_seq;
238 u32 end_seq;
239 };
240
241 /*These are used to set the sack_ok field in struct tcp_options_received */
242 #define TCP_SACK_SEEN (1 << 0) /*1 = peer is SACK capable, */
243 #define TCP_FACK_ENABLED (1 << 1) /*1 = FACK is enabled locally*/
244 #define TCP_DSACK_SEEN (1 << 2) /*1 = DSACK was received from peer*/
245
246 struct tcp_options_received {
247 /* PAWS/RTTM data */
248 long ts_recent_stamp;/* Time we stored ts_recent (for aging) */
249 u32 ts_recent; /* Time stamp to echo next */
250 u32 rcv_tsval; /* Time stamp value */
251 u32 rcv_tsecr; /* Time stamp echo reply */
252 u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */
253 tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */
254 dsack : 1, /* D-SACK is scheduled */
255 wscale_ok : 1, /* Wscale seen on SYN packet */
256 sack_ok : 4, /* SACK seen on SYN packet */
257 snd_wscale : 4, /* Window scaling received from sender */
258 rcv_wscale : 4; /* Window scaling to send to receiver */
259 u8 cookie_plus:6, /* bytes in authenticator/cookie option */
260 cookie_out_never:1,
261 cookie_in_always:1;
262 u8 num_sacks; /* Number of SACK blocks */
263 u16 user_mss; /* mss requested by user in ioctl */
264 u16 mss_clamp; /* Maximal mss, negotiated at connection setup */
265 };
266
tcp_clear_options(struct tcp_options_received * rx_opt)267 static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
268 {
269 rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
270 rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
271 rx_opt->cookie_plus = 0;
272 }
273
274 /* This is the max number of SACKS that we'll generate and process. It's safe
275 * to increase this, although since:
276 * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
277 * only four options will fit in a standard TCP header */
278 #define TCP_NUM_SACKS 4
279
280 struct tcp_cookie_values;
281 struct tcp_request_sock_ops;
282
283 struct tcp_request_sock {
284 struct inet_request_sock req;
285 #ifdef CONFIG_TCP_MD5SIG
286 /* Only used by TCP MD5 Signature so far. */
287 const struct tcp_request_sock_ops *af_specific;
288 #endif
289 u32 rcv_isn;
290 u32 snt_isn;
291 u32 snt_synack; /* synack sent time */
292 };
293
tcp_rsk(const struct request_sock * req)294 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
295 {
296 return (struct tcp_request_sock *)req;
297 }
298
299 struct tcp_sock {
300 /* inet_connection_sock has to be the first member of tcp_sock */
301 struct inet_connection_sock inet_conn;
302 u16 tcp_header_len; /* Bytes of tcp header to send */
303 u16 xmit_size_goal_segs; /* Goal for segmenting output packets */
304
305 /*
306 * Header prediction flags
307 * 0x5?10 << 16 + snd_wnd in net byte order
308 */
309 __be32 pred_flags;
310
311 /*
312 * RFC793 variables by their proper names. This means you can
313 * read the code and the spec side by side (and laugh ...)
314 * See RFC793 and RFC1122. The RFC writes these in capitals.
315 */
316 u32 rcv_nxt; /* What we want to receive next */
317 u32 copied_seq; /* Head of yet unread data */
318 u32 rcv_wup; /* rcv_nxt on last window update sent */
319 u32 snd_nxt; /* Next sequence we send */
320
321 u32 snd_una; /* First byte we want an ack for */
322 u32 snd_sml; /* Last byte of the most recently transmitted small packet */
323 u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */
324 u32 lsndtime; /* timestamp of last sent data packet (for restart window) */
325
326 /* Data for direct copy to user */
327 struct {
328 struct sk_buff_head prequeue;
329 struct task_struct *task;
330 struct iovec *iov;
331 int memory;
332 int len;
333 #ifdef CONFIG_NET_DMA
334 /* members for async copy */
335 struct dma_chan *dma_chan;
336 int wakeup;
337 struct dma_pinned_list *pinned_list;
338 dma_cookie_t dma_cookie;
339 #endif
340 } ucopy;
341
342 u32 snd_wl1; /* Sequence for window update */
343 u32 snd_wnd; /* The window we expect to receive */
344 u32 max_window; /* Maximal window ever seen from peer */
345 u32 mss_cache; /* Cached effective mss, not including SACKS */
346
347 u32 window_clamp; /* Maximal window to advertise */
348 u32 rcv_ssthresh; /* Current window clamp */
349
350 u32 frto_highmark; /* snd_nxt when RTO occurred */
351 u16 advmss; /* Advertised MSS */
352 u8 frto_counter; /* Number of new acks after RTO */
353 u8 nonagle : 4,/* Disable Nagle algorithm? */
354 thin_lto : 1,/* Use linear timeouts for thin streams */
355 thin_dupack : 1,/* Fast retransmit on first dupack */
356 unused : 2;
357
358 /* RTT measurement */
359 u32 srtt; /* smoothed round trip time << 3 */
360 u32 mdev; /* medium deviation */
361 u32 mdev_max; /* maximal mdev for the last rtt period */
362 u32 rttvar; /* smoothed mdev_max */
363 u32 rtt_seq; /* sequence number to update rttvar */
364
365 u32 packets_out; /* Packets which are "in flight" */
366 u32 retrans_out; /* Retransmitted packets out */
367
368 u16 urg_data; /* Saved octet of OOB data and control flags */
369 u8 ecn_flags; /* ECN status bits. */
370 u8 reordering; /* Packet reordering metric. */
371 u32 snd_up; /* Urgent pointer */
372
373 u8 keepalive_probes; /* num of allowed keep alive probes */
374 /*
375 * Options received (usually on last packet, some only on SYN packets).
376 */
377 struct tcp_options_received rx_opt;
378
379 /*
380 * Slow start and congestion control (see also Nagle, and Karn & Partridge)
381 */
382 u32 snd_ssthresh; /* Slow start size threshold */
383 u32 snd_cwnd; /* Sending congestion window */
384 u32 snd_cwnd_cnt; /* Linear increase counter */
385 u32 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
386 u32 snd_cwnd_used;
387 u32 snd_cwnd_stamp;
388 u32 prior_cwnd; /* Congestion window at start of Recovery. */
389 u32 prr_delivered; /* Number of newly delivered packets to
390 * receiver in Recovery. */
391 u32 prr_out; /* Total number of pkts sent during Recovery. */
392
393 u32 rcv_wnd; /* Current receiver window */
394 u32 write_seq; /* Tail(+1) of data held in tcp send buffer */
395 u32 pushed_seq; /* Last pushed seq, required to talk to windows */
396 u32 lost_out; /* Lost packets */
397 u32 sacked_out; /* SACK'd packets */
398 u32 fackets_out; /* FACK'd packets */
399 u32 tso_deferred;
400 u32 bytes_acked; /* Appropriate Byte Counting - RFC3465 */
401
402 /* from STCP, retrans queue hinting */
403 struct sk_buff* lost_skb_hint;
404 struct sk_buff *scoreboard_skb_hint;
405 struct sk_buff *retransmit_skb_hint;
406
407 struct sk_buff_head out_of_order_queue; /* Out of order segments go here */
408
409 /* SACKs data, these 2 need to be together (see tcp_build_and_update_options) */
410 struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
411 struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
412
413 struct tcp_sack_block recv_sack_cache[4];
414
415 struct sk_buff *highest_sack; /* skb just after the highest
416 * skb with SACKed bit set
417 * (validity guaranteed only if
418 * sacked_out > 0)
419 */
420
421 int lost_cnt_hint;
422 u32 retransmit_high; /* L-bits may be on up to this seqno */
423
424 u32 lost_retrans_low; /* Sent seq after any rxmit (lowest) */
425
426 u32 prior_ssthresh; /* ssthresh saved at recovery start */
427 u32 high_seq; /* snd_nxt at onset of congestion */
428
429 u32 retrans_stamp; /* Timestamp of the last retransmit,
430 * also used in SYN-SENT to remember stamp of
431 * the first SYN. */
432 u32 undo_marker; /* tracking retrans started here. */
433 int undo_retrans; /* number of undoable retransmissions. */
434 u32 total_retrans; /* Total retransmits for entire connection */
435
436 u32 urg_seq; /* Seq of received urgent pointer */
437 unsigned int keepalive_time; /* time before keep alive takes place */
438 unsigned int keepalive_intvl; /* time interval between keep alive probes */
439
440 int linger2;
441
442 /* Receiver side RTT estimation */
443 struct {
444 u32 rtt;
445 u32 seq;
446 u32 time;
447 } rcv_rtt_est;
448
449 /* Receiver queue space */
450 struct {
451 int space;
452 u32 seq;
453 u32 time;
454 } rcvq_space;
455
456 /* TCP-specific MTU probe information. */
457 struct {
458 u32 probe_seq_start;
459 u32 probe_seq_end;
460 } mtu_probe;
461
462 #ifdef CONFIG_TCP_MD5SIG
463 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
464 const struct tcp_sock_af_ops *af_specific;
465
466 /* TCP MD5 Signature Option information */
467 struct tcp_md5sig_info *md5sig_info;
468 #endif
469
470 /* When the cookie options are generated and exchanged, then this
471 * object holds a reference to them (cookie_values->kref). Also
472 * contains related tcp_cookie_transactions fields.
473 */
474 struct tcp_cookie_values *cookie_values;
475 };
476
tcp_sk(const struct sock * sk)477 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
478 {
479 return (struct tcp_sock *)sk;
480 }
481
482 struct tcp_timewait_sock {
483 struct inet_timewait_sock tw_sk;
484 u32 tw_rcv_nxt;
485 u32 tw_snd_nxt;
486 u32 tw_rcv_wnd;
487 u32 tw_ts_recent;
488 long tw_ts_recent_stamp;
489 #ifdef CONFIG_TCP_MD5SIG
490 u16 tw_md5_keylen;
491 u8 tw_md5_key[TCP_MD5SIG_MAXKEYLEN];
492 #endif
493 /* Few sockets in timewait have cookies; in that case, then this
494 * object holds a reference to them (tw_cookie_values->kref).
495 */
496 struct tcp_cookie_values *tw_cookie_values;
497 };
498
tcp_twsk(const struct sock * sk)499 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
500 {
501 return (struct tcp_timewait_sock *)sk;
502 }
503
504 #endif /* __KERNEL__ */
505
506 #endif /* _LINUX_TCP_H */
507