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