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> |
0c266898 | 44 | #include <net/dst.h> |
c752f073 | 45 | |
1da177e4 LT |
46 | #include <linux/seq_file.h> |
47 | ||
6e04e021 | 48 | extern struct inet_hashinfo tcp_hashinfo; |
1da177e4 | 49 | |
dd24c001 | 50 | extern struct percpu_counter tcp_orphan_count; |
1da177e4 | 51 | extern void tcp_time_wait(struct sock *sk, int state, int timeo); |
1da177e4 | 52 | |
1da177e4 | 53 | #define MAX_TCP_HEADER (128 + MAX_HEADER) |
33ad798c | 54 | #define MAX_TCP_OPTION_SPACE 40 |
1da177e4 LT |
55 | |
56 | /* | |
57 | * Never offer a window over 32767 without using window scaling. Some | |
58 | * poor stacks do signed 16bit maths! | |
59 | */ | |
60 | #define MAX_TCP_WINDOW 32767U | |
61 | ||
62 | /* Minimal accepted MSS. It is (60+60+8) - (20+20). */ | |
63 | #define TCP_MIN_MSS 88U | |
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 |
33ad798c | 186 | #define TCPOLEN_MSS_ALIGNED 4 |
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; |
519855c5 | 237 | extern int sysctl_tcp_cookie_size; |
1da177e4 LT |
238 | |
239 | extern atomic_t tcp_memory_allocated; | |
1748376b | 240 | extern struct percpu_counter tcp_sockets_allocated; |
1da177e4 LT |
241 | extern int tcp_memory_pressure; |
242 | ||
1da177e4 LT |
243 | /* |
244 | * The next routines deal with comparing 32 bit unsigned ints | |
245 | * and worry about wraparound (automatic with unsigned arithmetic). | |
246 | */ | |
247 | ||
248 | static inline int before(__u32 seq1, __u32 seq2) | |
249 | { | |
0d630cc0 | 250 | return (__s32)(seq1-seq2) < 0; |
1da177e4 | 251 | } |
9a036b9c | 252 | #define after(seq2, seq1) before(seq1, seq2) |
1da177e4 LT |
253 | |
254 | /* is s2<=s1<=s3 ? */ | |
255 | static inline int between(__u32 seq1, __u32 seq2, __u32 seq3) | |
256 | { | |
257 | return seq3 - seq2 >= seq1 - seq2; | |
258 | } | |
259 | ||
e4fd5da3 PE |
260 | static inline int tcp_too_many_orphans(struct sock *sk, int num) |
261 | { | |
262 | return (num > sysctl_tcp_max_orphans) || | |
263 | (sk->sk_wmem_queued > SOCK_MIN_SNDBUF && | |
264 | atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]); | |
265 | } | |
1da177e4 | 266 | |
a0f82f64 FW |
267 | /* syncookies: remember time of last synqueue overflow */ |
268 | static inline void tcp_synq_overflow(struct sock *sk) | |
269 | { | |
270 | tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies; | |
271 | } | |
272 | ||
273 | /* syncookies: no recent synqueue overflow on this listening socket? */ | |
274 | static inline int tcp_synq_no_recent_overflow(const struct sock *sk) | |
275 | { | |
276 | unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp; | |
277 | return time_after(jiffies, last_overflow + TCP_TIMEOUT_INIT); | |
278 | } | |
279 | ||
1da177e4 LT |
280 | extern struct proto tcp_prot; |
281 | ||
57ef42d5 PE |
282 | #define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field) |
283 | #define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field) | |
284 | #define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field) | |
285 | #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val) | |
1da177e4 | 286 | |
1da177e4 LT |
287 | extern void tcp_v4_err(struct sk_buff *skb, u32); |
288 | ||
289 | extern void tcp_shutdown (struct sock *sk, int how); | |
290 | ||
291 | extern int tcp_v4_rcv(struct sk_buff *skb); | |
292 | ||
293 | extern int tcp_v4_remember_stamp(struct sock *sk); | |
294 | ||
8feaf0c0 | 295 | extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw); |
1da177e4 | 296 | |
3516ffb0 | 297 | extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
298 | struct msghdr *msg, size_t size); |
299 | extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags); | |
300 | ||
301 | extern int tcp_ioctl(struct sock *sk, | |
302 | int cmd, | |
303 | unsigned long arg); | |
304 | ||
305 | extern int tcp_rcv_state_process(struct sock *sk, | |
306 | struct sk_buff *skb, | |
307 | struct tcphdr *th, | |
308 | unsigned len); | |
309 | ||
310 | extern int tcp_rcv_established(struct sock *sk, | |
311 | struct sk_buff *skb, | |
312 | struct tcphdr *th, | |
313 | unsigned len); | |
314 | ||
315 | extern void tcp_rcv_space_adjust(struct sock *sk); | |
316 | ||
0e4b4992 CL |
317 | extern void tcp_cleanup_rbuf(struct sock *sk, int copied); |
318 | ||
6d6ee43e ACM |
319 | extern int tcp_twsk_unique(struct sock *sk, |
320 | struct sock *sktw, void *twp); | |
321 | ||
cfb6eeb4 YH |
322 | extern void tcp_twsk_destructor(struct sock *sk); |
323 | ||
9c55e01c JA |
324 | extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos, |
325 | struct pipe_inode_info *pipe, size_t len, unsigned int flags); | |
326 | ||
463c84b9 ACM |
327 | static inline void tcp_dec_quickack_mode(struct sock *sk, |
328 | const unsigned int pkts) | |
1da177e4 | 329 | { |
463c84b9 | 330 | struct inet_connection_sock *icsk = inet_csk(sk); |
fc6415bc | 331 | |
463c84b9 ACM |
332 | if (icsk->icsk_ack.quick) { |
333 | if (pkts >= icsk->icsk_ack.quick) { | |
334 | icsk->icsk_ack.quick = 0; | |
fc6415bc | 335 | /* Leaving quickack mode we deflate ATO. */ |
463c84b9 | 336 | icsk->icsk_ack.ato = TCP_ATO_MIN; |
fc6415bc | 337 | } else |
463c84b9 | 338 | icsk->icsk_ack.quick -= pkts; |
1da177e4 LT |
339 | } |
340 | } | |
341 | ||
463c84b9 | 342 | extern void tcp_enter_quickack_mode(struct sock *sk); |
1da177e4 | 343 | |
bdf1ee5d IJ |
344 | #define TCP_ECN_OK 1 |
345 | #define TCP_ECN_QUEUE_CWR 2 | |
346 | #define TCP_ECN_DEMAND_CWR 4 | |
347 | ||
348 | static __inline__ void | |
349 | TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th) | |
350 | { | |
351 | if (sysctl_tcp_ecn && th->ece && th->cwr) | |
352 | inet_rsk(req)->ecn_ok = 1; | |
353 | } | |
354 | ||
fd2c3ef7 | 355 | enum tcp_tw_status { |
1da177e4 LT |
356 | TCP_TW_SUCCESS = 0, |
357 | TCP_TW_RST = 1, | |
358 | TCP_TW_ACK = 2, | |
359 | TCP_TW_SYN = 3 | |
360 | }; | |
361 | ||
362 | ||
8feaf0c0 | 363 | extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw, |
1da177e4 | 364 | struct sk_buff *skb, |
8feaf0c0 | 365 | const struct tcphdr *th); |
1da177e4 LT |
366 | |
367 | extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb, | |
60236fdd ACM |
368 | struct request_sock *req, |
369 | struct request_sock **prev); | |
1da177e4 LT |
370 | extern int tcp_child_process(struct sock *parent, |
371 | struct sock *child, | |
372 | struct sk_buff *skb); | |
46d0de4e | 373 | extern int tcp_use_frto(struct sock *sk); |
1da177e4 LT |
374 | extern void tcp_enter_frto(struct sock *sk); |
375 | extern void tcp_enter_loss(struct sock *sk, int how); | |
376 | extern void tcp_clear_retrans(struct tcp_sock *tp); | |
377 | extern void tcp_update_metrics(struct sock *sk); | |
378 | ||
379 | extern void tcp_close(struct sock *sk, | |
380 | long timeout); | |
1da177e4 LT |
381 | extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait); |
382 | ||
383 | extern int tcp_getsockopt(struct sock *sk, int level, | |
384 | int optname, | |
385 | char __user *optval, | |
386 | int __user *optlen); | |
387 | extern int tcp_setsockopt(struct sock *sk, int level, | |
388 | int optname, char __user *optval, | |
b7058842 | 389 | unsigned int optlen); |
3fdadf7d DM |
390 | extern int compat_tcp_getsockopt(struct sock *sk, |
391 | int level, int optname, | |
392 | char __user *optval, int __user *optlen); | |
393 | extern int compat_tcp_setsockopt(struct sock *sk, | |
394 | int level, int optname, | |
b7058842 | 395 | char __user *optval, unsigned int optlen); |
1da177e4 LT |
396 | extern void tcp_set_keepalive(struct sock *sk, int val); |
397 | extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, | |
398 | struct msghdr *msg, | |
399 | size_t len, int nonblock, | |
400 | int flags, int *addr_len); | |
401 | ||
1da177e4 LT |
402 | extern void tcp_parse_options(struct sk_buff *skb, |
403 | struct tcp_options_received *opt_rx, | |
022c3f7d GBY |
404 | int estab, |
405 | struct dst_entry *dst); | |
1da177e4 | 406 | |
7d5d5525 YH |
407 | extern u8 *tcp_parse_md5sig_option(struct tcphdr *th); |
408 | ||
1da177e4 LT |
409 | /* |
410 | * TCP v4 functions exported for the inet6 API | |
411 | */ | |
412 | ||
8292a17a | 413 | extern void tcp_v4_send_check(struct sock *sk, int len, |
1da177e4 LT |
414 | struct sk_buff *skb); |
415 | ||
416 | extern int tcp_v4_conn_request(struct sock *sk, | |
417 | struct sk_buff *skb); | |
418 | ||
419 | extern struct sock * tcp_create_openreq_child(struct sock *sk, | |
60236fdd | 420 | struct request_sock *req, |
1da177e4 LT |
421 | struct sk_buff *skb); |
422 | ||
423 | extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk, | |
424 | struct sk_buff *skb, | |
60236fdd | 425 | struct request_sock *req, |
1da177e4 LT |
426 | struct dst_entry *dst); |
427 | ||
428 | extern int tcp_v4_do_rcv(struct sock *sk, | |
429 | struct sk_buff *skb); | |
430 | ||
431 | extern int tcp_v4_connect(struct sock *sk, | |
432 | struct sockaddr *uaddr, | |
433 | int addr_len); | |
434 | ||
435 | extern int tcp_connect(struct sock *sk); | |
436 | ||
437 | extern struct sk_buff * tcp_make_synack(struct sock *sk, | |
438 | struct dst_entry *dst, | |
e6b4d113 WAS |
439 | struct request_sock *req, |
440 | struct request_values *rvp); | |
1da177e4 LT |
441 | |
442 | extern int tcp_disconnect(struct sock *sk, int flags); | |
443 | ||
1da177e4 | 444 | |
1da177e4 | 445 | /* From syncookies.c */ |
2051f11f | 446 | extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS]; |
1da177e4 LT |
447 | extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb, |
448 | struct ip_options *opt); | |
449 | extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb, | |
450 | __u16 *mss); | |
451 | ||
4dfc2817 FW |
452 | extern __u32 cookie_init_timestamp(struct request_sock *req); |
453 | extern void cookie_check_timestamp(struct tcp_options_received *tcp_opt); | |
454 | ||
c6aefafb GG |
455 | /* From net/ipv6/syncookies.c */ |
456 | extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb); | |
457 | extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb, | |
458 | __u16 *mss); | |
459 | ||
1da177e4 LT |
460 | /* tcp_output.c */ |
461 | ||
9e412ba7 IJ |
462 | extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss, |
463 | int nonagle); | |
464 | extern int tcp_may_send_now(struct sock *sk); | |
1da177e4 | 465 | extern int tcp_retransmit_skb(struct sock *, struct sk_buff *); |
f1ecd5d9 | 466 | extern void tcp_retransmit_timer(struct sock *sk); |
1da177e4 LT |
467 | extern void tcp_xmit_retransmit_queue(struct sock *); |
468 | extern void tcp_simple_retransmit(struct sock *); | |
469 | extern int tcp_trim_head(struct sock *, struct sk_buff *, u32); | |
6475be16 | 470 | extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int); |
1da177e4 LT |
471 | |
472 | extern void tcp_send_probe0(struct sock *); | |
473 | extern void tcp_send_partial(struct sock *); | |
474 | extern int tcp_write_wakeup(struct sock *); | |
475 | extern void tcp_send_fin(struct sock *sk); | |
dd0fc66f | 476 | extern void tcp_send_active_reset(struct sock *sk, gfp_t priority); |
1da177e4 | 477 | extern int tcp_send_synack(struct sock *); |
c1b4a7e6 | 478 | extern void tcp_push_one(struct sock *, unsigned int mss_now); |
1da177e4 LT |
479 | extern void tcp_send_ack(struct sock *sk); |
480 | extern void tcp_send_delayed_ack(struct sock *sk); | |
481 | ||
a762a980 DM |
482 | /* tcp_input.c */ |
483 | extern void tcp_cwnd_application_limited(struct sock *sk); | |
484 | ||
1da177e4 LT |
485 | /* tcp_timer.c */ |
486 | extern void tcp_init_xmit_timers(struct sock *); | |
463c84b9 ACM |
487 | static inline void tcp_clear_xmit_timers(struct sock *sk) |
488 | { | |
489 | inet_csk_clear_xmit_timers(sk); | |
490 | } | |
1da177e4 | 491 | |
1da177e4 | 492 | extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu); |
0c54b85f IJ |
493 | extern unsigned int tcp_current_mss(struct sock *sk); |
494 | ||
495 | /* Bound MSS / TSO packet size with the half of the window */ | |
496 | static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize) | |
497 | { | |
498 | if (tp->max_window && pktsize > (tp->max_window >> 1)) | |
499 | return max(tp->max_window >> 1, 68U - tp->tcp_header_len); | |
500 | else | |
501 | return pktsize; | |
502 | } | |
1da177e4 | 503 | |
17b085ea | 504 | /* tcp.c */ |
1da177e4 LT |
505 | extern void tcp_get_info(struct sock *, struct tcp_info *); |
506 | ||
507 | /* Read 'sendfile()'-style from a TCP socket */ | |
508 | typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *, | |
509 | unsigned int, size_t); | |
510 | extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc, | |
511 | sk_read_actor_t recv_actor); | |
512 | ||
40efc6fa | 513 | extern void tcp_initialize_rcv_mss(struct sock *sk); |
1da177e4 | 514 | |
5d424d5a JH |
515 | extern int tcp_mtu_to_mss(struct sock *sk, int pmtu); |
516 | extern int tcp_mss_to_mtu(struct sock *sk, int mss); | |
517 | extern void tcp_mtup_init(struct sock *sk); | |
518 | ||
f1ecd5d9 DL |
519 | static inline void tcp_bound_rto(const struct sock *sk) |
520 | { | |
521 | if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX) | |
522 | inet_csk(sk)->icsk_rto = TCP_RTO_MAX; | |
523 | } | |
524 | ||
525 | static inline u32 __tcp_set_rto(const struct tcp_sock *tp) | |
526 | { | |
527 | return (tp->srtt >> 3) + tp->rttvar; | |
528 | } | |
529 | ||
40efc6fa | 530 | static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd) |
1da177e4 LT |
531 | { |
532 | tp->pred_flags = htonl((tp->tcp_header_len << 26) | | |
533 | ntohl(TCP_FLAG_ACK) | | |
534 | snd_wnd); | |
535 | } | |
536 | ||
40efc6fa | 537 | static inline void tcp_fast_path_on(struct tcp_sock *tp) |
1da177e4 LT |
538 | { |
539 | __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale); | |
540 | } | |
541 | ||
9e412ba7 | 542 | static inline void tcp_fast_path_check(struct sock *sk) |
1da177e4 | 543 | { |
9e412ba7 IJ |
544 | struct tcp_sock *tp = tcp_sk(sk); |
545 | ||
b03efcfb | 546 | if (skb_queue_empty(&tp->out_of_order_queue) && |
1da177e4 LT |
547 | tp->rcv_wnd && |
548 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf && | |
549 | !tp->urg_data) | |
550 | tcp_fast_path_on(tp); | |
551 | } | |
552 | ||
0c266898 SS |
553 | /* Compute the actual rto_min value */ |
554 | static inline u32 tcp_rto_min(struct sock *sk) | |
555 | { | |
556 | struct dst_entry *dst = __sk_dst_get(sk); | |
557 | u32 rto_min = TCP_RTO_MIN; | |
558 | ||
559 | if (dst && dst_metric_locked(dst, RTAX_RTO_MIN)) | |
560 | rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN); | |
561 | return rto_min; | |
562 | } | |
563 | ||
1da177e4 LT |
564 | /* Compute the actual receive window we are currently advertising. |
565 | * Rcv_nxt can be after the window if our peer push more data | |
566 | * than the offered window. | |
567 | */ | |
40efc6fa | 568 | static inline u32 tcp_receive_window(const struct tcp_sock *tp) |
1da177e4 LT |
569 | { |
570 | s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt; | |
571 | ||
572 | if (win < 0) | |
573 | win = 0; | |
574 | return (u32) win; | |
575 | } | |
576 | ||
577 | /* Choose a new window, without checks for shrinking, and without | |
578 | * scaling applied to the result. The caller does these things | |
579 | * if necessary. This is a "raw" window selection. | |
580 | */ | |
581 | extern u32 __tcp_select_window(struct sock *sk); | |
582 | ||
583 | /* TCP timestamps are only 32-bits, this causes a slight | |
584 | * complication on 64-bit systems since we store a snapshot | |
31f34269 SH |
585 | * of jiffies in the buffer control blocks below. We decided |
586 | * to use only the low 32-bits of jiffies and hide the ugly | |
1da177e4 LT |
587 | * casts with the following macro. |
588 | */ | |
589 | #define tcp_time_stamp ((__u32)(jiffies)) | |
590 | ||
caa20d9a | 591 | /* This is what the send packet queuing engine uses to pass |
1da177e4 LT |
592 | * TCP per-packet control information to the transmission |
593 | * code. We also store the host-order sequence numbers in | |
594 | * here too. This is 36 bytes on 32-bit architectures, | |
595 | * 40 bytes on 64-bit machines, if this grows please adjust | |
596 | * skbuff.h:skbuff->cb[xxx] size appropriately. | |
597 | */ | |
598 | struct tcp_skb_cb { | |
599 | union { | |
600 | struct inet_skb_parm h4; | |
601 | #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE) | |
602 | struct inet6_skb_parm h6; | |
603 | #endif | |
604 | } header; /* For incoming frames */ | |
605 | __u32 seq; /* Starting sequence number */ | |
606 | __u32 end_seq; /* SEQ + FIN + SYN + datalen */ | |
607 | __u32 when; /* used to compute rtt's */ | |
608 | __u8 flags; /* TCP header flags. */ | |
609 | ||
610 | /* NOTE: These must match up to the flags byte in a | |
611 | * real TCP header. | |
612 | */ | |
613 | #define TCPCB_FLAG_FIN 0x01 | |
614 | #define TCPCB_FLAG_SYN 0x02 | |
615 | #define TCPCB_FLAG_RST 0x04 | |
616 | #define TCPCB_FLAG_PSH 0x08 | |
617 | #define TCPCB_FLAG_ACK 0x10 | |
618 | #define TCPCB_FLAG_URG 0x20 | |
619 | #define TCPCB_FLAG_ECE 0x40 | |
620 | #define TCPCB_FLAG_CWR 0x80 | |
621 | ||
622 | __u8 sacked; /* State flags for SACK/FACK. */ | |
623 | #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */ | |
624 | #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */ | |
625 | #define TCPCB_LOST 0x04 /* SKB is lost */ | |
626 | #define TCPCB_TAGBITS 0x07 /* All tag bits */ | |
627 | ||
628 | #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */ | |
629 | #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS) | |
630 | ||
1da177e4 LT |
631 | __u32 ack_seq; /* Sequence number ACK'd */ |
632 | }; | |
633 | ||
634 | #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0])) | |
635 | ||
1da177e4 LT |
636 | /* Due to TSO, an SKB can be composed of multiple actual |
637 | * packets. To keep these tracked properly, we use this. | |
638 | */ | |
639 | static inline int tcp_skb_pcount(const struct sk_buff *skb) | |
640 | { | |
7967168c | 641 | return skb_shinfo(skb)->gso_segs; |
1da177e4 LT |
642 | } |
643 | ||
644 | /* This is valid iff tcp_skb_pcount() > 1. */ | |
645 | static inline int tcp_skb_mss(const struct sk_buff *skb) | |
646 | { | |
7967168c | 647 | return skb_shinfo(skb)->gso_size; |
1da177e4 LT |
648 | } |
649 | ||
317a76f9 SH |
650 | /* Events passed to congestion control interface */ |
651 | enum tcp_ca_event { | |
652 | CA_EVENT_TX_START, /* first transmit when no packets in flight */ | |
653 | CA_EVENT_CWND_RESTART, /* congestion window restart */ | |
654 | CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */ | |
655 | CA_EVENT_FRTO, /* fast recovery timeout */ | |
656 | CA_EVENT_LOSS, /* loss timeout */ | |
657 | CA_EVENT_FAST_ACK, /* in sequence ack */ | |
658 | CA_EVENT_SLOW_ACK, /* other ack */ | |
659 | }; | |
660 | ||
661 | /* | |
662 | * Interface for adding new TCP congestion control handlers | |
663 | */ | |
664 | #define TCP_CA_NAME_MAX 16 | |
3ff825b2 SH |
665 | #define TCP_CA_MAX 128 |
666 | #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX) | |
667 | ||
164891aa SH |
668 | #define TCP_CONG_NON_RESTRICTED 0x1 |
669 | #define TCP_CONG_RTT_STAMP 0x2 | |
670 | ||
317a76f9 SH |
671 | struct tcp_congestion_ops { |
672 | struct list_head list; | |
164891aa | 673 | unsigned long flags; |
317a76f9 SH |
674 | |
675 | /* initialize private data (optional) */ | |
6687e988 | 676 | void (*init)(struct sock *sk); |
317a76f9 | 677 | /* cleanup private data (optional) */ |
6687e988 | 678 | void (*release)(struct sock *sk); |
317a76f9 SH |
679 | |
680 | /* return slow start threshold (required) */ | |
6687e988 | 681 | u32 (*ssthresh)(struct sock *sk); |
317a76f9 | 682 | /* lower bound for congestion window (optional) */ |
72dc5b92 | 683 | u32 (*min_cwnd)(const struct sock *sk); |
317a76f9 | 684 | /* do new cwnd calculation (required) */ |
c3a05c60 | 685 | void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight); |
317a76f9 | 686 | /* call before changing ca_state (optional) */ |
6687e988 | 687 | void (*set_state)(struct sock *sk, u8 new_state); |
317a76f9 | 688 | /* call when cwnd event occurs (optional) */ |
6687e988 | 689 | void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev); |
317a76f9 | 690 | /* new value of cwnd after loss (optional) */ |
6687e988 | 691 | u32 (*undo_cwnd)(struct sock *sk); |
317a76f9 | 692 | /* hook for packet ack accounting (optional) */ |
30cfd0ba | 693 | void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us); |
73c1f4a0 | 694 | /* get info for inet_diag (optional) */ |
6687e988 | 695 | void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb); |
317a76f9 SH |
696 | |
697 | char name[TCP_CA_NAME_MAX]; | |
698 | struct module *owner; | |
699 | }; | |
700 | ||
701 | extern int tcp_register_congestion_control(struct tcp_congestion_ops *type); | |
702 | extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type); | |
703 | ||
6687e988 ACM |
704 | extern void tcp_init_congestion_control(struct sock *sk); |
705 | extern void tcp_cleanup_congestion_control(struct sock *sk); | |
317a76f9 SH |
706 | extern int tcp_set_default_congestion_control(const char *name); |
707 | extern void tcp_get_default_congestion_control(char *name); | |
3ff825b2 | 708 | extern void tcp_get_available_congestion_control(char *buf, size_t len); |
ce7bc3bf SH |
709 | extern void tcp_get_allowed_congestion_control(char *buf, size_t len); |
710 | extern int tcp_set_allowed_congestion_control(char *allowed); | |
6687e988 | 711 | extern int tcp_set_congestion_control(struct sock *sk, const char *name); |
40efc6fa | 712 | extern void tcp_slow_start(struct tcp_sock *tp); |
758ce5c8 | 713 | extern void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w); |
317a76f9 | 714 | |
5f8ef48d | 715 | extern struct tcp_congestion_ops tcp_init_congestion_ops; |
6687e988 | 716 | extern u32 tcp_reno_ssthresh(struct sock *sk); |
c3a05c60 | 717 | extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight); |
72dc5b92 | 718 | extern u32 tcp_reno_min_cwnd(const struct sock *sk); |
a8acfbac | 719 | extern struct tcp_congestion_ops tcp_reno; |
317a76f9 | 720 | |
6687e988 | 721 | static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state) |
317a76f9 | 722 | { |
6687e988 ACM |
723 | struct inet_connection_sock *icsk = inet_csk(sk); |
724 | ||
725 | if (icsk->icsk_ca_ops->set_state) | |
726 | icsk->icsk_ca_ops->set_state(sk, ca_state); | |
727 | icsk->icsk_ca_state = ca_state; | |
317a76f9 SH |
728 | } |
729 | ||
6687e988 | 730 | static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event) |
317a76f9 | 731 | { |
6687e988 ACM |
732 | const struct inet_connection_sock *icsk = inet_csk(sk); |
733 | ||
734 | if (icsk->icsk_ca_ops->cwnd_event) | |
735 | icsk->icsk_ca_ops->cwnd_event(sk, event); | |
317a76f9 SH |
736 | } |
737 | ||
e60402d0 IJ |
738 | /* These functions determine how the current flow behaves in respect of SACK |
739 | * handling. SACK is negotiated with the peer, and therefore it can vary | |
740 | * between different flows. | |
741 | * | |
742 | * tcp_is_sack - SACK enabled | |
743 | * tcp_is_reno - No SACK | |
744 | * tcp_is_fack - FACK enabled, implies SACK enabled | |
745 | */ | |
746 | static inline int tcp_is_sack(const struct tcp_sock *tp) | |
747 | { | |
748 | return tp->rx_opt.sack_ok; | |
749 | } | |
750 | ||
751 | static inline int tcp_is_reno(const struct tcp_sock *tp) | |
752 | { | |
753 | return !tcp_is_sack(tp); | |
754 | } | |
755 | ||
756 | static inline int tcp_is_fack(const struct tcp_sock *tp) | |
757 | { | |
758 | return tp->rx_opt.sack_ok & 2; | |
759 | } | |
760 | ||
761 | static inline void tcp_enable_fack(struct tcp_sock *tp) | |
762 | { | |
763 | tp->rx_opt.sack_ok |= 2; | |
764 | } | |
765 | ||
83ae4088 IJ |
766 | static inline unsigned int tcp_left_out(const struct tcp_sock *tp) |
767 | { | |
768 | return tp->sacked_out + tp->lost_out; | |
769 | } | |
770 | ||
1da177e4 LT |
771 | /* This determines how many packets are "in the network" to the best |
772 | * of our knowledge. In many cases it is conservative, but where | |
773 | * detailed information is available from the receiver (via SACK | |
774 | * blocks etc.) we can make more aggressive calculations. | |
775 | * | |
776 | * Use this for decisions involving congestion control, use just | |
777 | * tp->packets_out to determine if the send queue is empty or not. | |
778 | * | |
779 | * Read this equation as: | |
780 | * | |
781 | * "Packets sent once on transmission queue" MINUS | |
782 | * "Packets left network, but not honestly ACKed yet" PLUS | |
783 | * "Packets fast retransmitted" | |
784 | */ | |
40efc6fa | 785 | static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp) |
1da177e4 | 786 | { |
83ae4088 | 787 | return tp->packets_out - tcp_left_out(tp) + tp->retrans_out; |
1da177e4 LT |
788 | } |
789 | ||
0b6a05c1 IJ |
790 | #define TCP_INFINITE_SSTHRESH 0x7fffffff |
791 | ||
792 | static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp) | |
793 | { | |
794 | return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH; | |
795 | } | |
796 | ||
1da177e4 LT |
797 | /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd. |
798 | * The exception is rate halving phase, when cwnd is decreasing towards | |
799 | * ssthresh. | |
800 | */ | |
6687e988 | 801 | static inline __u32 tcp_current_ssthresh(const struct sock *sk) |
1da177e4 | 802 | { |
6687e988 ACM |
803 | const struct tcp_sock *tp = tcp_sk(sk); |
804 | if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery)) | |
1da177e4 LT |
805 | return tp->snd_ssthresh; |
806 | else | |
807 | return max(tp->snd_ssthresh, | |
808 | ((tp->snd_cwnd >> 1) + | |
809 | (tp->snd_cwnd >> 2))); | |
810 | } | |
811 | ||
b9c4595b IJ |
812 | /* Use define here intentionally to get WARN_ON location shown at the caller */ |
813 | #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out) | |
1da177e4 | 814 | |
3cfe3baa | 815 | extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh); |
1da177e4 LT |
816 | extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst); |
817 | ||
818 | /* Slow start with delack produces 3 packets of burst, so that | |
dd9e0dda JH |
819 | * it is safe "de facto". This will be the default - same as |
820 | * the default reordering threshold - but if reordering increases, | |
821 | * we must be able to allow cwnd to burst at least this much in order | |
822 | * to not pull it back when holes are filled. | |
1da177e4 LT |
823 | */ |
824 | static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp) | |
825 | { | |
dd9e0dda | 826 | return tp->reordering; |
1da177e4 LT |
827 | } |
828 | ||
90840def IJ |
829 | /* Returns end sequence number of the receiver's advertised window */ |
830 | static inline u32 tcp_wnd_end(const struct tcp_sock *tp) | |
831 | { | |
832 | return tp->snd_una + tp->snd_wnd; | |
833 | } | |
cea14e0e | 834 | extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight); |
f4805ede | 835 | |
c1bd24b7 | 836 | static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss, |
40efc6fa | 837 | const struct sk_buff *skb) |
1da177e4 LT |
838 | { |
839 | if (skb->len < mss) | |
840 | tp->snd_sml = TCP_SKB_CB(skb)->end_seq; | |
841 | } | |
842 | ||
9e412ba7 | 843 | static inline void tcp_check_probe_timer(struct sock *sk) |
1da177e4 | 844 | { |
9e412ba7 | 845 | struct tcp_sock *tp = tcp_sk(sk); |
463c84b9 | 846 | const struct inet_connection_sock *icsk = inet_csk(sk); |
9e412ba7 | 847 | |
463c84b9 | 848 | if (!tp->packets_out && !icsk->icsk_pending) |
3f421baa ACM |
849 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, |
850 | icsk->icsk_rto, TCP_RTO_MAX); | |
1da177e4 LT |
851 | } |
852 | ||
9e412ba7 | 853 | static inline void tcp_push_pending_frames(struct sock *sk) |
1da177e4 | 854 | { |
9e412ba7 IJ |
855 | struct tcp_sock *tp = tcp_sk(sk); |
856 | ||
0c54b85f | 857 | __tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle); |
1da177e4 LT |
858 | } |
859 | ||
ee7537b6 | 860 | static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq) |
1da177e4 LT |
861 | { |
862 | tp->snd_wl1 = seq; | |
863 | } | |
864 | ||
ee7537b6 | 865 | static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq) |
1da177e4 LT |
866 | { |
867 | tp->snd_wl1 = seq; | |
868 | } | |
869 | ||
1da177e4 LT |
870 | /* |
871 | * Calculate(/check) TCP checksum | |
872 | */ | |
ba7808ea FD |
873 | static inline __sum16 tcp_v4_check(int len, __be32 saddr, |
874 | __be32 daddr, __wsum base) | |
1da177e4 LT |
875 | { |
876 | return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base); | |
877 | } | |
878 | ||
b51655b9 | 879 | static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 880 | { |
fb286bb2 | 881 | return __skb_checksum_complete(skb); |
1da177e4 LT |
882 | } |
883 | ||
40efc6fa | 884 | static inline int tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 885 | { |
60476372 | 886 | return !skb_csum_unnecessary(skb) && |
1da177e4 LT |
887 | __tcp_checksum_complete(skb); |
888 | } | |
889 | ||
890 | /* Prequeue for VJ style copy to user, combined with checksumming. */ | |
891 | ||
40efc6fa | 892 | static inline void tcp_prequeue_init(struct tcp_sock *tp) |
1da177e4 LT |
893 | { |
894 | tp->ucopy.task = NULL; | |
895 | tp->ucopy.len = 0; | |
896 | tp->ucopy.memory = 0; | |
897 | skb_queue_head_init(&tp->ucopy.prequeue); | |
97fc2f08 CL |
898 | #ifdef CONFIG_NET_DMA |
899 | tp->ucopy.dma_chan = NULL; | |
900 | tp->ucopy.wakeup = 0; | |
901 | tp->ucopy.pinned_list = NULL; | |
902 | tp->ucopy.dma_cookie = 0; | |
903 | #endif | |
1da177e4 LT |
904 | } |
905 | ||
906 | /* Packet is added to VJ-style prequeue for processing in process | |
907 | * context, if a reader task is waiting. Apparently, this exciting | |
908 | * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93) | |
909 | * failed somewhere. Latency? Burstiness? Well, at least now we will | |
910 | * see, why it failed. 8)8) --ANK | |
911 | * | |
912 | * NOTE: is this not too big to inline? | |
913 | */ | |
40efc6fa | 914 | static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
915 | { |
916 | struct tcp_sock *tp = tcp_sk(sk); | |
917 | ||
f5f8d86b ED |
918 | if (sysctl_tcp_low_latency || !tp->ucopy.task) |
919 | return 0; | |
920 | ||
921 | __skb_queue_tail(&tp->ucopy.prequeue, skb); | |
922 | tp->ucopy.memory += skb->truesize; | |
923 | if (tp->ucopy.memory > sk->sk_rcvbuf) { | |
924 | struct sk_buff *skb1; | |
925 | ||
926 | BUG_ON(sock_owned_by_user(sk)); | |
927 | ||
928 | while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) { | |
929 | sk_backlog_rcv(sk, skb1); | |
930 | NET_INC_STATS_BH(sock_net(sk), | |
931 | LINUX_MIB_TCPPREQUEUEDROPPED); | |
1da177e4 | 932 | } |
f5f8d86b ED |
933 | |
934 | tp->ucopy.memory = 0; | |
935 | } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) { | |
7aedec2a ED |
936 | wake_up_interruptible_poll(sk->sk_sleep, |
937 | POLLIN | POLLRDNORM | POLLRDBAND); | |
f5f8d86b ED |
938 | if (!inet_csk_ack_scheduled(sk)) |
939 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, | |
22f6dacd | 940 | (3 * tcp_rto_min(sk)) / 4, |
f5f8d86b | 941 | TCP_RTO_MAX); |
1da177e4 | 942 | } |
f5f8d86b | 943 | return 1; |
1da177e4 LT |
944 | } |
945 | ||
946 | ||
947 | #undef STATE_TRACE | |
948 | ||
949 | #ifdef STATE_TRACE | |
950 | static const char *statename[]={ | |
951 | "Unused","Established","Syn Sent","Syn Recv", | |
952 | "Fin Wait 1","Fin Wait 2","Time Wait", "Close", | |
953 | "Close Wait","Last ACK","Listen","Closing" | |
954 | }; | |
955 | #endif | |
490d5046 | 956 | extern void tcp_set_state(struct sock *sk, int state); |
1da177e4 | 957 | |
4ac02bab | 958 | extern void tcp_done(struct sock *sk); |
1da177e4 | 959 | |
40efc6fa | 960 | static inline void tcp_sack_reset(struct tcp_options_received *rx_opt) |
1da177e4 LT |
961 | { |
962 | rx_opt->dsack = 0; | |
1da177e4 LT |
963 | rx_opt->num_sacks = 0; |
964 | } | |
965 | ||
1da177e4 LT |
966 | /* Determine a window scaling and initial window to offer. */ |
967 | extern void tcp_select_initial_window(int __space, __u32 mss, | |
968 | __u32 *rcv_wnd, __u32 *window_clamp, | |
969 | int wscale_ok, __u8 *rcv_wscale); | |
970 | ||
971 | static inline int tcp_win_from_space(int space) | |
972 | { | |
973 | return sysctl_tcp_adv_win_scale<=0 ? | |
974 | (space>>(-sysctl_tcp_adv_win_scale)) : | |
975 | space - (space>>sysctl_tcp_adv_win_scale); | |
976 | } | |
977 | ||
978 | /* Note: caller must be prepared to deal with negative returns */ | |
979 | static inline int tcp_space(const struct sock *sk) | |
980 | { | |
981 | return tcp_win_from_space(sk->sk_rcvbuf - | |
982 | atomic_read(&sk->sk_rmem_alloc)); | |
983 | } | |
984 | ||
985 | static inline int tcp_full_space(const struct sock *sk) | |
986 | { | |
987 | return tcp_win_from_space(sk->sk_rcvbuf); | |
988 | } | |
989 | ||
40efc6fa SH |
990 | static inline void tcp_openreq_init(struct request_sock *req, |
991 | struct tcp_options_received *rx_opt, | |
992 | struct sk_buff *skb) | |
1da177e4 | 993 | { |
2e6599cb ACM |
994 | struct inet_request_sock *ireq = inet_rsk(req); |
995 | ||
1da177e4 | 996 | req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */ |
4dfc2817 | 997 | req->cookie_ts = 0; |
2e6599cb | 998 | tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq; |
1da177e4 LT |
999 | req->mss = rx_opt->mss_clamp; |
1000 | req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0; | |
2e6599cb ACM |
1001 | ireq->tstamp_ok = rx_opt->tstamp_ok; |
1002 | ireq->sack_ok = rx_opt->sack_ok; | |
1003 | ireq->snd_wscale = rx_opt->snd_wscale; | |
1004 | ireq->wscale_ok = rx_opt->wscale_ok; | |
1005 | ireq->acked = 0; | |
1006 | ireq->ecn_ok = 0; | |
aa8223c7 | 1007 | ireq->rmt_port = tcp_hdr(skb)->source; |
a3116ac5 | 1008 | ireq->loc_port = tcp_hdr(skb)->dest; |
1da177e4 LT |
1009 | } |
1010 | ||
5c52ba17 | 1011 | extern void tcp_enter_memory_pressure(struct sock *sk); |
1da177e4 | 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 | ||
df19a626 ED |
1023 | static inline int keepalive_probes(const struct tcp_sock *tp) |
1024 | { | |
1025 | return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes; | |
1026 | } | |
1027 | ||
463c84b9 | 1028 | static inline int tcp_fin_time(const struct sock *sk) |
1da177e4 | 1029 | { |
463c84b9 ACM |
1030 | int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout; |
1031 | const int rto = inet_csk(sk)->icsk_rto; | |
1da177e4 | 1032 | |
463c84b9 ACM |
1033 | if (fin_timeout < (rto << 2) - (rto >> 1)) |
1034 | fin_timeout = (rto << 2) - (rto >> 1); | |
1da177e4 LT |
1035 | |
1036 | return fin_timeout; | |
1037 | } | |
1038 | ||
c887e6d2 IJ |
1039 | static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, |
1040 | int paws_win) | |
1da177e4 | 1041 | { |
c887e6d2 IJ |
1042 | if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win) |
1043 | return 1; | |
1044 | if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)) | |
1045 | return 1; | |
1046 | ||
1047 | return 0; | |
1048 | } | |
1049 | ||
1050 | static inline int tcp_paws_reject(const struct tcp_options_received *rx_opt, | |
1051 | int rst) | |
1052 | { | |
1053 | if (tcp_paws_check(rx_opt, 0)) | |
1da177e4 LT |
1054 | return 0; |
1055 | ||
1056 | /* RST segments are not recommended to carry timestamp, | |
1057 | and, if they do, it is recommended to ignore PAWS because | |
1058 | "their cleanup function should take precedence over timestamps." | |
1059 | Certainly, it is mistake. It is necessary to understand the reasons | |
1060 | of this constraint to relax it: if peer reboots, clock may go | |
1061 | out-of-sync and half-open connections will not be reset. | |
1062 | Actually, the problem would be not existing if all | |
1063 | the implementations followed draft about maintaining clock | |
1064 | via reboots. Linux-2.2 DOES NOT! | |
1065 | ||
1066 | However, we can relax time bounds for RST segments to MSL. | |
1067 | */ | |
9d729f72 | 1068 | if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL) |
1da177e4 LT |
1069 | return 0; |
1070 | return 1; | |
1071 | } | |
1072 | ||
1da177e4 LT |
1073 | #define TCP_CHECK_TIMER(sk) do { } while (0) |
1074 | ||
a9c19329 | 1075 | static inline void tcp_mib_init(struct net *net) |
1da177e4 LT |
1076 | { |
1077 | /* See RFC 2012 */ | |
cf1100a7 PE |
1078 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1); |
1079 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ); | |
1080 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ); | |
1081 | TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1); | |
1da177e4 LT |
1082 | } |
1083 | ||
5af4ec23 | 1084 | /* from STCP */ |
ef9da47c | 1085 | static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp) |
0800f170 | 1086 | { |
6a438bbe SH |
1087 | tp->lost_skb_hint = NULL; |
1088 | tp->scoreboard_skb_hint = NULL; | |
ef9da47c IJ |
1089 | } |
1090 | ||
1091 | static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp) | |
1092 | { | |
1093 | tcp_clear_retrans_hints_partial(tp); | |
6a438bbe | 1094 | tp->retransmit_skb_hint = NULL; |
b7689205 IJ |
1095 | } |
1096 | ||
cfb6eeb4 YH |
1097 | /* MD5 Signature */ |
1098 | struct crypto_hash; | |
1099 | ||
1100 | /* - key database */ | |
1101 | struct tcp_md5sig_key { | |
1102 | u8 *key; | |
1103 | u8 keylen; | |
1104 | }; | |
1105 | ||
1106 | struct tcp4_md5sig_key { | |
f8ab18d2 | 1107 | struct tcp_md5sig_key base; |
cfb6eeb4 YH |
1108 | __be32 addr; |
1109 | }; | |
1110 | ||
1111 | struct tcp6_md5sig_key { | |
f8ab18d2 | 1112 | struct tcp_md5sig_key base; |
cfb6eeb4 YH |
1113 | #if 0 |
1114 | u32 scope_id; /* XXX */ | |
1115 | #endif | |
1116 | struct in6_addr addr; | |
1117 | }; | |
1118 | ||
1119 | /* - sock block */ | |
1120 | struct tcp_md5sig_info { | |
1121 | struct tcp4_md5sig_key *keys4; | |
1122 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) | |
1123 | struct tcp6_md5sig_key *keys6; | |
1124 | u32 entries6; | |
1125 | u32 alloced6; | |
1126 | #endif | |
1127 | u32 entries4; | |
1128 | u32 alloced4; | |
1129 | }; | |
1130 | ||
1131 | /* - pseudo header */ | |
1132 | struct tcp4_pseudohdr { | |
1133 | __be32 saddr; | |
1134 | __be32 daddr; | |
1135 | __u8 pad; | |
1136 | __u8 protocol; | |
1137 | __be16 len; | |
1138 | }; | |
1139 | ||
1140 | struct tcp6_pseudohdr { | |
1141 | struct in6_addr saddr; | |
1142 | struct in6_addr daddr; | |
1143 | __be32 len; | |
1144 | __be32 protocol; /* including padding */ | |
1145 | }; | |
1146 | ||
1147 | union tcp_md5sum_block { | |
1148 | struct tcp4_pseudohdr ip4; | |
1149 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) | |
1150 | struct tcp6_pseudohdr ip6; | |
1151 | #endif | |
1152 | }; | |
1153 | ||
1154 | /* - pool: digest algorithm, hash description and scratch buffer */ | |
1155 | struct tcp_md5sig_pool { | |
1156 | struct hash_desc md5_desc; | |
1157 | union tcp_md5sum_block md5_blk; | |
1158 | }; | |
1159 | ||
1160 | #define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */ | |
1161 | ||
1162 | /* - functions */ | |
49a72dfb AL |
1163 | extern int tcp_v4_md5_hash_skb(char *md5_hash, |
1164 | struct tcp_md5sig_key *key, | |
1165 | struct sock *sk, | |
1166 | struct request_sock *req, | |
1167 | struct sk_buff *skb); | |
1168 | ||
cfb6eeb4 YH |
1169 | extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, |
1170 | struct sock *addr_sk); | |
1171 | ||
1172 | extern int tcp_v4_md5_do_add(struct sock *sk, | |
1173 | __be32 addr, | |
1174 | u8 *newkey, | |
1175 | u8 newkeylen); | |
1176 | ||
1177 | extern int tcp_v4_md5_do_del(struct sock *sk, | |
8e5200f5 | 1178 | __be32 addr); |
cfb6eeb4 | 1179 | |
9501f972 YH |
1180 | #ifdef CONFIG_TCP_MD5SIG |
1181 | #define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_keylen ? \ | |
1182 | &(struct tcp_md5sig_key) { \ | |
1183 | .key = (twsk)->tw_md5_key, \ | |
1184 | .keylen = (twsk)->tw_md5_keylen, \ | |
1185 | } : NULL) | |
1186 | #else | |
1187 | #define tcp_twsk_md5_key(twsk) NULL | |
1188 | #endif | |
1189 | ||
aa133076 | 1190 | extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(struct sock *); |
cfb6eeb4 YH |
1191 | extern void tcp_free_md5sig_pool(void); |
1192 | ||
1193 | extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu); | |
1194 | extern void __tcp_put_md5sig_pool(void); | |
49a72dfb AL |
1195 | extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, struct tcphdr *); |
1196 | extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, struct sk_buff *, | |
1197 | unsigned header_len); | |
1198 | extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp, | |
1199 | struct tcp_md5sig_key *key); | |
cfb6eeb4 YH |
1200 | |
1201 | static inline | |
1202 | struct tcp_md5sig_pool *tcp_get_md5sig_pool(void) | |
1203 | { | |
1204 | int cpu = get_cpu(); | |
1205 | struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu); | |
1206 | if (!ret) | |
1207 | put_cpu(); | |
1208 | return ret; | |
1209 | } | |
1210 | ||
1211 | static inline void tcp_put_md5sig_pool(void) | |
1212 | { | |
1213 | __tcp_put_md5sig_pool(); | |
1214 | put_cpu(); | |
1215 | } | |
1216 | ||
fe067e8a DM |
1217 | /* write queue abstraction */ |
1218 | static inline void tcp_write_queue_purge(struct sock *sk) | |
1219 | { | |
1220 | struct sk_buff *skb; | |
1221 | ||
1222 | while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) | |
3ab224be HA |
1223 | sk_wmem_free_skb(sk, skb); |
1224 | sk_mem_reclaim(sk); | |
fe067e8a DM |
1225 | } |
1226 | ||
1227 | static inline struct sk_buff *tcp_write_queue_head(struct sock *sk) | |
1228 | { | |
cd07a8ea | 1229 | return skb_peek(&sk->sk_write_queue); |
fe067e8a DM |
1230 | } |
1231 | ||
1232 | static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk) | |
1233 | { | |
cd07a8ea | 1234 | return skb_peek_tail(&sk->sk_write_queue); |
fe067e8a DM |
1235 | } |
1236 | ||
1237 | static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb) | |
1238 | { | |
cd07a8ea | 1239 | return skb_queue_next(&sk->sk_write_queue, skb); |
fe067e8a DM |
1240 | } |
1241 | ||
832d11c5 IJ |
1242 | static inline struct sk_buff *tcp_write_queue_prev(struct sock *sk, struct sk_buff *skb) |
1243 | { | |
1244 | return skb_queue_prev(&sk->sk_write_queue, skb); | |
1245 | } | |
1246 | ||
fe067e8a | 1247 | #define tcp_for_write_queue(skb, sk) \ |
cd07a8ea | 1248 | skb_queue_walk(&(sk)->sk_write_queue, skb) |
fe067e8a DM |
1249 | |
1250 | #define tcp_for_write_queue_from(skb, sk) \ | |
cd07a8ea | 1251 | skb_queue_walk_from(&(sk)->sk_write_queue, skb) |
fe067e8a | 1252 | |
234b6860 | 1253 | #define tcp_for_write_queue_from_safe(skb, tmp, sk) \ |
cd07a8ea | 1254 | skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp) |
234b6860 | 1255 | |
5152fc7d DL |
1256 | /* This function calculates a "timeout" which is equivalent to the timeout of a |
1257 | * TCP connection after "boundary" unsucessful, exponentially backed-off | |
1258 | * retransmissions with an initial RTO of TCP_RTO_MIN. | |
1259 | */ | |
6fa12c85 DL |
1260 | static inline bool retransmits_timed_out(const struct sock *sk, |
1261 | unsigned int boundary) | |
1262 | { | |
5152fc7d DL |
1263 | unsigned int timeout, linear_backoff_thresh; |
1264 | ||
6fa12c85 DL |
1265 | if (!inet_csk(sk)->icsk_retransmits) |
1266 | return false; | |
1267 | ||
5152fc7d | 1268 | linear_backoff_thresh = ilog2(TCP_RTO_MAX/TCP_RTO_MIN); |
6fa12c85 | 1269 | |
5152fc7d DL |
1270 | if (boundary <= linear_backoff_thresh) |
1271 | timeout = ((2 << boundary) - 1) * TCP_RTO_MIN; | |
6fa12c85 | 1272 | else |
5152fc7d DL |
1273 | timeout = ((2 << linear_backoff_thresh) - 1) * TCP_RTO_MIN + |
1274 | (boundary - linear_backoff_thresh) * TCP_RTO_MAX; | |
6fa12c85 | 1275 | |
5152fc7d | 1276 | return (tcp_time_stamp - tcp_sk(sk)->retrans_stamp) >= timeout; |
6fa12c85 DL |
1277 | } |
1278 | ||
fe067e8a DM |
1279 | static inline struct sk_buff *tcp_send_head(struct sock *sk) |
1280 | { | |
1281 | return sk->sk_send_head; | |
1282 | } | |
1283 | ||
cd07a8ea DM |
1284 | static inline bool tcp_skb_is_last(const struct sock *sk, |
1285 | const struct sk_buff *skb) | |
1286 | { | |
1287 | return skb_queue_is_last(&sk->sk_write_queue, skb); | |
1288 | } | |
1289 | ||
fe067e8a DM |
1290 | static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb) |
1291 | { | |
cd07a8ea | 1292 | if (tcp_skb_is_last(sk, skb)) |
fe067e8a | 1293 | sk->sk_send_head = NULL; |
cd07a8ea DM |
1294 | else |
1295 | sk->sk_send_head = tcp_write_queue_next(sk, skb); | |
fe067e8a DM |
1296 | } |
1297 | ||
1298 | static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked) | |
1299 | { | |
1300 | if (sk->sk_send_head == skb_unlinked) | |
1301 | sk->sk_send_head = NULL; | |
1302 | } | |
1303 | ||
1304 | static inline void tcp_init_send_head(struct sock *sk) | |
1305 | { | |
1306 | sk->sk_send_head = NULL; | |
1307 | } | |
1308 | ||
1309 | static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1310 | { | |
1311 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1312 | } | |
1313 | ||
1314 | static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1315 | { | |
1316 | __tcp_add_write_queue_tail(sk, skb); | |
1317 | ||
1318 | /* Queue it, remembering where we must start sending. */ | |
6859d494 | 1319 | if (sk->sk_send_head == NULL) { |
fe067e8a | 1320 | sk->sk_send_head = skb; |
6859d494 IJ |
1321 | |
1322 | if (tcp_sk(sk)->highest_sack == NULL) | |
1323 | tcp_sk(sk)->highest_sack = skb; | |
1324 | } | |
fe067e8a DM |
1325 | } |
1326 | ||
1327 | static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb) | |
1328 | { | |
1329 | __skb_queue_head(&sk->sk_write_queue, skb); | |
1330 | } | |
1331 | ||
1332 | /* Insert buff after skb on the write queue of sk. */ | |
1333 | static inline void tcp_insert_write_queue_after(struct sk_buff *skb, | |
1334 | struct sk_buff *buff, | |
1335 | struct sock *sk) | |
1336 | { | |
7de6c033 | 1337 | __skb_queue_after(&sk->sk_write_queue, skb, buff); |
fe067e8a DM |
1338 | } |
1339 | ||
43f59c89 | 1340 | /* Insert new before skb on the write queue of sk. */ |
fe067e8a DM |
1341 | static inline void tcp_insert_write_queue_before(struct sk_buff *new, |
1342 | struct sk_buff *skb, | |
1343 | struct sock *sk) | |
1344 | { | |
43f59c89 | 1345 | __skb_queue_before(&sk->sk_write_queue, skb, new); |