[I/OAT]: Structure changes for TCP recv offload to I/OAT
[linux-2.6-block.git] / include / linux / tcp.h
CommitLineData
1da177e4
LT
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>
97fc2f08 21#include <linux/dmaengine.h>
1da177e4
LT
22#include <asm/byteorder.h>
23
24struct tcphdr {
25 __u16 source;
26 __u16 dest;
27 __u32 seq;
28 __u32 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 __u16 window;
55 __u16 check;
56 __u16 urg_ptr;
57};
58
1da177e4
LT
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 */
64union tcp_word_hdr {
65 struct tcphdr hdr;
66 __u32 words[5];
67};
68
69#define tcp_flag_word(tp) ( ((union tcp_word_hdr *)(tp))->words [3])
70
71enum {
72 TCP_FLAG_CWR = __constant_htonl(0x00800000),
73 TCP_FLAG_ECE = __constant_htonl(0x00400000),
74 TCP_FLAG_URG = __constant_htonl(0x00200000),
75 TCP_FLAG_ACK = __constant_htonl(0x00100000),
76 TCP_FLAG_PSH = __constant_htonl(0x00080000),
77 TCP_FLAG_RST = __constant_htonl(0x00040000),
78 TCP_FLAG_SYN = __constant_htonl(0x00020000),
79 TCP_FLAG_FIN = __constant_htonl(0x00010000),
80 TCP_RESERVED_BITS = __constant_htonl(0x0F000000),
81 TCP_DATA_OFFSET = __constant_htonl(0xF0000000)
82};
83
84/* TCP socket options */
85#define TCP_NODELAY 1 /* Turn off Nagle's algorithm. */
86#define TCP_MAXSEG 2 /* Limit MSS */
87#define TCP_CORK 3 /* Never send partially complete segments */
88#define TCP_KEEPIDLE 4 /* Start keeplives after this period */
89#define TCP_KEEPINTVL 5 /* Interval between keepalives */
90#define TCP_KEEPCNT 6 /* Number of keepalives before death */
91#define TCP_SYNCNT 7 /* Number of SYN retransmits */
92#define TCP_LINGER2 8 /* Life time of orphaned FIN-WAIT-2 state */
93#define TCP_DEFER_ACCEPT 9 /* Wake up listener only when data arrive */
94#define TCP_WINDOW_CLAMP 10 /* Bound advertised window */
95#define TCP_INFO 11 /* Information about this connection. */
96#define TCP_QUICKACK 12 /* Block/reenable quick acks */
5f8ef48d 97#define TCP_CONGESTION 13 /* Congestion control algorithm */
1da177e4
LT
98
99#define TCPI_OPT_TIMESTAMPS 1
100#define TCPI_OPT_SACK 2
101#define TCPI_OPT_WSCALE 4
102#define TCPI_OPT_ECN 8
103
104enum tcp_ca_state
105{
106 TCP_CA_Open = 0,
107#define TCPF_CA_Open (1<<TCP_CA_Open)
108 TCP_CA_Disorder = 1,
109#define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
110 TCP_CA_CWR = 2,
111#define TCPF_CA_CWR (1<<TCP_CA_CWR)
112 TCP_CA_Recovery = 3,
113#define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
114 TCP_CA_Loss = 4
115#define TCPF_CA_Loss (1<<TCP_CA_Loss)
116};
117
118struct tcp_info
119{
120 __u8 tcpi_state;
121 __u8 tcpi_ca_state;
122 __u8 tcpi_retransmits;
123 __u8 tcpi_probes;
124 __u8 tcpi_backoff;
125 __u8 tcpi_options;
126 __u8 tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
127
128 __u32 tcpi_rto;
129 __u32 tcpi_ato;
130 __u32 tcpi_snd_mss;
131 __u32 tcpi_rcv_mss;
132
133 __u32 tcpi_unacked;
134 __u32 tcpi_sacked;
135 __u32 tcpi_lost;
136 __u32 tcpi_retrans;
137 __u32 tcpi_fackets;
138
139 /* Times. */
140 __u32 tcpi_last_data_sent;
141 __u32 tcpi_last_ack_sent; /* Not remembered, sorry. */
142 __u32 tcpi_last_data_recv;
143 __u32 tcpi_last_ack_recv;
144
145 /* Metrics. */
146 __u32 tcpi_pmtu;
147 __u32 tcpi_rcv_ssthresh;
148 __u32 tcpi_rtt;
149 __u32 tcpi_rttvar;
150 __u32 tcpi_snd_ssthresh;
151 __u32 tcpi_snd_cwnd;
152 __u32 tcpi_advmss;
153 __u32 tcpi_reordering;
154
155 __u32 tcpi_rcv_rtt;
156 __u32 tcpi_rcv_space;
157
158 __u32 tcpi_total_retrans;
159};
160
161#ifdef __KERNEL__
162
163#include <linux/config.h>
164#include <linux/skbuff.h>
1da177e4 165#include <net/sock.h>
463c84b9 166#include <net/inet_connection_sock.h>
8feaf0c0 167#include <net/inet_timewait_sock.h>
1da177e4
LT
168
169/* This defines a selective acknowledgement block. */
170struct tcp_sack_block {
171 __u32 start_seq;
172 __u32 end_seq;
173};
174
1da177e4
LT
175struct tcp_options_received {
176/* PAWS/RTTM data */
177 long ts_recent_stamp;/* Time we stored ts_recent (for aging) */
178 __u32 ts_recent; /* Time stamp to echo next */
179 __u32 rcv_tsval; /* Time stamp value */
180 __u32 rcv_tsecr; /* Time stamp echo reply */
181 __u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */
182 tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */
183 dsack : 1, /* D-SACK is scheduled */
184 wscale_ok : 1, /* Wscale seen on SYN packet */
185 sack_ok : 4, /* SACK seen on SYN packet */
186 snd_wscale : 4, /* Window scaling received from sender */
187 rcv_wscale : 4; /* Window scaling to send to receiver */
188/* SACKs data */
189 __u8 eff_sacks; /* Size of SACK array to send with next packet */
190 __u8 num_sacks; /* Number of SACK blocks */
191 __u16 user_mss; /* mss requested by user in ioctl */
192 __u16 mss_clamp; /* Maximal mss, negotiated at connection setup */
193};
194
2e6599cb
ACM
195struct tcp_request_sock {
196 struct inet_request_sock req;
197 __u32 rcv_isn;
198 __u32 snt_isn;
199};
200
60236fdd 201static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
2e6599cb
ACM
202{
203 return (struct tcp_request_sock *)req;
204}
205
1da177e4 206struct tcp_sock {
463c84b9
ACM
207 /* inet_connection_sock has to be the first member of tcp_sock */
208 struct inet_connection_sock inet_conn;
1da177e4
LT
209 int tcp_header_len; /* Bytes of tcp header to send */
210
211/*
212 * Header prediction flags
213 * 0x5?10 << 16 + snd_wnd in net byte order
214 */
215 __u32 pred_flags;
216
217/*
218 * RFC793 variables by their proper names. This means you can
219 * read the code and the spec side by side (and laugh ...)
220 * See RFC793 and RFC1122. The RFC writes these in capitals.
221 */
222 __u32 rcv_nxt; /* What we want to receive next */
223 __u32 snd_nxt; /* Next sequence we send */
224
225 __u32 snd_una; /* First byte we want an ack for */
226 __u32 snd_sml; /* Last byte of the most recently transmitted small packet */
227 __u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */
228 __u32 lsndtime; /* timestamp of last sent data packet (for restart window) */
1da177e4
LT
229
230 /* Data for direct copy to user */
231 struct {
232 struct sk_buff_head prequeue;
233 struct task_struct *task;
234 struct iovec *iov;
235 int memory;
236 int len;
97fc2f08
CL
237#ifdef CONFIG_NET_DMA
238 /* members for async copy */
239 struct dma_chan *dma_chan;
240 int wakeup;
241 struct dma_pinned_list *pinned_list;
242 dma_cookie_t dma_cookie;
243#endif
1da177e4
LT
244 } ucopy;
245
246 __u32 snd_wl1; /* Sequence for window update */
247 __u32 snd_wnd; /* The window we expect to receive */
248 __u32 max_window; /* Maximal window ever seen from peer */
1da177e4 249 __u32 mss_cache; /* Cached effective mss, not including SACKS */
c1b4a7e6 250 __u16 xmit_size_goal; /* Goal for segmenting output packets */
d83d8461 251 /* XXX Two bytes hole, try to pack */
1da177e4 252
1da177e4
LT
253 __u32 window_clamp; /* Maximal window to advertise */
254 __u32 rcv_ssthresh; /* Current window clamp */
255
256 __u32 frto_highmark; /* snd_nxt when RTO occurred */
257 __u8 reordering; /* Packet reordering metric. */
258 __u8 frto_counter; /* Number of new acks after RTO */
463c84b9 259 __u8 nonagle; /* Disable Nagle algorithm? */
6687e988 260 __u8 keepalive_probes; /* num of allowed keep alive probes */
1da177e4
LT
261
262/* RTT measurement */
263 __u32 srtt; /* smoothed round trip time << 3 */
264 __u32 mdev; /* medium deviation */
265 __u32 mdev_max; /* maximal mdev for the last rtt period */
266 __u32 rttvar; /* smoothed mdev_max */
267 __u32 rtt_seq; /* sequence number to update rttvar */
1da177e4
LT
268
269 __u32 packets_out; /* Packets which are "in flight" */
270 __u32 left_out; /* Packets which leaved network */
271 __u32 retrans_out; /* Retransmitted packets out */
1da177e4
LT
272/*
273 * Options received (usually on last packet, some only on SYN packets).
274 */
1da177e4
LT
275 struct tcp_options_received rx_opt;
276
277/*
278 * Slow start and congestion control (see also Nagle, and Karn & Partridge)
279 */
280 __u32 snd_ssthresh; /* Slow start size threshold */
281 __u32 snd_cwnd; /* Sending congestion window */
282 __u16 snd_cwnd_cnt; /* Linear increase counter */
283 __u16 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
284 __u32 snd_cwnd_used;
285 __u32 snd_cwnd_stamp;
286
1da177e4
LT
287 struct sk_buff_head out_of_order_queue; /* Out of order segments go here */
288
1da177e4
LT
289 __u32 rcv_wnd; /* Current receiver window */
290 __u32 rcv_wup; /* rcv_nxt on last window update sent */
291 __u32 write_seq; /* Tail(+1) of data held in tcp send buffer */
292 __u32 pushed_seq; /* Last pushed seq, required to talk to windows */
293 __u32 copied_seq; /* Head of yet unread data */
294
295/* SACKs data */
296 struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
297 struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
298
6a438bbe
SH
299 struct tcp_sack_block recv_sack_cache[4];
300
301 /* from STCP, retrans queue hinting */
302 struct sk_buff* lost_skb_hint;
303
304 struct sk_buff *scoreboard_skb_hint;
305 struct sk_buff *retransmit_skb_hint;
306 struct sk_buff *forward_skb_hint;
307 struct sk_buff *fastpath_skb_hint;
308
309 int fastpath_cnt_hint;
310 int lost_cnt_hint;
311 int retransmit_cnt_hint;
312 int forward_cnt_hint;
313
6687e988 314 __u16 advmss; /* Advertised MSS */
1da177e4
LT
315 __u16 prior_ssthresh; /* ssthresh saved at recovery start */
316 __u32 lost_out; /* Lost packets */
317 __u32 sacked_out; /* SACK'd packets */
318 __u32 fackets_out; /* FACK'd packets */
319 __u32 high_seq; /* snd_nxt at onset of congestion */
320
321 __u32 retrans_stamp; /* Timestamp of the last retransmit,
322 * also used in SYN-SENT to remember stamp of
323 * the first SYN. */
324 __u32 undo_marker; /* tracking retrans started here. */
325 int undo_retrans; /* number of undoable retransmissions. */
326 __u32 urg_seq; /* Seq of received urgent pointer */
327 __u16 urg_data; /* Saved octet of OOB data and control flags */
1da177e4 328 __u8 urg_mode; /* In urgent mode */
6687e988 329 __u8 ecn_flags; /* ECN status bits. */
1da177e4
LT
330 __u32 snd_up; /* Urgent pointer */
331
332 __u32 total_retrans; /* Total retransmits for entire connection */
9772efb9 333 __u32 bytes_acked; /* Appropriate Byte Counting - RFC3465 */
1da177e4 334
1da177e4
LT
335 unsigned int keepalive_time; /* time before keep alive takes place */
336 unsigned int keepalive_intvl; /* time interval between keep alive probes */
337 int linger2;
338
339 unsigned long last_synq_overflow;
340
341/* Receiver side RTT estimation */
342 struct {
343 __u32 rtt;
344 __u32 seq;
345 __u32 time;
346 } rcv_rtt_est;
347
348/* Receiver queue space */
349 struct {
350 int space;
351 __u32 seq;
352 __u32 time;
353 } rcvq_space;
0e7b1368
JH
354
355/* TCP-specific MTU probe information. */
356 struct {
357 __u32 probe_seq_start;
358 __u32 probe_seq_end;
359 } mtu_probe;
1da177e4
LT
360};
361
362static inline struct tcp_sock *tcp_sk(const struct sock *sk)
363{
364 return (struct tcp_sock *)sk;
365}
366
8feaf0c0
ACM
367struct tcp_timewait_sock {
368 struct inet_timewait_sock tw_sk;
369 __u32 tw_rcv_nxt;
370 __u32 tw_snd_nxt;
371 __u32 tw_rcv_wnd;
372 __u32 tw_ts_recent;
373 long tw_ts_recent_stamp;
374};
375
376static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
377{
378 return (struct tcp_timewait_sock *)sk;
379}
380
1da177e4
LT
381#endif
382
383#endif /* _LINUX_TCP_H */