]>
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> |
c6aefafb | 32 | #include <linux/cryptohash.h> |
3f421baa ACM |
33 | |
34 | #include <net/inet_connection_sock.h> | |
295ff7ed | 35 | #include <net/inet_timewait_sock.h> |
77d8bf9c | 36 | #include <net/inet_hashtables.h> |
1da177e4 | 37 | #include <net/checksum.h> |
2e6599cb | 38 | #include <net/request_sock.h> |
1da177e4 LT |
39 | #include <net/sock.h> |
40 | #include <net/snmp.h> | |
41 | #include <net/ip.h> | |
c752f073 | 42 | #include <net/tcp_states.h> |
bdf1ee5d | 43 | #include <net/inet_ecn.h> |
c752f073 | 44 | |
1da177e4 LT |
45 | #include <linux/seq_file.h> |
46 | ||
6e04e021 | 47 | extern struct inet_hashinfo tcp_hashinfo; |
1da177e4 | 48 | |
1da177e4 | 49 | extern atomic_t tcp_orphan_count; |
1da177e4 | 50 | extern void tcp_time_wait(struct sock *sk, int state, int timeo); |
1da177e4 | 51 | |
1da177e4 LT |
52 | #define MAX_TCP_HEADER (128 + MAX_HEADER) |
53 | ||
54 | /* | |
55 | * Never offer a window over 32767 without using window scaling. Some | |
56 | * poor stacks do signed 16bit maths! | |
57 | */ | |
58 | #define MAX_TCP_WINDOW 32767U | |
59 | ||
60 | /* Minimal accepted MSS. It is (60+60+8) - (20+20). */ | |
61 | #define TCP_MIN_MSS 88U | |
62 | ||
63 | /* Minimal RCV_MSS. */ | |
64 | #define TCP_MIN_RCVMSS 536U | |
65 | ||
5d424d5a JH |
66 | /* The least MTU to use for probing */ |
67 | #define TCP_BASE_MSS 512 | |
68 | ||
1da177e4 LT |
69 | /* After receiving this amount of duplicate ACKs fast retransmit starts. */ |
70 | #define TCP_FASTRETRANS_THRESH 3 | |
71 | ||
72 | /* Maximal reordering. */ | |
73 | #define TCP_MAX_REORDERING 127 | |
74 | ||
75 | /* Maximal number of ACKs sent quickly to accelerate slow-start. */ | |
76 | #define TCP_MAX_QUICKACKS 16U | |
77 | ||
78 | /* urg_data states */ | |
79 | #define TCP_URG_VALID 0x0100 | |
80 | #define TCP_URG_NOTYET 0x0200 | |
81 | #define TCP_URG_READ 0x0400 | |
82 | ||
83 | #define TCP_RETR1 3 /* | |
84 | * This is how many retries it does before it | |
85 | * tries to figure out if the gateway is | |
86 | * down. Minimal RFC value is 3; it corresponds | |
87 | * to ~3sec-8min depending on RTO. | |
88 | */ | |
89 | ||
90 | #define TCP_RETR2 15 /* | |
91 | * This should take at least | |
92 | * 90 minutes to time out. | |
93 | * RFC1122 says that the limit is 100 sec. | |
94 | * 15 is ~13-30min depending on RTO. | |
95 | */ | |
96 | ||
97 | #define TCP_SYN_RETRIES 5 /* number of times to retry active opening a | |
caa20d9a | 98 | * connection: ~180sec is RFC minimum */ |
1da177e4 LT |
99 | |
100 | #define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a | |
caa20d9a | 101 | * connection: ~180sec is RFC minimum */ |
1da177e4 LT |
102 | |
103 | ||
104 | #define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned | |
105 | * socket. 7 is ~50sec-16min. | |
106 | */ | |
107 | ||
108 | ||
109 | #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT | |
110 | * state, about 60 seconds */ | |
111 | #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN | |
112 | /* BSD style FIN_WAIT2 deadlock breaker. | |
113 | * It used to be 3min, new value is 60sec, | |
114 | * to combine FIN-WAIT-2 timeout with | |
115 | * TIME-WAIT timer. | |
116 | */ | |
117 | ||
118 | #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */ | |
119 | #if HZ >= 100 | |
120 | #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */ | |
121 | #define TCP_ATO_MIN ((unsigned)(HZ/25)) | |
122 | #else | |
123 | #define TCP_DELACK_MIN 4U | |
124 | #define TCP_ATO_MIN 4U | |
125 | #endif | |
126 | #define TCP_RTO_MAX ((unsigned)(120*HZ)) | |
127 | #define TCP_RTO_MIN ((unsigned)(HZ/5)) | |
128 | #define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */ | |
129 | ||
130 | #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes | |
131 | * for local resources. | |
132 | */ | |
133 | ||
134 | #define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */ | |
135 | #define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */ | |
136 | #define TCP_KEEPALIVE_INTVL (75*HZ) | |
137 | ||
138 | #define MAX_TCP_KEEPIDLE 32767 | |
139 | #define MAX_TCP_KEEPINTVL 32767 | |
140 | #define MAX_TCP_KEEPCNT 127 | |
141 | #define MAX_TCP_SYNCNT 127 | |
142 | ||
143 | #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */ | |
1da177e4 LT |
144 | |
145 | #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24) | |
146 | #define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated | |
147 | * after this time. It should be equal | |
148 | * (or greater than) TCP_TIMEWAIT_LEN | |
149 | * to provide reliability equal to one | |
150 | * provided by timewait state. | |
151 | */ | |
152 | #define TCP_PAWS_WINDOW 1 /* Replay window for per-host | |
153 | * timestamps. It must be less than | |
154 | * minimal timewait lifetime. | |
155 | */ | |
1da177e4 LT |
156 | /* |
157 | * TCP option | |
158 | */ | |
159 | ||
160 | #define TCPOPT_NOP 1 /* Padding */ | |
161 | #define TCPOPT_EOL 0 /* End of options */ | |
162 | #define TCPOPT_MSS 2 /* Segment size negotiating */ | |
163 | #define TCPOPT_WINDOW 3 /* Window scaling */ | |
164 | #define TCPOPT_SACK_PERM 4 /* SACK Permitted */ | |
165 | #define TCPOPT_SACK 5 /* SACK Block */ | |
166 | #define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */ | |
cfb6eeb4 | 167 | #define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */ |
1da177e4 LT |
168 | |
169 | /* | |
170 | * TCP option lengths | |
171 | */ | |
172 | ||
173 | #define TCPOLEN_MSS 4 | |
174 | #define TCPOLEN_WINDOW 3 | |
175 | #define TCPOLEN_SACK_PERM 2 | |
176 | #define TCPOLEN_TIMESTAMP 10 | |
cfb6eeb4 | 177 | #define TCPOLEN_MD5SIG 18 |
1da177e4 LT |
178 | |
179 | /* But this is what stacks really send out. */ | |
180 | #define TCPOLEN_TSTAMP_ALIGNED 12 | |
181 | #define TCPOLEN_WSCALE_ALIGNED 4 | |
182 | #define TCPOLEN_SACKPERM_ALIGNED 4 | |
183 | #define TCPOLEN_SACK_BASE 2 | |
184 | #define TCPOLEN_SACK_BASE_ALIGNED 4 | |
185 | #define TCPOLEN_SACK_PERBLOCK 8 | |
cfb6eeb4 | 186 | #define TCPOLEN_MD5SIG_ALIGNED 20 |
1da177e4 | 187 | |
1da177e4 LT |
188 | /* Flags in tp->nonagle */ |
189 | #define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */ | |
190 | #define TCP_NAGLE_CORK 2 /* Socket is corked */ | |
caa20d9a | 191 | #define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */ |
1da177e4 | 192 | |
295ff7ed ACM |
193 | extern struct inet_timewait_death_row tcp_death_row; |
194 | ||
1da177e4 | 195 | /* sysctl variables for tcp */ |
1da177e4 LT |
196 | extern int sysctl_tcp_timestamps; |
197 | extern int sysctl_tcp_window_scaling; | |
198 | extern int sysctl_tcp_sack; | |
199 | extern int sysctl_tcp_fin_timeout; | |
1da177e4 LT |
200 | extern int sysctl_tcp_keepalive_time; |
201 | extern int sysctl_tcp_keepalive_probes; | |
202 | extern int sysctl_tcp_keepalive_intvl; | |
203 | extern int sysctl_tcp_syn_retries; | |
204 | extern int sysctl_tcp_synack_retries; | |
205 | extern int sysctl_tcp_retries1; | |
206 | extern int sysctl_tcp_retries2; | |
207 | extern int sysctl_tcp_orphan_retries; | |
208 | extern int sysctl_tcp_syncookies; | |
209 | extern int sysctl_tcp_retrans_collapse; | |
210 | extern int sysctl_tcp_stdurg; | |
211 | extern int sysctl_tcp_rfc1337; | |
212 | extern int sysctl_tcp_abort_on_overflow; | |
213 | extern int sysctl_tcp_max_orphans; | |
1da177e4 LT |
214 | extern int sysctl_tcp_fack; |
215 | extern int sysctl_tcp_reordering; | |
216 | extern int sysctl_tcp_ecn; | |
217 | extern int sysctl_tcp_dsack; | |
218 | extern int sysctl_tcp_mem[3]; | |
219 | extern int sysctl_tcp_wmem[3]; | |
220 | extern int sysctl_tcp_rmem[3]; | |
221 | extern int sysctl_tcp_app_win; | |
222 | extern int sysctl_tcp_adv_win_scale; | |
223 | extern int sysctl_tcp_tw_reuse; | |
224 | extern int sysctl_tcp_frto; | |
3cfe3baa | 225 | extern int sysctl_tcp_frto_response; |
1da177e4 | 226 | extern int sysctl_tcp_low_latency; |
95937825 | 227 | extern int sysctl_tcp_dma_copybreak; |
1da177e4 | 228 | extern int sysctl_tcp_nometrics_save; |
1da177e4 LT |
229 | extern int sysctl_tcp_moderate_rcvbuf; |
230 | extern int sysctl_tcp_tso_win_divisor; | |
9772efb9 | 231 | extern int sysctl_tcp_abc; |
5d424d5a JH |
232 | extern int sysctl_tcp_mtu_probing; |
233 | extern int sysctl_tcp_base_mss; | |
15d99e02 | 234 | extern int sysctl_tcp_workaround_signed_windows; |
35089bb2 | 235 | extern int sysctl_tcp_slow_start_after_idle; |
886236c1 | 236 | extern int sysctl_tcp_max_ssthresh; |
1da177e4 LT |
237 | |
238 | extern atomic_t tcp_memory_allocated; | |
239 | extern atomic_t tcp_sockets_allocated; | |
240 | extern int tcp_memory_pressure; | |
241 | ||
1da177e4 LT |
242 | /* |
243 | * The next routines deal with comparing 32 bit unsigned ints | |
244 | * and worry about wraparound (automatic with unsigned arithmetic). | |
245 | */ | |
246 | ||
247 | static inline int before(__u32 seq1, __u32 seq2) | |
248 | { | |
0d630cc0 | 249 | return (__s32)(seq1-seq2) < 0; |
1da177e4 | 250 | } |
9a036b9c | 251 | #define after(seq2, seq1) before(seq1, seq2) |
1da177e4 LT |
252 | |
253 | /* is s2<=s1<=s3 ? */ | |
254 | static inline int between(__u32 seq1, __u32 seq2, __u32 seq3) | |
255 | { | |
256 | return seq3 - seq2 >= seq1 - seq2; | |
257 | } | |
258 | ||
e4fd5da3 PE |
259 | static inline int tcp_too_many_orphans(struct sock *sk, int num) |
260 | { | |
261 | return (num > sysctl_tcp_max_orphans) || | |
262 | (sk->sk_wmem_queued > SOCK_MIN_SNDBUF && | |
263 | atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]); | |
264 | } | |
1da177e4 LT |
265 | |
266 | extern struct proto tcp_prot; | |
267 | ||
268 | DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics); | |
269 | #define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field) | |
270 | #define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field) | |
1da177e4 | 271 | #define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field) |
1da177e4 LT |
272 | #define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val) |
273 | ||
1da177e4 LT |
274 | extern void tcp_v4_err(struct sk_buff *skb, u32); |
275 | ||
276 | extern void tcp_shutdown (struct sock *sk, int how); | |
277 | ||
278 | extern int tcp_v4_rcv(struct sk_buff *skb); | |
279 | ||
280 | extern int tcp_v4_remember_stamp(struct sock *sk); | |
281 | ||
8feaf0c0 | 282 | extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw); |
1da177e4 | 283 | |
3516ffb0 | 284 | extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
285 | struct msghdr *msg, size_t size); |
286 | extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags); | |
287 | ||
288 | extern int tcp_ioctl(struct sock *sk, | |
289 | int cmd, | |
290 | unsigned long arg); | |
291 | ||
292 | extern int tcp_rcv_state_process(struct sock *sk, | |
293 | struct sk_buff *skb, | |
294 | struct tcphdr *th, | |
295 | unsigned len); | |
296 | ||
297 | extern int tcp_rcv_established(struct sock *sk, | |
298 | struct sk_buff *skb, | |
299 | struct tcphdr *th, | |
300 | unsigned len); | |
301 | ||
302 | extern void tcp_rcv_space_adjust(struct sock *sk); | |
303 | ||
0e4b4992 CL |
304 | extern void tcp_cleanup_rbuf(struct sock *sk, int copied); |
305 | ||
6d6ee43e ACM |
306 | extern int tcp_twsk_unique(struct sock *sk, |
307 | struct sock *sktw, void *twp); | |
308 | ||
cfb6eeb4 YH |
309 | extern void tcp_twsk_destructor(struct sock *sk); |
310 | ||
9c55e01c JA |
311 | extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos, |
312 | struct pipe_inode_info *pipe, size_t len, unsigned int flags); | |
313 | ||
463c84b9 ACM |
314 | static inline void tcp_dec_quickack_mode(struct sock *sk, |
315 | const unsigned int pkts) | |
1da177e4 | 316 | { |
463c84b9 | 317 | struct inet_connection_sock *icsk = inet_csk(sk); |
fc6415bc | 318 | |
463c84b9 ACM |
319 | if (icsk->icsk_ack.quick) { |
320 | if (pkts >= icsk->icsk_ack.quick) { | |
321 | icsk->icsk_ack.quick = 0; | |
fc6415bc | 322 | /* Leaving quickack mode we deflate ATO. */ |
463c84b9 | 323 | icsk->icsk_ack.ato = TCP_ATO_MIN; |
fc6415bc | 324 | } else |
463c84b9 | 325 | icsk->icsk_ack.quick -= pkts; |
1da177e4 LT |
326 | } |
327 | } | |
328 | ||
463c84b9 | 329 | extern void tcp_enter_quickack_mode(struct sock *sk); |
1da177e4 | 330 | |
1da177e4 LT |
331 | static inline void tcp_clear_options(struct tcp_options_received *rx_opt) |
332 | { | |
333 | rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0; | |
334 | } | |
335 | ||
bdf1ee5d IJ |
336 | #define TCP_ECN_OK 1 |
337 | #define TCP_ECN_QUEUE_CWR 2 | |
338 | #define TCP_ECN_DEMAND_CWR 4 | |
339 | ||
340 | static __inline__ void | |
341 | TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th) | |
342 | { | |
343 | if (sysctl_tcp_ecn && th->ece && th->cwr) | |
344 | inet_rsk(req)->ecn_ok = 1; | |
345 | } | |
346 | ||
1da177e4 LT |
347 | enum tcp_tw_status |
348 | { | |
349 | TCP_TW_SUCCESS = 0, | |
350 | TCP_TW_RST = 1, | |
351 | TCP_TW_ACK = 2, | |
352 | TCP_TW_SYN = 3 | |
353 | }; | |
354 | ||
355 | ||
8feaf0c0 | 356 | extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw, |
1da177e4 | 357 | struct sk_buff *skb, |
8feaf0c0 | 358 | const struct tcphdr *th); |
1da177e4 LT |
359 | |
360 | extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb, | |
60236fdd ACM |
361 | struct request_sock *req, |
362 | struct request_sock **prev); | |
1da177e4 LT |
363 | extern int tcp_child_process(struct sock *parent, |
364 | struct sock *child, | |
365 | struct sk_buff *skb); | |
46d0de4e | 366 | extern int tcp_use_frto(struct sock *sk); |
1da177e4 LT |
367 | extern void tcp_enter_frto(struct sock *sk); |
368 | extern void tcp_enter_loss(struct sock *sk, int how); | |
369 | extern void tcp_clear_retrans(struct tcp_sock *tp); | |
370 | extern void tcp_update_metrics(struct sock *sk); | |
371 | ||
372 | extern void tcp_close(struct sock *sk, | |
373 | long timeout); | |
1da177e4 LT |
374 | extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait); |
375 | ||
376 | extern int tcp_getsockopt(struct sock *sk, int level, | |
377 | int optname, | |
378 | char __user *optval, | |
379 | int __user *optlen); | |
380 | extern int tcp_setsockopt(struct sock *sk, int level, | |
381 | int optname, char __user *optval, | |
382 | int optlen); | |
3fdadf7d DM |
383 | extern int compat_tcp_getsockopt(struct sock *sk, |
384 | int level, int optname, | |
385 | char __user *optval, int __user *optlen); | |
386 | extern int compat_tcp_setsockopt(struct sock *sk, | |
387 | int level, int optname, | |
388 | char __user *optval, int optlen); | |
1da177e4 LT |
389 | extern void tcp_set_keepalive(struct sock *sk, int val); |
390 | extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, | |
391 | struct msghdr *msg, | |
392 | size_t len, int nonblock, | |
393 | int flags, int *addr_len); | |
394 | ||
1da177e4 LT |
395 | extern void tcp_parse_options(struct sk_buff *skb, |
396 | struct tcp_options_received *opt_rx, | |
397 | int estab); | |
398 | ||
7d5d5525 YH |
399 | extern u8 *tcp_parse_md5sig_option(struct tcphdr *th); |
400 | ||
1da177e4 LT |
401 | /* |
402 | * TCP v4 functions exported for the inet6 API | |
403 | */ | |
404 | ||
8292a17a | 405 | extern void tcp_v4_send_check(struct sock *sk, int len, |
1da177e4 LT |
406 | struct sk_buff *skb); |
407 | ||
408 | extern int tcp_v4_conn_request(struct sock *sk, | |
409 | struct sk_buff *skb); | |
410 | ||
411 | extern struct sock * tcp_create_openreq_child(struct sock *sk, | |
60236fdd | 412 | struct request_sock *req, |
1da177e4 LT |
413 | struct sk_buff *skb); |
414 | ||
415 | extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk, | |
416 | struct sk_buff *skb, | |
60236fdd | 417 | struct request_sock *req, |
1da177e4 LT |
418 | struct dst_entry *dst); |
419 | ||
420 | extern int tcp_v4_do_rcv(struct sock *sk, | |
421 | struct sk_buff *skb); | |
422 | ||
423 | extern int tcp_v4_connect(struct sock *sk, | |
424 | struct sockaddr *uaddr, | |
425 | int addr_len); | |
426 | ||
427 | extern int tcp_connect(struct sock *sk); | |
428 | ||
429 | extern struct sk_buff * tcp_make_synack(struct sock *sk, | |
430 | struct dst_entry *dst, | |
60236fdd | 431 | struct request_sock *req); |
1da177e4 LT |
432 | |
433 | extern int tcp_disconnect(struct sock *sk, int flags); | |
434 | ||
1da177e4 | 435 | |
1da177e4 | 436 | /* From syncookies.c */ |
2051f11f | 437 | extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS]; |
1da177e4 LT |
438 | extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb, |
439 | struct ip_options *opt); | |
440 | extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb, | |
441 | __u16 *mss); | |
442 | ||
4dfc2817 FW |
443 | extern __u32 cookie_init_timestamp(struct request_sock *req); |
444 | extern void cookie_check_timestamp(struct tcp_options_received *tcp_opt); | |
445 | ||
c6aefafb GG |
446 | /* From net/ipv6/syncookies.c */ |
447 | extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb); | |
448 | extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb, | |
449 | __u16 *mss); | |
450 | ||
1da177e4 LT |
451 | /* tcp_output.c */ |
452 | ||
9e412ba7 IJ |
453 | extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss, |
454 | int nonagle); | |
455 | extern int tcp_may_send_now(struct sock *sk); | |
1da177e4 LT |
456 | extern int tcp_retransmit_skb(struct sock *, struct sk_buff *); |
457 | extern void tcp_xmit_retransmit_queue(struct sock *); | |
458 | extern void tcp_simple_retransmit(struct sock *); | |
459 | extern int tcp_trim_head(struct sock *, struct sk_buff *, u32); | |
6475be16 | 460 | extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int); |
1da177e4 LT |
461 | |
462 | extern void tcp_send_probe0(struct sock *); | |
463 | extern void tcp_send_partial(struct sock *); | |
464 | extern int tcp_write_wakeup(struct sock *); | |
465 | extern void tcp_send_fin(struct sock *sk); | |
dd0fc66f | 466 | extern void tcp_send_active_reset(struct sock *sk, gfp_t priority); |
1da177e4 | 467 | extern int tcp_send_synack(struct sock *); |
c1b4a7e6 | 468 | extern void tcp_push_one(struct sock *, unsigned int mss_now); |
1da177e4 LT |
469 | extern void tcp_send_ack(struct sock *sk); |
470 | extern void tcp_send_delayed_ack(struct sock *sk); | |
471 | ||
a762a980 DM |
472 | /* tcp_input.c */ |
473 | extern void tcp_cwnd_application_limited(struct sock *sk); | |
474 | ||
1da177e4 LT |
475 | /* tcp_timer.c */ |
476 | extern void tcp_init_xmit_timers(struct sock *); | |
463c84b9 ACM |
477 | static inline void tcp_clear_xmit_timers(struct sock *sk) |
478 | { | |
479 | inet_csk_clear_xmit_timers(sk); | |
480 | } | |
1da177e4 | 481 | |
1da177e4 LT |
482 | extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu); |
483 | extern unsigned int tcp_current_mss(struct sock *sk, int large); | |
484 | ||
17b085ea | 485 | /* tcp.c */ |
1da177e4 LT |
486 | extern void tcp_get_info(struct sock *, struct tcp_info *); |
487 | ||
488 | /* Read 'sendfile()'-style from a TCP socket */ | |
489 | typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *, | |
490 | unsigned int, size_t); | |
491 | extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc, | |
492 | sk_read_actor_t recv_actor); | |
493 | ||
40efc6fa | 494 | extern void tcp_initialize_rcv_mss(struct sock *sk); |
1da177e4 | 495 | |
5d424d5a JH |
496 | extern int tcp_mtu_to_mss(struct sock *sk, int pmtu); |
497 | extern int tcp_mss_to_mtu(struct sock *sk, int mss); | |
498 | extern void tcp_mtup_init(struct sock *sk); | |
499 | ||
40efc6fa | 500 | static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd) |
1da177e4 LT |
501 | { |
502 | tp->pred_flags = htonl((tp->tcp_header_len << 26) | | |
503 | ntohl(TCP_FLAG_ACK) | | |
504 | snd_wnd); | |
505 | } | |
506 | ||
40efc6fa | 507 | static inline void tcp_fast_path_on(struct tcp_sock *tp) |
1da177e4 LT |
508 | { |
509 | __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale); | |
510 | } | |
511 | ||
9e412ba7 | 512 | static inline void tcp_fast_path_check(struct sock *sk) |
1da177e4 | 513 | { |
9e412ba7 IJ |
514 | struct tcp_sock *tp = tcp_sk(sk); |
515 | ||
b03efcfb | 516 | if (skb_queue_empty(&tp->out_of_order_queue) && |
1da177e4 LT |
517 | tp->rcv_wnd && |
518 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf && | |
519 | !tp->urg_data) | |
520 | tcp_fast_path_on(tp); | |
521 | } | |
522 | ||
523 | /* Compute the actual receive window we are currently advertising. | |
524 | * Rcv_nxt can be after the window if our peer push more data | |
525 | * than the offered window. | |
526 | */ | |
40efc6fa | 527 | static inline u32 tcp_receive_window(const struct tcp_sock *tp) |
1da177e4 LT |
528 | { |
529 | s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt; | |
530 | ||
531 | if (win < 0) | |
532 | win = 0; | |
533 | return (u32) win; | |
534 | } | |
535 | ||
536 | /* Choose a new window, without checks for shrinking, and without | |
537 | * scaling applied to the result. The caller does these things | |
538 | * if necessary. This is a "raw" window selection. | |
539 | */ | |
540 | extern u32 __tcp_select_window(struct sock *sk); | |
541 | ||
542 | /* TCP timestamps are only 32-bits, this causes a slight | |
543 | * complication on 64-bit systems since we store a snapshot | |
31f34269 SH |
544 | * of jiffies in the buffer control blocks below. We decided |
545 | * to use only the low 32-bits of jiffies and hide the ugly | |
1da177e4 LT |
546 | * casts with the following macro. |
547 | */ | |
548 | #define tcp_time_stamp ((__u32)(jiffies)) | |
549 | ||
caa20d9a | 550 | /* This is what the send packet queuing engine uses to pass |
1da177e4 LT |
551 | * TCP per-packet control information to the transmission |
552 | * code. We also store the host-order sequence numbers in | |
553 | * here too. This is 36 bytes on 32-bit architectures, | |
554 | * 40 bytes on 64-bit machines, if this grows please adjust | |
555 | * skbuff.h:skbuff->cb[xxx] size appropriately. | |
556 | */ | |
557 | struct tcp_skb_cb { | |
558 | union { | |
559 | struct inet_skb_parm h4; | |
560 | #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE) | |
561 | struct inet6_skb_parm h6; | |
562 | #endif | |
563 | } header; /* For incoming frames */ | |
564 | __u32 seq; /* Starting sequence number */ | |
565 | __u32 end_seq; /* SEQ + FIN + SYN + datalen */ | |
566 | __u32 when; /* used to compute rtt's */ | |
567 | __u8 flags; /* TCP header flags. */ | |
568 | ||
569 | /* NOTE: These must match up to the flags byte in a | |
570 | * real TCP header. | |
571 | */ | |
572 | #define TCPCB_FLAG_FIN 0x01 | |
573 | #define TCPCB_FLAG_SYN 0x02 | |
574 | #define TCPCB_FLAG_RST 0x04 | |
575 | #define TCPCB_FLAG_PSH 0x08 | |
576 | #define TCPCB_FLAG_ACK 0x10 | |
577 | #define TCPCB_FLAG_URG 0x20 | |
578 | #define TCPCB_FLAG_ECE 0x40 | |
579 | #define TCPCB_FLAG_CWR 0x80 | |
580 | ||
581 | __u8 sacked; /* State flags for SACK/FACK. */ | |
582 | #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */ | |
583 | #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */ | |
584 | #define TCPCB_LOST 0x04 /* SKB is lost */ | |
585 | #define TCPCB_TAGBITS 0x07 /* All tag bits */ | |
586 | ||
587 | #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */ | |
588 | #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS) | |
589 | ||
1da177e4 LT |
590 | __u16 urg_ptr; /* Valid w/URG flags is set. */ |
591 | __u32 ack_seq; /* Sequence number ACK'd */ | |
592 | }; | |
593 | ||
594 | #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0])) | |
595 | ||
1da177e4 LT |
596 | /* Due to TSO, an SKB can be composed of multiple actual |
597 | * packets. To keep these tracked properly, we use this. | |
598 | */ | |
599 | static inline int tcp_skb_pcount(const struct sk_buff *skb) | |
600 | { | |
7967168c | 601 | return skb_shinfo(skb)->gso_segs; |
1da177e4 LT |
602 | } |
603 | ||
604 | /* This is valid iff tcp_skb_pcount() > 1. */ | |
605 | static inline int tcp_skb_mss(const struct sk_buff *skb) | |
606 | { | |
7967168c | 607 | return skb_shinfo(skb)->gso_size; |
1da177e4 LT |
608 | } |
609 | ||
af610b4c | 610 | static inline void tcp_dec_pcount_approx_int(__u32 *count, const int decr) |
1da177e4 LT |
611 | { |
612 | if (*count) { | |
af610b4c | 613 | *count -= decr; |
1da177e4 LT |
614 | if ((int)*count < 0) |
615 | *count = 0; | |
616 | } | |
617 | } | |
618 | ||
af610b4c IJ |
619 | static inline void tcp_dec_pcount_approx(__u32 *count, |
620 | const struct sk_buff *skb) | |
621 | { | |
622 | tcp_dec_pcount_approx_int(count, tcp_skb_pcount(skb)); | |
623 | } | |
624 | ||
317a76f9 SH |
625 | /* Events passed to congestion control interface */ |
626 | enum tcp_ca_event { | |
627 | CA_EVENT_TX_START, /* first transmit when no packets in flight */ | |
628 | CA_EVENT_CWND_RESTART, /* congestion window restart */ | |
629 | CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */ | |
630 | CA_EVENT_FRTO, /* fast recovery timeout */ | |
631 | CA_EVENT_LOSS, /* loss timeout */ | |
632 | CA_EVENT_FAST_ACK, /* in sequence ack */ | |
633 | CA_EVENT_SLOW_ACK, /* other ack */ | |
634 | }; | |
635 | ||
636 | /* | |
637 | * Interface for adding new TCP congestion control handlers | |
638 | */ | |
639 | #define TCP_CA_NAME_MAX 16 | |
3ff825b2 SH |
640 | #define TCP_CA_MAX 128 |
641 | #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX) | |
642 | ||
164891aa SH |
643 | #define TCP_CONG_NON_RESTRICTED 0x1 |
644 | #define TCP_CONG_RTT_STAMP 0x2 | |
645 | ||
317a76f9 SH |
646 | struct tcp_congestion_ops { |
647 | struct list_head list; | |
164891aa | 648 | unsigned long flags; |
317a76f9 SH |
649 | |
650 | /* initialize private data (optional) */ | |
6687e988 | 651 | void (*init)(struct sock *sk); |
317a76f9 | 652 | /* cleanup private data (optional) */ |
6687e988 | 653 | void (*release)(struct sock *sk); |
317a76f9 SH |
654 | |
655 | /* return slow start threshold (required) */ | |
6687e988 | 656 | u32 (*ssthresh)(struct sock *sk); |
317a76f9 | 657 | /* lower bound for congestion window (optional) */ |
72dc5b92 | 658 | u32 (*min_cwnd)(const struct sock *sk); |
317a76f9 | 659 | /* do new cwnd calculation (required) */ |
c3a05c60 | 660 | void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight); |
317a76f9 | 661 | /* call before changing ca_state (optional) */ |
6687e988 | 662 | void (*set_state)(struct sock *sk, u8 new_state); |
317a76f9 | 663 | /* call when cwnd event occurs (optional) */ |
6687e988 | 664 | void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev); |
317a76f9 | 665 | /* new value of cwnd after loss (optional) */ |
6687e988 | 666 | u32 (*undo_cwnd)(struct sock *sk); |
317a76f9 | 667 | /* hook for packet ack accounting (optional) */ |
30cfd0ba | 668 | void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us); |
73c1f4a0 | 669 | /* get info for inet_diag (optional) */ |
6687e988 | 670 | void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb); |
317a76f9 SH |
671 | |
672 | char name[TCP_CA_NAME_MAX]; | |
673 | struct module *owner; | |
674 | }; | |
675 | ||
676 | extern int tcp_register_congestion_control(struct tcp_congestion_ops *type); | |
677 | extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type); | |
678 | ||
6687e988 ACM |
679 | extern void tcp_init_congestion_control(struct sock *sk); |
680 | extern void tcp_cleanup_congestion_control(struct sock *sk); | |
317a76f9 SH |
681 | extern int tcp_set_default_congestion_control(const char *name); |
682 | extern void tcp_get_default_congestion_control(char *name); | |
3ff825b2 | 683 | extern void tcp_get_available_congestion_control(char *buf, size_t len); |
ce7bc3bf SH |
684 | extern void tcp_get_allowed_congestion_control(char *buf, size_t len); |
685 | extern int tcp_set_allowed_congestion_control(char *allowed); | |
6687e988 | 686 | extern int tcp_set_congestion_control(struct sock *sk, const char *name); |
40efc6fa | 687 | extern void tcp_slow_start(struct tcp_sock *tp); |
317a76f9 | 688 | |
5f8ef48d | 689 | extern struct tcp_congestion_ops tcp_init_congestion_ops; |
6687e988 | 690 | extern u32 tcp_reno_ssthresh(struct sock *sk); |
c3a05c60 | 691 | extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight); |
72dc5b92 | 692 | extern u32 tcp_reno_min_cwnd(const struct sock *sk); |
a8acfbac | 693 | extern struct tcp_congestion_ops tcp_reno; |
317a76f9 | 694 | |
6687e988 | 695 | static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state) |
317a76f9 | 696 | { |
6687e988 ACM |
697 | struct inet_connection_sock *icsk = inet_csk(sk); |
698 | ||
699 | if (icsk->icsk_ca_ops->set_state) | |
700 | icsk->icsk_ca_ops->set_state(sk, ca_state); | |
701 | icsk->icsk_ca_state = ca_state; | |
317a76f9 SH |
702 | } |
703 | ||
6687e988 | 704 | static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event) |
317a76f9 | 705 | { |
6687e988 ACM |
706 | const struct inet_connection_sock *icsk = inet_csk(sk); |
707 | ||
708 | if (icsk->icsk_ca_ops->cwnd_event) | |
709 | icsk->icsk_ca_ops->cwnd_event(sk, event); | |
317a76f9 SH |
710 | } |
711 | ||
e60402d0 IJ |
712 | /* These functions determine how the current flow behaves in respect of SACK |
713 | * handling. SACK is negotiated with the peer, and therefore it can vary | |
714 | * between different flows. | |
715 | * | |
716 | * tcp_is_sack - SACK enabled | |
717 | * tcp_is_reno - No SACK | |
718 | * tcp_is_fack - FACK enabled, implies SACK enabled | |
719 | */ | |
720 | static inline int tcp_is_sack(const struct tcp_sock *tp) | |
721 | { | |
722 | return tp->rx_opt.sack_ok; | |
723 | } | |
724 | ||
725 | static inline int tcp_is_reno(const struct tcp_sock *tp) | |
726 | { | |
727 | return !tcp_is_sack(tp); | |
728 | } | |
729 | ||
730 | static inline int tcp_is_fack(const struct tcp_sock *tp) | |
731 | { | |
732 | return tp->rx_opt.sack_ok & 2; | |
733 | } | |
734 | ||
735 | static inline void tcp_enable_fack(struct tcp_sock *tp) | |
736 | { | |
737 | tp->rx_opt.sack_ok |= 2; | |
738 | } | |
739 | ||
83ae4088 IJ |
740 | static inline unsigned int tcp_left_out(const struct tcp_sock *tp) |
741 | { | |
742 | return tp->sacked_out + tp->lost_out; | |
743 | } | |
744 | ||
1da177e4 LT |
745 | /* This determines how many packets are "in the network" to the best |
746 | * of our knowledge. In many cases it is conservative, but where | |
747 | * detailed information is available from the receiver (via SACK | |
748 | * blocks etc.) we can make more aggressive calculations. | |
749 | * | |
750 | * Use this for decisions involving congestion control, use just | |
751 | * tp->packets_out to determine if the send queue is empty or not. | |
752 | * | |
753 | * Read this equation as: | |
754 | * | |
755 | * "Packets sent once on transmission queue" MINUS | |
756 | * "Packets left network, but not honestly ACKed yet" PLUS | |
757 | * "Packets fast retransmitted" | |
758 | */ | |
40efc6fa | 759 | static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp) |
1da177e4 | 760 | { |
83ae4088 | 761 | return tp->packets_out - tcp_left_out(tp) + tp->retrans_out; |
1da177e4 LT |
762 | } |
763 | ||
882bebaa IJ |
764 | extern int tcp_limit_reno_sacked(struct tcp_sock *tp); |
765 | ||
1da177e4 LT |
766 | /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd. |
767 | * The exception is rate halving phase, when cwnd is decreasing towards | |
768 | * ssthresh. | |
769 | */ | |
6687e988 | 770 | static inline __u32 tcp_current_ssthresh(const struct sock *sk) |
1da177e4 | 771 | { |
6687e988 ACM |
772 | const struct tcp_sock *tp = tcp_sk(sk); |
773 | if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery)) | |
1da177e4 LT |
774 | return tp->snd_ssthresh; |
775 | else | |
776 | return max(tp->snd_ssthresh, | |
777 | ((tp->snd_cwnd >> 1) + | |
778 | (tp->snd_cwnd >> 2))); | |
779 | } | |
780 | ||
b9c4595b IJ |
781 | /* Use define here intentionally to get WARN_ON location shown at the caller */ |
782 | #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out) | |
1da177e4 | 783 | |
3cfe3baa | 784 | extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh); |
1da177e4 LT |
785 | extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst); |
786 | ||
787 | /* Slow start with delack produces 3 packets of burst, so that | |
dd9e0dda JH |
788 | * it is safe "de facto". This will be the default - same as |
789 | * the default reordering threshold - but if reordering increases, | |
790 | * we must be able to allow cwnd to burst at least this much in order | |
791 | * to not pull it back when holes are filled. | |
1da177e4 LT |
792 | */ |
793 | static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp) | |
794 | { | |
dd9e0dda | 795 | return tp->reordering; |
1da177e4 LT |
796 | } |
797 | ||
90840def IJ |
798 | /* Returns end sequence number of the receiver's advertised window */ |
799 | static inline u32 tcp_wnd_end(const struct tcp_sock *tp) | |
800 | { | |
801 | return tp->snd_una + tp->snd_wnd; | |
802 | } | |
cea14e0e | 803 | extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight); |
f4805ede | 804 | |
c1bd24b7 | 805 | static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss, |
40efc6fa | 806 | const struct sk_buff *skb) |
1da177e4 LT |
807 | { |
808 | if (skb->len < mss) | |
809 | tp->snd_sml = TCP_SKB_CB(skb)->end_seq; | |
810 | } | |
811 | ||
9e412ba7 | 812 | static inline void tcp_check_probe_timer(struct sock *sk) |
1da177e4 | 813 | { |
9e412ba7 | 814 | struct tcp_sock *tp = tcp_sk(sk); |
463c84b9 | 815 | const struct inet_connection_sock *icsk = inet_csk(sk); |
9e412ba7 | 816 | |
463c84b9 | 817 | if (!tp->packets_out && !icsk->icsk_pending) |
3f421baa ACM |
818 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, |
819 | icsk->icsk_rto, TCP_RTO_MAX); | |
1da177e4 LT |
820 | } |
821 | ||
9e412ba7 | 822 | static inline void tcp_push_pending_frames(struct sock *sk) |
1da177e4 | 823 | { |
9e412ba7 IJ |
824 | struct tcp_sock *tp = tcp_sk(sk); |
825 | ||
826 | __tcp_push_pending_frames(sk, tcp_current_mss(sk, 1), tp->nonagle); | |
1da177e4 LT |
827 | } |
828 | ||
40efc6fa | 829 | static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq) |
1da177e4 LT |
830 | { |
831 | tp->snd_wl1 = seq; | |
832 | } | |
833 | ||
40efc6fa | 834 | static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq) |
1da177e4 LT |
835 | { |
836 | tp->snd_wl1 = seq; | |
837 | } | |
838 | ||
1da177e4 LT |
839 | /* |
840 | * Calculate(/check) TCP checksum | |
841 | */ | |
ba7808ea FD |
842 | static inline __sum16 tcp_v4_check(int len, __be32 saddr, |
843 | __be32 daddr, __wsum base) | |
1da177e4 LT |
844 | { |
845 | return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base); | |
846 | } | |
847 | ||
b51655b9 | 848 | static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 849 | { |
fb286bb2 | 850 | return __skb_checksum_complete(skb); |
1da177e4 LT |
851 | } |
852 | ||
40efc6fa | 853 | static inline int tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 854 | { |
60476372 | 855 | return !skb_csum_unnecessary(skb) && |
1da177e4 LT |
856 | __tcp_checksum_complete(skb); |
857 | } | |
858 | ||
859 | /* Prequeue for VJ style copy to user, combined with checksumming. */ | |
860 | ||
40efc6fa | 861 | static inline void tcp_prequeue_init(struct tcp_sock *tp) |
1da177e4 LT |
862 | { |
863 | tp->ucopy.task = NULL; | |
864 | tp->ucopy.len = 0; | |
865 | tp->ucopy.memory = 0; | |
866 | skb_queue_head_init(&tp->ucopy.prequeue); | |
97fc2f08 CL |
867 | #ifdef CONFIG_NET_DMA |
868 | tp->ucopy.dma_chan = NULL; | |
869 | tp->ucopy.wakeup = 0; | |
870 | tp->ucopy.pinned_list = NULL; | |
871 | tp->ucopy.dma_cookie = 0; | |
872 | #endif | |
1da177e4 LT |
873 | } |
874 | ||
875 | /* Packet is added to VJ-style prequeue for processing in process | |
876 | * context, if a reader task is waiting. Apparently, this exciting | |
877 | * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93) | |
878 | * failed somewhere. Latency? Burstiness? Well, at least now we will | |
879 | * see, why it failed. 8)8) --ANK | |
880 | * | |
881 | * NOTE: is this not too big to inline? | |
882 | */ | |
40efc6fa | 883 | static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
884 | { |
885 | struct tcp_sock *tp = tcp_sk(sk); | |
886 | ||
887 | if (!sysctl_tcp_low_latency && tp->ucopy.task) { | |
888 | __skb_queue_tail(&tp->ucopy.prequeue, skb); | |
889 | tp->ucopy.memory += skb->truesize; | |
890 | if (tp->ucopy.memory > sk->sk_rcvbuf) { | |
891 | struct sk_buff *skb1; | |
892 | ||
893 | BUG_ON(sock_owned_by_user(sk)); | |
894 | ||
895 | while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) { | |
896 | sk->sk_backlog_rcv(sk, skb1); | |
897 | NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED); | |
898 | } | |
899 | ||
900 | tp->ucopy.memory = 0; | |
901 | } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) { | |
902 | wake_up_interruptible(sk->sk_sleep); | |
463c84b9 ACM |
903 | if (!inet_csk_ack_scheduled(sk)) |
904 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, | |
3f421baa ACM |
905 | (3 * TCP_RTO_MIN) / 4, |
906 | TCP_RTO_MAX); | |
1da177e4 LT |
907 | } |
908 | return 1; | |
909 | } | |
910 | return 0; | |
911 | } | |
912 | ||
913 | ||
914 | #undef STATE_TRACE | |
915 | ||
916 | #ifdef STATE_TRACE | |
917 | static const char *statename[]={ | |
918 | "Unused","Established","Syn Sent","Syn Recv", | |
919 | "Fin Wait 1","Fin Wait 2","Time Wait", "Close", | |
920 | "Close Wait","Last ACK","Listen","Closing" | |
921 | }; | |
922 | #endif | |
490d5046 | 923 | extern void tcp_set_state(struct sock *sk, int state); |
1da177e4 | 924 | |
4ac02bab | 925 | extern void tcp_done(struct sock *sk); |
1da177e4 | 926 | |
40efc6fa | 927 | static inline void tcp_sack_reset(struct tcp_options_received *rx_opt) |
1da177e4 LT |
928 | { |
929 | rx_opt->dsack = 0; | |
930 | rx_opt->eff_sacks = 0; | |
931 | rx_opt->num_sacks = 0; | |
932 | } | |
933 | ||
1da177e4 LT |
934 | /* Determine a window scaling and initial window to offer. */ |
935 | extern void tcp_select_initial_window(int __space, __u32 mss, | |
936 | __u32 *rcv_wnd, __u32 *window_clamp, | |
937 | int wscale_ok, __u8 *rcv_wscale); | |
938 | ||
939 | static inline int tcp_win_from_space(int space) | |
940 | { | |
941 | return sysctl_tcp_adv_win_scale<=0 ? | |
942 | (space>>(-sysctl_tcp_adv_win_scale)) : | |
943 | space - (space>>sysctl_tcp_adv_win_scale); | |
944 | } | |
945 | ||
946 | /* Note: caller must be prepared to deal with negative returns */ | |
947 | static inline int tcp_space(const struct sock *sk) | |
948 | { | |
949 | return tcp_win_from_space(sk->sk_rcvbuf - | |
950 | atomic_read(&sk->sk_rmem_alloc)); | |
951 | } | |
952 | ||
953 | static inline int tcp_full_space(const struct sock *sk) | |
954 | { | |
955 | return tcp_win_from_space(sk->sk_rcvbuf); | |
956 | } | |
957 | ||
40efc6fa SH |
958 | static inline void tcp_openreq_init(struct request_sock *req, |
959 | struct tcp_options_received *rx_opt, | |
960 | struct sk_buff *skb) | |
1da177e4 | 961 | { |
2e6599cb ACM |
962 | struct inet_request_sock *ireq = inet_rsk(req); |
963 | ||
1da177e4 | 964 | req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */ |
4dfc2817 | 965 | req->cookie_ts = 0; |
2e6599cb | 966 | tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq; |
1da177e4 LT |
967 | req->mss = rx_opt->mss_clamp; |
968 | req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0; | |
2e6599cb ACM |
969 | ireq->tstamp_ok = rx_opt->tstamp_ok; |
970 | ireq->sack_ok = rx_opt->sack_ok; | |
971 | ireq->snd_wscale = rx_opt->snd_wscale; | |
972 | ireq->wscale_ok = rx_opt->wscale_ok; | |
973 | ireq->acked = 0; | |
974 | ireq->ecn_ok = 0; | |
aa8223c7 | 975 | ireq->rmt_port = tcp_hdr(skb)->source; |
1da177e4 LT |
976 | } |
977 | ||
978 | extern void tcp_enter_memory_pressure(void); | |
979 | ||
1da177e4 LT |
980 | static inline int keepalive_intvl_when(const struct tcp_sock *tp) |
981 | { | |
982 | return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl; | |
983 | } | |
984 | ||
985 | static inline int keepalive_time_when(const struct tcp_sock *tp) | |
986 | { | |
987 | return tp->keepalive_time ? : sysctl_tcp_keepalive_time; | |
988 | } | |
989 | ||
463c84b9 | 990 | static inline int tcp_fin_time(const struct sock *sk) |
1da177e4 | 991 | { |
463c84b9 ACM |
992 | int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout; |
993 | const int rto = inet_csk(sk)->icsk_rto; | |
1da177e4 | 994 | |
463c84b9 ACM |
995 | if (fin_timeout < (rto << 2) - (rto >> 1)) |
996 | fin_timeout = (rto << 2) - (rto >> 1); | |
1da177e4 LT |
997 | |
998 | return fin_timeout; | |
999 | } | |
1000 | ||
1001 | static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst) | |
1002 | { | |
1003 | if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0) | |
1004 | return 0; | |
9d729f72 | 1005 | if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS) |
1da177e4 LT |
1006 | return 0; |
1007 | ||
1008 | /* RST segments are not recommended to carry timestamp, | |
1009 | and, if they do, it is recommended to ignore PAWS because | |
1010 | "their cleanup function should take precedence over timestamps." | |
1011 | Certainly, it is mistake. It is necessary to understand the reasons | |
1012 | of this constraint to relax it: if peer reboots, clock may go | |
1013 | out-of-sync and half-open connections will not be reset. | |
1014 | Actually, the problem would be not existing if all | |
1015 | the implementations followed draft about maintaining clock | |
1016 | via reboots. Linux-2.2 DOES NOT! | |
1017 | ||
1018 | However, we can relax time bounds for RST segments to MSL. | |
1019 | */ | |
9d729f72 | 1020 | if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL) |
1da177e4 LT |
1021 | return 0; |
1022 | return 1; | |
1023 | } | |
1024 | ||
1da177e4 LT |
1025 | #define TCP_CHECK_TIMER(sk) do { } while (0) |
1026 | ||
1da177e4 LT |
1027 | static inline void tcp_mib_init(void) |
1028 | { | |
1029 | /* See RFC 2012 */ | |
1030 | TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1); | |
1031 | TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ); | |
1032 | TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ); | |
1033 | TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1); | |
1034 | } | |
1035 | ||
5af4ec23 | 1036 | /* from STCP */ |
0800f170 DM |
1037 | static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp) |
1038 | { | |
6a438bbe SH |
1039 | tp->lost_skb_hint = NULL; |
1040 | tp->scoreboard_skb_hint = NULL; | |
1041 | tp->retransmit_skb_hint = NULL; | |
1042 | tp->forward_skb_hint = NULL; | |
b7689205 IJ |
1043 | } |
1044 | ||
0800f170 DM |
1045 | static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp) |
1046 | { | |
b7689205 | 1047 | tcp_clear_retrans_hints_partial(tp); |
6a438bbe SH |
1048 | } |
1049 | ||
cfb6eeb4 YH |
1050 | /* MD5 Signature */ |
1051 | struct crypto_hash; | |
1052 | ||
1053 | /* - key database */ | |
1054 | struct tcp_md5sig_key { | |
1055 | u8 *key; | |
1056 | u8 keylen; | |
1057 | }; | |
1058 | ||
1059 | struct tcp4_md5sig_key { | |
f8ab18d2 | 1060 | struct tcp_md5sig_key base; |
cfb6eeb4 YH |
1061 | __be32 addr; |
1062 | }; | |
1063 | ||
1064 | struct tcp6_md5sig_key { | |
f8ab18d2 | 1065 | struct tcp_md5sig_key base; |
cfb6eeb4 YH |
1066 | #if 0 |
1067 | u32 scope_id; /* XXX */ | |
1068 | #endif | |
1069 | struct in6_addr addr; | |
1070 | }; | |
1071 | ||
1072 | /* - sock block */ | |
1073 | struct tcp_md5sig_info { | |
1074 | struct tcp4_md5sig_key *keys4; | |
1075 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) | |
1076 | struct tcp6_md5sig_key *keys6; | |
1077 | u32 entries6; | |
1078 | u32 alloced6; | |
1079 | #endif | |
1080 | u32 entries4; | |
1081 | u32 alloced4; | |
1082 | }; | |
1083 | ||
1084 | /* - pseudo header */ | |
1085 | struct tcp4_pseudohdr { | |
1086 | __be32 saddr; | |
1087 | __be32 daddr; | |
1088 | __u8 pad; | |
1089 | __u8 protocol; | |
1090 | __be16 len; | |
1091 | }; | |
1092 | ||
1093 | struct tcp6_pseudohdr { | |
1094 | struct in6_addr saddr; | |
1095 | struct in6_addr daddr; | |
1096 | __be32 len; | |
1097 | __be32 protocol; /* including padding */ | |
1098 | }; | |
1099 | ||
1100 | union tcp_md5sum_block { | |
1101 | struct tcp4_pseudohdr ip4; | |
1102 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) | |
1103 | struct tcp6_pseudohdr ip6; | |
1104 | #endif | |
1105 | }; | |
1106 | ||
1107 | /* - pool: digest algorithm, hash description and scratch buffer */ | |
1108 | struct tcp_md5sig_pool { | |
1109 | struct hash_desc md5_desc; | |
1110 | union tcp_md5sum_block md5_blk; | |
1111 | }; | |
1112 | ||
1113 | #define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */ | |
1114 | ||
1115 | /* - functions */ | |
8d26d76d YH |
1116 | extern int tcp_calc_md5_hash(char *md5_hash, |
1117 | struct tcp_md5sig_key *key, | |
1118 | int bplen, | |
1119 | struct tcphdr *th, | |
1120 | unsigned int tcplen, | |
1121 | struct tcp_md5sig_pool *hp); | |
1122 | ||
cfb6eeb4 YH |
1123 | extern int tcp_v4_calc_md5_hash(char *md5_hash, |
1124 | struct tcp_md5sig_key *key, | |
1125 | struct sock *sk, | |
1126 | struct dst_entry *dst, | |
1127 | struct request_sock *req, | |
1128 | struct tcphdr *th, | |
9cb5734e | 1129 | unsigned int tcplen); |
cfb6eeb4 YH |
1130 | extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, |
1131 | struct sock *addr_sk); | |
1132 | ||
1133 | extern int tcp_v4_md5_do_add(struct sock *sk, | |
1134 | __be32 addr, | |
1135 | u8 *newkey, | |
1136 | u8 newkeylen); | |
1137 | ||
1138 | extern int tcp_v4_md5_do_del(struct sock *sk, | |
8e5200f5 | 1139 | __be32 addr); |
cfb6eeb4 | 1140 | |
9501f972 YH |
1141 | #ifdef CONFIG_TCP_MD5SIG |
1142 | #define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_keylen ? \ | |
1143 | &(struct tcp_md5sig_key) { \ | |
1144 | .key = (twsk)->tw_md5_key, \ | |
1145 | .keylen = (twsk)->tw_md5_keylen, \ | |
1146 | } : NULL) | |
1147 | #else | |
1148 | #define tcp_twsk_md5_key(twsk) NULL | |
1149 | #endif | |
1150 | ||
cfb6eeb4 YH |
1151 | extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void); |
1152 | extern void tcp_free_md5sig_pool(void); | |
1153 | ||
1154 | extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu); | |
1155 | extern void __tcp_put_md5sig_pool(void); | |
1156 | ||
1157 | static inline | |
1158 | struct tcp_md5sig_pool *tcp_get_md5sig_pool(void) | |
1159 | { | |
1160 | int cpu = get_cpu(); | |
1161 | struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu); | |
1162 | if (!ret) | |
1163 | put_cpu(); | |
1164 | return ret; | |
1165 | } | |
1166 | ||
1167 | static inline void tcp_put_md5sig_pool(void) | |
1168 | { | |
1169 | __tcp_put_md5sig_pool(); | |
1170 | put_cpu(); | |
1171 | } | |
1172 | ||
fe067e8a DM |
1173 | /* write queue abstraction */ |
1174 | static inline void tcp_write_queue_purge(struct sock *sk) | |
1175 | { | |
1176 | struct sk_buff *skb; | |
1177 | ||
1178 | while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) | |
3ab224be HA |
1179 | sk_wmem_free_skb(sk, skb); |
1180 | sk_mem_reclaim(sk); | |
fe067e8a DM |
1181 | } |
1182 | ||
1183 | static inline struct sk_buff *tcp_write_queue_head(struct sock *sk) | |
1184 | { | |
1185 | struct sk_buff *skb = sk->sk_write_queue.next; | |
1186 | if (skb == (struct sk_buff *) &sk->sk_write_queue) | |
1187 | return NULL; | |
1188 | return skb; | |
1189 | } | |
1190 | ||
1191 | static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk) | |
1192 | { | |
1193 | struct sk_buff *skb = sk->sk_write_queue.prev; | |
1194 | if (skb == (struct sk_buff *) &sk->sk_write_queue) | |
1195 | return NULL; | |
1196 | return skb; | |
1197 | } | |
1198 | ||
1199 | static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb) | |
1200 | { | |
1201 | return skb->next; | |
1202 | } | |
1203 | ||
1204 | #define tcp_for_write_queue(skb, sk) \ | |
1205 | for (skb = (sk)->sk_write_queue.next; \ | |
1206 | (skb != (struct sk_buff *)&(sk)->sk_write_queue); \ | |
1207 | skb = skb->next) | |
1208 | ||
1209 | #define tcp_for_write_queue_from(skb, sk) \ | |
1210 | for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\ | |
1211 | skb = skb->next) | |
1212 | ||
234b6860 IJ |
1213 | #define tcp_for_write_queue_from_safe(skb, tmp, sk) \ |
1214 | for (tmp = skb->next; \ | |
1215 | (skb != (struct sk_buff *)&(sk)->sk_write_queue); \ | |
1216 | skb = tmp, tmp = skb->next) | |
1217 | ||
fe067e8a DM |
1218 | static inline struct sk_buff *tcp_send_head(struct sock *sk) |
1219 | { | |
1220 | return sk->sk_send_head; | |
1221 | } | |
1222 | ||
1223 | static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb) | |
1224 | { | |
1225 | sk->sk_send_head = skb->next; | |
1226 | if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue) | |
1227 | sk->sk_send_head = NULL; | |
1228 | } | |
1229 | ||
1230 | static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked) | |
1231 | { | |
1232 | if (sk->sk_send_head == skb_unlinked) | |
1233 | sk->sk_send_head = NULL; | |
1234 | } | |
1235 | ||
1236 | static inline void tcp_init_send_head(struct sock *sk) | |
1237 | { | |
1238 | sk->sk_send_head = NULL; | |
1239 | } | |
1240 | ||
1241 | static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1242 | { | |
1243 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1244 | } | |
1245 | ||
1246 | static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1247 | { | |
1248 | __tcp_add_write_queue_tail(sk, skb); | |
1249 | ||
1250 | /* Queue it, remembering where we must start sending. */ | |
6859d494 | 1251 | if (sk->sk_send_head == NULL) { |
fe067e8a | 1252 | sk->sk_send_head = skb; |
6859d494 IJ |
1253 | |
1254 | if (tcp_sk(sk)->highest_sack == NULL) | |
1255 | tcp_sk(sk)->highest_sack = skb; | |
1256 | } | |
fe067e8a DM |
1257 | } |
1258 | ||
1259 | static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb) | |
1260 | { | |
1261 | __skb_queue_head(&sk->sk_write_queue, skb); | |
1262 | } | |
1263 | ||
1264 | /* Insert buff after skb on the write queue of sk. */ | |
1265 | static inline void tcp_insert_write_queue_after(struct sk_buff *skb, | |
1266 | struct sk_buff *buff, | |
1267 | struct sock *sk) | |
1268 | { | |
7de6c033 | 1269 | __skb_queue_after(&sk->sk_write_queue, skb, buff); |
fe067e8a DM |
1270 | } |
1271 | ||
1272 | /* Insert skb between prev and next on the write queue of sk. */ | |
1273 | static inline void tcp_insert_write_queue_before(struct sk_buff *new, | |
1274 | struct sk_buff *skb, | |
1275 | struct sock *sk) | |
1276 | { | |
1277 | __skb_insert(new, skb->prev, skb, &sk->sk_write_queue); | |