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 | 41 | #include <net/tcp_states.h> |
bdf1ee5d | 42 | #include <net/inet_ecn.h> |
c752f073 | 43 | |
1da177e4 LT |
44 | #include <linux/seq_file.h> |
45 | ||
6e04e021 | 46 | extern struct inet_hashinfo tcp_hashinfo; |
1da177e4 | 47 | |
1da177e4 | 48 | extern atomic_t tcp_orphan_count; |
1da177e4 | 49 | extern void tcp_time_wait(struct sock *sk, int state, int timeo); |
1da177e4 | 50 | |
1da177e4 LT |
51 | #define MAX_TCP_HEADER (128 + MAX_HEADER) |
52 | ||
53 | /* | |
54 | * Never offer a window over 32767 without using window scaling. Some | |
55 | * poor stacks do signed 16bit maths! | |
56 | */ | |
57 | #define MAX_TCP_WINDOW 32767U | |
58 | ||
59 | /* Minimal accepted MSS. It is (60+60+8) - (20+20). */ | |
60 | #define TCP_MIN_MSS 88U | |
61 | ||
62 | /* Minimal RCV_MSS. */ | |
63 | #define TCP_MIN_RCVMSS 536U | |
64 | ||
5d424d5a JH |
65 | /* The least MTU to use for probing */ |
66 | #define TCP_BASE_MSS 512 | |
67 | ||
1da177e4 LT |
68 | /* After receiving this amount of duplicate ACKs fast retransmit starts. */ |
69 | #define TCP_FASTRETRANS_THRESH 3 | |
70 | ||
71 | /* Maximal reordering. */ | |
72 | #define TCP_MAX_REORDERING 127 | |
73 | ||
74 | /* Maximal number of ACKs sent quickly to accelerate slow-start. */ | |
75 | #define TCP_MAX_QUICKACKS 16U | |
76 | ||
77 | /* urg_data states */ | |
78 | #define TCP_URG_VALID 0x0100 | |
79 | #define TCP_URG_NOTYET 0x0200 | |
80 | #define TCP_URG_READ 0x0400 | |
81 | ||
82 | #define TCP_RETR1 3 /* | |
83 | * This is how many retries it does before it | |
84 | * tries to figure out if the gateway is | |
85 | * down. Minimal RFC value is 3; it corresponds | |
86 | * to ~3sec-8min depending on RTO. | |
87 | */ | |
88 | ||
89 | #define TCP_RETR2 15 /* | |
90 | * This should take at least | |
91 | * 90 minutes to time out. | |
92 | * RFC1122 says that the limit is 100 sec. | |
93 | * 15 is ~13-30min depending on RTO. | |
94 | */ | |
95 | ||
96 | #define TCP_SYN_RETRIES 5 /* number of times to retry active opening a | |
caa20d9a | 97 | * connection: ~180sec is RFC minimum */ |
1da177e4 LT |
98 | |
99 | #define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a | |
caa20d9a | 100 | * connection: ~180sec is RFC minimum */ |
1da177e4 LT |
101 | |
102 | ||
103 | #define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned | |
104 | * socket. 7 is ~50sec-16min. | |
105 | */ | |
106 | ||
107 | ||
108 | #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT | |
109 | * state, about 60 seconds */ | |
110 | #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN | |
111 | /* BSD style FIN_WAIT2 deadlock breaker. | |
112 | * It used to be 3min, new value is 60sec, | |
113 | * to combine FIN-WAIT-2 timeout with | |
114 | * TIME-WAIT timer. | |
115 | */ | |
116 | ||
117 | #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */ | |
118 | #if HZ >= 100 | |
119 | #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */ | |
120 | #define TCP_ATO_MIN ((unsigned)(HZ/25)) | |
121 | #else | |
122 | #define TCP_DELACK_MIN 4U | |
123 | #define TCP_ATO_MIN 4U | |
124 | #endif | |
125 | #define TCP_RTO_MAX ((unsigned)(120*HZ)) | |
126 | #define TCP_RTO_MIN ((unsigned)(HZ/5)) | |
127 | #define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */ | |
128 | ||
129 | #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes | |
130 | * for local resources. | |
131 | */ | |
132 | ||
133 | #define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */ | |
134 | #define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */ | |
135 | #define TCP_KEEPALIVE_INTVL (75*HZ) | |
136 | ||
137 | #define MAX_TCP_KEEPIDLE 32767 | |
138 | #define MAX_TCP_KEEPINTVL 32767 | |
139 | #define MAX_TCP_KEEPCNT 127 | |
140 | #define MAX_TCP_SYNCNT 127 | |
141 | ||
142 | #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */ | |
1da177e4 LT |
143 | |
144 | #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24) | |
145 | #define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated | |
146 | * after this time. It should be equal | |
147 | * (or greater than) TCP_TIMEWAIT_LEN | |
148 | * to provide reliability equal to one | |
149 | * provided by timewait state. | |
150 | */ | |
151 | #define TCP_PAWS_WINDOW 1 /* Replay window for per-host | |
152 | * timestamps. It must be less than | |
153 | * minimal timewait lifetime. | |
154 | */ | |
1da177e4 LT |
155 | /* |
156 | * TCP option | |
157 | */ | |
158 | ||
159 | #define TCPOPT_NOP 1 /* Padding */ | |
160 | #define TCPOPT_EOL 0 /* End of options */ | |
161 | #define TCPOPT_MSS 2 /* Segment size negotiating */ | |
162 | #define TCPOPT_WINDOW 3 /* Window scaling */ | |
163 | #define TCPOPT_SACK_PERM 4 /* SACK Permitted */ | |
164 | #define TCPOPT_SACK 5 /* SACK Block */ | |
165 | #define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */ | |
cfb6eeb4 | 166 | #define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */ |
1da177e4 LT |
167 | |
168 | /* | |
169 | * TCP option lengths | |
170 | */ | |
171 | ||
172 | #define TCPOLEN_MSS 4 | |
173 | #define TCPOLEN_WINDOW 3 | |
174 | #define TCPOLEN_SACK_PERM 2 | |
175 | #define TCPOLEN_TIMESTAMP 10 | |
cfb6eeb4 | 176 | #define TCPOLEN_MD5SIG 18 |
1da177e4 LT |
177 | |
178 | /* But this is what stacks really send out. */ | |
179 | #define TCPOLEN_TSTAMP_ALIGNED 12 | |
180 | #define TCPOLEN_WSCALE_ALIGNED 4 | |
181 | #define TCPOLEN_SACKPERM_ALIGNED 4 | |
182 | #define TCPOLEN_SACK_BASE 2 | |
183 | #define TCPOLEN_SACK_BASE_ALIGNED 4 | |
184 | #define TCPOLEN_SACK_PERBLOCK 8 | |
cfb6eeb4 | 185 | #define TCPOLEN_MD5SIG_ALIGNED 20 |
1da177e4 | 186 | |
1da177e4 LT |
187 | /* Flags in tp->nonagle */ |
188 | #define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */ | |
189 | #define TCP_NAGLE_CORK 2 /* Socket is corked */ | |
caa20d9a | 190 | #define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */ |
1da177e4 | 191 | |
295ff7ed ACM |
192 | extern struct inet_timewait_death_row tcp_death_row; |
193 | ||
1da177e4 | 194 | /* sysctl variables for tcp */ |
1da177e4 LT |
195 | extern int sysctl_tcp_timestamps; |
196 | extern int sysctl_tcp_window_scaling; | |
197 | extern int sysctl_tcp_sack; | |
198 | extern int sysctl_tcp_fin_timeout; | |
1da177e4 LT |
199 | extern int sysctl_tcp_keepalive_time; |
200 | extern int sysctl_tcp_keepalive_probes; | |
201 | extern int sysctl_tcp_keepalive_intvl; | |
202 | extern int sysctl_tcp_syn_retries; | |
203 | extern int sysctl_tcp_synack_retries; | |
204 | extern int sysctl_tcp_retries1; | |
205 | extern int sysctl_tcp_retries2; | |
206 | extern int sysctl_tcp_orphan_retries; | |
207 | extern int sysctl_tcp_syncookies; | |
208 | extern int sysctl_tcp_retrans_collapse; | |
209 | extern int sysctl_tcp_stdurg; | |
210 | extern int sysctl_tcp_rfc1337; | |
211 | extern int sysctl_tcp_abort_on_overflow; | |
212 | extern int sysctl_tcp_max_orphans; | |
1da177e4 LT |
213 | extern int sysctl_tcp_fack; |
214 | extern int sysctl_tcp_reordering; | |
215 | extern int sysctl_tcp_ecn; | |
216 | extern int sysctl_tcp_dsack; | |
217 | extern int sysctl_tcp_mem[3]; | |
218 | extern int sysctl_tcp_wmem[3]; | |
219 | extern int sysctl_tcp_rmem[3]; | |
220 | extern int sysctl_tcp_app_win; | |
221 | extern int sysctl_tcp_adv_win_scale; | |
222 | extern int sysctl_tcp_tw_reuse; | |
223 | extern int sysctl_tcp_frto; | |
3cfe3baa | 224 | extern int sysctl_tcp_frto_response; |
1da177e4 | 225 | extern int sysctl_tcp_low_latency; |
95937825 | 226 | extern int sysctl_tcp_dma_copybreak; |
1da177e4 | 227 | extern int sysctl_tcp_nometrics_save; |
1da177e4 LT |
228 | extern int sysctl_tcp_moderate_rcvbuf; |
229 | extern int sysctl_tcp_tso_win_divisor; | |
9772efb9 | 230 | extern int sysctl_tcp_abc; |
5d424d5a JH |
231 | extern int sysctl_tcp_mtu_probing; |
232 | extern int sysctl_tcp_base_mss; | |
15d99e02 | 233 | extern int sysctl_tcp_workaround_signed_windows; |
35089bb2 | 234 | extern int sysctl_tcp_slow_start_after_idle; |
886236c1 | 235 | extern int sysctl_tcp_max_ssthresh; |
1da177e4 LT |
236 | |
237 | extern atomic_t tcp_memory_allocated; | |
238 | extern atomic_t tcp_sockets_allocated; | |
239 | extern int tcp_memory_pressure; | |
240 | ||
1da177e4 LT |
241 | /* |
242 | * The next routines deal with comparing 32 bit unsigned ints | |
243 | * and worry about wraparound (automatic with unsigned arithmetic). | |
244 | */ | |
245 | ||
246 | static inline int before(__u32 seq1, __u32 seq2) | |
247 | { | |
0d630cc0 | 248 | return (__s32)(seq1-seq2) < 0; |
1da177e4 | 249 | } |
9a036b9c | 250 | #define after(seq2, seq1) before(seq1, seq2) |
1da177e4 LT |
251 | |
252 | /* is s2<=s1<=s3 ? */ | |
253 | static inline int between(__u32 seq1, __u32 seq2, __u32 seq3) | |
254 | { | |
255 | return seq3 - seq2 >= seq1 - seq2; | |
256 | } | |
257 | ||
e4fd5da3 PE |
258 | static inline int tcp_too_many_orphans(struct sock *sk, int num) |
259 | { | |
260 | return (num > sysctl_tcp_max_orphans) || | |
261 | (sk->sk_wmem_queued > SOCK_MIN_SNDBUF && | |
262 | atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]); | |
263 | } | |
1da177e4 LT |
264 | |
265 | extern struct proto tcp_prot; | |
266 | ||
267 | DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics); | |
268 | #define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field) | |
269 | #define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field) | |
270 | #define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field) | |
271 | #define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field) | |
272 | #define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val) | |
273 | #define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val) | |
274 | ||
1da177e4 LT |
275 | extern void tcp_v4_err(struct sk_buff *skb, u32); |
276 | ||
277 | extern void tcp_shutdown (struct sock *sk, int how); | |
278 | ||
279 | extern int tcp_v4_rcv(struct sk_buff *skb); | |
280 | ||
281 | extern int tcp_v4_remember_stamp(struct sock *sk); | |
282 | ||
8feaf0c0 | 283 | extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw); |
1da177e4 | 284 | |
3516ffb0 | 285 | extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
286 | struct msghdr *msg, size_t size); |
287 | extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags); | |
288 | ||
289 | extern int tcp_ioctl(struct sock *sk, | |
290 | int cmd, | |
291 | unsigned long arg); | |
292 | ||
293 | extern int tcp_rcv_state_process(struct sock *sk, | |
294 | struct sk_buff *skb, | |
295 | struct tcphdr *th, | |
296 | unsigned len); | |
297 | ||
298 | extern int tcp_rcv_established(struct sock *sk, | |
299 | struct sk_buff *skb, | |
300 | struct tcphdr *th, | |
301 | unsigned len); | |
302 | ||
303 | extern void tcp_rcv_space_adjust(struct sock *sk); | |
304 | ||
0e4b4992 CL |
305 | extern void tcp_cleanup_rbuf(struct sock *sk, int copied); |
306 | ||
6d6ee43e ACM |
307 | extern int tcp_twsk_unique(struct sock *sk, |
308 | struct sock *sktw, void *twp); | |
309 | ||
cfb6eeb4 YH |
310 | extern void tcp_twsk_destructor(struct sock *sk); |
311 | ||
463c84b9 ACM |
312 | static inline void tcp_dec_quickack_mode(struct sock *sk, |
313 | const unsigned int pkts) | |
1da177e4 | 314 | { |
463c84b9 | 315 | struct inet_connection_sock *icsk = inet_csk(sk); |
fc6415bc | 316 | |
463c84b9 ACM |
317 | if (icsk->icsk_ack.quick) { |
318 | if (pkts >= icsk->icsk_ack.quick) { | |
319 | icsk->icsk_ack.quick = 0; | |
fc6415bc | 320 | /* Leaving quickack mode we deflate ATO. */ |
463c84b9 | 321 | icsk->icsk_ack.ato = TCP_ATO_MIN; |
fc6415bc | 322 | } else |
463c84b9 | 323 | icsk->icsk_ack.quick -= pkts; |
1da177e4 LT |
324 | } |
325 | } | |
326 | ||
463c84b9 | 327 | extern void tcp_enter_quickack_mode(struct sock *sk); |
1da177e4 | 328 | |
1da177e4 LT |
329 | static inline void tcp_clear_options(struct tcp_options_received *rx_opt) |
330 | { | |
331 | rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0; | |
332 | } | |
333 | ||
bdf1ee5d IJ |
334 | #define TCP_ECN_OK 1 |
335 | #define TCP_ECN_QUEUE_CWR 2 | |
336 | #define TCP_ECN_DEMAND_CWR 4 | |
337 | ||
338 | static __inline__ void | |
339 | TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th) | |
340 | { | |
341 | if (sysctl_tcp_ecn && th->ece && th->cwr) | |
342 | inet_rsk(req)->ecn_ok = 1; | |
343 | } | |
344 | ||
1da177e4 LT |
345 | enum tcp_tw_status |
346 | { | |
347 | TCP_TW_SUCCESS = 0, | |
348 | TCP_TW_RST = 1, | |
349 | TCP_TW_ACK = 2, | |
350 | TCP_TW_SYN = 3 | |
351 | }; | |
352 | ||
353 | ||
8feaf0c0 | 354 | extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw, |
1da177e4 | 355 | struct sk_buff *skb, |
8feaf0c0 | 356 | const struct tcphdr *th); |
1da177e4 LT |
357 | |
358 | extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb, | |
60236fdd ACM |
359 | struct request_sock *req, |
360 | struct request_sock **prev); | |
1da177e4 LT |
361 | extern int tcp_child_process(struct sock *parent, |
362 | struct sock *child, | |
363 | struct sk_buff *skb); | |
46d0de4e | 364 | extern int tcp_use_frto(struct sock *sk); |
1da177e4 LT |
365 | extern void tcp_enter_frto(struct sock *sk); |
366 | extern void tcp_enter_loss(struct sock *sk, int how); | |
367 | extern void tcp_clear_retrans(struct tcp_sock *tp); | |
368 | extern void tcp_update_metrics(struct sock *sk); | |
369 | ||
370 | extern void tcp_close(struct sock *sk, | |
371 | long timeout); | |
1da177e4 LT |
372 | extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait); |
373 | ||
374 | extern int tcp_getsockopt(struct sock *sk, int level, | |
375 | int optname, | |
376 | char __user *optval, | |
377 | int __user *optlen); | |
378 | extern int tcp_setsockopt(struct sock *sk, int level, | |
379 | int optname, char __user *optval, | |
380 | int optlen); | |
3fdadf7d DM |
381 | extern int compat_tcp_getsockopt(struct sock *sk, |
382 | int level, int optname, | |
383 | char __user *optval, int __user *optlen); | |
384 | extern int compat_tcp_setsockopt(struct sock *sk, | |
385 | int level, int optname, | |
386 | char __user *optval, int optlen); | |
1da177e4 LT |
387 | extern void tcp_set_keepalive(struct sock *sk, int val); |
388 | extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, | |
389 | struct msghdr *msg, | |
390 | size_t len, int nonblock, | |
391 | int flags, int *addr_len); | |
392 | ||
1da177e4 LT |
393 | extern void tcp_parse_options(struct sk_buff *skb, |
394 | struct tcp_options_received *opt_rx, | |
395 | int estab); | |
396 | ||
397 | /* | |
398 | * TCP v4 functions exported for the inet6 API | |
399 | */ | |
400 | ||
8292a17a | 401 | extern void tcp_v4_send_check(struct sock *sk, int len, |
1da177e4 LT |
402 | struct sk_buff *skb); |
403 | ||
404 | extern int tcp_v4_conn_request(struct sock *sk, | |
405 | struct sk_buff *skb); | |
406 | ||
407 | extern struct sock * tcp_create_openreq_child(struct sock *sk, | |
60236fdd | 408 | struct request_sock *req, |
1da177e4 LT |
409 | struct sk_buff *skb); |
410 | ||
411 | extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk, | |
412 | struct sk_buff *skb, | |
60236fdd | 413 | struct request_sock *req, |
1da177e4 LT |
414 | struct dst_entry *dst); |
415 | ||
416 | extern int tcp_v4_do_rcv(struct sock *sk, | |
417 | struct sk_buff *skb); | |
418 | ||
419 | extern int tcp_v4_connect(struct sock *sk, | |
420 | struct sockaddr *uaddr, | |
421 | int addr_len); | |
422 | ||
423 | extern int tcp_connect(struct sock *sk); | |
424 | ||
425 | extern struct sk_buff * tcp_make_synack(struct sock *sk, | |
426 | struct dst_entry *dst, | |
60236fdd | 427 | struct request_sock *req); |
1da177e4 LT |
428 | |
429 | extern int tcp_disconnect(struct sock *sk, int flags); | |
430 | ||
431 | extern void tcp_unhash(struct sock *sk); | |
432 | ||
1da177e4 LT |
433 | /* From syncookies.c */ |
434 | extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb, | |
435 | struct ip_options *opt); | |
436 | extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb, | |
437 | __u16 *mss); | |
438 | ||
439 | /* tcp_output.c */ | |
440 | ||
9e412ba7 IJ |
441 | extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss, |
442 | int nonagle); | |
443 | extern int tcp_may_send_now(struct sock *sk); | |
1da177e4 LT |
444 | extern int tcp_retransmit_skb(struct sock *, struct sk_buff *); |
445 | extern void tcp_xmit_retransmit_queue(struct sock *); | |
446 | extern void tcp_simple_retransmit(struct sock *); | |
447 | extern int tcp_trim_head(struct sock *, struct sk_buff *, u32); | |
6475be16 | 448 | extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int); |
1da177e4 LT |
449 | |
450 | extern void tcp_send_probe0(struct sock *); | |
451 | extern void tcp_send_partial(struct sock *); | |
452 | extern int tcp_write_wakeup(struct sock *); | |
453 | extern void tcp_send_fin(struct sock *sk); | |
dd0fc66f | 454 | extern void tcp_send_active_reset(struct sock *sk, gfp_t priority); |
1da177e4 | 455 | extern int tcp_send_synack(struct sock *); |
c1b4a7e6 | 456 | extern void tcp_push_one(struct sock *, unsigned int mss_now); |
1da177e4 LT |
457 | extern void tcp_send_ack(struct sock *sk); |
458 | extern void tcp_send_delayed_ack(struct sock *sk); | |
459 | ||
a762a980 DM |
460 | /* tcp_input.c */ |
461 | extern void tcp_cwnd_application_limited(struct sock *sk); | |
462 | ||
1da177e4 LT |
463 | /* tcp_timer.c */ |
464 | extern void tcp_init_xmit_timers(struct sock *); | |
463c84b9 ACM |
465 | static inline void tcp_clear_xmit_timers(struct sock *sk) |
466 | { | |
467 | inet_csk_clear_xmit_timers(sk); | |
468 | } | |
1da177e4 | 469 | |
1da177e4 LT |
470 | extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu); |
471 | extern unsigned int tcp_current_mss(struct sock *sk, int large); | |
472 | ||
17b085ea | 473 | /* tcp.c */ |
1da177e4 LT |
474 | extern void tcp_get_info(struct sock *, struct tcp_info *); |
475 | ||
476 | /* Read 'sendfile()'-style from a TCP socket */ | |
477 | typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *, | |
478 | unsigned int, size_t); | |
479 | extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc, | |
480 | sk_read_actor_t recv_actor); | |
481 | ||
40efc6fa | 482 | extern void tcp_initialize_rcv_mss(struct sock *sk); |
1da177e4 | 483 | |
5d424d5a JH |
484 | extern int tcp_mtu_to_mss(struct sock *sk, int pmtu); |
485 | extern int tcp_mss_to_mtu(struct sock *sk, int mss); | |
486 | extern void tcp_mtup_init(struct sock *sk); | |
487 | ||
40efc6fa | 488 | static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd) |
1da177e4 LT |
489 | { |
490 | tp->pred_flags = htonl((tp->tcp_header_len << 26) | | |
491 | ntohl(TCP_FLAG_ACK) | | |
492 | snd_wnd); | |
493 | } | |
494 | ||
40efc6fa | 495 | static inline void tcp_fast_path_on(struct tcp_sock *tp) |
1da177e4 LT |
496 | { |
497 | __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale); | |
498 | } | |
499 | ||
9e412ba7 | 500 | static inline void tcp_fast_path_check(struct sock *sk) |
1da177e4 | 501 | { |
9e412ba7 IJ |
502 | struct tcp_sock *tp = tcp_sk(sk); |
503 | ||
b03efcfb | 504 | if (skb_queue_empty(&tp->out_of_order_queue) && |
1da177e4 LT |
505 | tp->rcv_wnd && |
506 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf && | |
507 | !tp->urg_data) | |
508 | tcp_fast_path_on(tp); | |
509 | } | |
510 | ||
511 | /* Compute the actual receive window we are currently advertising. | |
512 | * Rcv_nxt can be after the window if our peer push more data | |
513 | * than the offered window. | |
514 | */ | |
40efc6fa | 515 | static inline u32 tcp_receive_window(const struct tcp_sock *tp) |
1da177e4 LT |
516 | { |
517 | s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt; | |
518 | ||
519 | if (win < 0) | |
520 | win = 0; | |
521 | return (u32) win; | |
522 | } | |
523 | ||
524 | /* Choose a new window, without checks for shrinking, and without | |
525 | * scaling applied to the result. The caller does these things | |
526 | * if necessary. This is a "raw" window selection. | |
527 | */ | |
528 | extern u32 __tcp_select_window(struct sock *sk); | |
529 | ||
530 | /* TCP timestamps are only 32-bits, this causes a slight | |
531 | * complication on 64-bit systems since we store a snapshot | |
31f34269 SH |
532 | * of jiffies in the buffer control blocks below. We decided |
533 | * to use only the low 32-bits of jiffies and hide the ugly | |
1da177e4 LT |
534 | * casts with the following macro. |
535 | */ | |
536 | #define tcp_time_stamp ((__u32)(jiffies)) | |
537 | ||
caa20d9a | 538 | /* This is what the send packet queuing engine uses to pass |
1da177e4 LT |
539 | * TCP per-packet control information to the transmission |
540 | * code. We also store the host-order sequence numbers in | |
541 | * here too. This is 36 bytes on 32-bit architectures, | |
542 | * 40 bytes on 64-bit machines, if this grows please adjust | |
543 | * skbuff.h:skbuff->cb[xxx] size appropriately. | |
544 | */ | |
545 | struct tcp_skb_cb { | |
546 | union { | |
547 | struct inet_skb_parm h4; | |
548 | #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE) | |
549 | struct inet6_skb_parm h6; | |
550 | #endif | |
551 | } header; /* For incoming frames */ | |
552 | __u32 seq; /* Starting sequence number */ | |
553 | __u32 end_seq; /* SEQ + FIN + SYN + datalen */ | |
554 | __u32 when; /* used to compute rtt's */ | |
555 | __u8 flags; /* TCP header flags. */ | |
556 | ||
557 | /* NOTE: These must match up to the flags byte in a | |
558 | * real TCP header. | |
559 | */ | |
560 | #define TCPCB_FLAG_FIN 0x01 | |
561 | #define TCPCB_FLAG_SYN 0x02 | |
562 | #define TCPCB_FLAG_RST 0x04 | |
563 | #define TCPCB_FLAG_PSH 0x08 | |
564 | #define TCPCB_FLAG_ACK 0x10 | |
565 | #define TCPCB_FLAG_URG 0x20 | |
566 | #define TCPCB_FLAG_ECE 0x40 | |
567 | #define TCPCB_FLAG_CWR 0x80 | |
568 | ||
569 | __u8 sacked; /* State flags for SACK/FACK. */ | |
570 | #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */ | |
571 | #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */ | |
572 | #define TCPCB_LOST 0x04 /* SKB is lost */ | |
573 | #define TCPCB_TAGBITS 0x07 /* All tag bits */ | |
574 | ||
575 | #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */ | |
576 | #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS) | |
577 | ||
caa20d9a | 578 | #define TCPCB_URG 0x20 /* Urgent pointer advanced here */ |
1da177e4 LT |
579 | |
580 | #define TCPCB_AT_TAIL (TCPCB_URG) | |
581 | ||
582 | __u16 urg_ptr; /* Valid w/URG flags is set. */ | |
583 | __u32 ack_seq; /* Sequence number ACK'd */ | |
584 | }; | |
585 | ||
586 | #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0])) | |
587 | ||
1da177e4 LT |
588 | /* Due to TSO, an SKB can be composed of multiple actual |
589 | * packets. To keep these tracked properly, we use this. | |
590 | */ | |
591 | static inline int tcp_skb_pcount(const struct sk_buff *skb) | |
592 | { | |
7967168c | 593 | return skb_shinfo(skb)->gso_segs; |
1da177e4 LT |
594 | } |
595 | ||
596 | /* This is valid iff tcp_skb_pcount() > 1. */ | |
597 | static inline int tcp_skb_mss(const struct sk_buff *skb) | |
598 | { | |
7967168c | 599 | return skb_shinfo(skb)->gso_size; |
1da177e4 LT |
600 | } |
601 | ||
af610b4c | 602 | static inline void tcp_dec_pcount_approx_int(__u32 *count, const int decr) |
1da177e4 LT |
603 | { |
604 | if (*count) { | |
af610b4c | 605 | *count -= decr; |
1da177e4 LT |
606 | if ((int)*count < 0) |
607 | *count = 0; | |
608 | } | |
609 | } | |
610 | ||
af610b4c IJ |
611 | static inline void tcp_dec_pcount_approx(__u32 *count, |
612 | const struct sk_buff *skb) | |
613 | { | |
614 | tcp_dec_pcount_approx_int(count, tcp_skb_pcount(skb)); | |
615 | } | |
616 | ||
317a76f9 SH |
617 | /* Events passed to congestion control interface */ |
618 | enum tcp_ca_event { | |
619 | CA_EVENT_TX_START, /* first transmit when no packets in flight */ | |
620 | CA_EVENT_CWND_RESTART, /* congestion window restart */ | |
621 | CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */ | |
622 | CA_EVENT_FRTO, /* fast recovery timeout */ | |
623 | CA_EVENT_LOSS, /* loss timeout */ | |
624 | CA_EVENT_FAST_ACK, /* in sequence ack */ | |
625 | CA_EVENT_SLOW_ACK, /* other ack */ | |
626 | }; | |
627 | ||
628 | /* | |
629 | * Interface for adding new TCP congestion control handlers | |
630 | */ | |
631 | #define TCP_CA_NAME_MAX 16 | |
3ff825b2 SH |
632 | #define TCP_CA_MAX 128 |
633 | #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX) | |
634 | ||
164891aa SH |
635 | #define TCP_CONG_NON_RESTRICTED 0x1 |
636 | #define TCP_CONG_RTT_STAMP 0x2 | |
637 | ||
317a76f9 SH |
638 | struct tcp_congestion_ops { |
639 | struct list_head list; | |
164891aa | 640 | unsigned long flags; |
317a76f9 SH |
641 | |
642 | /* initialize private data (optional) */ | |
6687e988 | 643 | void (*init)(struct sock *sk); |
317a76f9 | 644 | /* cleanup private data (optional) */ |
6687e988 | 645 | void (*release)(struct sock *sk); |
317a76f9 SH |
646 | |
647 | /* return slow start threshold (required) */ | |
6687e988 | 648 | u32 (*ssthresh)(struct sock *sk); |
317a76f9 | 649 | /* lower bound for congestion window (optional) */ |
72dc5b92 | 650 | u32 (*min_cwnd)(const struct sock *sk); |
317a76f9 | 651 | /* do new cwnd calculation (required) */ |
16751347 | 652 | void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight, int good_ack); |
317a76f9 | 653 | /* call before changing ca_state (optional) */ |
6687e988 | 654 | void (*set_state)(struct sock *sk, u8 new_state); |
317a76f9 | 655 | /* call when cwnd event occurs (optional) */ |
6687e988 | 656 | void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev); |
317a76f9 | 657 | /* new value of cwnd after loss (optional) */ |
6687e988 | 658 | u32 (*undo_cwnd)(struct sock *sk); |
317a76f9 | 659 | /* hook for packet ack accounting (optional) */ |
30cfd0ba | 660 | void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us); |
73c1f4a0 | 661 | /* get info for inet_diag (optional) */ |
6687e988 | 662 | void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb); |
317a76f9 SH |
663 | |
664 | char name[TCP_CA_NAME_MAX]; | |
665 | struct module *owner; | |
666 | }; | |
667 | ||
668 | extern int tcp_register_congestion_control(struct tcp_congestion_ops *type); | |
669 | extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type); | |
670 | ||
6687e988 ACM |
671 | extern void tcp_init_congestion_control(struct sock *sk); |
672 | extern void tcp_cleanup_congestion_control(struct sock *sk); | |
317a76f9 SH |
673 | extern int tcp_set_default_congestion_control(const char *name); |
674 | extern void tcp_get_default_congestion_control(char *name); | |
3ff825b2 | 675 | extern void tcp_get_available_congestion_control(char *buf, size_t len); |
ce7bc3bf SH |
676 | extern void tcp_get_allowed_congestion_control(char *buf, size_t len); |
677 | extern int tcp_set_allowed_congestion_control(char *allowed); | |
6687e988 | 678 | extern int tcp_set_congestion_control(struct sock *sk, const char *name); |
40efc6fa | 679 | extern void tcp_slow_start(struct tcp_sock *tp); |
317a76f9 | 680 | |
5f8ef48d | 681 | extern struct tcp_congestion_ops tcp_init_congestion_ops; |
6687e988 | 682 | extern u32 tcp_reno_ssthresh(struct sock *sk); |
16751347 | 683 | extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight, int flag); |
72dc5b92 | 684 | extern u32 tcp_reno_min_cwnd(const struct sock *sk); |
a8acfbac | 685 | extern struct tcp_congestion_ops tcp_reno; |
317a76f9 | 686 | |
6687e988 | 687 | static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state) |
317a76f9 | 688 | { |
6687e988 ACM |
689 | struct inet_connection_sock *icsk = inet_csk(sk); |
690 | ||
691 | if (icsk->icsk_ca_ops->set_state) | |
692 | icsk->icsk_ca_ops->set_state(sk, ca_state); | |
693 | icsk->icsk_ca_state = ca_state; | |
317a76f9 SH |
694 | } |
695 | ||
6687e988 | 696 | static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event) |
317a76f9 | 697 | { |
6687e988 ACM |
698 | const struct inet_connection_sock *icsk = inet_csk(sk); |
699 | ||
700 | if (icsk->icsk_ca_ops->cwnd_event) | |
701 | icsk->icsk_ca_ops->cwnd_event(sk, event); | |
317a76f9 SH |
702 | } |
703 | ||
e60402d0 IJ |
704 | /* These functions determine how the current flow behaves in respect of SACK |
705 | * handling. SACK is negotiated with the peer, and therefore it can vary | |
706 | * between different flows. | |
707 | * | |
708 | * tcp_is_sack - SACK enabled | |
709 | * tcp_is_reno - No SACK | |
710 | * tcp_is_fack - FACK enabled, implies SACK enabled | |
711 | */ | |
712 | static inline int tcp_is_sack(const struct tcp_sock *tp) | |
713 | { | |
714 | return tp->rx_opt.sack_ok; | |
715 | } | |
716 | ||
717 | static inline int tcp_is_reno(const struct tcp_sock *tp) | |
718 | { | |
719 | return !tcp_is_sack(tp); | |
720 | } | |
721 | ||
722 | static inline int tcp_is_fack(const struct tcp_sock *tp) | |
723 | { | |
724 | return tp->rx_opt.sack_ok & 2; | |
725 | } | |
726 | ||
727 | static inline void tcp_enable_fack(struct tcp_sock *tp) | |
728 | { | |
729 | tp->rx_opt.sack_ok |= 2; | |
730 | } | |
731 | ||
83ae4088 IJ |
732 | static inline unsigned int tcp_left_out(const struct tcp_sock *tp) |
733 | { | |
734 | return tp->sacked_out + tp->lost_out; | |
735 | } | |
736 | ||
1da177e4 LT |
737 | /* This determines how many packets are "in the network" to the best |
738 | * of our knowledge. In many cases it is conservative, but where | |
739 | * detailed information is available from the receiver (via SACK | |
740 | * blocks etc.) we can make more aggressive calculations. | |
741 | * | |
742 | * Use this for decisions involving congestion control, use just | |
743 | * tp->packets_out to determine if the send queue is empty or not. | |
744 | * | |
745 | * Read this equation as: | |
746 | * | |
747 | * "Packets sent once on transmission queue" MINUS | |
748 | * "Packets left network, but not honestly ACKed yet" PLUS | |
749 | * "Packets fast retransmitted" | |
750 | */ | |
40efc6fa | 751 | static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp) |
1da177e4 | 752 | { |
83ae4088 | 753 | return tp->packets_out - tcp_left_out(tp) + tp->retrans_out; |
1da177e4 LT |
754 | } |
755 | ||
1da177e4 LT |
756 | /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd. |
757 | * The exception is rate halving phase, when cwnd is decreasing towards | |
758 | * ssthresh. | |
759 | */ | |
6687e988 | 760 | static inline __u32 tcp_current_ssthresh(const struct sock *sk) |
1da177e4 | 761 | { |
6687e988 ACM |
762 | const struct tcp_sock *tp = tcp_sk(sk); |
763 | if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery)) | |
1da177e4 LT |
764 | return tp->snd_ssthresh; |
765 | else | |
766 | return max(tp->snd_ssthresh, | |
767 | ((tp->snd_cwnd >> 1) + | |
768 | (tp->snd_cwnd >> 2))); | |
769 | } | |
770 | ||
b9c4595b IJ |
771 | /* Use define here intentionally to get WARN_ON location shown at the caller */ |
772 | #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out) | |
1da177e4 | 773 | |
3cfe3baa | 774 | extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh); |
1da177e4 LT |
775 | extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst); |
776 | ||
777 | /* Slow start with delack produces 3 packets of burst, so that | |
778 | * it is safe "de facto". | |
779 | */ | |
780 | static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp) | |
781 | { | |
782 | return 3; | |
783 | } | |
784 | ||
f4805ede SH |
785 | /* RFC2861 Check whether we are limited by application or congestion window |
786 | * This is the inverse of cwnd check in tcp_tso_should_defer | |
787 | */ | |
788 | static inline int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight) | |
789 | { | |
790 | const struct tcp_sock *tp = tcp_sk(sk); | |
791 | u32 left; | |
792 | ||
793 | if (in_flight >= tp->snd_cwnd) | |
794 | return 1; | |
795 | ||
bcd76111 | 796 | if (!sk_can_gso(sk)) |
f4805ede SH |
797 | return 0; |
798 | ||
799 | left = tp->snd_cwnd - in_flight; | |
800 | if (sysctl_tcp_tso_win_divisor) | |
801 | return left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd; | |
802 | else | |
803 | return left <= tcp_max_burst(tp); | |
804 | } | |
805 | ||
c1bd24b7 | 806 | static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss, |
40efc6fa | 807 | const struct sk_buff *skb) |
1da177e4 LT |
808 | { |
809 | if (skb->len < mss) | |
810 | tp->snd_sml = TCP_SKB_CB(skb)->end_seq; | |
811 | } | |
812 | ||
9e412ba7 | 813 | static inline void tcp_check_probe_timer(struct sock *sk) |
1da177e4 | 814 | { |
9e412ba7 | 815 | struct tcp_sock *tp = tcp_sk(sk); |
463c84b9 | 816 | const struct inet_connection_sock *icsk = inet_csk(sk); |
9e412ba7 | 817 | |
463c84b9 | 818 | if (!tp->packets_out && !icsk->icsk_pending) |
3f421baa ACM |
819 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, |
820 | icsk->icsk_rto, TCP_RTO_MAX); | |
1da177e4 LT |
821 | } |
822 | ||
9e412ba7 | 823 | static inline void tcp_push_pending_frames(struct sock *sk) |
1da177e4 | 824 | { |
9e412ba7 IJ |
825 | struct tcp_sock *tp = tcp_sk(sk); |
826 | ||
827 | __tcp_push_pending_frames(sk, tcp_current_mss(sk, 1), tp->nonagle); | |
1da177e4 LT |
828 | } |
829 | ||
40efc6fa | 830 | static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq) |
1da177e4 LT |
831 | { |
832 | tp->snd_wl1 = seq; | |
833 | } | |
834 | ||
40efc6fa | 835 | static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq) |
1da177e4 LT |
836 | { |
837 | tp->snd_wl1 = seq; | |
838 | } | |
839 | ||
1da177e4 LT |
840 | /* |
841 | * Calculate(/check) TCP checksum | |
842 | */ | |
ba7808ea FD |
843 | static inline __sum16 tcp_v4_check(int len, __be32 saddr, |
844 | __be32 daddr, __wsum base) | |
1da177e4 LT |
845 | { |
846 | return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base); | |
847 | } | |
848 | ||
b51655b9 | 849 | static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 850 | { |
fb286bb2 | 851 | return __skb_checksum_complete(skb); |
1da177e4 LT |
852 | } |
853 | ||
40efc6fa | 854 | static inline int tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 855 | { |
60476372 | 856 | return !skb_csum_unnecessary(skb) && |
1da177e4 LT |
857 | __tcp_checksum_complete(skb); |
858 | } | |
859 | ||
860 | /* Prequeue for VJ style copy to user, combined with checksumming. */ | |
861 | ||
40efc6fa | 862 | static inline void tcp_prequeue_init(struct tcp_sock *tp) |
1da177e4 LT |
863 | { |
864 | tp->ucopy.task = NULL; | |
865 | tp->ucopy.len = 0; | |
866 | tp->ucopy.memory = 0; | |
867 | skb_queue_head_init(&tp->ucopy.prequeue); | |
97fc2f08 CL |
868 | #ifdef CONFIG_NET_DMA |
869 | tp->ucopy.dma_chan = NULL; | |
870 | tp->ucopy.wakeup = 0; | |
871 | tp->ucopy.pinned_list = NULL; | |
872 | tp->ucopy.dma_cookie = 0; | |
873 | #endif | |
1da177e4 LT |
874 | } |
875 | ||
876 | /* Packet is added to VJ-style prequeue for processing in process | |
877 | * context, if a reader task is waiting. Apparently, this exciting | |
878 | * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93) | |
879 | * failed somewhere. Latency? Burstiness? Well, at least now we will | |
880 | * see, why it failed. 8)8) --ANK | |
881 | * | |
882 | * NOTE: is this not too big to inline? | |
883 | */ | |
40efc6fa | 884 | static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
885 | { |
886 | struct tcp_sock *tp = tcp_sk(sk); | |
887 | ||
888 | if (!sysctl_tcp_low_latency && tp->ucopy.task) { | |
889 | __skb_queue_tail(&tp->ucopy.prequeue, skb); | |
890 | tp->ucopy.memory += skb->truesize; | |
891 | if (tp->ucopy.memory > sk->sk_rcvbuf) { | |
892 | struct sk_buff *skb1; | |
893 | ||
894 | BUG_ON(sock_owned_by_user(sk)); | |
895 | ||
896 | while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) { | |
897 | sk->sk_backlog_rcv(sk, skb1); | |
898 | NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED); | |
899 | } | |
900 | ||
901 | tp->ucopy.memory = 0; | |
902 | } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) { | |
903 | wake_up_interruptible(sk->sk_sleep); | |
463c84b9 ACM |
904 | if (!inet_csk_ack_scheduled(sk)) |
905 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, | |
3f421baa ACM |
906 | (3 * TCP_RTO_MIN) / 4, |
907 | TCP_RTO_MAX); | |
1da177e4 LT |
908 | } |
909 | return 1; | |
910 | } | |
911 | return 0; | |
912 | } | |
913 | ||
914 | ||
915 | #undef STATE_TRACE | |
916 | ||
917 | #ifdef STATE_TRACE | |
918 | static const char *statename[]={ | |
919 | "Unused","Established","Syn Sent","Syn Recv", | |
920 | "Fin Wait 1","Fin Wait 2","Time Wait", "Close", | |
921 | "Close Wait","Last ACK","Listen","Closing" | |
922 | }; | |
923 | #endif | |
924 | ||
40efc6fa | 925 | static inline void tcp_set_state(struct sock *sk, int state) |
1da177e4 LT |
926 | { |
927 | int oldstate = sk->sk_state; | |
928 | ||
929 | switch (state) { | |
930 | case TCP_ESTABLISHED: | |
931 | if (oldstate != TCP_ESTABLISHED) | |
932 | TCP_INC_STATS(TCP_MIB_CURRESTAB); | |
933 | break; | |
934 | ||
935 | case TCP_CLOSE: | |
936 | if (oldstate == TCP_CLOSE_WAIT || oldstate == TCP_ESTABLISHED) | |
937 | TCP_INC_STATS(TCP_MIB_ESTABRESETS); | |
938 | ||
939 | sk->sk_prot->unhash(sk); | |
463c84b9 | 940 | if (inet_csk(sk)->icsk_bind_hash && |
1da177e4 | 941 | !(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) |
2d8c4ce5 | 942 | inet_put_port(&tcp_hashinfo, sk); |
1da177e4 LT |
943 | /* fall through */ |
944 | default: | |
945 | if (oldstate==TCP_ESTABLISHED) | |
946 | TCP_DEC_STATS(TCP_MIB_CURRESTAB); | |
947 | } | |
948 | ||
949 | /* Change state AFTER socket is unhashed to avoid closed | |
950 | * socket sitting in hash tables. | |
951 | */ | |
952 | sk->sk_state = state; | |
953 | ||
954 | #ifdef STATE_TRACE | |
955 | SOCK_DEBUG(sk, "TCP sk=%p, State %s -> %s\n",sk, statename[oldstate],statename[state]); | |
956 | #endif | |
957 | } | |
958 | ||
4ac02bab | 959 | extern void tcp_done(struct sock *sk); |
1da177e4 | 960 | |
40efc6fa | 961 | static inline void tcp_sack_reset(struct tcp_options_received *rx_opt) |
1da177e4 LT |
962 | { |
963 | rx_opt->dsack = 0; | |
964 | rx_opt->eff_sacks = 0; | |
965 | rx_opt->num_sacks = 0; | |
966 | } | |
967 | ||
1da177e4 LT |
968 | /* Determine a window scaling and initial window to offer. */ |
969 | extern void tcp_select_initial_window(int __space, __u32 mss, | |
970 | __u32 *rcv_wnd, __u32 *window_clamp, | |
971 | int wscale_ok, __u8 *rcv_wscale); | |
972 | ||
973 | static inline int tcp_win_from_space(int space) | |
974 | { | |
975 | return sysctl_tcp_adv_win_scale<=0 ? | |
976 | (space>>(-sysctl_tcp_adv_win_scale)) : | |
977 | space - (space>>sysctl_tcp_adv_win_scale); | |
978 | } | |
979 | ||
980 | /* Note: caller must be prepared to deal with negative returns */ | |
981 | static inline int tcp_space(const struct sock *sk) | |
982 | { | |
983 | return tcp_win_from_space(sk->sk_rcvbuf - | |
984 | atomic_read(&sk->sk_rmem_alloc)); | |
985 | } | |
986 | ||
987 | static inline int tcp_full_space(const struct sock *sk) | |
988 | { | |
989 | return tcp_win_from_space(sk->sk_rcvbuf); | |
990 | } | |
991 | ||
40efc6fa SH |
992 | static inline void tcp_openreq_init(struct request_sock *req, |
993 | struct tcp_options_received *rx_opt, | |
994 | struct sk_buff *skb) | |
1da177e4 | 995 | { |
2e6599cb ACM |
996 | struct inet_request_sock *ireq = inet_rsk(req); |
997 | ||
1da177e4 | 998 | req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */ |
2e6599cb | 999 | tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq; |
1da177e4 LT |
1000 | req->mss = rx_opt->mss_clamp; |
1001 | req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0; | |
2e6599cb ACM |
1002 | ireq->tstamp_ok = rx_opt->tstamp_ok; |
1003 | ireq->sack_ok = rx_opt->sack_ok; | |
1004 | ireq->snd_wscale = rx_opt->snd_wscale; | |
1005 | ireq->wscale_ok = rx_opt->wscale_ok; | |
1006 | ireq->acked = 0; | |
1007 | ireq->ecn_ok = 0; | |
aa8223c7 | 1008 | ireq->rmt_port = tcp_hdr(skb)->source; |
1da177e4 LT |
1009 | } |
1010 | ||
1011 | extern void tcp_enter_memory_pressure(void); | |
1012 | ||
1da177e4 LT |
1013 | static inline int keepalive_intvl_when(const struct tcp_sock *tp) |
1014 | { | |
1015 | return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl; | |
1016 | } | |
1017 | ||
1018 | static inline int keepalive_time_when(const struct tcp_sock *tp) | |
1019 | { | |
1020 | return tp->keepalive_time ? : sysctl_tcp_keepalive_time; | |
1021 | } | |
1022 | ||
463c84b9 | 1023 | static inline int tcp_fin_time(const struct sock *sk) |
1da177e4 | 1024 | { |
463c84b9 ACM |
1025 | int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout; |
1026 | const int rto = inet_csk(sk)->icsk_rto; | |
1da177e4 | 1027 | |
463c84b9 ACM |
1028 | if (fin_timeout < (rto << 2) - (rto >> 1)) |
1029 | fin_timeout = (rto << 2) - (rto >> 1); | |
1da177e4 LT |
1030 | |
1031 | return fin_timeout; | |
1032 | } | |
1033 | ||
1034 | static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst) | |
1035 | { | |
1036 | if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0) | |
1037 | return 0; | |
9d729f72 | 1038 | if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS) |
1da177e4 LT |
1039 | return 0; |
1040 | ||
1041 | /* RST segments are not recommended to carry timestamp, | |
1042 | and, if they do, it is recommended to ignore PAWS because | |
1043 | "their cleanup function should take precedence over timestamps." | |
1044 | Certainly, it is mistake. It is necessary to understand the reasons | |
1045 | of this constraint to relax it: if peer reboots, clock may go | |
1046 | out-of-sync and half-open connections will not be reset. | |
1047 | Actually, the problem would be not existing if all | |
1048 | the implementations followed draft about maintaining clock | |
1049 | via reboots. Linux-2.2 DOES NOT! | |
1050 | ||
1051 | However, we can relax time bounds for RST segments to MSL. | |
1052 | */ | |
9d729f72 | 1053 | if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL) |
1da177e4 LT |
1054 | return 0; |
1055 | return 1; | |
1056 | } | |
1057 | ||
1da177e4 LT |
1058 | #define TCP_CHECK_TIMER(sk) do { } while (0) |
1059 | ||
1da177e4 LT |
1060 | static inline void tcp_mib_init(void) |
1061 | { | |
1062 | /* See RFC 2012 */ | |
1063 | TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1); | |
1064 | TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ); | |
1065 | TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ); | |
1066 | TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1); | |
1067 | } | |
1068 | ||
5af4ec23 | 1069 | /* from STCP */ |
0800f170 DM |
1070 | static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp) |
1071 | { | |
6a438bbe SH |
1072 | tp->lost_skb_hint = NULL; |
1073 | tp->scoreboard_skb_hint = NULL; | |
1074 | tp->retransmit_skb_hint = NULL; | |
1075 | tp->forward_skb_hint = NULL; | |
b7689205 IJ |
1076 | } |
1077 | ||
0800f170 DM |
1078 | static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp) |
1079 | { | |
b7689205 | 1080 | tcp_clear_retrans_hints_partial(tp); |
6a438bbe SH |
1081 | tp->fastpath_skb_hint = NULL; |
1082 | } | |
1083 | ||
cfb6eeb4 YH |
1084 | /* MD5 Signature */ |
1085 | struct crypto_hash; | |
1086 | ||
1087 | /* - key database */ | |
1088 | struct tcp_md5sig_key { | |
1089 | u8 *key; | |
1090 | u8 keylen; | |
1091 | }; | |
1092 | ||
1093 | struct tcp4_md5sig_key { | |
f8ab18d2 | 1094 | struct tcp_md5sig_key base; |
cfb6eeb4 YH |
1095 | __be32 addr; |
1096 | }; | |
1097 | ||
1098 | struct tcp6_md5sig_key { | |
f8ab18d2 | 1099 | struct tcp_md5sig_key base; |
cfb6eeb4 YH |
1100 | #if 0 |
1101 | u32 scope_id; /* XXX */ | |
1102 | #endif | |
1103 | struct in6_addr addr; | |
1104 | }; | |
1105 | ||
1106 | /* - sock block */ | |
1107 | struct tcp_md5sig_info { | |
1108 | struct tcp4_md5sig_key *keys4; | |
1109 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) | |
1110 | struct tcp6_md5sig_key *keys6; | |
1111 | u32 entries6; | |
1112 | u32 alloced6; | |
1113 | #endif | |
1114 | u32 entries4; | |
1115 | u32 alloced4; | |
1116 | }; | |
1117 | ||
1118 | /* - pseudo header */ | |
1119 | struct tcp4_pseudohdr { | |
1120 | __be32 saddr; | |
1121 | __be32 daddr; | |
1122 | __u8 pad; | |
1123 | __u8 protocol; | |
1124 | __be16 len; | |
1125 | }; | |
1126 | ||
1127 | struct tcp6_pseudohdr { | |
1128 | struct in6_addr saddr; | |
1129 | struct in6_addr daddr; | |
1130 | __be32 len; | |
1131 | __be32 protocol; /* including padding */ | |
1132 | }; | |
1133 | ||
1134 | union tcp_md5sum_block { | |
1135 | struct tcp4_pseudohdr ip4; | |
1136 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) | |
1137 | struct tcp6_pseudohdr ip6; | |
1138 | #endif | |
1139 | }; | |
1140 | ||
1141 | /* - pool: digest algorithm, hash description and scratch buffer */ | |
1142 | struct tcp_md5sig_pool { | |
1143 | struct hash_desc md5_desc; | |
1144 | union tcp_md5sum_block md5_blk; | |
1145 | }; | |
1146 | ||
1147 | #define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */ | |
1148 | ||
1149 | /* - functions */ | |
1150 | extern int tcp_v4_calc_md5_hash(char *md5_hash, | |
1151 | struct tcp_md5sig_key *key, | |
1152 | struct sock *sk, | |
1153 | struct dst_entry *dst, | |
1154 | struct request_sock *req, | |
1155 | struct tcphdr *th, | |
1156 | int protocol, int tcplen); | |
1157 | extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, | |
1158 | struct sock *addr_sk); | |
1159 | ||
1160 | extern int tcp_v4_md5_do_add(struct sock *sk, | |
1161 | __be32 addr, | |
1162 | u8 *newkey, | |
1163 | u8 newkeylen); | |
1164 | ||
1165 | extern int tcp_v4_md5_do_del(struct sock *sk, | |
8e5200f5 | 1166 | __be32 addr); |
cfb6eeb4 YH |
1167 | |
1168 | extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void); | |
1169 | extern void tcp_free_md5sig_pool(void); | |
1170 | ||
1171 | extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu); | |
1172 | extern void __tcp_put_md5sig_pool(void); | |
1173 | ||
1174 | static inline | |
1175 | struct tcp_md5sig_pool *tcp_get_md5sig_pool(void) | |
1176 | { | |
1177 | int cpu = get_cpu(); | |
1178 | struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu); | |
1179 | if (!ret) | |
1180 | put_cpu(); | |
1181 | return ret; | |
1182 | } | |
1183 | ||
1184 | static inline void tcp_put_md5sig_pool(void) | |
1185 | { | |
1186 | __tcp_put_md5sig_pool(); | |
1187 | put_cpu(); | |
1188 | } | |
1189 | ||
fe067e8a DM |
1190 | /* write queue abstraction */ |
1191 | static inline void tcp_write_queue_purge(struct sock *sk) | |
1192 | { | |
1193 | struct sk_buff *skb; | |
1194 | ||
1195 | while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) | |
1196 | sk_stream_free_skb(sk, skb); | |
1197 | sk_stream_mem_reclaim(sk); | |
1198 | } | |
1199 | ||
1200 | static inline struct sk_buff *tcp_write_queue_head(struct sock *sk) | |
1201 | { | |
1202 | struct sk_buff *skb = sk->sk_write_queue.next; | |
1203 | if (skb == (struct sk_buff *) &sk->sk_write_queue) | |
1204 | return NULL; | |
1205 | return skb; | |
1206 | } | |
1207 | ||
1208 | static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk) | |
1209 | { | |
1210 | struct sk_buff *skb = sk->sk_write_queue.prev; | |
1211 | if (skb == (struct sk_buff *) &sk->sk_write_queue) | |
1212 | return NULL; | |
1213 | return skb; | |
1214 | } | |
1215 | ||
1216 | static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb) | |
1217 | { | |
1218 | return skb->next; | |
1219 | } | |
1220 | ||
1221 | #define tcp_for_write_queue(skb, sk) \ | |
1222 | for (skb = (sk)->sk_write_queue.next; \ | |
1223 | (skb != (struct sk_buff *)&(sk)->sk_write_queue); \ | |
1224 | skb = skb->next) | |
1225 | ||
1226 | #define tcp_for_write_queue_from(skb, sk) \ | |
1227 | for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\ | |
1228 | skb = skb->next) | |
1229 | ||
1230 | static inline struct sk_buff *tcp_send_head(struct sock *sk) | |
1231 | { | |
1232 | return sk->sk_send_head; | |
1233 | } | |
1234 | ||
1235 | static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb) | |
1236 | { | |
d551e454 IJ |
1237 | struct tcp_sock *tp = tcp_sk(sk); |
1238 | ||
fe067e8a DM |
1239 | sk->sk_send_head = skb->next; |
1240 | if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue) | |
1241 | sk->sk_send_head = NULL; | |
d551e454 IJ |
1242 | /* Don't override Nagle indefinately with F-RTO */ |
1243 | if (tp->frto_counter == 2) | |
1244 | tp->frto_counter = 3; | |
fe067e8a DM |
1245 | } |
1246 | ||
1247 | static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked) | |
1248 | { | |
1249 | if (sk->sk_send_head == skb_unlinked) | |
1250 | sk->sk_send_head = NULL; | |
1251 | } | |
1252 | ||
1253 | static inline void tcp_init_send_head(struct sock *sk) | |
1254 | { | |
1255 | sk->sk_send_head = NULL; | |
1256 | } | |
1257 | ||
1258 | static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1259 | { | |
1260 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1261 | } | |
1262 | ||
1263 | static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1264 | { | |
1265 | __tcp_add_write_queue_tail(sk, skb); | |
1266 | ||
1267 | /* Queue it, remembering where we must start sending. */ | |
1268 | if (sk->sk_send_head == NULL) | |
1269 | sk->sk_send_head = skb; | |
1270 | } | |
1271 | ||
1272 | static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb) | |
1273 | { | |
1274 | __skb_queue_head(&sk->sk_write_queue, skb); | |
1275 | } | |
1276 | ||
1277 | /* Insert buff after skb on the write queue of sk. */ | |
1278 | static inline void tcp_insert_write_queue_after(struct sk_buff *skb, | |
1279 | struct sk_buff *buff, | |
1280 | struct sock *sk) | |
1281 | { | |
1282 | __skb_append(skb, buff, &sk->sk_write_queue); | |
1283 | } | |
1284 | ||
1285 | /* Insert skb between prev and next on the write queue of sk. */ | |
1286 | static inline void tcp_insert_write_queue_before(struct sk_buff *new, | |
1287 | struct sk_buff *skb, | |
1288 | struct sock *sk) | |
1289 | { | |
1290 | __skb_insert(new, skb->prev, skb, &sk->sk_write_queue); |