]>
Commit | Line | Data |
---|---|---|
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 module. | |
7 | * | |
8 | * Version: @(#)tcp.h 1.0.5 05/23/93 | |
9 | * | |
02c30a84 | 10 | * Authors: Ross Biro |
1da177e4 LT |
11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
12 | * | |
13 | * This program is free software; you can redistribute it and/or | |
14 | * modify it under the terms of the GNU General Public License | |
15 | * as published by the Free Software Foundation; either version | |
16 | * 2 of the License, or (at your option) any later version. | |
17 | */ | |
18 | #ifndef _TCP_H | |
19 | #define _TCP_H | |
20 | ||
21 | #define TCP_DEBUG 1 | |
22 | #define FASTRETRANS_DEBUG 1 | |
23 | ||
1da177e4 LT |
24 | #include <linux/list.h> |
25 | #include <linux/tcp.h> | |
26 | #include <linux/slab.h> | |
27 | #include <linux/cache.h> | |
28 | #include <linux/percpu.h> | |
fb286bb2 | 29 | #include <linux/skbuff.h> |
97fc2f08 | 30 | #include <linux/dmaengine.h> |
cfb6eeb4 | 31 | #include <linux/crypto.h> |
3f421baa ACM |
32 | |
33 | #include <net/inet_connection_sock.h> | |
295ff7ed | 34 | #include <net/inet_timewait_sock.h> |
77d8bf9c | 35 | #include <net/inet_hashtables.h> |
1da177e4 | 36 | #include <net/checksum.h> |
2e6599cb | 37 | #include <net/request_sock.h> |
1da177e4 LT |
38 | #include <net/sock.h> |
39 | #include <net/snmp.h> | |
40 | #include <net/ip.h> | |
c752f073 ACM |
41 | #include <net/tcp_states.h> |
42 | ||
1da177e4 LT |
43 | #include <linux/seq_file.h> |
44 | ||
6e04e021 | 45 | extern struct inet_hashinfo tcp_hashinfo; |
1da177e4 | 46 | |
1da177e4 | 47 | extern atomic_t tcp_orphan_count; |
1da177e4 | 48 | extern void tcp_time_wait(struct sock *sk, int state, int timeo); |
1da177e4 | 49 | |
1da177e4 LT |
50 | #define MAX_TCP_HEADER (128 + MAX_HEADER) |
51 | ||
52 | /* | |
53 | * Never offer a window over 32767 without using window scaling. Some | |
54 | * poor stacks do signed 16bit maths! | |
55 | */ | |
56 | #define MAX_TCP_WINDOW 32767U | |
57 | ||
58 | /* Minimal accepted MSS. It is (60+60+8) - (20+20). */ | |
59 | #define TCP_MIN_MSS 88U | |
60 | ||
61 | /* Minimal RCV_MSS. */ | |
62 | #define TCP_MIN_RCVMSS 536U | |
63 | ||
5d424d5a JH |
64 | /* The least MTU to use for probing */ |
65 | #define TCP_BASE_MSS 512 | |
66 | ||
1da177e4 LT |
67 | /* After receiving this amount of duplicate ACKs fast retransmit starts. */ |
68 | #define TCP_FASTRETRANS_THRESH 3 | |
69 | ||
70 | /* Maximal reordering. */ | |
71 | #define TCP_MAX_REORDERING 127 | |
72 | ||
73 | /* Maximal number of ACKs sent quickly to accelerate slow-start. */ | |
74 | #define TCP_MAX_QUICKACKS 16U | |
75 | ||
76 | /* urg_data states */ | |
77 | #define TCP_URG_VALID 0x0100 | |
78 | #define TCP_URG_NOTYET 0x0200 | |
79 | #define TCP_URG_READ 0x0400 | |
80 | ||
81 | #define TCP_RETR1 3 /* | |
82 | * This is how many retries it does before it | |
83 | * tries to figure out if the gateway is | |
84 | * down. Minimal RFC value is 3; it corresponds | |
85 | * to ~3sec-8min depending on RTO. | |
86 | */ | |
87 | ||
88 | #define TCP_RETR2 15 /* | |
89 | * This should take at least | |
90 | * 90 minutes to time out. | |
91 | * RFC1122 says that the limit is 100 sec. | |
92 | * 15 is ~13-30min depending on RTO. | |
93 | */ | |
94 | ||
95 | #define TCP_SYN_RETRIES 5 /* number of times to retry active opening a | |
caa20d9a | 96 | * connection: ~180sec is RFC minimum */ |
1da177e4 LT |
97 | |
98 | #define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a | |
caa20d9a | 99 | * connection: ~180sec is RFC minimum */ |
1da177e4 LT |
100 | |
101 | ||
102 | #define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned | |
103 | * socket. 7 is ~50sec-16min. | |
104 | */ | |
105 | ||
106 | ||
107 | #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT | |
108 | * state, about 60 seconds */ | |
109 | #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN | |
110 | /* BSD style FIN_WAIT2 deadlock breaker. | |
111 | * It used to be 3min, new value is 60sec, | |
112 | * to combine FIN-WAIT-2 timeout with | |
113 | * TIME-WAIT timer. | |
114 | */ | |
115 | ||
116 | #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */ | |
117 | #if HZ >= 100 | |
118 | #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */ | |
119 | #define TCP_ATO_MIN ((unsigned)(HZ/25)) | |
120 | #else | |
121 | #define TCP_DELACK_MIN 4U | |
122 | #define TCP_ATO_MIN 4U | |
123 | #endif | |
124 | #define TCP_RTO_MAX ((unsigned)(120*HZ)) | |
125 | #define TCP_RTO_MIN ((unsigned)(HZ/5)) | |
126 | #define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */ | |
127 | ||
128 | #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes | |
129 | * for local resources. | |
130 | */ | |
131 | ||
132 | #define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */ | |
133 | #define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */ | |
134 | #define TCP_KEEPALIVE_INTVL (75*HZ) | |
135 | ||
136 | #define MAX_TCP_KEEPIDLE 32767 | |
137 | #define MAX_TCP_KEEPINTVL 32767 | |
138 | #define MAX_TCP_KEEPCNT 127 | |
139 | #define MAX_TCP_SYNCNT 127 | |
140 | ||
141 | #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */ | |
1da177e4 LT |
142 | |
143 | #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24) | |
144 | #define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated | |
145 | * after this time. It should be equal | |
146 | * (or greater than) TCP_TIMEWAIT_LEN | |
147 | * to provide reliability equal to one | |
148 | * provided by timewait state. | |
149 | */ | |
150 | #define TCP_PAWS_WINDOW 1 /* Replay window for per-host | |
151 | * timestamps. It must be less than | |
152 | * minimal timewait lifetime. | |
153 | */ | |
1da177e4 LT |
154 | /* |
155 | * TCP option | |
156 | */ | |
157 | ||
158 | #define TCPOPT_NOP 1 /* Padding */ | |
159 | #define TCPOPT_EOL 0 /* End of options */ | |
160 | #define TCPOPT_MSS 2 /* Segment size negotiating */ | |
161 | #define TCPOPT_WINDOW 3 /* Window scaling */ | |
162 | #define TCPOPT_SACK_PERM 4 /* SACK Permitted */ | |
163 | #define TCPOPT_SACK 5 /* SACK Block */ | |
164 | #define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */ | |
cfb6eeb4 | 165 | #define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */ |
1da177e4 LT |
166 | |
167 | /* | |
168 | * TCP option lengths | |
169 | */ | |
170 | ||
171 | #define TCPOLEN_MSS 4 | |
172 | #define TCPOLEN_WINDOW 3 | |
173 | #define TCPOLEN_SACK_PERM 2 | |
174 | #define TCPOLEN_TIMESTAMP 10 | |
cfb6eeb4 | 175 | #define TCPOLEN_MD5SIG 18 |
1da177e4 LT |
176 | |
177 | /* But this is what stacks really send out. */ | |
178 | #define TCPOLEN_TSTAMP_ALIGNED 12 | |
179 | #define TCPOLEN_WSCALE_ALIGNED 4 | |
180 | #define TCPOLEN_SACKPERM_ALIGNED 4 | |
181 | #define TCPOLEN_SACK_BASE 2 | |
182 | #define TCPOLEN_SACK_BASE_ALIGNED 4 | |
183 | #define TCPOLEN_SACK_PERBLOCK 8 | |
cfb6eeb4 | 184 | #define TCPOLEN_MD5SIG_ALIGNED 20 |
1da177e4 | 185 | |
1da177e4 LT |
186 | /* Flags in tp->nonagle */ |
187 | #define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */ | |
188 | #define TCP_NAGLE_CORK 2 /* Socket is corked */ | |
caa20d9a | 189 | #define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */ |
1da177e4 | 190 | |
295ff7ed ACM |
191 | extern struct inet_timewait_death_row tcp_death_row; |
192 | ||
1da177e4 | 193 | /* sysctl variables for tcp */ |
1da177e4 LT |
194 | extern int sysctl_tcp_timestamps; |
195 | extern int sysctl_tcp_window_scaling; | |
196 | extern int sysctl_tcp_sack; | |
197 | extern int sysctl_tcp_fin_timeout; | |
1da177e4 LT |
198 | extern int sysctl_tcp_keepalive_time; |
199 | extern int sysctl_tcp_keepalive_probes; | |
200 | extern int sysctl_tcp_keepalive_intvl; | |
201 | extern int sysctl_tcp_syn_retries; | |
202 | extern int sysctl_tcp_synack_retries; | |
203 | extern int sysctl_tcp_retries1; | |
204 | extern int sysctl_tcp_retries2; | |
205 | extern int sysctl_tcp_orphan_retries; | |
206 | extern int sysctl_tcp_syncookies; | |
207 | extern int sysctl_tcp_retrans_collapse; | |
208 | extern int sysctl_tcp_stdurg; | |
209 | extern int sysctl_tcp_rfc1337; | |
210 | extern int sysctl_tcp_abort_on_overflow; | |
211 | extern int sysctl_tcp_max_orphans; | |
1da177e4 LT |
212 | extern int sysctl_tcp_fack; |
213 | extern int sysctl_tcp_reordering; | |
214 | extern int sysctl_tcp_ecn; | |
215 | extern int sysctl_tcp_dsack; | |
216 | extern int sysctl_tcp_mem[3]; | |
217 | extern int sysctl_tcp_wmem[3]; | |
218 | extern int sysctl_tcp_rmem[3]; | |
219 | extern int sysctl_tcp_app_win; | |
220 | extern int sysctl_tcp_adv_win_scale; | |
221 | extern int sysctl_tcp_tw_reuse; | |
222 | extern int sysctl_tcp_frto; | |
223 | extern int sysctl_tcp_low_latency; | |
95937825 | 224 | extern int sysctl_tcp_dma_copybreak; |
1da177e4 | 225 | extern int sysctl_tcp_nometrics_save; |
1da177e4 LT |
226 | extern int sysctl_tcp_moderate_rcvbuf; |
227 | extern int sysctl_tcp_tso_win_divisor; | |
9772efb9 | 228 | extern int sysctl_tcp_abc; |
5d424d5a JH |
229 | extern int sysctl_tcp_mtu_probing; |
230 | extern int sysctl_tcp_base_mss; | |
15d99e02 | 231 | extern int sysctl_tcp_workaround_signed_windows; |
35089bb2 | 232 | extern int sysctl_tcp_slow_start_after_idle; |
1da177e4 LT |
233 | |
234 | extern atomic_t tcp_memory_allocated; | |
235 | extern atomic_t tcp_sockets_allocated; | |
236 | extern int tcp_memory_pressure; | |
237 | ||
1da177e4 LT |
238 | /* |
239 | * The next routines deal with comparing 32 bit unsigned ints | |
240 | * and worry about wraparound (automatic with unsigned arithmetic). | |
241 | */ | |
242 | ||
243 | static inline int before(__u32 seq1, __u32 seq2) | |
244 | { | |
0d630cc0 | 245 | return (__s32)(seq1-seq2) < 0; |
1da177e4 | 246 | } |
9a036b9c | 247 | #define after(seq2, seq1) before(seq1, seq2) |
1da177e4 LT |
248 | |
249 | /* is s2<=s1<=s3 ? */ | |
250 | static inline int between(__u32 seq1, __u32 seq2, __u32 seq3) | |
251 | { | |
252 | return seq3 - seq2 >= seq1 - seq2; | |
253 | } | |
254 | ||
255 | ||
256 | extern struct proto tcp_prot; | |
257 | ||
258 | DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics); | |
259 | #define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field) | |
260 | #define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field) | |
261 | #define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field) | |
262 | #define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field) | |
263 | #define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val) | |
264 | #define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val) | |
265 | ||
1da177e4 LT |
266 | extern void tcp_v4_err(struct sk_buff *skb, u32); |
267 | ||
268 | extern void tcp_shutdown (struct sock *sk, int how); | |
269 | ||
270 | extern int tcp_v4_rcv(struct sk_buff *skb); | |
271 | ||
272 | extern int tcp_v4_remember_stamp(struct sock *sk); | |
273 | ||
8feaf0c0 | 274 | extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw); |
1da177e4 LT |
275 | |
276 | extern int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, | |
277 | struct msghdr *msg, size_t size); | |
278 | extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags); | |
279 | ||
280 | extern int tcp_ioctl(struct sock *sk, | |
281 | int cmd, | |
282 | unsigned long arg); | |
283 | ||
284 | extern int tcp_rcv_state_process(struct sock *sk, | |
285 | struct sk_buff *skb, | |
286 | struct tcphdr *th, | |
287 | unsigned len); | |
288 | ||
289 | extern int tcp_rcv_established(struct sock *sk, | |
290 | struct sk_buff *skb, | |
291 | struct tcphdr *th, | |
292 | unsigned len); | |
293 | ||
294 | extern void tcp_rcv_space_adjust(struct sock *sk); | |
295 | ||
0e4b4992 CL |
296 | extern void tcp_cleanup_rbuf(struct sock *sk, int copied); |
297 | ||
6d6ee43e ACM |
298 | extern int tcp_twsk_unique(struct sock *sk, |
299 | struct sock *sktw, void *twp); | |
300 | ||
cfb6eeb4 YH |
301 | extern void tcp_twsk_destructor(struct sock *sk); |
302 | ||
463c84b9 ACM |
303 | static inline void tcp_dec_quickack_mode(struct sock *sk, |
304 | const unsigned int pkts) | |
1da177e4 | 305 | { |
463c84b9 | 306 | struct inet_connection_sock *icsk = inet_csk(sk); |
fc6415bc | 307 | |
463c84b9 ACM |
308 | if (icsk->icsk_ack.quick) { |
309 | if (pkts >= icsk->icsk_ack.quick) { | |
310 | icsk->icsk_ack.quick = 0; | |
fc6415bc | 311 | /* Leaving quickack mode we deflate ATO. */ |
463c84b9 | 312 | icsk->icsk_ack.ato = TCP_ATO_MIN; |
fc6415bc | 313 | } else |
463c84b9 | 314 | icsk->icsk_ack.quick -= pkts; |
1da177e4 LT |
315 | } |
316 | } | |
317 | ||
463c84b9 | 318 | extern void tcp_enter_quickack_mode(struct sock *sk); |
1da177e4 | 319 | |
1da177e4 LT |
320 | static inline void tcp_clear_options(struct tcp_options_received *rx_opt) |
321 | { | |
322 | rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0; | |
323 | } | |
324 | ||
325 | enum tcp_tw_status | |
326 | { | |
327 | TCP_TW_SUCCESS = 0, | |
328 | TCP_TW_RST = 1, | |
329 | TCP_TW_ACK = 2, | |
330 | TCP_TW_SYN = 3 | |
331 | }; | |
332 | ||
333 | ||
8feaf0c0 | 334 | extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw, |
1da177e4 | 335 | struct sk_buff *skb, |
8feaf0c0 | 336 | const struct tcphdr *th); |
1da177e4 LT |
337 | |
338 | extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb, | |
60236fdd ACM |
339 | struct request_sock *req, |
340 | struct request_sock **prev); | |
1da177e4 LT |
341 | extern int tcp_child_process(struct sock *parent, |
342 | struct sock *child, | |
343 | struct sk_buff *skb); | |
344 | extern void tcp_enter_frto(struct sock *sk); | |
345 | extern void tcp_enter_loss(struct sock *sk, int how); | |
346 | extern void tcp_clear_retrans(struct tcp_sock *tp); | |
347 | extern void tcp_update_metrics(struct sock *sk); | |
348 | ||
349 | extern void tcp_close(struct sock *sk, | |
350 | long timeout); | |
1da177e4 LT |
351 | extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait); |
352 | ||
353 | extern int tcp_getsockopt(struct sock *sk, int level, | |
354 | int optname, | |
355 | char __user *optval, | |
356 | int __user *optlen); | |
357 | extern int tcp_setsockopt(struct sock *sk, int level, | |
358 | int optname, char __user *optval, | |
359 | int optlen); | |
3fdadf7d DM |
360 | extern int compat_tcp_getsockopt(struct sock *sk, |
361 | int level, int optname, | |
362 | char __user *optval, int __user *optlen); | |
363 | extern int compat_tcp_setsockopt(struct sock *sk, | |
364 | int level, int optname, | |
365 | char __user *optval, int optlen); | |
1da177e4 LT |
366 | extern void tcp_set_keepalive(struct sock *sk, int val); |
367 | extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, | |
368 | struct msghdr *msg, | |
369 | size_t len, int nonblock, | |
370 | int flags, int *addr_len); | |
371 | ||
1da177e4 LT |
372 | extern void tcp_parse_options(struct sk_buff *skb, |
373 | struct tcp_options_received *opt_rx, | |
374 | int estab); | |
375 | ||
376 | /* | |
377 | * TCP v4 functions exported for the inet6 API | |
378 | */ | |
379 | ||
8292a17a | 380 | extern void tcp_v4_send_check(struct sock *sk, int len, |
1da177e4 LT |
381 | struct sk_buff *skb); |
382 | ||
383 | extern int tcp_v4_conn_request(struct sock *sk, | |
384 | struct sk_buff *skb); | |
385 | ||
386 | extern struct sock * tcp_create_openreq_child(struct sock *sk, | |
60236fdd | 387 | struct request_sock *req, |
1da177e4 LT |
388 | struct sk_buff *skb); |
389 | ||
390 | extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk, | |
391 | struct sk_buff *skb, | |
60236fdd | 392 | struct request_sock *req, |
1da177e4 LT |
393 | struct dst_entry *dst); |
394 | ||
395 | extern int tcp_v4_do_rcv(struct sock *sk, | |
396 | struct sk_buff *skb); | |
397 | ||
398 | extern int tcp_v4_connect(struct sock *sk, | |
399 | struct sockaddr *uaddr, | |
400 | int addr_len); | |
401 | ||
402 | extern int tcp_connect(struct sock *sk); | |
403 | ||
404 | extern struct sk_buff * tcp_make_synack(struct sock *sk, | |
405 | struct dst_entry *dst, | |
60236fdd | 406 | struct request_sock *req); |
1da177e4 LT |
407 | |
408 | extern int tcp_disconnect(struct sock *sk, int flags); | |
409 | ||
410 | extern void tcp_unhash(struct sock *sk); | |
411 | ||
1da177e4 LT |
412 | /* From syncookies.c */ |
413 | extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb, | |
414 | struct ip_options *opt); | |
415 | extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb, | |
416 | __u16 *mss); | |
417 | ||
418 | /* tcp_output.c */ | |
419 | ||
f6302d1d | 420 | extern void __tcp_push_pending_frames(struct sock *sk, struct tcp_sock *tp, |
a2e2a59c | 421 | unsigned int cur_mss, int nonagle); |
f6302d1d | 422 | extern int tcp_may_send_now(struct sock *sk, struct tcp_sock *tp); |
1da177e4 LT |
423 | extern int tcp_retransmit_skb(struct sock *, struct sk_buff *); |
424 | extern void tcp_xmit_retransmit_queue(struct sock *); | |
425 | extern void tcp_simple_retransmit(struct sock *); | |
426 | extern int tcp_trim_head(struct sock *, struct sk_buff *, u32); | |
6475be16 | 427 | extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int); |
1da177e4 LT |
428 | |
429 | extern void tcp_send_probe0(struct sock *); | |
430 | extern void tcp_send_partial(struct sock *); | |
431 | extern int tcp_write_wakeup(struct sock *); | |
432 | extern void tcp_send_fin(struct sock *sk); | |
dd0fc66f | 433 | extern void tcp_send_active_reset(struct sock *sk, gfp_t priority); |
1da177e4 | 434 | extern int tcp_send_synack(struct sock *); |
c1b4a7e6 | 435 | extern void tcp_push_one(struct sock *, unsigned int mss_now); |
1da177e4 LT |
436 | extern void tcp_send_ack(struct sock *sk); |
437 | extern void tcp_send_delayed_ack(struct sock *sk); | |
438 | ||
a762a980 DM |
439 | /* tcp_input.c */ |
440 | extern void tcp_cwnd_application_limited(struct sock *sk); | |
441 | ||
1da177e4 LT |
442 | /* tcp_timer.c */ |
443 | extern void tcp_init_xmit_timers(struct sock *); | |
463c84b9 ACM |
444 | static inline void tcp_clear_xmit_timers(struct sock *sk) |
445 | { | |
446 | inet_csk_clear_xmit_timers(sk); | |
447 | } | |
1da177e4 | 448 | |
1da177e4 LT |
449 | extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu); |
450 | extern unsigned int tcp_current_mss(struct sock *sk, int large); | |
451 | ||
17b085ea | 452 | /* tcp.c */ |
1da177e4 LT |
453 | extern void tcp_get_info(struct sock *, struct tcp_info *); |
454 | ||
455 | /* Read 'sendfile()'-style from a TCP socket */ | |
456 | typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *, | |
457 | unsigned int, size_t); | |
458 | extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc, | |
459 | sk_read_actor_t recv_actor); | |
460 | ||
40efc6fa | 461 | extern void tcp_initialize_rcv_mss(struct sock *sk); |
1da177e4 | 462 | |
5d424d5a JH |
463 | extern int tcp_mtu_to_mss(struct sock *sk, int pmtu); |
464 | extern int tcp_mss_to_mtu(struct sock *sk, int mss); | |
465 | extern void tcp_mtup_init(struct sock *sk); | |
466 | ||
40efc6fa | 467 | static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd) |
1da177e4 LT |
468 | { |
469 | tp->pred_flags = htonl((tp->tcp_header_len << 26) | | |
470 | ntohl(TCP_FLAG_ACK) | | |
471 | snd_wnd); | |
472 | } | |
473 | ||
40efc6fa | 474 | static inline void tcp_fast_path_on(struct tcp_sock *tp) |
1da177e4 LT |
475 | { |
476 | __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale); | |
477 | } | |
478 | ||
479 | static inline void tcp_fast_path_check(struct sock *sk, struct tcp_sock *tp) | |
480 | { | |
b03efcfb | 481 | if (skb_queue_empty(&tp->out_of_order_queue) && |
1da177e4 LT |
482 | tp->rcv_wnd && |
483 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf && | |
484 | !tp->urg_data) | |
485 | tcp_fast_path_on(tp); | |
486 | } | |
487 | ||
488 | /* Compute the actual receive window we are currently advertising. | |
489 | * Rcv_nxt can be after the window if our peer push more data | |
490 | * than the offered window. | |
491 | */ | |
40efc6fa | 492 | static inline u32 tcp_receive_window(const struct tcp_sock *tp) |
1da177e4 LT |
493 | { |
494 | s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt; | |
495 | ||
496 | if (win < 0) | |
497 | win = 0; | |
498 | return (u32) win; | |
499 | } | |
500 | ||
501 | /* Choose a new window, without checks for shrinking, and without | |
502 | * scaling applied to the result. The caller does these things | |
503 | * if necessary. This is a "raw" window selection. | |
504 | */ | |
505 | extern u32 __tcp_select_window(struct sock *sk); | |
506 | ||
507 | /* TCP timestamps are only 32-bits, this causes a slight | |
508 | * complication on 64-bit systems since we store a snapshot | |
31f34269 SH |
509 | * of jiffies in the buffer control blocks below. We decided |
510 | * to use only the low 32-bits of jiffies and hide the ugly | |
1da177e4 LT |
511 | * casts with the following macro. |
512 | */ | |
513 | #define tcp_time_stamp ((__u32)(jiffies)) | |
514 | ||
caa20d9a | 515 | /* This is what the send packet queuing engine uses to pass |
1da177e4 LT |
516 | * TCP per-packet control information to the transmission |
517 | * code. We also store the host-order sequence numbers in | |
518 | * here too. This is 36 bytes on 32-bit architectures, | |
519 | * 40 bytes on 64-bit machines, if this grows please adjust | |
520 | * skbuff.h:skbuff->cb[xxx] size appropriately. | |
521 | */ | |
522 | struct tcp_skb_cb { | |
523 | union { | |
524 | struct inet_skb_parm h4; | |
525 | #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE) | |
526 | struct inet6_skb_parm h6; | |
527 | #endif | |
528 | } header; /* For incoming frames */ | |
529 | __u32 seq; /* Starting sequence number */ | |
530 | __u32 end_seq; /* SEQ + FIN + SYN + datalen */ | |
531 | __u32 when; /* used to compute rtt's */ | |
532 | __u8 flags; /* TCP header flags. */ | |
533 | ||
534 | /* NOTE: These must match up to the flags byte in a | |
535 | * real TCP header. | |
536 | */ | |
537 | #define TCPCB_FLAG_FIN 0x01 | |
538 | #define TCPCB_FLAG_SYN 0x02 | |
539 | #define TCPCB_FLAG_RST 0x04 | |
540 | #define TCPCB_FLAG_PSH 0x08 | |
541 | #define TCPCB_FLAG_ACK 0x10 | |
542 | #define TCPCB_FLAG_URG 0x20 | |
543 | #define TCPCB_FLAG_ECE 0x40 | |
544 | #define TCPCB_FLAG_CWR 0x80 | |
545 | ||
546 | __u8 sacked; /* State flags for SACK/FACK. */ | |
547 | #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */ | |
548 | #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */ | |
549 | #define TCPCB_LOST 0x04 /* SKB is lost */ | |
550 | #define TCPCB_TAGBITS 0x07 /* All tag bits */ | |
551 | ||
552 | #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */ | |
553 | #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS) | |
554 | ||
caa20d9a | 555 | #define TCPCB_URG 0x20 /* Urgent pointer advanced here */ |
1da177e4 LT |
556 | |
557 | #define TCPCB_AT_TAIL (TCPCB_URG) | |
558 | ||
559 | __u16 urg_ptr; /* Valid w/URG flags is set. */ | |
560 | __u32 ack_seq; /* Sequence number ACK'd */ | |
561 | }; | |
562 | ||
563 | #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0])) | |
564 | ||
565 | #include <net/tcp_ecn.h> | |
566 | ||
567 | /* Due to TSO, an SKB can be composed of multiple actual | |
568 | * packets. To keep these tracked properly, we use this. | |
569 | */ | |
570 | static inline int tcp_skb_pcount(const struct sk_buff *skb) | |
571 | { | |
7967168c | 572 | return skb_shinfo(skb)->gso_segs; |
1da177e4 LT |
573 | } |
574 | ||
575 | /* This is valid iff tcp_skb_pcount() > 1. */ | |
576 | static inline int tcp_skb_mss(const struct sk_buff *skb) | |
577 | { | |
7967168c | 578 | return skb_shinfo(skb)->gso_size; |
1da177e4 LT |
579 | } |
580 | ||
581 | static inline void tcp_dec_pcount_approx(__u32 *count, | |
582 | const struct sk_buff *skb) | |
583 | { | |
584 | if (*count) { | |
585 | *count -= tcp_skb_pcount(skb); | |
586 | if ((int)*count < 0) | |
587 | *count = 0; | |
588 | } | |
589 | } | |
590 | ||
591 | static inline void tcp_packets_out_inc(struct sock *sk, | |
592 | struct tcp_sock *tp, | |
593 | const struct sk_buff *skb) | |
594 | { | |
595 | int orig = tp->packets_out; | |
596 | ||
597 | tp->packets_out += tcp_skb_pcount(skb); | |
598 | if (!orig) | |
3f421baa ACM |
599 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, |
600 | inet_csk(sk)->icsk_rto, TCP_RTO_MAX); | |
1da177e4 LT |
601 | } |
602 | ||
603 | static inline void tcp_packets_out_dec(struct tcp_sock *tp, | |
604 | const struct sk_buff *skb) | |
605 | { | |
606 | tp->packets_out -= tcp_skb_pcount(skb); | |
607 | } | |
608 | ||
317a76f9 SH |
609 | /* Events passed to congestion control interface */ |
610 | enum tcp_ca_event { | |
611 | CA_EVENT_TX_START, /* first transmit when no packets in flight */ | |
612 | CA_EVENT_CWND_RESTART, /* congestion window restart */ | |
613 | CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */ | |
614 | CA_EVENT_FRTO, /* fast recovery timeout */ | |
615 | CA_EVENT_LOSS, /* loss timeout */ | |
616 | CA_EVENT_FAST_ACK, /* in sequence ack */ | |
617 | CA_EVENT_SLOW_ACK, /* other ack */ | |
618 | }; | |
619 | ||
620 | /* | |
621 | * Interface for adding new TCP congestion control handlers | |
622 | */ | |
623 | #define TCP_CA_NAME_MAX 16 | |
3ff825b2 SH |
624 | #define TCP_CA_MAX 128 |
625 | #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX) | |
626 | ||
317a76f9 SH |
627 | struct tcp_congestion_ops { |
628 | struct list_head list; | |
ce7bc3bf | 629 | int non_restricted; |
317a76f9 SH |
630 | |
631 | /* initialize private data (optional) */ | |
6687e988 | 632 | void (*init)(struct sock *sk); |
317a76f9 | 633 | /* cleanup private data (optional) */ |
6687e988 | 634 | void (*release)(struct sock *sk); |
317a76f9 SH |
635 | |
636 | /* return slow start threshold (required) */ | |
6687e988 | 637 | u32 (*ssthresh)(struct sock *sk); |
317a76f9 | 638 | /* lower bound for congestion window (optional) */ |
72dc5b92 | 639 | u32 (*min_cwnd)(const struct sock *sk); |
317a76f9 | 640 | /* do new cwnd calculation (required) */ |
6687e988 | 641 | void (*cong_avoid)(struct sock *sk, u32 ack, |
317a76f9 SH |
642 | u32 rtt, u32 in_flight, int good_ack); |
643 | /* round trip time sample per acked packet (optional) */ | |
6687e988 | 644 | void (*rtt_sample)(struct sock *sk, u32 usrtt); |
317a76f9 | 645 | /* call before changing ca_state (optional) */ |
6687e988 | 646 | void (*set_state)(struct sock *sk, u8 new_state); |
317a76f9 | 647 | /* call when cwnd event occurs (optional) */ |
6687e988 | 648 | void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev); |
317a76f9 | 649 | /* new value of cwnd after loss (optional) */ |
6687e988 | 650 | u32 (*undo_cwnd)(struct sock *sk); |
317a76f9 | 651 | /* hook for packet ack accounting (optional) */ |
6687e988 | 652 | void (*pkts_acked)(struct sock *sk, u32 num_acked); |
73c1f4a0 | 653 | /* get info for inet_diag (optional) */ |
6687e988 | 654 | void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb); |
317a76f9 SH |
655 | |
656 | char name[TCP_CA_NAME_MAX]; | |
657 | struct module *owner; | |
658 | }; | |
659 | ||
660 | extern int tcp_register_congestion_control(struct tcp_congestion_ops *type); | |
661 | extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type); | |
662 | ||
6687e988 ACM |
663 | extern void tcp_init_congestion_control(struct sock *sk); |
664 | extern void tcp_cleanup_congestion_control(struct sock *sk); | |
317a76f9 SH |
665 | extern int tcp_set_default_congestion_control(const char *name); |
666 | extern void tcp_get_default_congestion_control(char *name); | |
3ff825b2 | 667 | extern void tcp_get_available_congestion_control(char *buf, size_t len); |
ce7bc3bf SH |
668 | extern void tcp_get_allowed_congestion_control(char *buf, size_t len); |
669 | extern int tcp_set_allowed_congestion_control(char *allowed); | |
6687e988 | 670 | extern int tcp_set_congestion_control(struct sock *sk, const char *name); |
40efc6fa | 671 | extern void tcp_slow_start(struct tcp_sock *tp); |
317a76f9 | 672 | |
5f8ef48d | 673 | extern struct tcp_congestion_ops tcp_init_congestion_ops; |
6687e988 ACM |
674 | extern u32 tcp_reno_ssthresh(struct sock *sk); |
675 | extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, | |
317a76f9 | 676 | u32 rtt, u32 in_flight, int flag); |
72dc5b92 | 677 | extern u32 tcp_reno_min_cwnd(const struct sock *sk); |
a8acfbac | 678 | extern struct tcp_congestion_ops tcp_reno; |
317a76f9 | 679 | |
6687e988 | 680 | static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state) |
317a76f9 | 681 | { |
6687e988 ACM |
682 | struct inet_connection_sock *icsk = inet_csk(sk); |
683 | ||
684 | if (icsk->icsk_ca_ops->set_state) | |
685 | icsk->icsk_ca_ops->set_state(sk, ca_state); | |
686 | icsk->icsk_ca_state = ca_state; | |
317a76f9 SH |
687 | } |
688 | ||
6687e988 | 689 | static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event) |
317a76f9 | 690 | { |
6687e988 ACM |
691 | const struct inet_connection_sock *icsk = inet_csk(sk); |
692 | ||
693 | if (icsk->icsk_ca_ops->cwnd_event) | |
694 | icsk->icsk_ca_ops->cwnd_event(sk, event); | |
317a76f9 SH |
695 | } |
696 | ||
1da177e4 LT |
697 | /* This determines how many packets are "in the network" to the best |
698 | * of our knowledge. In many cases it is conservative, but where | |
699 | * detailed information is available from the receiver (via SACK | |
700 | * blocks etc.) we can make more aggressive calculations. | |
701 | * | |
702 | * Use this for decisions involving congestion control, use just | |
703 | * tp->packets_out to determine if the send queue is empty or not. | |
704 | * | |
705 | * Read this equation as: | |
706 | * | |
707 | * "Packets sent once on transmission queue" MINUS | |
708 | * "Packets left network, but not honestly ACKed yet" PLUS | |
709 | * "Packets fast retransmitted" | |
710 | */ | |
40efc6fa | 711 | static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp) |
1da177e4 LT |
712 | { |
713 | return (tp->packets_out - tp->left_out + tp->retrans_out); | |
714 | } | |
715 | ||
1da177e4 LT |
716 | /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd. |
717 | * The exception is rate halving phase, when cwnd is decreasing towards | |
718 | * ssthresh. | |
719 | */ | |
6687e988 | 720 | static inline __u32 tcp_current_ssthresh(const struct sock *sk) |
1da177e4 | 721 | { |
6687e988 ACM |
722 | const struct tcp_sock *tp = tcp_sk(sk); |
723 | if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery)) | |
1da177e4 LT |
724 | return tp->snd_ssthresh; |
725 | else | |
726 | return max(tp->snd_ssthresh, | |
727 | ((tp->snd_cwnd >> 1) + | |
728 | (tp->snd_cwnd >> 2))); | |
729 | } | |
730 | ||
731 | static inline void tcp_sync_left_out(struct tcp_sock *tp) | |
732 | { | |
733 | if (tp->rx_opt.sack_ok && | |
734 | (tp->sacked_out >= tp->packets_out - tp->lost_out)) | |
735 | tp->sacked_out = tp->packets_out - tp->lost_out; | |
736 | tp->left_out = tp->sacked_out + tp->lost_out; | |
737 | } | |
738 | ||
40efc6fa | 739 | extern void tcp_enter_cwr(struct sock *sk); |
1da177e4 LT |
740 | extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst); |
741 | ||
742 | /* Slow start with delack produces 3 packets of burst, so that | |
743 | * it is safe "de facto". | |
744 | */ | |
745 | static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp) | |
746 | { | |
747 | return 3; | |
748 | } | |
749 | ||
f4805ede SH |
750 | /* RFC2861 Check whether we are limited by application or congestion window |
751 | * This is the inverse of cwnd check in tcp_tso_should_defer | |
752 | */ | |
753 | static inline int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight) | |
754 | { | |
755 | const struct tcp_sock *tp = tcp_sk(sk); | |
756 | u32 left; | |
757 | ||
758 | if (in_flight >= tp->snd_cwnd) | |
759 | return 1; | |
760 | ||
bcd76111 | 761 | if (!sk_can_gso(sk)) |
f4805ede SH |
762 | return 0; |
763 | ||
764 | left = tp->snd_cwnd - in_flight; | |
765 | if (sysctl_tcp_tso_win_divisor) | |
766 | return left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd; | |
767 | else | |
768 | return left <= tcp_max_burst(tp); | |
769 | } | |
770 | ||
40efc6fa SH |
771 | static inline void tcp_minshall_update(struct tcp_sock *tp, int mss, |
772 | const struct sk_buff *skb) | |
1da177e4 LT |
773 | { |
774 | if (skb->len < mss) | |
775 | tp->snd_sml = TCP_SKB_CB(skb)->end_seq; | |
776 | } | |
777 | ||
40efc6fa | 778 | static inline void tcp_check_probe_timer(struct sock *sk, struct tcp_sock *tp) |
1da177e4 | 779 | { |
463c84b9 ACM |
780 | const struct inet_connection_sock *icsk = inet_csk(sk); |
781 | if (!tp->packets_out && !icsk->icsk_pending) | |
3f421baa ACM |
782 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, |
783 | icsk->icsk_rto, TCP_RTO_MAX); | |
1da177e4 LT |
784 | } |
785 | ||
40efc6fa SH |
786 | static inline void tcp_push_pending_frames(struct sock *sk, |
787 | struct tcp_sock *tp) | |
1da177e4 LT |
788 | { |
789 | __tcp_push_pending_frames(sk, tp, tcp_current_mss(sk, 1), tp->nonagle); | |
790 | } | |
791 | ||
40efc6fa | 792 | static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq) |
1da177e4 LT |
793 | { |
794 | tp->snd_wl1 = seq; | |
795 | } | |
796 | ||
40efc6fa | 797 | static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq) |
1da177e4 LT |
798 | { |
799 | tp->snd_wl1 = seq; | |
800 | } | |
801 | ||
1da177e4 LT |
802 | /* |
803 | * Calculate(/check) TCP checksum | |
804 | */ | |
6b11687e AV |
805 | static inline __sum16 tcp_v4_check(struct tcphdr *th, int len, |
806 | __be32 saddr, __be32 daddr, | |
807 | __wsum base) | |
1da177e4 LT |
808 | { |
809 | return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base); | |
810 | } | |
811 | ||
b51655b9 | 812 | static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 813 | { |
fb286bb2 | 814 | return __skb_checksum_complete(skb); |
1da177e4 LT |
815 | } |
816 | ||
40efc6fa | 817 | static inline int tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 LT |
818 | { |
819 | return skb->ip_summed != CHECKSUM_UNNECESSARY && | |
820 | __tcp_checksum_complete(skb); | |
821 | } | |
822 | ||
823 | /* Prequeue for VJ style copy to user, combined with checksumming. */ | |
824 | ||
40efc6fa | 825 | static inline void tcp_prequeue_init(struct tcp_sock *tp) |
1da177e4 LT |
826 | { |
827 | tp->ucopy.task = NULL; | |
828 | tp->ucopy.len = 0; | |
829 | tp->ucopy.memory = 0; | |
830 | skb_queue_head_init(&tp->ucopy.prequeue); | |
97fc2f08 CL |
831 | #ifdef CONFIG_NET_DMA |
832 | tp->ucopy.dma_chan = NULL; | |
833 | tp->ucopy.wakeup = 0; | |
834 | tp->ucopy.pinned_list = NULL; | |
835 | tp->ucopy.dma_cookie = 0; | |
836 | #endif | |
1da177e4 LT |
837 | } |
838 | ||
839 | /* Packet is added to VJ-style prequeue for processing in process | |
840 | * context, if a reader task is waiting. Apparently, this exciting | |
841 | * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93) | |
842 | * failed somewhere. Latency? Burstiness? Well, at least now we will | |
843 | * see, why it failed. 8)8) --ANK | |
844 | * | |
845 | * NOTE: is this not too big to inline? | |
846 | */ | |
40efc6fa | 847 | static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
848 | { |
849 | struct tcp_sock *tp = tcp_sk(sk); | |
850 | ||
851 | if (!sysctl_tcp_low_latency && tp->ucopy.task) { | |
852 | __skb_queue_tail(&tp->ucopy.prequeue, skb); | |
853 | tp->ucopy.memory += skb->truesize; | |
854 | if (tp->ucopy.memory > sk->sk_rcvbuf) { | |
855 | struct sk_buff *skb1; | |
856 | ||
857 | BUG_ON(sock_owned_by_user(sk)); | |
858 | ||
859 | while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) { | |
860 | sk->sk_backlog_rcv(sk, skb1); | |
861 | NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED); | |
862 | } | |
863 | ||
864 | tp->ucopy.memory = 0; | |
865 | } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) { | |
866 | wake_up_interruptible(sk->sk_sleep); | |
463c84b9 ACM |
867 | if (!inet_csk_ack_scheduled(sk)) |
868 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, | |
3f421baa ACM |
869 | (3 * TCP_RTO_MIN) / 4, |
870 | TCP_RTO_MAX); | |
1da177e4 LT |
871 | } |
872 | return 1; | |
873 | } | |
874 | return 0; | |
875 | } | |
876 | ||
877 | ||
878 | #undef STATE_TRACE | |
879 | ||
880 | #ifdef STATE_TRACE | |
881 | static const char *statename[]={ | |
882 | "Unused","Established","Syn Sent","Syn Recv", | |
883 | "Fin Wait 1","Fin Wait 2","Time Wait", "Close", | |
884 | "Close Wait","Last ACK","Listen","Closing" | |
885 | }; | |
886 | #endif | |
887 | ||
40efc6fa | 888 | static inline void tcp_set_state(struct sock *sk, int state) |
1da177e4 LT |
889 | { |
890 | int oldstate = sk->sk_state; | |
891 | ||
892 | switch (state) { | |
893 | case TCP_ESTABLISHED: | |
894 | if (oldstate != TCP_ESTABLISHED) | |
895 | TCP_INC_STATS(TCP_MIB_CURRESTAB); | |
896 | break; | |
897 | ||
898 | case TCP_CLOSE: | |
899 | if (oldstate == TCP_CLOSE_WAIT || oldstate == TCP_ESTABLISHED) | |
900 | TCP_INC_STATS(TCP_MIB_ESTABRESETS); | |
901 | ||
902 | sk->sk_prot->unhash(sk); | |
463c84b9 | 903 | if (inet_csk(sk)->icsk_bind_hash && |
1da177e4 | 904 | !(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) |
2d8c4ce5 | 905 | inet_put_port(&tcp_hashinfo, sk); |
1da177e4 LT |
906 | /* fall through */ |
907 | default: | |
908 | if (oldstate==TCP_ESTABLISHED) | |
909 | TCP_DEC_STATS(TCP_MIB_CURRESTAB); | |
910 | } | |
911 | ||
912 | /* Change state AFTER socket is unhashed to avoid closed | |
913 | * socket sitting in hash tables. | |
914 | */ | |
915 | sk->sk_state = state; | |
916 | ||
917 | #ifdef STATE_TRACE | |
918 | SOCK_DEBUG(sk, "TCP sk=%p, State %s -> %s\n",sk, statename[oldstate],statename[state]); | |
919 | #endif | |
920 | } | |
921 | ||
40efc6fa | 922 | static inline void tcp_done(struct sock *sk) |
1da177e4 | 923 | { |
3687b1dc WY |
924 | if(sk->sk_state == TCP_SYN_SENT || sk->sk_state == TCP_SYN_RECV) |
925 | TCP_INC_STATS_BH(TCP_MIB_ATTEMPTFAILS); | |
926 | ||
1da177e4 LT |
927 | tcp_set_state(sk, TCP_CLOSE); |
928 | tcp_clear_xmit_timers(sk); | |
929 | ||
930 | sk->sk_shutdown = SHUTDOWN_MASK; | |
931 | ||
932 | if (!sock_flag(sk, SOCK_DEAD)) | |
933 | sk->sk_state_change(sk); | |
934 | else | |
0a5578cf | 935 | inet_csk_destroy_sock(sk); |
1da177e4 LT |
936 | } |
937 | ||
40efc6fa | 938 | static inline void tcp_sack_reset(struct tcp_options_received *rx_opt) |
1da177e4 LT |
939 | { |
940 | rx_opt->dsack = 0; | |
941 | rx_opt->eff_sacks = 0; | |
942 | rx_opt->num_sacks = 0; | |
943 | } | |
944 | ||
1da177e4 LT |
945 | /* Determine a window scaling and initial window to offer. */ |
946 | extern void tcp_select_initial_window(int __space, __u32 mss, | |
947 | __u32 *rcv_wnd, __u32 *window_clamp, | |
948 | int wscale_ok, __u8 *rcv_wscale); | |
949 | ||
950 | static inline int tcp_win_from_space(int space) | |
951 | { | |
952 | return sysctl_tcp_adv_win_scale<=0 ? | |
953 | (space>>(-sysctl_tcp_adv_win_scale)) : | |
954 | space - (space>>sysctl_tcp_adv_win_scale); | |
955 | } | |
956 | ||
957 | /* Note: caller must be prepared to deal with negative returns */ | |
958 | static inline int tcp_space(const struct sock *sk) | |
959 | { | |
960 | return tcp_win_from_space(sk->sk_rcvbuf - | |
961 | atomic_read(&sk->sk_rmem_alloc)); | |
962 | } | |
963 | ||
964 | static inline int tcp_full_space(const struct sock *sk) | |
965 | { | |
966 | return tcp_win_from_space(sk->sk_rcvbuf); | |
967 | } | |
968 | ||
40efc6fa SH |
969 | static inline void tcp_openreq_init(struct request_sock *req, |
970 | struct tcp_options_received *rx_opt, | |
971 | struct sk_buff *skb) | |
1da177e4 | 972 | { |
2e6599cb ACM |
973 | struct inet_request_sock *ireq = inet_rsk(req); |
974 | ||
1da177e4 | 975 | req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */ |
2e6599cb | 976 | tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq; |
1da177e4 LT |
977 | req->mss = rx_opt->mss_clamp; |
978 | req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0; | |
2e6599cb ACM |
979 | ireq->tstamp_ok = rx_opt->tstamp_ok; |
980 | ireq->sack_ok = rx_opt->sack_ok; | |
981 | ireq->snd_wscale = rx_opt->snd_wscale; | |
982 | ireq->wscale_ok = rx_opt->wscale_ok; | |
983 | ireq->acked = 0; | |
984 | ireq->ecn_ok = 0; | |
985 | ireq->rmt_port = skb->h.th->source; | |
1da177e4 LT |
986 | } |
987 | ||
988 | extern void tcp_enter_memory_pressure(void); | |
989 | ||
1da177e4 LT |
990 | static inline int keepalive_intvl_when(const struct tcp_sock *tp) |
991 | { | |
992 | return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl; | |
993 | } | |
994 | ||
995 | static inline int keepalive_time_when(const struct tcp_sock *tp) | |
996 | { | |
997 | return tp->keepalive_time ? : sysctl_tcp_keepalive_time; | |
998 | } | |
999 | ||
463c84b9 | 1000 | static inline int tcp_fin_time(const struct sock *sk) |
1da177e4 | 1001 | { |
463c84b9 ACM |
1002 | int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout; |
1003 | const int rto = inet_csk(sk)->icsk_rto; | |
1da177e4 | 1004 | |
463c84b9 ACM |
1005 | if (fin_timeout < (rto << 2) - (rto >> 1)) |
1006 | fin_timeout = (rto << 2) - (rto >> 1); | |
1da177e4 LT |
1007 | |
1008 | return fin_timeout; | |
1009 | } | |
1010 | ||
1011 | static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst) | |
1012 | { | |
1013 | if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0) | |
1014 | return 0; | |
1015 | if (xtime.tv_sec >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS) | |
1016 | return 0; | |
1017 | ||
1018 | /* RST segments are not recommended to carry timestamp, | |
1019 | and, if they do, it is recommended to ignore PAWS because | |
1020 | "their cleanup function should take precedence over timestamps." | |
1021 | Certainly, it is mistake. It is necessary to understand the reasons | |
1022 | of this constraint to relax it: if peer reboots, clock may go | |
1023 | out-of-sync and half-open connections will not be reset. | |
1024 | Actually, the problem would be not existing if all | |
1025 | the implementations followed draft about maintaining clock | |
1026 | via reboots. Linux-2.2 DOES NOT! | |
1027 | ||
1028 | However, we can relax time bounds for RST segments to MSL. | |
1029 | */ | |
1030 | if (rst && xtime.tv_sec >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL) | |
1031 | return 0; | |
1032 | return 1; | |
1033 | } | |
1034 | ||
1da177e4 LT |
1035 | #define TCP_CHECK_TIMER(sk) do { } while (0) |
1036 | ||
1037 | static inline int tcp_use_frto(const struct sock *sk) | |
1038 | { | |
1039 | const struct tcp_sock *tp = tcp_sk(sk); | |
1040 | ||
1041 | /* F-RTO must be activated in sysctl and there must be some | |
1042 | * unsent new data, and the advertised window should allow | |
1043 | * sending it. | |
1044 | */ | |
1045 | return (sysctl_tcp_frto && sk->sk_send_head && | |
1046 | !after(TCP_SKB_CB(sk->sk_send_head)->end_seq, | |
1047 | tp->snd_una + tp->snd_wnd)); | |
1048 | } | |
1049 | ||
1050 | static inline void tcp_mib_init(void) | |
1051 | { | |
1052 | /* See RFC 2012 */ | |
1053 | TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1); | |
1054 | TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ); | |
1055 | TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ); | |
1056 | TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1); | |
1057 | } | |
1058 | ||
6a438bbe SH |
1059 | /*from STCP */ |
1060 | static inline void clear_all_retrans_hints(struct tcp_sock *tp){ | |
1061 | tp->lost_skb_hint = NULL; | |
1062 | tp->scoreboard_skb_hint = NULL; | |
1063 | tp->retransmit_skb_hint = NULL; | |
1064 | tp->forward_skb_hint = NULL; | |
1065 | tp->fastpath_skb_hint = NULL; | |
1066 | } | |
1067 | ||
cfb6eeb4 YH |
1068 | /* MD5 Signature */ |
1069 | struct crypto_hash; | |
1070 | ||
1071 | /* - key database */ | |
1072 | struct tcp_md5sig_key { | |
1073 | u8 *key; | |
1074 | u8 keylen; | |
1075 | }; | |
1076 | ||
1077 | struct tcp4_md5sig_key { | |
1078 | u8 *key; | |
1079 | u16 keylen; | |
1080 | __be32 addr; | |
1081 | }; | |
1082 | ||
1083 | struct tcp6_md5sig_key { | |
1084 | u8 *key; | |
1085 | u16 keylen; | |
1086 | #if 0 | |
1087 | u32 scope_id; /* XXX */ | |
1088 | #endif | |
1089 | struct in6_addr addr; | |
1090 | }; | |
1091 | ||
1092 | /* - sock block */ | |
1093 | struct tcp_md5sig_info { | |
1094 | struct tcp4_md5sig_key *keys4; | |
1095 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) | |
1096 | struct tcp6_md5sig_key *keys6; | |
1097 | u32 entries6; | |
1098 | u32 alloced6; | |
1099 | #endif | |
1100 | u32 entries4; | |
1101 | u32 alloced4; | |
1102 | }; | |
1103 | ||
1104 | /* - pseudo header */ | |
1105 | struct tcp4_pseudohdr { | |
1106 | __be32 saddr; | |
1107 | __be32 daddr; | |
1108 | __u8 pad; | |
1109 | __u8 protocol; | |
1110 | __be16 len; | |
1111 | }; | |
1112 | ||
1113 | struct tcp6_pseudohdr { | |
1114 | struct in6_addr saddr; | |
1115 | struct in6_addr daddr; | |
1116 | __be32 len; | |
1117 | __be32 protocol; /* including padding */ | |
1118 | }; | |
1119 | ||
1120 | union tcp_md5sum_block { | |
1121 | struct tcp4_pseudohdr ip4; | |
1122 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) | |
1123 | struct tcp6_pseudohdr ip6; | |
1124 | #endif | |
1125 | }; | |
1126 | ||
1127 | /* - pool: digest algorithm, hash description and scratch buffer */ | |
1128 | struct tcp_md5sig_pool { | |
1129 | struct hash_desc md5_desc; | |
1130 | union tcp_md5sum_block md5_blk; | |
1131 | }; | |
1132 | ||
1133 | #define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */ | |
1134 | ||
1135 | /* - functions */ | |
1136 | extern int tcp_v4_calc_md5_hash(char *md5_hash, | |
1137 | struct tcp_md5sig_key *key, | |
1138 | struct sock *sk, | |
1139 | struct dst_entry *dst, | |
1140 | struct request_sock *req, | |
1141 | struct tcphdr *th, | |
1142 | int protocol, int tcplen); | |
1143 | extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, | |
1144 | struct sock *addr_sk); | |
1145 | ||
1146 | extern int tcp_v4_md5_do_add(struct sock *sk, | |
1147 | __be32 addr, | |
1148 | u8 *newkey, | |
1149 | u8 newkeylen); | |
1150 | ||
1151 | extern int tcp_v4_md5_do_del(struct sock *sk, | |
8e5200f5 | 1152 | __be32 addr); |
cfb6eeb4 YH |
1153 | |
1154 | extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void); | |
1155 | extern void tcp_free_md5sig_pool(void); | |
1156 | ||
1157 | extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu); | |
1158 | extern void __tcp_put_md5sig_pool(void); | |
1159 | ||
1160 | static inline | |
1161 | struct tcp_md5sig_pool *tcp_get_md5sig_pool(void) | |
1162 | { | |
1163 | int cpu = get_cpu(); | |
1164 | struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu); | |
1165 | if (!ret) | |
1166 | put_cpu(); | |
1167 | return ret; | |
1168 | } | |
1169 | ||
1170 | static inline void tcp_put_md5sig_pool(void) | |
1171 | { | |
1172 | __tcp_put_md5sig_pool(); | |
1173 | put_cpu(); | |
1174 | } | |
1175 | ||
1da177e4 LT |
1176 | /* /proc */ |
1177 | enum tcp_seq_states { | |
1178 | TCP_SEQ_STATE_LISTENING, | |
1179 | TCP_SEQ_STATE_OPENREQ, | |
1180 | TCP_SEQ_STATE_ESTABLISHED, | |
1181 | TCP_SEQ_STATE_TIME_WAIT, | |
1182 | }; | |
1183 | ||
1184 | struct tcp_seq_afinfo { | |
1185 | struct module *owner; | |
1186 | char *name; | |
1187 | sa_family_t family; | |
1188 | int (*seq_show) (struct seq_file *m, void *v); | |
1189 | struct file_operations *seq_fops; | |
1190 | }; | |
1191 | ||
1192 | struct tcp_iter_state { | |
1193 | sa_family_t family; | |
1194 | enum tcp_seq_states state; | |
1195 | struct sock *syn_wait_sk; | |
1196 | int bucket, sbucket, num, uid; | |
1197 | struct seq_operations seq_ops; | |
1198 | }; | |
1199 | ||
1200 | extern int tcp_proc_register(struct tcp_seq_afinfo *afinfo); | |
1201 | extern void tcp_proc_unregister(struct tcp_seq_afinfo *afinfo); | |
1202 | ||
20380731 ACM |
1203 | extern struct request_sock_ops tcp_request_sock_ops; |
1204 | ||
1205 | extern int tcp_v4_destroy_sock(struct sock *sk); | |
1206 | ||
a430a43d | 1207 | extern int tcp_v4_gso_send_check(struct sk_buff *skb); |
576a30eb | 1208 | extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features); |
f4c50d99 | 1209 | |
20380731 ACM |
1210 | #ifdef CONFIG_PROC_FS |
1211 | extern int tcp4_proc_init(void); | |
1212 | extern void tcp4_proc_exit(void); | |
1213 | #endif | |
1214 | ||
cfb6eeb4 YH |
1215 | /* TCP af-specific functions */ |
1216 | struct tcp_sock_af_ops { | |
1217 | #ifdef CONFIG_TCP_MD5SIG | |
1218 | struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk, | |
1219 | struct sock *addr_sk); | |
1220 | int (*calc_md5_hash) (char *location, | |
1221 | struct tcp_md5sig_key *md5, | |
1222 | struct sock *sk, | |
1223 | struct dst_entry *dst, | |
1224 | struct request_sock *req, | |
1225 | struct tcphdr *th, | |
1226 | int protocol, int len); | |
1227 | int (*md5_add) (struct sock *sk, | |
1228 | struct sock *addr_sk, | |
1229 | u8 *newkey, | |
1230 | u8 len); | |
1231 | int (*md5_parse) (struct sock *sk, | |
1232 | char __user *optval, | |
1233 | int optlen); | |
1234 | #endif | |
1235 | }; | |
1236 | ||
1237 | struct tcp_request_sock_ops { | |
1238 | #ifdef CONFIG_TCP_MD5SIG | |
1239 | struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk, | |
1240 | struct request_sock *req); | |
1241 | #endif | |
1242 | }; | |
1243 | ||
20380731 ACM |
1244 | extern void tcp_v4_init(struct net_proto_family *ops); |
1245 | extern void tcp_init(void); | |
1246 | ||
1da177e4 | 1247 | #endif /* _TCP_H */ |