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 | ||
1da177e4 LT |
21 | #define FASTRETRANS_DEBUG 1 |
22 | ||
1da177e4 LT |
23 | #include <linux/list.h> |
24 | #include <linux/tcp.h> | |
187f1882 | 25 | #include <linux/bug.h> |
1da177e4 LT |
26 | #include <linux/slab.h> |
27 | #include <linux/cache.h> | |
28 | #include <linux/percpu.h> | |
fb286bb2 | 29 | #include <linux/skbuff.h> |
c6aefafb | 30 | #include <linux/cryptohash.h> |
435cf559 | 31 | #include <linux/kref.h> |
740b0f18 | 32 | #include <linux/ktime.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 | 46 | #include <linux/seq_file.h> |
180d8cd9 | 47 | #include <linux/memcontrol.h> |
1da177e4 | 48 | |
6e04e021 | 49 | extern struct inet_hashinfo tcp_hashinfo; |
1da177e4 | 50 | |
dd24c001 | 51 | extern struct percpu_counter tcp_orphan_count; |
5c9f3023 | 52 | void tcp_time_wait(struct sock *sk, int state, int timeo); |
1da177e4 | 53 | |
1da177e4 | 54 | #define MAX_TCP_HEADER (128 + MAX_HEADER) |
33ad798c | 55 | #define MAX_TCP_OPTION_SPACE 40 |
1da177e4 | 56 | |
105970f6 | 57 | /* |
1da177e4 | 58 | * Never offer a window over 32767 without using window scaling. Some |
105970f6 | 59 | * poor stacks do signed 16bit maths! |
1da177e4 LT |
60 | */ |
61 | #define MAX_TCP_WINDOW 32767U | |
62 | ||
63 | /* Minimal accepted MSS. It is (60+60+8) - (20+20). */ | |
64 | #define TCP_MIN_MSS 88U | |
65 | ||
5d424d5a | 66 | /* The least MTU to use for probing */ |
dcd8fb85 | 67 | #define TCP_BASE_MSS 1024 |
5d424d5a | 68 | |
05cbc0db FD |
69 | /* probing interval, default to 10 minutes as per RFC4821 */ |
70 | #define TCP_PROBE_INTERVAL 600 | |
71 | ||
6b58e0a5 FD |
72 | /* Specify interval when tcp mtu probing will stop */ |
73 | #define TCP_PROBE_THRESHOLD 8 | |
74 | ||
1da177e4 LT |
75 | /* After receiving this amount of duplicate ACKs fast retransmit starts. */ |
76 | #define TCP_FASTRETRANS_THRESH 3 | |
77 | ||
1da177e4 LT |
78 | /* Maximal number of ACKs sent quickly to accelerate slow-start. */ |
79 | #define TCP_MAX_QUICKACKS 16U | |
80 | ||
81 | /* urg_data states */ | |
82 | #define TCP_URG_VALID 0x0100 | |
83 | #define TCP_URG_NOTYET 0x0200 | |
84 | #define TCP_URG_READ 0x0400 | |
85 | ||
86 | #define TCP_RETR1 3 /* | |
87 | * This is how many retries it does before it | |
88 | * tries to figure out if the gateway is | |
89 | * down. Minimal RFC value is 3; it corresponds | |
90 | * to ~3sec-8min depending on RTO. | |
91 | */ | |
92 | ||
93 | #define TCP_RETR2 15 /* | |
94 | * This should take at least | |
95 | * 90 minutes to time out. | |
96 | * RFC1122 says that the limit is 100 sec. | |
97 | * 15 is ~13-30min depending on RTO. | |
98 | */ | |
99 | ||
6c9ff979 AB |
100 | #define TCP_SYN_RETRIES 6 /* This is how many retries are done |
101 | * when active opening a connection. | |
102 | * RFC1122 says the minimum retry MUST | |
103 | * be at least 180secs. Nevertheless | |
104 | * this value is corresponding to | |
105 | * 63secs of retransmission with the | |
106 | * current initial RTO. | |
107 | */ | |
1da177e4 | 108 | |
6c9ff979 AB |
109 | #define TCP_SYNACK_RETRIES 5 /* This is how may retries are done |
110 | * when passive opening a connection. | |
111 | * This is corresponding to 31secs of | |
112 | * retransmission with the current | |
113 | * initial RTO. | |
114 | */ | |
1da177e4 | 115 | |
1da177e4 LT |
116 | #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT |
117 | * state, about 60 seconds */ | |
118 | #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN | |
119 | /* BSD style FIN_WAIT2 deadlock breaker. | |
120 | * It used to be 3min, new value is 60sec, | |
121 | * to combine FIN-WAIT-2 timeout with | |
122 | * TIME-WAIT timer. | |
123 | */ | |
124 | ||
125 | #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */ | |
126 | #if HZ >= 100 | |
127 | #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */ | |
128 | #define TCP_ATO_MIN ((unsigned)(HZ/25)) | |
129 | #else | |
130 | #define TCP_DELACK_MIN 4U | |
131 | #define TCP_ATO_MIN 4U | |
132 | #endif | |
133 | #define TCP_RTO_MAX ((unsigned)(120*HZ)) | |
134 | #define TCP_RTO_MIN ((unsigned)(HZ/5)) | |
fd4f2cea | 135 | #define TCP_TIMEOUT_INIT ((unsigned)(1*HZ)) /* RFC6298 2.1 initial RTO value */ |
9ad7c049 JC |
136 | #define TCP_TIMEOUT_FALLBACK ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value, now |
137 | * used as a fallback RTO for the | |
138 | * initial data transmission if no | |
139 | * valid RTT sample has been acquired, | |
140 | * most likely due to retrans in 3WHS. | |
141 | */ | |
1da177e4 LT |
142 | |
143 | #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes | |
144 | * for local resources. | |
145 | */ | |
146 | ||
147 | #define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */ | |
148 | #define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */ | |
149 | #define TCP_KEEPALIVE_INTVL (75*HZ) | |
150 | ||
151 | #define MAX_TCP_KEEPIDLE 32767 | |
152 | #define MAX_TCP_KEEPINTVL 32767 | |
153 | #define MAX_TCP_KEEPCNT 127 | |
154 | #define MAX_TCP_SYNCNT 127 | |
155 | ||
156 | #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */ | |
1da177e4 LT |
157 | |
158 | #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24) | |
159 | #define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated | |
160 | * after this time. It should be equal | |
161 | * (or greater than) TCP_TIMEWAIT_LEN | |
162 | * to provide reliability equal to one | |
163 | * provided by timewait state. | |
164 | */ | |
165 | #define TCP_PAWS_WINDOW 1 /* Replay window for per-host | |
166 | * timestamps. It must be less than | |
167 | * minimal timewait lifetime. | |
168 | */ | |
1da177e4 LT |
169 | /* |
170 | * TCP option | |
171 | */ | |
105970f6 | 172 | |
1da177e4 LT |
173 | #define TCPOPT_NOP 1 /* Padding */ |
174 | #define TCPOPT_EOL 0 /* End of options */ | |
175 | #define TCPOPT_MSS 2 /* Segment size negotiating */ | |
176 | #define TCPOPT_WINDOW 3 /* Window scaling */ | |
177 | #define TCPOPT_SACK_PERM 4 /* SACK Permitted */ | |
178 | #define TCPOPT_SACK 5 /* SACK Block */ | |
179 | #define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */ | |
cfb6eeb4 | 180 | #define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */ |
7f9b838b | 181 | #define TCPOPT_FASTOPEN 34 /* Fast open (RFC7413) */ |
2100c8d2 YC |
182 | #define TCPOPT_EXP 254 /* Experimental */ |
183 | /* Magic number to be after the option value for sharing TCP | |
184 | * experimental options. See draft-ietf-tcpm-experimental-options-00.txt | |
185 | */ | |
186 | #define TCPOPT_FASTOPEN_MAGIC 0xF989 | |
1da177e4 LT |
187 | |
188 | /* | |
189 | * TCP option lengths | |
190 | */ | |
191 | ||
192 | #define TCPOLEN_MSS 4 | |
193 | #define TCPOLEN_WINDOW 3 | |
194 | #define TCPOLEN_SACK_PERM 2 | |
195 | #define TCPOLEN_TIMESTAMP 10 | |
cfb6eeb4 | 196 | #define TCPOLEN_MD5SIG 18 |
7f9b838b | 197 | #define TCPOLEN_FASTOPEN_BASE 2 |
2100c8d2 | 198 | #define TCPOLEN_EXP_FASTOPEN_BASE 4 |
1da177e4 LT |
199 | |
200 | /* But this is what stacks really send out. */ | |
201 | #define TCPOLEN_TSTAMP_ALIGNED 12 | |
202 | #define TCPOLEN_WSCALE_ALIGNED 4 | |
203 | #define TCPOLEN_SACKPERM_ALIGNED 4 | |
204 | #define TCPOLEN_SACK_BASE 2 | |
205 | #define TCPOLEN_SACK_BASE_ALIGNED 4 | |
206 | #define TCPOLEN_SACK_PERBLOCK 8 | |
cfb6eeb4 | 207 | #define TCPOLEN_MD5SIG_ALIGNED 20 |
33ad798c | 208 | #define TCPOLEN_MSS_ALIGNED 4 |
1da177e4 | 209 | |
1da177e4 LT |
210 | /* Flags in tp->nonagle */ |
211 | #define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */ | |
212 | #define TCP_NAGLE_CORK 2 /* Socket is corked */ | |
caa20d9a | 213 | #define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */ |
1da177e4 | 214 | |
36e31b0a AP |
215 | /* TCP thin-stream limits */ |
216 | #define TCP_THIN_LINEAR_RETRIES 6 /* After 6 linear retries, do exp. backoff */ | |
217 | ||
21603fc4 | 218 | /* TCP initial congestion window as per rfc6928 */ |
442b9635 DM |
219 | #define TCP_INIT_CWND 10 |
220 | ||
cf60af03 YC |
221 | /* Bit Flags for sysctl_tcp_fastopen */ |
222 | #define TFO_CLIENT_ENABLE 1 | |
10467163 | 223 | #define TFO_SERVER_ENABLE 2 |
67da22d2 | 224 | #define TFO_CLIENT_NO_COOKIE 4 /* Data in SYN w/o cookie option */ |
cf60af03 | 225 | |
10467163 JC |
226 | /* Accept SYN data w/o any cookie option */ |
227 | #define TFO_SERVER_COOKIE_NOT_REQD 0x200 | |
228 | ||
229 | /* Force enable TFO on all listeners, i.e., not requiring the | |
230 | * TCP_FASTOPEN socket option. SOCKOPT1/2 determine how to set max_qlen. | |
231 | */ | |
232 | #define TFO_SERVER_WO_SOCKOPT1 0x400 | |
233 | #define TFO_SERVER_WO_SOCKOPT2 0x800 | |
10467163 | 234 | |
295ff7ed ACM |
235 | extern struct inet_timewait_death_row tcp_death_row; |
236 | ||
1da177e4 | 237 | /* sysctl variables for tcp */ |
1da177e4 LT |
238 | extern int sysctl_tcp_timestamps; |
239 | extern int sysctl_tcp_window_scaling; | |
240 | extern int sysctl_tcp_sack; | |
2100c8d2 | 241 | extern int sysctl_tcp_fastopen; |
1da177e4 LT |
242 | extern int sysctl_tcp_retrans_collapse; |
243 | extern int sysctl_tcp_stdurg; | |
244 | extern int sysctl_tcp_rfc1337; | |
245 | extern int sysctl_tcp_abort_on_overflow; | |
246 | extern int sysctl_tcp_max_orphans; | |
1da177e4 LT |
247 | extern int sysctl_tcp_fack; |
248 | extern int sysctl_tcp_reordering; | |
dca145ff | 249 | extern int sysctl_tcp_max_reordering; |
1da177e4 | 250 | extern int sysctl_tcp_dsack; |
a4fe34bf | 251 | extern long sysctl_tcp_mem[3]; |
1da177e4 LT |
252 | extern int sysctl_tcp_wmem[3]; |
253 | extern int sysctl_tcp_rmem[3]; | |
254 | extern int sysctl_tcp_app_win; | |
255 | extern int sysctl_tcp_adv_win_scale; | |
256 | extern int sysctl_tcp_tw_reuse; | |
257 | extern int sysctl_tcp_frto; | |
258 | extern int sysctl_tcp_low_latency; | |
1da177e4 | 259 | extern int sysctl_tcp_nometrics_save; |
1da177e4 LT |
260 | extern int sysctl_tcp_moderate_rcvbuf; |
261 | extern int sysctl_tcp_tso_win_divisor; | |
15d99e02 | 262 | extern int sysctl_tcp_workaround_signed_windows; |
35089bb2 | 263 | extern int sysctl_tcp_slow_start_after_idle; |
36e31b0a | 264 | extern int sysctl_tcp_thin_linear_timeouts; |
7e380175 | 265 | extern int sysctl_tcp_thin_dupack; |
eed530b6 | 266 | extern int sysctl_tcp_early_retrans; |
46d3ceab | 267 | extern int sysctl_tcp_limit_output_bytes; |
282f23c6 | 268 | extern int sysctl_tcp_challenge_ack_limit; |
95bd09eb | 269 | extern int sysctl_tcp_min_tso_segs; |
f6722583 | 270 | extern int sysctl_tcp_min_rtt_wlen; |
f54b3111 | 271 | extern int sysctl_tcp_autocorking; |
032ee423 | 272 | extern int sysctl_tcp_invalid_ratelimit; |
43e122b0 ED |
273 | extern int sysctl_tcp_pacing_ss_ratio; |
274 | extern int sysctl_tcp_pacing_ca_ratio; | |
1da177e4 | 275 | |
8d987e5c | 276 | extern atomic_long_t tcp_memory_allocated; |
1748376b | 277 | extern struct percpu_counter tcp_sockets_allocated; |
1da177e4 LT |
278 | extern int tcp_memory_pressure; |
279 | ||
b8da51eb ED |
280 | /* optimized version of sk_under_memory_pressure() for TCP sockets */ |
281 | static inline bool tcp_under_memory_pressure(const struct sock *sk) | |
282 | { | |
baac50bb JW |
283 | if (mem_cgroup_sockets_enabled && sk->sk_memcg && |
284 | mem_cgroup_under_socket_pressure(sk->sk_memcg)) | |
e805605c | 285 | return true; |
b8da51eb ED |
286 | |
287 | return tcp_memory_pressure; | |
288 | } | |
1da177e4 LT |
289 | /* |
290 | * The next routines deal with comparing 32 bit unsigned ints | |
291 | * and worry about wraparound (automatic with unsigned arithmetic). | |
292 | */ | |
293 | ||
a2a385d6 | 294 | static inline bool before(__u32 seq1, __u32 seq2) |
1da177e4 | 295 | { |
0d630cc0 | 296 | return (__s32)(seq1-seq2) < 0; |
1da177e4 | 297 | } |
9a036b9c | 298 | #define after(seq2, seq1) before(seq1, seq2) |
1da177e4 LT |
299 | |
300 | /* is s2<=s1<=s3 ? */ | |
a2a385d6 | 301 | static inline bool between(__u32 seq1, __u32 seq2, __u32 seq3) |
1da177e4 LT |
302 | { |
303 | return seq3 - seq2 >= seq1 - seq2; | |
304 | } | |
305 | ||
efcdbf24 AS |
306 | static inline bool tcp_out_of_memory(struct sock *sk) |
307 | { | |
308 | if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF && | |
309 | sk_memory_allocated(sk) > sk_prot_mem_limits(sk, 2)) | |
310 | return true; | |
311 | return false; | |
312 | } | |
313 | ||
a6c5ea4c ED |
314 | void sk_forced_mem_schedule(struct sock *sk, int size); |
315 | ||
ad1af0fe | 316 | static inline bool tcp_too_many_orphans(struct sock *sk, int shift) |
e4fd5da3 | 317 | { |
ad1af0fe DM |
318 | struct percpu_counter *ocp = sk->sk_prot->orphan_count; |
319 | int orphans = percpu_counter_read_positive(ocp); | |
320 | ||
321 | if (orphans << shift > sysctl_tcp_max_orphans) { | |
322 | orphans = percpu_counter_sum_positive(ocp); | |
323 | if (orphans << shift > sysctl_tcp_max_orphans) | |
324 | return true; | |
325 | } | |
ad1af0fe | 326 | return false; |
e4fd5da3 | 327 | } |
1da177e4 | 328 | |
5c9f3023 | 329 | bool tcp_check_oom(struct sock *sk, int shift); |
efcdbf24 | 330 | |
a0f82f64 | 331 | |
1da177e4 LT |
332 | extern struct proto tcp_prot; |
333 | ||
57ef42d5 PE |
334 | #define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field) |
335 | #define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field) | |
336 | #define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field) | |
337 | #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val) | |
aa2ea058 | 338 | #define TCP_ADD_STATS(net, field, val) SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val) |
1da177e4 | 339 | |
5c9f3023 JP |
340 | void tcp_tasklet_init(void); |
341 | ||
342 | void tcp_v4_err(struct sk_buff *skb, u32); | |
343 | ||
344 | void tcp_shutdown(struct sock *sk, int how); | |
345 | ||
346 | void tcp_v4_early_demux(struct sk_buff *skb); | |
347 | int tcp_v4_rcv(struct sk_buff *skb); | |
348 | ||
349 | int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw); | |
1b784140 | 350 | int tcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size); |
5c9f3023 JP |
351 | int tcp_sendpage(struct sock *sk, struct page *page, int offset, size_t size, |
352 | int flags); | |
353 | void tcp_release_cb(struct sock *sk); | |
354 | void tcp_wfree(struct sk_buff *skb); | |
355 | void tcp_write_timer_handler(struct sock *sk); | |
356 | void tcp_delack_timer_handler(struct sock *sk); | |
357 | int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg); | |
72ab4a86 | 358 | int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb); |
5c9f3023 JP |
359 | void tcp_rcv_established(struct sock *sk, struct sk_buff *skb, |
360 | const struct tcphdr *th, unsigned int len); | |
361 | void tcp_rcv_space_adjust(struct sock *sk); | |
5c9f3023 JP |
362 | int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp); |
363 | void tcp_twsk_destructor(struct sock *sk); | |
364 | ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos, | |
365 | struct pipe_inode_info *pipe, size_t len, | |
366 | unsigned int flags); | |
9c55e01c | 367 | |
463c84b9 ACM |
368 | static inline void tcp_dec_quickack_mode(struct sock *sk, |
369 | const unsigned int pkts) | |
1da177e4 | 370 | { |
463c84b9 | 371 | struct inet_connection_sock *icsk = inet_csk(sk); |
fc6415bc | 372 | |
463c84b9 ACM |
373 | if (icsk->icsk_ack.quick) { |
374 | if (pkts >= icsk->icsk_ack.quick) { | |
375 | icsk->icsk_ack.quick = 0; | |
fc6415bc | 376 | /* Leaving quickack mode we deflate ATO. */ |
463c84b9 | 377 | icsk->icsk_ack.ato = TCP_ATO_MIN; |
fc6415bc | 378 | } else |
463c84b9 | 379 | icsk->icsk_ack.quick -= pkts; |
1da177e4 LT |
380 | } |
381 | } | |
382 | ||
bdf1ee5d IJ |
383 | #define TCP_ECN_OK 1 |
384 | #define TCP_ECN_QUEUE_CWR 2 | |
385 | #define TCP_ECN_DEMAND_CWR 4 | |
7a269ffa | 386 | #define TCP_ECN_SEEN 8 |
bdf1ee5d | 387 | |
fd2c3ef7 | 388 | enum tcp_tw_status { |
1da177e4 LT |
389 | TCP_TW_SUCCESS = 0, |
390 | TCP_TW_RST = 1, | |
391 | TCP_TW_ACK = 2, | |
392 | TCP_TW_SYN = 3 | |
393 | }; | |
394 | ||
395 | ||
5c9f3023 JP |
396 | enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw, |
397 | struct sk_buff *skb, | |
398 | const struct tcphdr *th); | |
399 | struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb, | |
52452c54 | 400 | struct request_sock *req, bool fastopen); |
5c9f3023 JP |
401 | int tcp_child_process(struct sock *parent, struct sock *child, |
402 | struct sk_buff *skb); | |
5ae344c9 | 403 | void tcp_enter_loss(struct sock *sk); |
5c9f3023 JP |
404 | void tcp_clear_retrans(struct tcp_sock *tp); |
405 | void tcp_update_metrics(struct sock *sk); | |
406 | void tcp_init_metrics(struct sock *sk); | |
407 | void tcp_metrics_init(void); | |
408 | bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst, | |
a26552af | 409 | bool paws_check, bool timestamps); |
5c9f3023 JP |
410 | bool tcp_remember_stamp(struct sock *sk); |
411 | bool tcp_tw_remember_stamp(struct inet_timewait_sock *tw); | |
412 | void tcp_fetch_timewait_stamp(struct sock *sk, struct dst_entry *dst); | |
413 | void tcp_disable_fack(struct tcp_sock *tp); | |
414 | void tcp_close(struct sock *sk, long timeout); | |
415 | void tcp_init_sock(struct sock *sk); | |
416 | unsigned int tcp_poll(struct file *file, struct socket *sock, | |
417 | struct poll_table_struct *wait); | |
418 | int tcp_getsockopt(struct sock *sk, int level, int optname, | |
419 | char __user *optval, int __user *optlen); | |
420 | int tcp_setsockopt(struct sock *sk, int level, int optname, | |
421 | char __user *optval, unsigned int optlen); | |
422 | int compat_tcp_getsockopt(struct sock *sk, int level, int optname, | |
53d3176b | 423 | char __user *optval, int __user *optlen); |
5c9f3023 | 424 | int compat_tcp_setsockopt(struct sock *sk, int level, int optname, |
53d3176b | 425 | char __user *optval, unsigned int optlen); |
5c9f3023 | 426 | void tcp_set_keepalive(struct sock *sk, int val); |
42cb80a2 | 427 | void tcp_syn_ack_timeout(const struct request_sock *req); |
1b784140 YX |
428 | int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock, |
429 | int flags, int *addr_len); | |
5c9f3023 JP |
430 | void tcp_parse_options(const struct sk_buff *skb, |
431 | struct tcp_options_received *opt_rx, | |
432 | int estab, struct tcp_fastopen_cookie *foc); | |
433 | const u8 *tcp_parse_md5sig_option(const struct tcphdr *th); | |
7d5d5525 | 434 | |
1da177e4 LT |
435 | /* |
436 | * TCP v4 functions exported for the inet6 API | |
437 | */ | |
438 | ||
5c9f3023 | 439 | void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb); |
4fab9071 | 440 | void tcp_v4_mtu_reduced(struct sock *sk); |
9cf74903 | 441 | void tcp_req_err(struct sock *sk, u32 seq, bool abort); |
5c9f3023 | 442 | int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb); |
c28c6f04 | 443 | struct sock *tcp_create_openreq_child(const struct sock *sk, |
5c9f3023 JP |
444 | struct request_sock *req, |
445 | struct sk_buff *skb); | |
81164413 | 446 | void tcp_ca_openreq_child(struct sock *sk, const struct dst_entry *dst); |
0c27171e | 447 | struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb, |
5c9f3023 | 448 | struct request_sock *req, |
5e0724d0 ED |
449 | struct dst_entry *dst, |
450 | struct request_sock *req_unhash, | |
451 | bool *own_req); | |
5c9f3023 JP |
452 | int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb); |
453 | int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len); | |
454 | int tcp_connect(struct sock *sk); | |
5d062de7 | 455 | struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst, |
5c9f3023 | 456 | struct request_sock *req, |
ca6fb065 ED |
457 | struct tcp_fastopen_cookie *foc, |
458 | bool attach_req); | |
5c9f3023 | 459 | int tcp_disconnect(struct sock *sk, int flags); |
1da177e4 | 460 | |
370816ae | 461 | void tcp_finish_connect(struct sock *sk, struct sk_buff *skb); |
292e8d8c | 462 | int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size); |
63d02d15 | 463 | void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb); |
1da177e4 | 464 | |
1da177e4 | 465 | /* From syncookies.c */ |
b80c0e78 ED |
466 | struct sock *tcp_get_cookie_sock(struct sock *sk, struct sk_buff *skb, |
467 | struct request_sock *req, | |
468 | struct dst_entry *dst); | |
5c9f3023 JP |
469 | int __cookie_v4_check(const struct iphdr *iph, const struct tcphdr *th, |
470 | u32 cookie); | |
461b74c3 | 471 | struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb); |
e05c82d3 | 472 | #ifdef CONFIG_SYN_COOKIES |
8c27bd75 | 473 | |
63262315 | 474 | /* Syncookies use a monotonic timer which increments every 60 seconds. |
8c27bd75 FW |
475 | * This counter is used both as a hash input and partially encoded into |
476 | * the cookie value. A cookie is only validated further if the delta | |
477 | * between the current counter value and the encoded one is less than this, | |
63262315 | 478 | * i.e. a sent cookie is valid only at most for 2*60 seconds (or less if |
8c27bd75 FW |
479 | * the counter advances immediately after a cookie is generated). |
480 | */ | |
264ea103 ED |
481 | #define MAX_SYNCOOKIE_AGE 2 |
482 | #define TCP_SYNCOOKIE_PERIOD (60 * HZ) | |
483 | #define TCP_SYNCOOKIE_VALID (MAX_SYNCOOKIE_AGE * TCP_SYNCOOKIE_PERIOD) | |
484 | ||
485 | /* syncookies: remember time of last synqueue overflow | |
486 | * But do not dirty this field too often (once per second is enough) | |
3f684b4b | 487 | * It is racy as we do not hold a lock, but race is very minor. |
264ea103 | 488 | */ |
3f684b4b | 489 | static inline void tcp_synq_overflow(const struct sock *sk) |
264ea103 ED |
490 | { |
491 | unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp; | |
492 | unsigned long now = jiffies; | |
493 | ||
494 | if (time_after(now, last_overflow + HZ)) | |
495 | tcp_sk(sk)->rx_opt.ts_recent_stamp = now; | |
496 | } | |
497 | ||
498 | /* syncookies: no recent synqueue overflow on this listening socket? */ | |
499 | static inline bool tcp_synq_no_recent_overflow(const struct sock *sk) | |
500 | { | |
501 | unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp; | |
502 | ||
503 | return time_after(jiffies, last_overflow + TCP_SYNCOOKIE_VALID); | |
504 | } | |
8c27bd75 FW |
505 | |
506 | static inline u32 tcp_cookie_time(void) | |
507 | { | |
63262315 ED |
508 | u64 val = get_jiffies_64(); |
509 | ||
264ea103 | 510 | do_div(val, TCP_SYNCOOKIE_PERIOD); |
63262315 | 511 | return val; |
8c27bd75 FW |
512 | } |
513 | ||
5c9f3023 JP |
514 | u32 __cookie_v4_init_sequence(const struct iphdr *iph, const struct tcphdr *th, |
515 | u16 *mssp); | |
3f684b4b | 516 | __u32 cookie_v4_init_sequence(const struct sk_buff *skb, __u16 *mss); |
5c9f3023 | 517 | __u32 cookie_init_timestamp(struct request_sock *req); |
f1673381 FW |
518 | bool cookie_timestamp_decode(struct tcp_options_received *opt); |
519 | bool cookie_ecn_ok(const struct tcp_options_received *opt, | |
f7b3bec6 | 520 | const struct net *net, const struct dst_entry *dst); |
4dfc2817 | 521 | |
c6aefafb | 522 | /* From net/ipv6/syncookies.c */ |
5c9f3023 JP |
523 | int __cookie_v6_check(const struct ipv6hdr *iph, const struct tcphdr *th, |
524 | u32 cookie); | |
525 | struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb); | |
f1673381 | 526 | |
5c9f3023 JP |
527 | u32 __cookie_v6_init_sequence(const struct ipv6hdr *iph, |
528 | const struct tcphdr *th, u16 *mssp); | |
3f684b4b | 529 | __u32 cookie_v6_init_sequence(const struct sk_buff *skb, __u16 *mss); |
e05c82d3 | 530 | #endif |
1da177e4 LT |
531 | /* tcp_output.c */ |
532 | ||
5c9f3023 JP |
533 | void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss, |
534 | int nonagle); | |
535 | bool tcp_may_send_now(struct sock *sk); | |
536 | int __tcp_retransmit_skb(struct sock *, struct sk_buff *); | |
537 | int tcp_retransmit_skb(struct sock *, struct sk_buff *); | |
538 | void tcp_retransmit_timer(struct sock *sk); | |
539 | void tcp_xmit_retransmit_queue(struct sock *); | |
540 | void tcp_simple_retransmit(struct sock *); | |
541 | int tcp_trim_head(struct sock *, struct sk_buff *, u32); | |
6cc55e09 | 542 | int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int, gfp_t); |
5c9f3023 JP |
543 | |
544 | void tcp_send_probe0(struct sock *); | |
545 | void tcp_send_partial(struct sock *); | |
e520af48 | 546 | int tcp_write_wakeup(struct sock *, int mib); |
5c9f3023 JP |
547 | void tcp_send_fin(struct sock *sk); |
548 | void tcp_send_active_reset(struct sock *sk, gfp_t priority); | |
549 | int tcp_send_synack(struct sock *); | |
5c9f3023 JP |
550 | void tcp_push_one(struct sock *, unsigned int mss_now); |
551 | void tcp_send_ack(struct sock *sk); | |
552 | void tcp_send_delayed_ack(struct sock *sk); | |
553 | void tcp_send_loss_probe(struct sock *sk); | |
554 | bool tcp_schedule_loss_probe(struct sock *sk); | |
1da177e4 | 555 | |
a762a980 | 556 | /* tcp_input.c */ |
5c9f3023 JP |
557 | void tcp_resume_early_retransmit(struct sock *sk); |
558 | void tcp_rearm_rto(struct sock *sk); | |
0f1c28ae | 559 | void tcp_synack_rtt_meas(struct sock *sk, struct request_sock *req); |
5c9f3023 | 560 | void tcp_reset(struct sock *sk); |
4f41b1c5 | 561 | void tcp_skb_mark_lost_uncond_verify(struct tcp_sock *tp, struct sk_buff *skb); |
e3e17b77 | 562 | void tcp_fin(struct sock *sk); |
a762a980 | 563 | |
1da177e4 | 564 | /* tcp_timer.c */ |
5c9f3023 | 565 | void tcp_init_xmit_timers(struct sock *); |
463c84b9 ACM |
566 | static inline void tcp_clear_xmit_timers(struct sock *sk) |
567 | { | |
568 | inet_csk_clear_xmit_timers(sk); | |
569 | } | |
1da177e4 | 570 | |
5c9f3023 JP |
571 | unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu); |
572 | unsigned int tcp_current_mss(struct sock *sk); | |
0c54b85f IJ |
573 | |
574 | /* Bound MSS / TSO packet size with the half of the window */ | |
575 | static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize) | |
576 | { | |
01f83d69 AK |
577 | int cutoff; |
578 | ||
579 | /* When peer uses tiny windows, there is no use in packetizing | |
580 | * to sub-MSS pieces for the sake of SWS or making sure there | |
581 | * are enough packets in the pipe for fast recovery. | |
582 | * | |
583 | * On the other hand, for extremely large MSS devices, handling | |
584 | * smaller than MSS windows in this way does make sense. | |
585 | */ | |
586 | if (tp->max_window >= 512) | |
587 | cutoff = (tp->max_window >> 1); | |
588 | else | |
589 | cutoff = tp->max_window; | |
590 | ||
591 | if (cutoff && pktsize > cutoff) | |
592 | return max_t(int, cutoff, 68U - tp->tcp_header_len); | |
0c54b85f IJ |
593 | else |
594 | return pktsize; | |
595 | } | |
1da177e4 | 596 | |
17b085ea | 597 | /* tcp.c */ |
0df48c26 | 598 | void tcp_get_info(struct sock *, struct tcp_info *); |
1da177e4 LT |
599 | |
600 | /* Read 'sendfile()'-style from a TCP socket */ | |
601 | typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *, | |
602 | unsigned int, size_t); | |
5c9f3023 JP |
603 | int tcp_read_sock(struct sock *sk, read_descriptor_t *desc, |
604 | sk_read_actor_t recv_actor); | |
1da177e4 | 605 | |
5c9f3023 | 606 | void tcp_initialize_rcv_mss(struct sock *sk); |
1da177e4 | 607 | |
5c9f3023 JP |
608 | int tcp_mtu_to_mss(struct sock *sk, int pmtu); |
609 | int tcp_mss_to_mtu(struct sock *sk, int mss); | |
610 | void tcp_mtup_init(struct sock *sk); | |
611 | void tcp_init_buffer_space(struct sock *sk); | |
5d424d5a | 612 | |
f1ecd5d9 DL |
613 | static inline void tcp_bound_rto(const struct sock *sk) |
614 | { | |
615 | if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX) | |
616 | inet_csk(sk)->icsk_rto = TCP_RTO_MAX; | |
617 | } | |
618 | ||
619 | static inline u32 __tcp_set_rto(const struct tcp_sock *tp) | |
620 | { | |
740b0f18 | 621 | return usecs_to_jiffies((tp->srtt_us >> 3) + tp->rttvar_us); |
f1ecd5d9 DL |
622 | } |
623 | ||
40efc6fa | 624 | static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd) |
1da177e4 LT |
625 | { |
626 | tp->pred_flags = htonl((tp->tcp_header_len << 26) | | |
627 | ntohl(TCP_FLAG_ACK) | | |
628 | snd_wnd); | |
629 | } | |
630 | ||
40efc6fa | 631 | static inline void tcp_fast_path_on(struct tcp_sock *tp) |
1da177e4 LT |
632 | { |
633 | __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale); | |
634 | } | |
635 | ||
9e412ba7 | 636 | static inline void tcp_fast_path_check(struct sock *sk) |
1da177e4 | 637 | { |
9e412ba7 IJ |
638 | struct tcp_sock *tp = tcp_sk(sk); |
639 | ||
b03efcfb | 640 | if (skb_queue_empty(&tp->out_of_order_queue) && |
1da177e4 LT |
641 | tp->rcv_wnd && |
642 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf && | |
643 | !tp->urg_data) | |
644 | tcp_fast_path_on(tp); | |
645 | } | |
646 | ||
0c266898 SS |
647 | /* Compute the actual rto_min value */ |
648 | static inline u32 tcp_rto_min(struct sock *sk) | |
649 | { | |
cf533ea5 | 650 | const struct dst_entry *dst = __sk_dst_get(sk); |
0c266898 SS |
651 | u32 rto_min = TCP_RTO_MIN; |
652 | ||
653 | if (dst && dst_metric_locked(dst, RTAX_RTO_MIN)) | |
654 | rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN); | |
655 | return rto_min; | |
656 | } | |
657 | ||
740b0f18 ED |
658 | static inline u32 tcp_rto_min_us(struct sock *sk) |
659 | { | |
660 | return jiffies_to_usecs(tcp_rto_min(sk)); | |
661 | } | |
662 | ||
81164413 DB |
663 | static inline bool tcp_ca_dst_locked(const struct dst_entry *dst) |
664 | { | |
665 | return dst_metric_locked(dst, RTAX_CC_ALGO); | |
666 | } | |
667 | ||
f6722583 YC |
668 | /* Minimum RTT in usec. ~0 means not available. */ |
669 | static inline u32 tcp_min_rtt(const struct tcp_sock *tp) | |
670 | { | |
671 | return tp->rtt_min[0].rtt; | |
672 | } | |
673 | ||
1da177e4 LT |
674 | /* Compute the actual receive window we are currently advertising. |
675 | * Rcv_nxt can be after the window if our peer push more data | |
676 | * than the offered window. | |
677 | */ | |
40efc6fa | 678 | static inline u32 tcp_receive_window(const struct tcp_sock *tp) |
1da177e4 LT |
679 | { |
680 | s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt; | |
681 | ||
682 | if (win < 0) | |
683 | win = 0; | |
684 | return (u32) win; | |
685 | } | |
686 | ||
687 | /* Choose a new window, without checks for shrinking, and without | |
688 | * scaling applied to the result. The caller does these things | |
689 | * if necessary. This is a "raw" window selection. | |
690 | */ | |
5c9f3023 | 691 | u32 __tcp_select_window(struct sock *sk); |
1da177e4 | 692 | |
ee995283 PE |
693 | void tcp_send_window_probe(struct sock *sk); |
694 | ||
1da177e4 LT |
695 | /* TCP timestamps are only 32-bits, this causes a slight |
696 | * complication on 64-bit systems since we store a snapshot | |
31f34269 SH |
697 | * of jiffies in the buffer control blocks below. We decided |
698 | * to use only the low 32-bits of jiffies and hide the ugly | |
1da177e4 LT |
699 | * casts with the following macro. |
700 | */ | |
701 | #define tcp_time_stamp ((__u32)(jiffies)) | |
702 | ||
7faee5c0 ED |
703 | static inline u32 tcp_skb_timestamp(const struct sk_buff *skb) |
704 | { | |
705 | return skb->skb_mstamp.stamp_jiffies; | |
706 | } | |
707 | ||
708 | ||
a3433f35 CG |
709 | #define tcp_flag_byte(th) (((u_int8_t *)th)[13]) |
710 | ||
711 | #define TCPHDR_FIN 0x01 | |
712 | #define TCPHDR_SYN 0x02 | |
713 | #define TCPHDR_RST 0x04 | |
714 | #define TCPHDR_PSH 0x08 | |
715 | #define TCPHDR_ACK 0x10 | |
716 | #define TCPHDR_URG 0x20 | |
717 | #define TCPHDR_ECE 0x40 | |
718 | #define TCPHDR_CWR 0x80 | |
719 | ||
49213555 DB |
720 | #define TCPHDR_SYN_ECN (TCPHDR_SYN | TCPHDR_ECE | TCPHDR_CWR) |
721 | ||
caa20d9a | 722 | /* This is what the send packet queuing engine uses to pass |
f86586fa ED |
723 | * TCP per-packet control information to the transmission code. |
724 | * We also store the host-order sequence numbers in here too. | |
725 | * This is 44 bytes if IPV6 is enabled. | |
726 | * If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately. | |
1da177e4 LT |
727 | */ |
728 | struct tcp_skb_cb { | |
1da177e4 LT |
729 | __u32 seq; /* Starting sequence number */ |
730 | __u32 end_seq; /* SEQ + FIN + SYN + datalen */ | |
cd7d8498 ED |
731 | union { |
732 | /* Note : tcp_tw_isn is used in input path only | |
733 | * (isn chosen by tcp_timewait_state_process()) | |
734 | * | |
f69ad292 ED |
735 | * tcp_gso_segs/size are used in write queue only, |
736 | * cf tcp_skb_pcount()/tcp_skb_mss() | |
cd7d8498 ED |
737 | */ |
738 | __u32 tcp_tw_isn; | |
f69ad292 ED |
739 | struct { |
740 | u16 tcp_gso_segs; | |
741 | u16 tcp_gso_size; | |
742 | }; | |
cd7d8498 | 743 | }; |
4de075e0 | 744 | __u8 tcp_flags; /* TCP header flags. (tcp[13]) */ |
f4f9f6e7 | 745 | |
1da177e4 LT |
746 | __u8 sacked; /* State flags for SACK/FACK. */ |
747 | #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */ | |
748 | #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */ | |
749 | #define TCPCB_LOST 0x04 /* SKB is lost */ | |
750 | #define TCPCB_TAGBITS 0x07 /* All tag bits */ | |
9d186cac | 751 | #define TCPCB_REPAIRED 0x10 /* SKB repaired (no skb_mstamp) */ |
1da177e4 | 752 | #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */ |
9d186cac AV |
753 | #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS| \ |
754 | TCPCB_REPAIRED) | |
1da177e4 | 755 | |
f4f9f6e7 NC |
756 | __u8 ip_dsfield; /* IPv4 tos or IPv6 dsfield */ |
757 | /* 1 byte hole */ | |
1da177e4 | 758 | __u32 ack_seq; /* Sequence number ACK'd */ |
971f10ec ED |
759 | union { |
760 | struct inet_skb_parm h4; | |
761 | #if IS_ENABLED(CONFIG_IPV6) | |
762 | struct inet6_skb_parm h6; | |
763 | #endif | |
764 | } header; /* For incoming frames */ | |
1da177e4 LT |
765 | }; |
766 | ||
767 | #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0])) | |
768 | ||
870c3151 | 769 | |
815afe17 | 770 | #if IS_ENABLED(CONFIG_IPV6) |
870c3151 ED |
771 | /* This is the variant of inet6_iif() that must be used by TCP, |
772 | * as TCP moves IP6CB into a different location in skb->cb[] | |
773 | */ | |
774 | static inline int tcp_v6_iif(const struct sk_buff *skb) | |
775 | { | |
776 | return TCP_SKB_CB(skb)->header.h6.iif; | |
777 | } | |
815afe17 | 778 | #endif |
870c3151 | 779 | |
1da177e4 LT |
780 | /* Due to TSO, an SKB can be composed of multiple actual |
781 | * packets. To keep these tracked properly, we use this. | |
bd14b1b2 | 782 | */ |
1da177e4 | 783 | static inline int tcp_skb_pcount(const struct sk_buff *skb) |
bd14b1b2 | 784 | { |
cd7d8498 ED |
785 | return TCP_SKB_CB(skb)->tcp_gso_segs; |
786 | } | |
bd14b1b2 | 787 | |
cd7d8498 ED |
788 | static inline void tcp_skb_pcount_set(struct sk_buff *skb, int segs) |
789 | { | |
790 | TCP_SKB_CB(skb)->tcp_gso_segs = segs; | |
bd14b1b2 ED |
791 | } |
792 | ||
cd7d8498 | 793 | static inline void tcp_skb_pcount_add(struct sk_buff *skb, int segs) |
1da177e4 | 794 | { |
cd7d8498 | 795 | TCP_SKB_CB(skb)->tcp_gso_segs += segs; |
1da177e4 LT |
796 | } |
797 | ||
f69ad292 | 798 | /* This is valid iff skb is in write queue and tcp_skb_pcount() > 1. */ |
1da177e4 LT |
799 | static inline int tcp_skb_mss(const struct sk_buff *skb) |
800 | { | |
f69ad292 | 801 | return TCP_SKB_CB(skb)->tcp_gso_size; |
1da177e4 LT |
802 | } |
803 | ||
317a76f9 SH |
804 | /* Events passed to congestion control interface */ |
805 | enum tcp_ca_event { | |
806 | CA_EVENT_TX_START, /* first transmit when no packets in flight */ | |
807 | CA_EVENT_CWND_RESTART, /* congestion window restart */ | |
808 | CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */ | |
317a76f9 | 809 | CA_EVENT_LOSS, /* loss timeout */ |
9890092e FW |
810 | CA_EVENT_ECN_NO_CE, /* ECT set, but not CE marked */ |
811 | CA_EVENT_ECN_IS_CE, /* received CE marked IP packet */ | |
812 | CA_EVENT_DELAYED_ACK, /* Delayed ack is sent */ | |
813 | CA_EVENT_NON_DELAYED_ACK, | |
7354c8c3 FW |
814 | }; |
815 | ||
9890092e | 816 | /* Information about inbound ACK, passed to cong_ops->in_ack_event() */ |
7354c8c3 | 817 | enum tcp_ca_ack_event_flags { |
9890092e FW |
818 | CA_ACK_SLOWPATH = (1 << 0), /* In slow path processing */ |
819 | CA_ACK_WIN_UPDATE = (1 << 1), /* ACK updated window */ | |
820 | CA_ACK_ECE = (1 << 2), /* ECE bit is set on ack */ | |
317a76f9 SH |
821 | }; |
822 | ||
823 | /* | |
824 | * Interface for adding new TCP congestion control handlers | |
825 | */ | |
826 | #define TCP_CA_NAME_MAX 16 | |
3ff825b2 SH |
827 | #define TCP_CA_MAX 128 |
828 | #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX) | |
829 | ||
c5c6a8ab DB |
830 | #define TCP_CA_UNSPEC 0 |
831 | ||
30e502a3 | 832 | /* Algorithm can be set on socket without CAP_NET_ADMIN privileges */ |
164891aa | 833 | #define TCP_CONG_NON_RESTRICTED 0x1 |
30e502a3 DB |
834 | /* Requires ECN/ECT set on all packets */ |
835 | #define TCP_CONG_NEEDS_ECN 0x2 | |
164891aa | 836 | |
64f40ff5 ED |
837 | union tcp_cc_info; |
838 | ||
317a76f9 SH |
839 | struct tcp_congestion_ops { |
840 | struct list_head list; | |
c5c6a8ab DB |
841 | u32 key; |
842 | u32 flags; | |
317a76f9 SH |
843 | |
844 | /* initialize private data (optional) */ | |
6687e988 | 845 | void (*init)(struct sock *sk); |
317a76f9 | 846 | /* cleanup private data (optional) */ |
6687e988 | 847 | void (*release)(struct sock *sk); |
317a76f9 SH |
848 | |
849 | /* return slow start threshold (required) */ | |
6687e988 | 850 | u32 (*ssthresh)(struct sock *sk); |
317a76f9 | 851 | /* do new cwnd calculation (required) */ |
24901551 | 852 | void (*cong_avoid)(struct sock *sk, u32 ack, u32 acked); |
317a76f9 | 853 | /* call before changing ca_state (optional) */ |
6687e988 | 854 | void (*set_state)(struct sock *sk, u8 new_state); |
317a76f9 | 855 | /* call when cwnd event occurs (optional) */ |
6687e988 | 856 | void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev); |
7354c8c3 FW |
857 | /* call when ack arrives (optional) */ |
858 | void (*in_ack_event)(struct sock *sk, u32 flags); | |
317a76f9 | 859 | /* new value of cwnd after loss (optional) */ |
6687e988 | 860 | u32 (*undo_cwnd)(struct sock *sk); |
317a76f9 | 861 | /* hook for packet ack accounting (optional) */ |
30cfd0ba | 862 | void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us); |
73c1f4a0 | 863 | /* get info for inet_diag (optional) */ |
64f40ff5 ED |
864 | size_t (*get_info)(struct sock *sk, u32 ext, int *attr, |
865 | union tcp_cc_info *info); | |
317a76f9 SH |
866 | |
867 | char name[TCP_CA_NAME_MAX]; | |
868 | struct module *owner; | |
869 | }; | |
870 | ||
5c9f3023 JP |
871 | int tcp_register_congestion_control(struct tcp_congestion_ops *type); |
872 | void tcp_unregister_congestion_control(struct tcp_congestion_ops *type); | |
317a76f9 | 873 | |
55d8694f | 874 | void tcp_assign_congestion_control(struct sock *sk); |
5c9f3023 JP |
875 | void tcp_init_congestion_control(struct sock *sk); |
876 | void tcp_cleanup_congestion_control(struct sock *sk); | |
877 | int tcp_set_default_congestion_control(const char *name); | |
878 | void tcp_get_default_congestion_control(char *name); | |
879 | void tcp_get_available_congestion_control(char *buf, size_t len); | |
880 | void tcp_get_allowed_congestion_control(char *buf, size_t len); | |
881 | int tcp_set_allowed_congestion_control(char *allowed); | |
882 | int tcp_set_congestion_control(struct sock *sk, const char *name); | |
e73ebb08 NC |
883 | u32 tcp_slow_start(struct tcp_sock *tp, u32 acked); |
884 | void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w, u32 acked); | |
317a76f9 | 885 | |
5c9f3023 | 886 | u32 tcp_reno_ssthresh(struct sock *sk); |
24901551 | 887 | void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 acked); |
a8acfbac | 888 | extern struct tcp_congestion_ops tcp_reno; |
317a76f9 | 889 | |
c5c6a8ab | 890 | struct tcp_congestion_ops *tcp_ca_find_key(u32 key); |
c3a8d947 | 891 | u32 tcp_ca_get_key_by_name(const char *name, bool *ecn_ca); |
ea697639 | 892 | #ifdef CONFIG_INET |
c5c6a8ab | 893 | char *tcp_ca_get_name_by_key(u32 key, char *buffer); |
ea697639 DB |
894 | #else |
895 | static inline char *tcp_ca_get_name_by_key(u32 key, char *buffer) | |
896 | { | |
897 | return NULL; | |
898 | } | |
899 | #endif | |
c5c6a8ab | 900 | |
30e502a3 DB |
901 | static inline bool tcp_ca_needs_ecn(const struct sock *sk) |
902 | { | |
903 | const struct inet_connection_sock *icsk = inet_csk(sk); | |
904 | ||
905 | return icsk->icsk_ca_ops->flags & TCP_CONG_NEEDS_ECN; | |
906 | } | |
907 | ||
6687e988 | 908 | static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state) |
317a76f9 | 909 | { |
6687e988 ACM |
910 | struct inet_connection_sock *icsk = inet_csk(sk); |
911 | ||
912 | if (icsk->icsk_ca_ops->set_state) | |
913 | icsk->icsk_ca_ops->set_state(sk, ca_state); | |
914 | icsk->icsk_ca_state = ca_state; | |
317a76f9 SH |
915 | } |
916 | ||
6687e988 | 917 | static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event) |
317a76f9 | 918 | { |
6687e988 ACM |
919 | const struct inet_connection_sock *icsk = inet_csk(sk); |
920 | ||
921 | if (icsk->icsk_ca_ops->cwnd_event) | |
922 | icsk->icsk_ca_ops->cwnd_event(sk, event); | |
317a76f9 SH |
923 | } |
924 | ||
e60402d0 IJ |
925 | /* These functions determine how the current flow behaves in respect of SACK |
926 | * handling. SACK is negotiated with the peer, and therefore it can vary | |
927 | * between different flows. | |
928 | * | |
929 | * tcp_is_sack - SACK enabled | |
930 | * tcp_is_reno - No SACK | |
931 | * tcp_is_fack - FACK enabled, implies SACK enabled | |
932 | */ | |
933 | static inline int tcp_is_sack(const struct tcp_sock *tp) | |
934 | { | |
935 | return tp->rx_opt.sack_ok; | |
936 | } | |
937 | ||
a2a385d6 | 938 | static inline bool tcp_is_reno(const struct tcp_sock *tp) |
e60402d0 IJ |
939 | { |
940 | return !tcp_is_sack(tp); | |
941 | } | |
942 | ||
a2a385d6 | 943 | static inline bool tcp_is_fack(const struct tcp_sock *tp) |
e60402d0 | 944 | { |
ab56222a | 945 | return tp->rx_opt.sack_ok & TCP_FACK_ENABLED; |
e60402d0 IJ |
946 | } |
947 | ||
948 | static inline void tcp_enable_fack(struct tcp_sock *tp) | |
949 | { | |
ab56222a | 950 | tp->rx_opt.sack_ok |= TCP_FACK_ENABLED; |
e60402d0 IJ |
951 | } |
952 | ||
eed530b6 YC |
953 | /* TCP early-retransmit (ER) is similar to but more conservative than |
954 | * the thin-dupack feature. Enable ER only if thin-dupack is disabled. | |
955 | */ | |
956 | static inline void tcp_enable_early_retrans(struct tcp_sock *tp) | |
957 | { | |
1043e25f NB |
958 | struct net *net = sock_net((struct sock *)tp); |
959 | ||
eed530b6 | 960 | tp->do_early_retrans = sysctl_tcp_early_retrans && |
6ba8a3b1 | 961 | sysctl_tcp_early_retrans < 4 && !sysctl_tcp_thin_dupack && |
1043e25f | 962 | net->ipv4.sysctl_tcp_reordering == 3; |
eed530b6 YC |
963 | } |
964 | ||
965 | static inline void tcp_disable_early_retrans(struct tcp_sock *tp) | |
966 | { | |
967 | tp->do_early_retrans = 0; | |
968 | } | |
969 | ||
83ae4088 IJ |
970 | static inline unsigned int tcp_left_out(const struct tcp_sock *tp) |
971 | { | |
972 | return tp->sacked_out + tp->lost_out; | |
973 | } | |
974 | ||
1da177e4 LT |
975 | /* This determines how many packets are "in the network" to the best |
976 | * of our knowledge. In many cases it is conservative, but where | |
977 | * detailed information is available from the receiver (via SACK | |
978 | * blocks etc.) we can make more aggressive calculations. | |
979 | * | |
980 | * Use this for decisions involving congestion control, use just | |
981 | * tp->packets_out to determine if the send queue is empty or not. | |
982 | * | |
983 | * Read this equation as: | |
984 | * | |
985 | * "Packets sent once on transmission queue" MINUS | |
986 | * "Packets left network, but not honestly ACKed yet" PLUS | |
987 | * "Packets fast retransmitted" | |
988 | */ | |
40efc6fa | 989 | static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp) |
1da177e4 | 990 | { |
83ae4088 | 991 | return tp->packets_out - tcp_left_out(tp) + tp->retrans_out; |
1da177e4 LT |
992 | } |
993 | ||
0b6a05c1 IJ |
994 | #define TCP_INFINITE_SSTHRESH 0x7fffffff |
995 | ||
071d5080 YC |
996 | static inline bool tcp_in_slow_start(const struct tcp_sock *tp) |
997 | { | |
76174004 | 998 | return tp->snd_cwnd < tp->snd_ssthresh; |
071d5080 YC |
999 | } |
1000 | ||
0b6a05c1 IJ |
1001 | static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp) |
1002 | { | |
1003 | return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH; | |
1004 | } | |
1005 | ||
684bad11 YC |
1006 | static inline bool tcp_in_cwnd_reduction(const struct sock *sk) |
1007 | { | |
1008 | return (TCPF_CA_CWR | TCPF_CA_Recovery) & | |
1009 | (1 << inet_csk(sk)->icsk_ca_state); | |
1010 | } | |
1011 | ||
1da177e4 | 1012 | /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd. |
684bad11 | 1013 | * The exception is cwnd reduction phase, when cwnd is decreasing towards |
1da177e4 LT |
1014 | * ssthresh. |
1015 | */ | |
6687e988 | 1016 | static inline __u32 tcp_current_ssthresh(const struct sock *sk) |
1da177e4 | 1017 | { |
6687e988 | 1018 | const struct tcp_sock *tp = tcp_sk(sk); |
cf533ea5 | 1019 | |
684bad11 | 1020 | if (tcp_in_cwnd_reduction(sk)) |
1da177e4 LT |
1021 | return tp->snd_ssthresh; |
1022 | else | |
1023 | return max(tp->snd_ssthresh, | |
1024 | ((tp->snd_cwnd >> 1) + | |
1025 | (tp->snd_cwnd >> 2))); | |
1026 | } | |
1027 | ||
b9c4595b IJ |
1028 | /* Use define here intentionally to get WARN_ON location shown at the caller */ |
1029 | #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out) | |
1da177e4 | 1030 | |
5ee2c941 | 1031 | void tcp_enter_cwr(struct sock *sk); |
5c9f3023 | 1032 | __u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst); |
1da177e4 | 1033 | |
6b5a5c0d NC |
1034 | /* The maximum number of MSS of available cwnd for which TSO defers |
1035 | * sending if not using sysctl_tcp_tso_win_divisor. | |
1036 | */ | |
1037 | static inline __u32 tcp_max_tso_deferred_mss(const struct tcp_sock *tp) | |
1038 | { | |
1039 | return 3; | |
1040 | } | |
1041 | ||
1da177e4 | 1042 | /* Slow start with delack produces 3 packets of burst, so that |
dd9e0dda JH |
1043 | * it is safe "de facto". This will be the default - same as |
1044 | * the default reordering threshold - but if reordering increases, | |
1045 | * we must be able to allow cwnd to burst at least this much in order | |
1046 | * to not pull it back when holes are filled. | |
1da177e4 LT |
1047 | */ |
1048 | static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp) | |
1049 | { | |
dd9e0dda | 1050 | return tp->reordering; |
1da177e4 LT |
1051 | } |
1052 | ||
90840def IJ |
1053 | /* Returns end sequence number of the receiver's advertised window */ |
1054 | static inline u32 tcp_wnd_end(const struct tcp_sock *tp) | |
1055 | { | |
1056 | return tp->snd_una + tp->snd_wnd; | |
1057 | } | |
e114a710 ED |
1058 | |
1059 | /* We follow the spirit of RFC2861 to validate cwnd but implement a more | |
1060 | * flexible approach. The RFC suggests cwnd should not be raised unless | |
ca8a2263 NC |
1061 | * it was fully used previously. And that's exactly what we do in |
1062 | * congestion avoidance mode. But in slow start we allow cwnd to grow | |
1063 | * as long as the application has used half the cwnd. | |
e114a710 ED |
1064 | * Example : |
1065 | * cwnd is 10 (IW10), but application sends 9 frames. | |
1066 | * We allow cwnd to reach 18 when all frames are ACKed. | |
1067 | * This check is safe because it's as aggressive as slow start which already | |
1068 | * risks 100% overshoot. The advantage is that we discourage application to | |
1069 | * either send more filler packets or data to artificially blow up the cwnd | |
1070 | * usage, and allow application-limited process to probe bw more aggressively. | |
e114a710 | 1071 | */ |
24901551 | 1072 | static inline bool tcp_is_cwnd_limited(const struct sock *sk) |
e114a710 ED |
1073 | { |
1074 | const struct tcp_sock *tp = tcp_sk(sk); | |
1075 | ||
ca8a2263 | 1076 | /* If in slow start, ensure cwnd grows to twice what was ACKed. */ |
071d5080 | 1077 | if (tcp_in_slow_start(tp)) |
ca8a2263 NC |
1078 | return tp->snd_cwnd < 2 * tp->max_packets_out; |
1079 | ||
1080 | return tp->is_cwnd_limited; | |
e114a710 | 1081 | } |
f4805ede | 1082 | |
21c8fe99 ED |
1083 | /* Something is really bad, we could not queue an additional packet, |
1084 | * because qdisc is full or receiver sent a 0 window. | |
1085 | * We do not want to add fuel to the fire, or abort too early, | |
1086 | * so make sure the timer we arm now is at least 200ms in the future, | |
1087 | * regardless of current icsk_rto value (as it could be ~2ms) | |
1088 | */ | |
1089 | static inline unsigned long tcp_probe0_base(const struct sock *sk) | |
1da177e4 | 1090 | { |
21c8fe99 ED |
1091 | return max_t(unsigned long, inet_csk(sk)->icsk_rto, TCP_RTO_MIN); |
1092 | } | |
9e412ba7 | 1093 | |
21c8fe99 ED |
1094 | /* Variant of inet_csk_rto_backoff() used for zero window probes */ |
1095 | static inline unsigned long tcp_probe0_when(const struct sock *sk, | |
1096 | unsigned long max_when) | |
1097 | { | |
1098 | u64 when = (u64)tcp_probe0_base(sk) << inet_csk(sk)->icsk_backoff; | |
1099 | ||
1100 | return (unsigned long)min_t(u64, when, max_when); | |
1101 | } | |
1102 | ||
1103 | static inline void tcp_check_probe_timer(struct sock *sk) | |
1104 | { | |
1105 | if (!tcp_sk(sk)->packets_out && !inet_csk(sk)->icsk_pending) | |
3f421baa | 1106 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, |
21c8fe99 | 1107 | tcp_probe0_base(sk), TCP_RTO_MAX); |
1da177e4 LT |
1108 | } |
1109 | ||
ee7537b6 | 1110 | static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq) |
1da177e4 LT |
1111 | { |
1112 | tp->snd_wl1 = seq; | |
1113 | } | |
1114 | ||
ee7537b6 | 1115 | static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq) |
1da177e4 LT |
1116 | { |
1117 | tp->snd_wl1 = seq; | |
1118 | } | |
1119 | ||
1da177e4 LT |
1120 | /* |
1121 | * Calculate(/check) TCP checksum | |
1122 | */ | |
ba7808ea FD |
1123 | static inline __sum16 tcp_v4_check(int len, __be32 saddr, |
1124 | __be32 daddr, __wsum base) | |
1da177e4 LT |
1125 | { |
1126 | return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base); | |
1127 | } | |
1128 | ||
b51655b9 | 1129 | static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 1130 | { |
fb286bb2 | 1131 | return __skb_checksum_complete(skb); |
1da177e4 LT |
1132 | } |
1133 | ||
a2a385d6 | 1134 | static inline bool tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 1135 | { |
60476372 | 1136 | return !skb_csum_unnecessary(skb) && |
1da177e4 LT |
1137 | __tcp_checksum_complete(skb); |
1138 | } | |
1139 | ||
1140 | /* Prequeue for VJ style copy to user, combined with checksumming. */ | |
1141 | ||
40efc6fa | 1142 | static inline void tcp_prequeue_init(struct tcp_sock *tp) |
1da177e4 LT |
1143 | { |
1144 | tp->ucopy.task = NULL; | |
1145 | tp->ucopy.len = 0; | |
1146 | tp->ucopy.memory = 0; | |
1147 | skb_queue_head_init(&tp->ucopy.prequeue); | |
1148 | } | |
1149 | ||
5c9f3023 | 1150 | bool tcp_prequeue(struct sock *sk, struct sk_buff *skb); |
1da177e4 LT |
1151 | |
1152 | #undef STATE_TRACE | |
1153 | ||
1154 | #ifdef STATE_TRACE | |
1155 | static const char *statename[]={ | |
1156 | "Unused","Established","Syn Sent","Syn Recv", | |
1157 | "Fin Wait 1","Fin Wait 2","Time Wait", "Close", | |
1158 | "Close Wait","Last ACK","Listen","Closing" | |
1159 | }; | |
1160 | #endif | |
5c9f3023 | 1161 | void tcp_set_state(struct sock *sk, int state); |
1da177e4 | 1162 | |
5c9f3023 | 1163 | void tcp_done(struct sock *sk); |
1da177e4 | 1164 | |
c1e64e29 LC |
1165 | int tcp_abort(struct sock *sk, int err); |
1166 | ||
40efc6fa | 1167 | static inline void tcp_sack_reset(struct tcp_options_received *rx_opt) |
1da177e4 LT |
1168 | { |
1169 | rx_opt->dsack = 0; | |
1da177e4 LT |
1170 | rx_opt->num_sacks = 0; |
1171 | } | |
1172 | ||
5c9f3023 | 1173 | u32 tcp_default_init_rwnd(u32 mss); |
6f021c62 ED |
1174 | void tcp_cwnd_restart(struct sock *sk, s32 delta); |
1175 | ||
1176 | static inline void tcp_slow_start_after_idle_check(struct sock *sk) | |
1177 | { | |
1178 | struct tcp_sock *tp = tcp_sk(sk); | |
1179 | s32 delta; | |
1180 | ||
1181 | if (!sysctl_tcp_slow_start_after_idle || tp->packets_out) | |
1182 | return; | |
1183 | delta = tcp_time_stamp - tp->lsndtime; | |
1184 | if (delta > inet_csk(sk)->icsk_rto) | |
1185 | tcp_cwnd_restart(sk, delta); | |
1186 | } | |
85f16525 | 1187 | |
1da177e4 | 1188 | /* Determine a window scaling and initial window to offer. */ |
5c9f3023 JP |
1189 | void tcp_select_initial_window(int __space, __u32 mss, __u32 *rcv_wnd, |
1190 | __u32 *window_clamp, int wscale_ok, | |
1191 | __u8 *rcv_wscale, __u32 init_rcv_wnd); | |
1da177e4 LT |
1192 | |
1193 | static inline int tcp_win_from_space(int space) | |
1194 | { | |
1195 | return sysctl_tcp_adv_win_scale<=0 ? | |
1196 | (space>>(-sysctl_tcp_adv_win_scale)) : | |
1197 | space - (space>>sysctl_tcp_adv_win_scale); | |
1198 | } | |
1199 | ||
105970f6 | 1200 | /* Note: caller must be prepared to deal with negative returns */ |
1da177e4 LT |
1201 | static inline int tcp_space(const struct sock *sk) |
1202 | { | |
1203 | return tcp_win_from_space(sk->sk_rcvbuf - | |
1204 | atomic_read(&sk->sk_rmem_alloc)); | |
105970f6 | 1205 | } |
1da177e4 LT |
1206 | |
1207 | static inline int tcp_full_space(const struct sock *sk) | |
1208 | { | |
105970f6 | 1209 | return tcp_win_from_space(sk->sk_rcvbuf); |
1da177e4 LT |
1210 | } |
1211 | ||
843f4a55 | 1212 | extern void tcp_openreq_init_rwin(struct request_sock *req, |
b1964b5f ED |
1213 | const struct sock *sk_listener, |
1214 | const struct dst_entry *dst); | |
843f4a55 | 1215 | |
5c9f3023 | 1216 | void tcp_enter_memory_pressure(struct sock *sk); |
1da177e4 | 1217 | |
1da177e4 LT |
1218 | static inline int keepalive_intvl_when(const struct tcp_sock *tp) |
1219 | { | |
b840d15d NB |
1220 | struct net *net = sock_net((struct sock *)tp); |
1221 | ||
1222 | return tp->keepalive_intvl ? : net->ipv4.sysctl_tcp_keepalive_intvl; | |
1da177e4 LT |
1223 | } |
1224 | ||
1225 | static inline int keepalive_time_when(const struct tcp_sock *tp) | |
1226 | { | |
13b287e8 NB |
1227 | struct net *net = sock_net((struct sock *)tp); |
1228 | ||
1229 | return tp->keepalive_time ? : net->ipv4.sysctl_tcp_keepalive_time; | |
1da177e4 LT |
1230 | } |
1231 | ||
df19a626 ED |
1232 | static inline int keepalive_probes(const struct tcp_sock *tp) |
1233 | { | |
9bd6861b NB |
1234 | struct net *net = sock_net((struct sock *)tp); |
1235 | ||
1236 | return tp->keepalive_probes ? : net->ipv4.sysctl_tcp_keepalive_probes; | |
df19a626 ED |
1237 | } |
1238 | ||
6c37e5de FL |
1239 | static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp) |
1240 | { | |
1241 | const struct inet_connection_sock *icsk = &tp->inet_conn; | |
1242 | ||
1243 | return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime, | |
1244 | tcp_time_stamp - tp->rcv_tstamp); | |
1245 | } | |
1246 | ||
463c84b9 | 1247 | static inline int tcp_fin_time(const struct sock *sk) |
1da177e4 | 1248 | { |
1e579caa | 1249 | int fin_timeout = tcp_sk(sk)->linger2 ? : sock_net(sk)->ipv4.sysctl_tcp_fin_timeout; |
463c84b9 | 1250 | const int rto = inet_csk(sk)->icsk_rto; |
1da177e4 | 1251 | |
463c84b9 ACM |
1252 | if (fin_timeout < (rto << 2) - (rto >> 1)) |
1253 | fin_timeout = (rto << 2) - (rto >> 1); | |
1da177e4 LT |
1254 | |
1255 | return fin_timeout; | |
1256 | } | |
1257 | ||
a2a385d6 ED |
1258 | static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt, |
1259 | int paws_win) | |
1da177e4 | 1260 | { |
c887e6d2 | 1261 | if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win) |
a2a385d6 | 1262 | return true; |
c887e6d2 | 1263 | if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)) |
a2a385d6 | 1264 | return true; |
bc2ce894 ED |
1265 | /* |
1266 | * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0, | |
1267 | * then following tcp messages have valid values. Ignore 0 value, | |
1268 | * or else 'negative' tsval might forbid us to accept their packets. | |
1269 | */ | |
1270 | if (!rx_opt->ts_recent) | |
a2a385d6 ED |
1271 | return true; |
1272 | return false; | |
c887e6d2 IJ |
1273 | } |
1274 | ||
a2a385d6 ED |
1275 | static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt, |
1276 | int rst) | |
c887e6d2 IJ |
1277 | { |
1278 | if (tcp_paws_check(rx_opt, 0)) | |
a2a385d6 | 1279 | return false; |
1da177e4 LT |
1280 | |
1281 | /* RST segments are not recommended to carry timestamp, | |
1282 | and, if they do, it is recommended to ignore PAWS because | |
1283 | "their cleanup function should take precedence over timestamps." | |
1284 | Certainly, it is mistake. It is necessary to understand the reasons | |
1285 | of this constraint to relax it: if peer reboots, clock may go | |
1286 | out-of-sync and half-open connections will not be reset. | |
1287 | Actually, the problem would be not existing if all | |
1288 | the implementations followed draft about maintaining clock | |
1289 | via reboots. Linux-2.2 DOES NOT! | |
1290 | ||
1291 | However, we can relax time bounds for RST segments to MSL. | |
1292 | */ | |
9d729f72 | 1293 | if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL) |
a2a385d6 ED |
1294 | return false; |
1295 | return true; | |
1da177e4 LT |
1296 | } |
1297 | ||
7970ddc8 ED |
1298 | bool tcp_oow_rate_limited(struct net *net, const struct sk_buff *skb, |
1299 | int mib_idx, u32 *last_oow_ack_time); | |
032ee423 | 1300 | |
a9c19329 | 1301 | static inline void tcp_mib_init(struct net *net) |
1da177e4 LT |
1302 | { |
1303 | /* See RFC 2012 */ | |
cf1100a7 PE |
1304 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1); |
1305 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ); | |
1306 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ); | |
1307 | TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1); | |
1da177e4 LT |
1308 | } |
1309 | ||
5af4ec23 | 1310 | /* from STCP */ |
ef9da47c | 1311 | static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp) |
0800f170 | 1312 | { |
6a438bbe | 1313 | tp->lost_skb_hint = NULL; |
ef9da47c IJ |
1314 | } |
1315 | ||
1316 | static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp) | |
1317 | { | |
1318 | tcp_clear_retrans_hints_partial(tp); | |
6a438bbe | 1319 | tp->retransmit_skb_hint = NULL; |
b7689205 IJ |
1320 | } |
1321 | ||
a915da9b ED |
1322 | union tcp_md5_addr { |
1323 | struct in_addr a4; | |
1324 | #if IS_ENABLED(CONFIG_IPV6) | |
1325 | struct in6_addr a6; | |
1326 | #endif | |
1327 | }; | |
1328 | ||
cfb6eeb4 YH |
1329 | /* - key database */ |
1330 | struct tcp_md5sig_key { | |
a915da9b | 1331 | struct hlist_node node; |
cfb6eeb4 | 1332 | u8 keylen; |
a915da9b ED |
1333 | u8 family; /* AF_INET or AF_INET6 */ |
1334 | union tcp_md5_addr addr; | |
1335 | u8 key[TCP_MD5SIG_MAXKEYLEN]; | |
1336 | struct rcu_head rcu; | |
cfb6eeb4 YH |
1337 | }; |
1338 | ||
1339 | /* - sock block */ | |
1340 | struct tcp_md5sig_info { | |
a915da9b | 1341 | struct hlist_head head; |
a8afca03 | 1342 | struct rcu_head rcu; |
cfb6eeb4 YH |
1343 | }; |
1344 | ||
1345 | /* - pseudo header */ | |
1346 | struct tcp4_pseudohdr { | |
1347 | __be32 saddr; | |
1348 | __be32 daddr; | |
1349 | __u8 pad; | |
1350 | __u8 protocol; | |
1351 | __be16 len; | |
1352 | }; | |
1353 | ||
1354 | struct tcp6_pseudohdr { | |
1355 | struct in6_addr saddr; | |
1356 | struct in6_addr daddr; | |
1357 | __be32 len; | |
1358 | __be32 protocol; /* including padding */ | |
1359 | }; | |
1360 | ||
1361 | union tcp_md5sum_block { | |
1362 | struct tcp4_pseudohdr ip4; | |
dfd56b8b | 1363 | #if IS_ENABLED(CONFIG_IPV6) |
cfb6eeb4 YH |
1364 | struct tcp6_pseudohdr ip6; |
1365 | #endif | |
1366 | }; | |
1367 | ||
1368 | /* - pool: digest algorithm, hash description and scratch buffer */ | |
1369 | struct tcp_md5sig_pool { | |
cf80e0e4 | 1370 | struct ahash_request *md5_req; |
cfb6eeb4 YH |
1371 | union tcp_md5sum_block md5_blk; |
1372 | }; | |
1373 | ||
cfb6eeb4 | 1374 | /* - functions */ |
39f8e58e ED |
1375 | int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key, |
1376 | const struct sock *sk, const struct sk_buff *skb); | |
5c9f3023 JP |
1377 | int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr, |
1378 | int family, const u8 *newkey, u8 newkeylen, gfp_t gfp); | |
1379 | int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, | |
1380 | int family); | |
b83e3deb | 1381 | struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk, |
fd3a154a | 1382 | const struct sock *addr_sk); |
cfb6eeb4 | 1383 | |
9501f972 | 1384 | #ifdef CONFIG_TCP_MD5SIG |
b83e3deb | 1385 | struct tcp_md5sig_key *tcp_md5_do_lookup(const struct sock *sk, |
5c9f3023 JP |
1386 | const union tcp_md5_addr *addr, |
1387 | int family); | |
a915da9b | 1388 | #define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_key) |
9501f972 | 1389 | #else |
b83e3deb | 1390 | static inline struct tcp_md5sig_key *tcp_md5_do_lookup(const struct sock *sk, |
a915da9b ED |
1391 | const union tcp_md5_addr *addr, |
1392 | int family) | |
1393 | { | |
1394 | return NULL; | |
1395 | } | |
9501f972 YH |
1396 | #define tcp_twsk_md5_key(twsk) NULL |
1397 | #endif | |
1398 | ||
5c9f3023 | 1399 | bool tcp_alloc_md5sig_pool(void); |
cfb6eeb4 | 1400 | |
5c9f3023 | 1401 | struct tcp_md5sig_pool *tcp_get_md5sig_pool(void); |
71cea17e ED |
1402 | static inline void tcp_put_md5sig_pool(void) |
1403 | { | |
1404 | local_bh_enable(); | |
1405 | } | |
35790c04 | 1406 | |
5c9f3023 JP |
1407 | int tcp_md5_hash_header(struct tcp_md5sig_pool *, const struct tcphdr *); |
1408 | int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *, | |
1409 | unsigned int header_len); | |
1410 | int tcp_md5_hash_key(struct tcp_md5sig_pool *hp, | |
1411 | const struct tcp_md5sig_key *key); | |
cfb6eeb4 | 1412 | |
10467163 | 1413 | /* From tcp_fastopen.c */ |
5c9f3023 JP |
1414 | void tcp_fastopen_cache_get(struct sock *sk, u16 *mss, |
1415 | struct tcp_fastopen_cookie *cookie, int *syn_loss, | |
1416 | unsigned long *last_syn_loss); | |
1417 | void tcp_fastopen_cache_set(struct sock *sk, u16 mss, | |
2646c831 DL |
1418 | struct tcp_fastopen_cookie *cookie, bool syn_lost, |
1419 | u16 try_exp); | |
783237e8 YC |
1420 | struct tcp_fastopen_request { |
1421 | /* Fast Open cookie. Size 0 means a cookie request */ | |
1422 | struct tcp_fastopen_cookie cookie; | |
1423 | struct msghdr *data; /* data in MSG_FASTOPEN */ | |
f5ddcbbb ED |
1424 | size_t size; |
1425 | int copied; /* queued in tcp_connect() */ | |
783237e8 | 1426 | }; |
783237e8 YC |
1427 | void tcp_free_fastopen_req(struct tcp_sock *tp); |
1428 | ||
10467163 JC |
1429 | extern struct tcp_fastopen_context __rcu *tcp_fastopen_ctx; |
1430 | int tcp_fastopen_reset_cipher(void *key, unsigned int len); | |
61d2bcae | 1431 | void tcp_fastopen_add_skb(struct sock *sk, struct sk_buff *skb); |
7c85af88 ED |
1432 | struct sock *tcp_try_fastopen(struct sock *sk, struct sk_buff *skb, |
1433 | struct request_sock *req, | |
1434 | struct tcp_fastopen_cookie *foc, | |
1435 | struct dst_entry *dst); | |
222e83d2 | 1436 | void tcp_fastopen_init_key_once(bool publish); |
10467163 JC |
1437 | #define TCP_FASTOPEN_KEY_LENGTH 16 |
1438 | ||
1439 | /* Fastopen key context */ | |
1440 | struct tcp_fastopen_context { | |
7ae8639c ED |
1441 | struct crypto_cipher *tfm; |
1442 | __u8 key[TCP_FASTOPEN_KEY_LENGTH]; | |
1443 | struct rcu_head rcu; | |
10467163 JC |
1444 | }; |
1445 | ||
fe067e8a DM |
1446 | /* write queue abstraction */ |
1447 | static inline void tcp_write_queue_purge(struct sock *sk) | |
1448 | { | |
1449 | struct sk_buff *skb; | |
1450 | ||
1451 | while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) | |
3ab224be HA |
1452 | sk_wmem_free_skb(sk, skb); |
1453 | sk_mem_reclaim(sk); | |
8818a9d8 | 1454 | tcp_clear_all_retrans_hints(tcp_sk(sk)); |
fe067e8a DM |
1455 | } |
1456 | ||
cf533ea5 | 1457 | static inline struct sk_buff *tcp_write_queue_head(const struct sock *sk) |
fe067e8a | 1458 | { |
cd07a8ea | 1459 | return skb_peek(&sk->sk_write_queue); |
fe067e8a DM |
1460 | } |
1461 | ||
cf533ea5 | 1462 | static inline struct sk_buff *tcp_write_queue_tail(const struct sock *sk) |
fe067e8a | 1463 | { |
cd07a8ea | 1464 | return skb_peek_tail(&sk->sk_write_queue); |
fe067e8a DM |
1465 | } |
1466 | ||
cf533ea5 ED |
1467 | static inline struct sk_buff *tcp_write_queue_next(const struct sock *sk, |
1468 | const struct sk_buff *skb) | |
fe067e8a | 1469 | { |
cd07a8ea | 1470 | return skb_queue_next(&sk->sk_write_queue, skb); |
fe067e8a DM |
1471 | } |
1472 | ||
cf533ea5 ED |
1473 | static inline struct sk_buff *tcp_write_queue_prev(const struct sock *sk, |
1474 | const struct sk_buff *skb) | |
832d11c5 IJ |
1475 | { |
1476 | return skb_queue_prev(&sk->sk_write_queue, skb); | |
1477 | } | |
1478 | ||
fe067e8a | 1479 | #define tcp_for_write_queue(skb, sk) \ |
cd07a8ea | 1480 | skb_queue_walk(&(sk)->sk_write_queue, skb) |
fe067e8a DM |
1481 | |
1482 | #define tcp_for_write_queue_from(skb, sk) \ | |
cd07a8ea | 1483 | skb_queue_walk_from(&(sk)->sk_write_queue, skb) |
fe067e8a | 1484 | |
234b6860 | 1485 | #define tcp_for_write_queue_from_safe(skb, tmp, sk) \ |
cd07a8ea | 1486 | skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp) |
234b6860 | 1487 | |
cf533ea5 | 1488 | static inline struct sk_buff *tcp_send_head(const struct sock *sk) |
fe067e8a DM |
1489 | { |
1490 | return sk->sk_send_head; | |
1491 | } | |
1492 | ||
cd07a8ea DM |
1493 | static inline bool tcp_skb_is_last(const struct sock *sk, |
1494 | const struct sk_buff *skb) | |
1495 | { | |
1496 | return skb_queue_is_last(&sk->sk_write_queue, skb); | |
1497 | } | |
1498 | ||
cf533ea5 | 1499 | static inline void tcp_advance_send_head(struct sock *sk, const struct sk_buff *skb) |
fe067e8a | 1500 | { |
cd07a8ea | 1501 | if (tcp_skb_is_last(sk, skb)) |
fe067e8a | 1502 | sk->sk_send_head = NULL; |
cd07a8ea DM |
1503 | else |
1504 | sk->sk_send_head = tcp_write_queue_next(sk, skb); | |
fe067e8a DM |
1505 | } |
1506 | ||
1507 | static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked) | |
1508 | { | |
1509 | if (sk->sk_send_head == skb_unlinked) | |
1510 | sk->sk_send_head = NULL; | |
1511 | } | |
1512 | ||
1513 | static inline void tcp_init_send_head(struct sock *sk) | |
1514 | { | |
1515 | sk->sk_send_head = NULL; | |
1516 | } | |
1517 | ||
1518 | static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1519 | { | |
1520 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1521 | } | |
1522 | ||
1523 | static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1524 | { | |
1525 | __tcp_add_write_queue_tail(sk, skb); | |
1526 | ||
1527 | /* Queue it, remembering where we must start sending. */ | |
6859d494 | 1528 | if (sk->sk_send_head == NULL) { |
fe067e8a | 1529 | sk->sk_send_head = skb; |
6859d494 IJ |
1530 | |
1531 | if (tcp_sk(sk)->highest_sack == NULL) | |
1532 | tcp_sk(sk)->highest_sack = skb; | |
1533 | } | |
fe067e8a DM |
1534 | } |
1535 | ||
1536 | static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb) | |
1537 | { | |
1538 | __skb_queue_head(&sk->sk_write_queue, skb); | |
1539 | } | |
1540 | ||
1541 | /* Insert buff after skb on the write queue of sk. */ | |
1542 | static inline void tcp_insert_write_queue_after(struct sk_buff *skb, | |
1543 | struct sk_buff *buff, | |
1544 | struct sock *sk) | |
1545 | { | |
7de6c033 | 1546 | __skb_queue_after(&sk->sk_write_queue, skb, buff); |
fe067e8a DM |
1547 | } |
1548 | ||
43f59c89 | 1549 | /* Insert new before skb on the write queue of sk. */ |
fe067e8a DM |
1550 | static inline void tcp_insert_write_queue_before(struct sk_buff *new, |
1551 | struct sk_buff *skb, | |
1552 | struct sock *sk) | |
1553 | { | |
43f59c89 | 1554 | __skb_queue_before(&sk->sk_write_queue, skb, new); |