]>
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> |
cfb6eeb4 | 30 | #include <linux/crypto.h> |
c6aefafb | 31 | #include <linux/cryptohash.h> |
435cf559 | 32 | #include <linux/kref.h> |
740b0f18 | 33 | #include <linux/ktime.h> |
3f421baa ACM |
34 | |
35 | #include <net/inet_connection_sock.h> | |
295ff7ed | 36 | #include <net/inet_timewait_sock.h> |
77d8bf9c | 37 | #include <net/inet_hashtables.h> |
1da177e4 | 38 | #include <net/checksum.h> |
2e6599cb | 39 | #include <net/request_sock.h> |
1da177e4 LT |
40 | #include <net/sock.h> |
41 | #include <net/snmp.h> | |
42 | #include <net/ip.h> | |
c752f073 | 43 | #include <net/tcp_states.h> |
bdf1ee5d | 44 | #include <net/inet_ecn.h> |
0c266898 | 45 | #include <net/dst.h> |
c752f073 | 46 | |
1da177e4 | 47 | #include <linux/seq_file.h> |
180d8cd9 | 48 | #include <linux/memcontrol.h> |
1da177e4 | 49 | |
6e04e021 | 50 | extern struct inet_hashinfo tcp_hashinfo; |
1da177e4 | 51 | |
dd24c001 | 52 | extern struct percpu_counter tcp_orphan_count; |
5c9f3023 | 53 | void tcp_time_wait(struct sock *sk, int state, int timeo); |
1da177e4 | 54 | |
1da177e4 | 55 | #define MAX_TCP_HEADER (128 + MAX_HEADER) |
33ad798c | 56 | #define MAX_TCP_OPTION_SPACE 40 |
1da177e4 | 57 | |
105970f6 | 58 | /* |
1da177e4 | 59 | * Never offer a window over 32767 without using window scaling. Some |
105970f6 | 60 | * poor stacks do signed 16bit maths! |
1da177e4 LT |
61 | */ |
62 | #define MAX_TCP_WINDOW 32767U | |
63 | ||
64 | /* Minimal accepted MSS. It is (60+60+8) - (20+20). */ | |
65 | #define TCP_MIN_MSS 88U | |
66 | ||
5d424d5a | 67 | /* The least MTU to use for probing */ |
dcd8fb85 | 68 | #define TCP_BASE_MSS 1024 |
5d424d5a | 69 | |
05cbc0db FD |
70 | /* probing interval, default to 10 minutes as per RFC4821 */ |
71 | #define TCP_PROBE_INTERVAL 600 | |
72 | ||
6b58e0a5 FD |
73 | /* Specify interval when tcp mtu probing will stop */ |
74 | #define TCP_PROBE_THRESHOLD 8 | |
75 | ||
1da177e4 LT |
76 | /* After receiving this amount of duplicate ACKs fast retransmit starts. */ |
77 | #define TCP_FASTRETRANS_THRESH 3 | |
78 | ||
1da177e4 LT |
79 | /* Maximal number of ACKs sent quickly to accelerate slow-start. */ |
80 | #define TCP_MAX_QUICKACKS 16U | |
81 | ||
82 | /* urg_data states */ | |
83 | #define TCP_URG_VALID 0x0100 | |
84 | #define TCP_URG_NOTYET 0x0200 | |
85 | #define TCP_URG_READ 0x0400 | |
86 | ||
87 | #define TCP_RETR1 3 /* | |
88 | * This is how many retries it does before it | |
89 | * tries to figure out if the gateway is | |
90 | * down. Minimal RFC value is 3; it corresponds | |
91 | * to ~3sec-8min depending on RTO. | |
92 | */ | |
93 | ||
94 | #define TCP_RETR2 15 /* | |
95 | * This should take at least | |
96 | * 90 minutes to time out. | |
97 | * RFC1122 says that the limit is 100 sec. | |
98 | * 15 is ~13-30min depending on RTO. | |
99 | */ | |
100 | ||
6c9ff979 AB |
101 | #define TCP_SYN_RETRIES 6 /* This is how many retries are done |
102 | * when active opening a connection. | |
103 | * RFC1122 says the minimum retry MUST | |
104 | * be at least 180secs. Nevertheless | |
105 | * this value is corresponding to | |
106 | * 63secs of retransmission with the | |
107 | * current initial RTO. | |
108 | */ | |
1da177e4 | 109 | |
6c9ff979 AB |
110 | #define TCP_SYNACK_RETRIES 5 /* This is how may retries are done |
111 | * when passive opening a connection. | |
112 | * This is corresponding to 31secs of | |
113 | * retransmission with the current | |
114 | * initial RTO. | |
115 | */ | |
1da177e4 | 116 | |
1da177e4 LT |
117 | #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT |
118 | * state, about 60 seconds */ | |
119 | #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN | |
120 | /* BSD style FIN_WAIT2 deadlock breaker. | |
121 | * It used to be 3min, new value is 60sec, | |
122 | * to combine FIN-WAIT-2 timeout with | |
123 | * TIME-WAIT timer. | |
124 | */ | |
125 | ||
126 | #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */ | |
127 | #if HZ >= 100 | |
128 | #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */ | |
129 | #define TCP_ATO_MIN ((unsigned)(HZ/25)) | |
130 | #else | |
131 | #define TCP_DELACK_MIN 4U | |
132 | #define TCP_ATO_MIN 4U | |
133 | #endif | |
134 | #define TCP_RTO_MAX ((unsigned)(120*HZ)) | |
135 | #define TCP_RTO_MIN ((unsigned)(HZ/5)) | |
fd4f2cea | 136 | #define TCP_TIMEOUT_INIT ((unsigned)(1*HZ)) /* RFC6298 2.1 initial RTO value */ |
9ad7c049 JC |
137 | #define TCP_TIMEOUT_FALLBACK ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value, now |
138 | * used as a fallback RTO for the | |
139 | * initial data transmission if no | |
140 | * valid RTT sample has been acquired, | |
141 | * most likely due to retrans in 3WHS. | |
142 | */ | |
1da177e4 LT |
143 | |
144 | #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes | |
145 | * for local resources. | |
146 | */ | |
147 | ||
148 | #define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */ | |
149 | #define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */ | |
150 | #define TCP_KEEPALIVE_INTVL (75*HZ) | |
151 | ||
152 | #define MAX_TCP_KEEPIDLE 32767 | |
153 | #define MAX_TCP_KEEPINTVL 32767 | |
154 | #define MAX_TCP_KEEPCNT 127 | |
155 | #define MAX_TCP_SYNCNT 127 | |
156 | ||
157 | #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */ | |
1da177e4 LT |
158 | |
159 | #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24) | |
160 | #define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated | |
161 | * after this time. It should be equal | |
162 | * (or greater than) TCP_TIMEWAIT_LEN | |
163 | * to provide reliability equal to one | |
164 | * provided by timewait state. | |
165 | */ | |
166 | #define TCP_PAWS_WINDOW 1 /* Replay window for per-host | |
167 | * timestamps. It must be less than | |
168 | * minimal timewait lifetime. | |
169 | */ | |
1da177e4 LT |
170 | /* |
171 | * TCP option | |
172 | */ | |
105970f6 | 173 | |
1da177e4 LT |
174 | #define TCPOPT_NOP 1 /* Padding */ |
175 | #define TCPOPT_EOL 0 /* End of options */ | |
176 | #define TCPOPT_MSS 2 /* Segment size negotiating */ | |
177 | #define TCPOPT_WINDOW 3 /* Window scaling */ | |
178 | #define TCPOPT_SACK_PERM 4 /* SACK Permitted */ | |
179 | #define TCPOPT_SACK 5 /* SACK Block */ | |
180 | #define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */ | |
cfb6eeb4 | 181 | #define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */ |
7f9b838b | 182 | #define TCPOPT_FASTOPEN 34 /* Fast open (RFC7413) */ |
2100c8d2 YC |
183 | #define TCPOPT_EXP 254 /* Experimental */ |
184 | /* Magic number to be after the option value for sharing TCP | |
185 | * experimental options. See draft-ietf-tcpm-experimental-options-00.txt | |
186 | */ | |
187 | #define TCPOPT_FASTOPEN_MAGIC 0xF989 | |
1da177e4 LT |
188 | |
189 | /* | |
190 | * TCP option lengths | |
191 | */ | |
192 | ||
193 | #define TCPOLEN_MSS 4 | |
194 | #define TCPOLEN_WINDOW 3 | |
195 | #define TCPOLEN_SACK_PERM 2 | |
196 | #define TCPOLEN_TIMESTAMP 10 | |
cfb6eeb4 | 197 | #define TCPOLEN_MD5SIG 18 |
7f9b838b | 198 | #define TCPOLEN_FASTOPEN_BASE 2 |
2100c8d2 | 199 | #define TCPOLEN_EXP_FASTOPEN_BASE 4 |
1da177e4 LT |
200 | |
201 | /* But this is what stacks really send out. */ | |
202 | #define TCPOLEN_TSTAMP_ALIGNED 12 | |
203 | #define TCPOLEN_WSCALE_ALIGNED 4 | |
204 | #define TCPOLEN_SACKPERM_ALIGNED 4 | |
205 | #define TCPOLEN_SACK_BASE 2 | |
206 | #define TCPOLEN_SACK_BASE_ALIGNED 4 | |
207 | #define TCPOLEN_SACK_PERBLOCK 8 | |
cfb6eeb4 | 208 | #define TCPOLEN_MD5SIG_ALIGNED 20 |
33ad798c | 209 | #define TCPOLEN_MSS_ALIGNED 4 |
1da177e4 | 210 | |
1da177e4 LT |
211 | /* Flags in tp->nonagle */ |
212 | #define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */ | |
213 | #define TCP_NAGLE_CORK 2 /* Socket is corked */ | |
caa20d9a | 214 | #define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */ |
1da177e4 | 215 | |
36e31b0a AP |
216 | /* TCP thin-stream limits */ |
217 | #define TCP_THIN_LINEAR_RETRIES 6 /* After 6 linear retries, do exp. backoff */ | |
218 | ||
7eb38527 | 219 | /* TCP initial congestion window as per draft-hkchu-tcpm-initcwnd-01 */ |
442b9635 DM |
220 | #define TCP_INIT_CWND 10 |
221 | ||
cf60af03 YC |
222 | /* Bit Flags for sysctl_tcp_fastopen */ |
223 | #define TFO_CLIENT_ENABLE 1 | |
10467163 | 224 | #define TFO_SERVER_ENABLE 2 |
67da22d2 | 225 | #define TFO_CLIENT_NO_COOKIE 4 /* Data in SYN w/o cookie option */ |
cf60af03 | 226 | |
10467163 JC |
227 | /* Accept SYN data w/o any cookie option */ |
228 | #define TFO_SERVER_COOKIE_NOT_REQD 0x200 | |
229 | ||
230 | /* Force enable TFO on all listeners, i.e., not requiring the | |
231 | * TCP_FASTOPEN socket option. SOCKOPT1/2 determine how to set max_qlen. | |
232 | */ | |
233 | #define TFO_SERVER_WO_SOCKOPT1 0x400 | |
234 | #define TFO_SERVER_WO_SOCKOPT2 0x800 | |
10467163 | 235 | |
295ff7ed ACM |
236 | extern struct inet_timewait_death_row tcp_death_row; |
237 | ||
1da177e4 | 238 | /* sysctl variables for tcp */ |
1da177e4 LT |
239 | extern int sysctl_tcp_timestamps; |
240 | extern int sysctl_tcp_window_scaling; | |
241 | extern int sysctl_tcp_sack; | |
242 | extern int sysctl_tcp_fin_timeout; | |
1da177e4 LT |
243 | extern int sysctl_tcp_keepalive_time; |
244 | extern int sysctl_tcp_keepalive_probes; | |
245 | extern int sysctl_tcp_keepalive_intvl; | |
246 | extern int sysctl_tcp_syn_retries; | |
247 | extern int sysctl_tcp_synack_retries; | |
248 | extern int sysctl_tcp_retries1; | |
249 | extern int sysctl_tcp_retries2; | |
250 | extern int sysctl_tcp_orphan_retries; | |
251 | extern int sysctl_tcp_syncookies; | |
2100c8d2 | 252 | extern int sysctl_tcp_fastopen; |
1da177e4 LT |
253 | extern int sysctl_tcp_retrans_collapse; |
254 | extern int sysctl_tcp_stdurg; | |
255 | extern int sysctl_tcp_rfc1337; | |
256 | extern int sysctl_tcp_abort_on_overflow; | |
257 | extern int sysctl_tcp_max_orphans; | |
1da177e4 LT |
258 | extern int sysctl_tcp_fack; |
259 | extern int sysctl_tcp_reordering; | |
dca145ff | 260 | extern int sysctl_tcp_max_reordering; |
1da177e4 | 261 | extern int sysctl_tcp_dsack; |
a4fe34bf | 262 | extern long sysctl_tcp_mem[3]; |
1da177e4 LT |
263 | extern int sysctl_tcp_wmem[3]; |
264 | extern int sysctl_tcp_rmem[3]; | |
265 | extern int sysctl_tcp_app_win; | |
266 | extern int sysctl_tcp_adv_win_scale; | |
267 | extern int sysctl_tcp_tw_reuse; | |
268 | extern int sysctl_tcp_frto; | |
269 | extern int sysctl_tcp_low_latency; | |
1da177e4 | 270 | extern int sysctl_tcp_nometrics_save; |
1da177e4 LT |
271 | extern int sysctl_tcp_moderate_rcvbuf; |
272 | extern int sysctl_tcp_tso_win_divisor; | |
15d99e02 | 273 | extern int sysctl_tcp_workaround_signed_windows; |
35089bb2 | 274 | extern int sysctl_tcp_slow_start_after_idle; |
36e31b0a | 275 | extern int sysctl_tcp_thin_linear_timeouts; |
7e380175 | 276 | extern int sysctl_tcp_thin_dupack; |
eed530b6 | 277 | extern int sysctl_tcp_early_retrans; |
46d3ceab | 278 | extern int sysctl_tcp_limit_output_bytes; |
282f23c6 | 279 | extern int sysctl_tcp_challenge_ack_limit; |
c9bee3b7 | 280 | extern unsigned int sysctl_tcp_notsent_lowat; |
95bd09eb | 281 | extern int sysctl_tcp_min_tso_segs; |
f54b3111 | 282 | extern int sysctl_tcp_autocorking; |
032ee423 | 283 | extern int sysctl_tcp_invalid_ratelimit; |
43e122b0 ED |
284 | extern int sysctl_tcp_pacing_ss_ratio; |
285 | extern int sysctl_tcp_pacing_ca_ratio; | |
1da177e4 | 286 | |
8d987e5c | 287 | extern atomic_long_t tcp_memory_allocated; |
1748376b | 288 | extern struct percpu_counter tcp_sockets_allocated; |
1da177e4 LT |
289 | extern int tcp_memory_pressure; |
290 | ||
b8da51eb ED |
291 | /* optimized version of sk_under_memory_pressure() for TCP sockets */ |
292 | static inline bool tcp_under_memory_pressure(const struct sock *sk) | |
293 | { | |
294 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) | |
295 | return !!sk->sk_cgrp->memory_pressure; | |
296 | ||
297 | return tcp_memory_pressure; | |
298 | } | |
1da177e4 LT |
299 | /* |
300 | * The next routines deal with comparing 32 bit unsigned ints | |
301 | * and worry about wraparound (automatic with unsigned arithmetic). | |
302 | */ | |
303 | ||
a2a385d6 | 304 | static inline bool before(__u32 seq1, __u32 seq2) |
1da177e4 | 305 | { |
0d630cc0 | 306 | return (__s32)(seq1-seq2) < 0; |
1da177e4 | 307 | } |
9a036b9c | 308 | #define after(seq2, seq1) before(seq1, seq2) |
1da177e4 LT |
309 | |
310 | /* is s2<=s1<=s3 ? */ | |
a2a385d6 | 311 | static inline bool between(__u32 seq1, __u32 seq2, __u32 seq3) |
1da177e4 LT |
312 | { |
313 | return seq3 - seq2 >= seq1 - seq2; | |
314 | } | |
315 | ||
efcdbf24 AS |
316 | static inline bool tcp_out_of_memory(struct sock *sk) |
317 | { | |
318 | if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF && | |
319 | sk_memory_allocated(sk) > sk_prot_mem_limits(sk, 2)) | |
320 | return true; | |
321 | return false; | |
322 | } | |
323 | ||
a6c5ea4c ED |
324 | void sk_forced_mem_schedule(struct sock *sk, int size); |
325 | ||
ad1af0fe | 326 | static inline bool tcp_too_many_orphans(struct sock *sk, int shift) |
e4fd5da3 | 327 | { |
ad1af0fe DM |
328 | struct percpu_counter *ocp = sk->sk_prot->orphan_count; |
329 | int orphans = percpu_counter_read_positive(ocp); | |
330 | ||
331 | if (orphans << shift > sysctl_tcp_max_orphans) { | |
332 | orphans = percpu_counter_sum_positive(ocp); | |
333 | if (orphans << shift > sysctl_tcp_max_orphans) | |
334 | return true; | |
335 | } | |
ad1af0fe | 336 | return false; |
e4fd5da3 | 337 | } |
1da177e4 | 338 | |
5c9f3023 | 339 | bool tcp_check_oom(struct sock *sk, int shift); |
efcdbf24 | 340 | |
a0f82f64 | 341 | |
1da177e4 LT |
342 | extern struct proto tcp_prot; |
343 | ||
57ef42d5 PE |
344 | #define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field) |
345 | #define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field) | |
346 | #define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field) | |
347 | #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val) | |
aa2ea058 | 348 | #define TCP_ADD_STATS(net, field, val) SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val) |
1da177e4 | 349 | |
5c9f3023 JP |
350 | void tcp_tasklet_init(void); |
351 | ||
352 | void tcp_v4_err(struct sk_buff *skb, u32); | |
353 | ||
354 | void tcp_shutdown(struct sock *sk, int how); | |
355 | ||
356 | void tcp_v4_early_demux(struct sk_buff *skb); | |
357 | int tcp_v4_rcv(struct sk_buff *skb); | |
358 | ||
359 | int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw); | |
1b784140 | 360 | int tcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size); |
5c9f3023 JP |
361 | int tcp_sendpage(struct sock *sk, struct page *page, int offset, size_t size, |
362 | int flags); | |
363 | void tcp_release_cb(struct sock *sk); | |
364 | void tcp_wfree(struct sk_buff *skb); | |
365 | void tcp_write_timer_handler(struct sock *sk); | |
366 | void tcp_delack_timer_handler(struct sock *sk); | |
367 | int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg); | |
368 | int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, | |
369 | const struct tcphdr *th, unsigned int len); | |
370 | void tcp_rcv_established(struct sock *sk, struct sk_buff *skb, | |
371 | const struct tcphdr *th, unsigned int len); | |
372 | void tcp_rcv_space_adjust(struct sock *sk); | |
5c9f3023 JP |
373 | int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp); |
374 | void tcp_twsk_destructor(struct sock *sk); | |
375 | ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos, | |
376 | struct pipe_inode_info *pipe, size_t len, | |
377 | unsigned int flags); | |
9c55e01c | 378 | |
463c84b9 ACM |
379 | static inline void tcp_dec_quickack_mode(struct sock *sk, |
380 | const unsigned int pkts) | |
1da177e4 | 381 | { |
463c84b9 | 382 | struct inet_connection_sock *icsk = inet_csk(sk); |
fc6415bc | 383 | |
463c84b9 ACM |
384 | if (icsk->icsk_ack.quick) { |
385 | if (pkts >= icsk->icsk_ack.quick) { | |
386 | icsk->icsk_ack.quick = 0; | |
fc6415bc | 387 | /* Leaving quickack mode we deflate ATO. */ |
463c84b9 | 388 | icsk->icsk_ack.ato = TCP_ATO_MIN; |
fc6415bc | 389 | } else |
463c84b9 | 390 | icsk->icsk_ack.quick -= pkts; |
1da177e4 LT |
391 | } |
392 | } | |
393 | ||
bdf1ee5d IJ |
394 | #define TCP_ECN_OK 1 |
395 | #define TCP_ECN_QUEUE_CWR 2 | |
396 | #define TCP_ECN_DEMAND_CWR 4 | |
7a269ffa | 397 | #define TCP_ECN_SEEN 8 |
bdf1ee5d | 398 | |
fd2c3ef7 | 399 | enum tcp_tw_status { |
1da177e4 LT |
400 | TCP_TW_SUCCESS = 0, |
401 | TCP_TW_RST = 1, | |
402 | TCP_TW_ACK = 2, | |
403 | TCP_TW_SYN = 3 | |
404 | }; | |
405 | ||
406 | ||
5c9f3023 JP |
407 | enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw, |
408 | struct sk_buff *skb, | |
409 | const struct tcphdr *th); | |
410 | struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb, | |
52452c54 | 411 | struct request_sock *req, bool fastopen); |
5c9f3023 JP |
412 | int tcp_child_process(struct sock *parent, struct sock *child, |
413 | struct sk_buff *skb); | |
5ae344c9 | 414 | void tcp_enter_loss(struct sock *sk); |
5c9f3023 JP |
415 | void tcp_clear_retrans(struct tcp_sock *tp); |
416 | void tcp_update_metrics(struct sock *sk); | |
417 | void tcp_init_metrics(struct sock *sk); | |
418 | void tcp_metrics_init(void); | |
419 | bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst, | |
a26552af | 420 | bool paws_check, bool timestamps); |
5c9f3023 JP |
421 | bool tcp_remember_stamp(struct sock *sk); |
422 | bool tcp_tw_remember_stamp(struct inet_timewait_sock *tw); | |
423 | void tcp_fetch_timewait_stamp(struct sock *sk, struct dst_entry *dst); | |
424 | void tcp_disable_fack(struct tcp_sock *tp); | |
425 | void tcp_close(struct sock *sk, long timeout); | |
426 | void tcp_init_sock(struct sock *sk); | |
427 | unsigned int tcp_poll(struct file *file, struct socket *sock, | |
428 | struct poll_table_struct *wait); | |
429 | int tcp_getsockopt(struct sock *sk, int level, int optname, | |
430 | char __user *optval, int __user *optlen); | |
431 | int tcp_setsockopt(struct sock *sk, int level, int optname, | |
432 | char __user *optval, unsigned int optlen); | |
433 | int compat_tcp_getsockopt(struct sock *sk, int level, int optname, | |
53d3176b | 434 | char __user *optval, int __user *optlen); |
5c9f3023 | 435 | int compat_tcp_setsockopt(struct sock *sk, int level, int optname, |
53d3176b | 436 | char __user *optval, unsigned int optlen); |
5c9f3023 | 437 | void tcp_set_keepalive(struct sock *sk, int val); |
42cb80a2 | 438 | void tcp_syn_ack_timeout(const struct request_sock *req); |
1b784140 YX |
439 | int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock, |
440 | int flags, int *addr_len); | |
5c9f3023 JP |
441 | void tcp_parse_options(const struct sk_buff *skb, |
442 | struct tcp_options_received *opt_rx, | |
443 | int estab, struct tcp_fastopen_cookie *foc); | |
444 | const u8 *tcp_parse_md5sig_option(const struct tcphdr *th); | |
7d5d5525 | 445 | |
1da177e4 LT |
446 | /* |
447 | * TCP v4 functions exported for the inet6 API | |
448 | */ | |
449 | ||
5c9f3023 | 450 | void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb); |
4fab9071 | 451 | void tcp_v4_mtu_reduced(struct sock *sk); |
26e37360 | 452 | void tcp_req_err(struct sock *sk, u32 seq); |
5c9f3023 JP |
453 | int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb); |
454 | struct sock *tcp_create_openreq_child(struct sock *sk, | |
455 | struct request_sock *req, | |
456 | struct sk_buff *skb); | |
81164413 | 457 | void tcp_ca_openreq_child(struct sock *sk, const struct dst_entry *dst); |
5c9f3023 JP |
458 | struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb, |
459 | struct request_sock *req, | |
460 | struct dst_entry *dst); | |
461 | int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb); | |
462 | int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len); | |
463 | int tcp_connect(struct sock *sk); | |
464 | struct sk_buff *tcp_make_synack(struct sock *sk, struct dst_entry *dst, | |
465 | struct request_sock *req, | |
466 | struct tcp_fastopen_cookie *foc); | |
467 | int tcp_disconnect(struct sock *sk, int flags); | |
1da177e4 | 468 | |
370816ae | 469 | void tcp_finish_connect(struct sock *sk, struct sk_buff *skb); |
292e8d8c | 470 | int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size); |
63d02d15 | 471 | void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb); |
1da177e4 | 472 | |
1da177e4 | 473 | /* From syncookies.c */ |
b80c0e78 ED |
474 | struct sock *tcp_get_cookie_sock(struct sock *sk, struct sk_buff *skb, |
475 | struct request_sock *req, | |
476 | struct dst_entry *dst); | |
5c9f3023 JP |
477 | int __cookie_v4_check(const struct iphdr *iph, const struct tcphdr *th, |
478 | u32 cookie); | |
461b74c3 | 479 | struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb); |
e05c82d3 | 480 | #ifdef CONFIG_SYN_COOKIES |
8c27bd75 | 481 | |
63262315 | 482 | /* Syncookies use a monotonic timer which increments every 60 seconds. |
8c27bd75 FW |
483 | * This counter is used both as a hash input and partially encoded into |
484 | * the cookie value. A cookie is only validated further if the delta | |
485 | * between the current counter value and the encoded one is less than this, | |
63262315 | 486 | * i.e. a sent cookie is valid only at most for 2*60 seconds (or less if |
8c27bd75 FW |
487 | * the counter advances immediately after a cookie is generated). |
488 | */ | |
264ea103 ED |
489 | #define MAX_SYNCOOKIE_AGE 2 |
490 | #define TCP_SYNCOOKIE_PERIOD (60 * HZ) | |
491 | #define TCP_SYNCOOKIE_VALID (MAX_SYNCOOKIE_AGE * TCP_SYNCOOKIE_PERIOD) | |
492 | ||
493 | /* syncookies: remember time of last synqueue overflow | |
494 | * But do not dirty this field too often (once per second is enough) | |
495 | */ | |
496 | static inline void tcp_synq_overflow(struct sock *sk) | |
497 | { | |
498 | unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp; | |
499 | unsigned long now = jiffies; | |
500 | ||
501 | if (time_after(now, last_overflow + HZ)) | |
502 | tcp_sk(sk)->rx_opt.ts_recent_stamp = now; | |
503 | } | |
504 | ||
505 | /* syncookies: no recent synqueue overflow on this listening socket? */ | |
506 | static inline bool tcp_synq_no_recent_overflow(const struct sock *sk) | |
507 | { | |
508 | unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp; | |
509 | ||
510 | return time_after(jiffies, last_overflow + TCP_SYNCOOKIE_VALID); | |
511 | } | |
8c27bd75 FW |
512 | |
513 | static inline u32 tcp_cookie_time(void) | |
514 | { | |
63262315 ED |
515 | u64 val = get_jiffies_64(); |
516 | ||
264ea103 | 517 | do_div(val, TCP_SYNCOOKIE_PERIOD); |
63262315 | 518 | return val; |
8c27bd75 FW |
519 | } |
520 | ||
5c9f3023 JP |
521 | u32 __cookie_v4_init_sequence(const struct iphdr *iph, const struct tcphdr *th, |
522 | u16 *mssp); | |
57b47553 OP |
523 | __u32 cookie_v4_init_sequence(struct sock *sk, const struct sk_buff *skb, |
524 | __u16 *mss); | |
5c9f3023 | 525 | __u32 cookie_init_timestamp(struct request_sock *req); |
f1673381 FW |
526 | bool cookie_timestamp_decode(struct tcp_options_received *opt); |
527 | bool cookie_ecn_ok(const struct tcp_options_received *opt, | |
f7b3bec6 | 528 | const struct net *net, const struct dst_entry *dst); |
4dfc2817 | 529 | |
c6aefafb | 530 | /* From net/ipv6/syncookies.c */ |
5c9f3023 JP |
531 | int __cookie_v6_check(const struct ipv6hdr *iph, const struct tcphdr *th, |
532 | u32 cookie); | |
533 | struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb); | |
f1673381 | 534 | |
5c9f3023 JP |
535 | u32 __cookie_v6_init_sequence(const struct ipv6hdr *iph, |
536 | const struct tcphdr *th, u16 *mssp); | |
537 | __u32 cookie_v6_init_sequence(struct sock *sk, const struct sk_buff *skb, | |
538 | __u16 *mss); | |
e05c82d3 | 539 | #endif |
1da177e4 LT |
540 | /* tcp_output.c */ |
541 | ||
5c9f3023 JP |
542 | void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss, |
543 | int nonagle); | |
544 | bool tcp_may_send_now(struct sock *sk); | |
545 | int __tcp_retransmit_skb(struct sock *, struct sk_buff *); | |
546 | int tcp_retransmit_skb(struct sock *, struct sk_buff *); | |
547 | void tcp_retransmit_timer(struct sock *sk); | |
548 | void tcp_xmit_retransmit_queue(struct sock *); | |
549 | void tcp_simple_retransmit(struct sock *); | |
550 | int tcp_trim_head(struct sock *, struct sk_buff *, u32); | |
6cc55e09 | 551 | int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int, gfp_t); |
5c9f3023 JP |
552 | |
553 | void tcp_send_probe0(struct sock *); | |
554 | void tcp_send_partial(struct sock *); | |
e520af48 | 555 | int tcp_write_wakeup(struct sock *, int mib); |
5c9f3023 JP |
556 | void tcp_send_fin(struct sock *sk); |
557 | void tcp_send_active_reset(struct sock *sk, gfp_t priority); | |
558 | int tcp_send_synack(struct sock *); | |
5c9f3023 JP |
559 | void tcp_push_one(struct sock *, unsigned int mss_now); |
560 | void tcp_send_ack(struct sock *sk); | |
561 | void tcp_send_delayed_ack(struct sock *sk); | |
562 | void tcp_send_loss_probe(struct sock *sk); | |
563 | bool tcp_schedule_loss_probe(struct sock *sk); | |
1da177e4 | 564 | |
a762a980 | 565 | /* tcp_input.c */ |
5c9f3023 JP |
566 | void tcp_resume_early_retransmit(struct sock *sk); |
567 | void tcp_rearm_rto(struct sock *sk); | |
568 | void tcp_reset(struct sock *sk); | |
a762a980 | 569 | |
1da177e4 | 570 | /* tcp_timer.c */ |
5c9f3023 | 571 | void tcp_init_xmit_timers(struct sock *); |
463c84b9 ACM |
572 | static inline void tcp_clear_xmit_timers(struct sock *sk) |
573 | { | |
574 | inet_csk_clear_xmit_timers(sk); | |
575 | } | |
1da177e4 | 576 | |
5c9f3023 JP |
577 | unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu); |
578 | unsigned int tcp_current_mss(struct sock *sk); | |
0c54b85f IJ |
579 | |
580 | /* Bound MSS / TSO packet size with the half of the window */ | |
581 | static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize) | |
582 | { | |
01f83d69 AK |
583 | int cutoff; |
584 | ||
585 | /* When peer uses tiny windows, there is no use in packetizing | |
586 | * to sub-MSS pieces for the sake of SWS or making sure there | |
587 | * are enough packets in the pipe for fast recovery. | |
588 | * | |
589 | * On the other hand, for extremely large MSS devices, handling | |
590 | * smaller than MSS windows in this way does make sense. | |
591 | */ | |
592 | if (tp->max_window >= 512) | |
593 | cutoff = (tp->max_window >> 1); | |
594 | else | |
595 | cutoff = tp->max_window; | |
596 | ||
597 | if (cutoff && pktsize > cutoff) | |
598 | return max_t(int, cutoff, 68U - tp->tcp_header_len); | |
0c54b85f IJ |
599 | else |
600 | return pktsize; | |
601 | } | |
1da177e4 | 602 | |
17b085ea | 603 | /* tcp.c */ |
0df48c26 | 604 | void tcp_get_info(struct sock *, struct tcp_info *); |
1da177e4 LT |
605 | |
606 | /* Read 'sendfile()'-style from a TCP socket */ | |
607 | typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *, | |
608 | unsigned int, size_t); | |
5c9f3023 JP |
609 | int tcp_read_sock(struct sock *sk, read_descriptor_t *desc, |
610 | sk_read_actor_t recv_actor); | |
1da177e4 | 611 | |
5c9f3023 | 612 | void tcp_initialize_rcv_mss(struct sock *sk); |
1da177e4 | 613 | |
5c9f3023 JP |
614 | int tcp_mtu_to_mss(struct sock *sk, int pmtu); |
615 | int tcp_mss_to_mtu(struct sock *sk, int mss); | |
616 | void tcp_mtup_init(struct sock *sk); | |
617 | void tcp_init_buffer_space(struct sock *sk); | |
5d424d5a | 618 | |
f1ecd5d9 DL |
619 | static inline void tcp_bound_rto(const struct sock *sk) |
620 | { | |
621 | if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX) | |
622 | inet_csk(sk)->icsk_rto = TCP_RTO_MAX; | |
623 | } | |
624 | ||
625 | static inline u32 __tcp_set_rto(const struct tcp_sock *tp) | |
626 | { | |
740b0f18 | 627 | return usecs_to_jiffies((tp->srtt_us >> 3) + tp->rttvar_us); |
f1ecd5d9 DL |
628 | } |
629 | ||
40efc6fa | 630 | static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd) |
1da177e4 LT |
631 | { |
632 | tp->pred_flags = htonl((tp->tcp_header_len << 26) | | |
633 | ntohl(TCP_FLAG_ACK) | | |
634 | snd_wnd); | |
635 | } | |
636 | ||
40efc6fa | 637 | static inline void tcp_fast_path_on(struct tcp_sock *tp) |
1da177e4 LT |
638 | { |
639 | __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale); | |
640 | } | |
641 | ||
9e412ba7 | 642 | static inline void tcp_fast_path_check(struct sock *sk) |
1da177e4 | 643 | { |
9e412ba7 IJ |
644 | struct tcp_sock *tp = tcp_sk(sk); |
645 | ||
b03efcfb | 646 | if (skb_queue_empty(&tp->out_of_order_queue) && |
1da177e4 LT |
647 | tp->rcv_wnd && |
648 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf && | |
649 | !tp->urg_data) | |
650 | tcp_fast_path_on(tp); | |
651 | } | |
652 | ||
0c266898 SS |
653 | /* Compute the actual rto_min value */ |
654 | static inline u32 tcp_rto_min(struct sock *sk) | |
655 | { | |
cf533ea5 | 656 | const struct dst_entry *dst = __sk_dst_get(sk); |
0c266898 SS |
657 | u32 rto_min = TCP_RTO_MIN; |
658 | ||
659 | if (dst && dst_metric_locked(dst, RTAX_RTO_MIN)) | |
660 | rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN); | |
661 | return rto_min; | |
662 | } | |
663 | ||
740b0f18 ED |
664 | static inline u32 tcp_rto_min_us(struct sock *sk) |
665 | { | |
666 | return jiffies_to_usecs(tcp_rto_min(sk)); | |
667 | } | |
668 | ||
81164413 DB |
669 | static inline bool tcp_ca_dst_locked(const struct dst_entry *dst) |
670 | { | |
671 | return dst_metric_locked(dst, RTAX_CC_ALGO); | |
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 | { | |
958 | tp->do_early_retrans = sysctl_tcp_early_retrans && | |
6ba8a3b1 ND |
959 | sysctl_tcp_early_retrans < 4 && !sysctl_tcp_thin_dupack && |
960 | sysctl_tcp_reordering == 3; | |
eed530b6 YC |
961 | } |
962 | ||
963 | static inline void tcp_disable_early_retrans(struct tcp_sock *tp) | |
964 | { | |
965 | tp->do_early_retrans = 0; | |
966 | } | |
967 | ||
83ae4088 IJ |
968 | static inline unsigned int tcp_left_out(const struct tcp_sock *tp) |
969 | { | |
970 | return tp->sacked_out + tp->lost_out; | |
971 | } | |
972 | ||
1da177e4 LT |
973 | /* This determines how many packets are "in the network" to the best |
974 | * of our knowledge. In many cases it is conservative, but where | |
975 | * detailed information is available from the receiver (via SACK | |
976 | * blocks etc.) we can make more aggressive calculations. | |
977 | * | |
978 | * Use this for decisions involving congestion control, use just | |
979 | * tp->packets_out to determine if the send queue is empty or not. | |
980 | * | |
981 | * Read this equation as: | |
982 | * | |
983 | * "Packets sent once on transmission queue" MINUS | |
984 | * "Packets left network, but not honestly ACKed yet" PLUS | |
985 | * "Packets fast retransmitted" | |
986 | */ | |
40efc6fa | 987 | static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp) |
1da177e4 | 988 | { |
83ae4088 | 989 | return tp->packets_out - tcp_left_out(tp) + tp->retrans_out; |
1da177e4 LT |
990 | } |
991 | ||
0b6a05c1 IJ |
992 | #define TCP_INFINITE_SSTHRESH 0x7fffffff |
993 | ||
071d5080 YC |
994 | static inline bool tcp_in_slow_start(const struct tcp_sock *tp) |
995 | { | |
76174004 | 996 | return tp->snd_cwnd < tp->snd_ssthresh; |
071d5080 YC |
997 | } |
998 | ||
0b6a05c1 IJ |
999 | static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp) |
1000 | { | |
1001 | return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH; | |
1002 | } | |
1003 | ||
684bad11 YC |
1004 | static inline bool tcp_in_cwnd_reduction(const struct sock *sk) |
1005 | { | |
1006 | return (TCPF_CA_CWR | TCPF_CA_Recovery) & | |
1007 | (1 << inet_csk(sk)->icsk_ca_state); | |
1008 | } | |
1009 | ||
1da177e4 | 1010 | /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd. |
684bad11 | 1011 | * The exception is cwnd reduction phase, when cwnd is decreasing towards |
1da177e4 LT |
1012 | * ssthresh. |
1013 | */ | |
6687e988 | 1014 | static inline __u32 tcp_current_ssthresh(const struct sock *sk) |
1da177e4 | 1015 | { |
6687e988 | 1016 | const struct tcp_sock *tp = tcp_sk(sk); |
cf533ea5 | 1017 | |
684bad11 | 1018 | if (tcp_in_cwnd_reduction(sk)) |
1da177e4 LT |
1019 | return tp->snd_ssthresh; |
1020 | else | |
1021 | return max(tp->snd_ssthresh, | |
1022 | ((tp->snd_cwnd >> 1) + | |
1023 | (tp->snd_cwnd >> 2))); | |
1024 | } | |
1025 | ||
b9c4595b IJ |
1026 | /* Use define here intentionally to get WARN_ON location shown at the caller */ |
1027 | #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out) | |
1da177e4 | 1028 | |
5ee2c941 | 1029 | void tcp_enter_cwr(struct sock *sk); |
5c9f3023 | 1030 | __u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst); |
1da177e4 | 1031 | |
6b5a5c0d NC |
1032 | /* The maximum number of MSS of available cwnd for which TSO defers |
1033 | * sending if not using sysctl_tcp_tso_win_divisor. | |
1034 | */ | |
1035 | static inline __u32 tcp_max_tso_deferred_mss(const struct tcp_sock *tp) | |
1036 | { | |
1037 | return 3; | |
1038 | } | |
1039 | ||
1da177e4 | 1040 | /* Slow start with delack produces 3 packets of burst, so that |
dd9e0dda JH |
1041 | * it is safe "de facto". This will be the default - same as |
1042 | * the default reordering threshold - but if reordering increases, | |
1043 | * we must be able to allow cwnd to burst at least this much in order | |
1044 | * to not pull it back when holes are filled. | |
1da177e4 LT |
1045 | */ |
1046 | static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp) | |
1047 | { | |
dd9e0dda | 1048 | return tp->reordering; |
1da177e4 LT |
1049 | } |
1050 | ||
90840def IJ |
1051 | /* Returns end sequence number of the receiver's advertised window */ |
1052 | static inline u32 tcp_wnd_end(const struct tcp_sock *tp) | |
1053 | { | |
1054 | return tp->snd_una + tp->snd_wnd; | |
1055 | } | |
e114a710 ED |
1056 | |
1057 | /* We follow the spirit of RFC2861 to validate cwnd but implement a more | |
1058 | * flexible approach. The RFC suggests cwnd should not be raised unless | |
ca8a2263 NC |
1059 | * it was fully used previously. And that's exactly what we do in |
1060 | * congestion avoidance mode. But in slow start we allow cwnd to grow | |
1061 | * as long as the application has used half the cwnd. | |
e114a710 ED |
1062 | * Example : |
1063 | * cwnd is 10 (IW10), but application sends 9 frames. | |
1064 | * We allow cwnd to reach 18 when all frames are ACKed. | |
1065 | * This check is safe because it's as aggressive as slow start which already | |
1066 | * risks 100% overshoot. The advantage is that we discourage application to | |
1067 | * either send more filler packets or data to artificially blow up the cwnd | |
1068 | * usage, and allow application-limited process to probe bw more aggressively. | |
e114a710 | 1069 | */ |
24901551 | 1070 | static inline bool tcp_is_cwnd_limited(const struct sock *sk) |
e114a710 ED |
1071 | { |
1072 | const struct tcp_sock *tp = tcp_sk(sk); | |
1073 | ||
ca8a2263 | 1074 | /* If in slow start, ensure cwnd grows to twice what was ACKed. */ |
071d5080 | 1075 | if (tcp_in_slow_start(tp)) |
ca8a2263 NC |
1076 | return tp->snd_cwnd < 2 * tp->max_packets_out; |
1077 | ||
1078 | return tp->is_cwnd_limited; | |
e114a710 | 1079 | } |
f4805ede | 1080 | |
21c8fe99 ED |
1081 | /* Something is really bad, we could not queue an additional packet, |
1082 | * because qdisc is full or receiver sent a 0 window. | |
1083 | * We do not want to add fuel to the fire, or abort too early, | |
1084 | * so make sure the timer we arm now is at least 200ms in the future, | |
1085 | * regardless of current icsk_rto value (as it could be ~2ms) | |
1086 | */ | |
1087 | static inline unsigned long tcp_probe0_base(const struct sock *sk) | |
1da177e4 | 1088 | { |
21c8fe99 ED |
1089 | return max_t(unsigned long, inet_csk(sk)->icsk_rto, TCP_RTO_MIN); |
1090 | } | |
9e412ba7 | 1091 | |
21c8fe99 ED |
1092 | /* Variant of inet_csk_rto_backoff() used for zero window probes */ |
1093 | static inline unsigned long tcp_probe0_when(const struct sock *sk, | |
1094 | unsigned long max_when) | |
1095 | { | |
1096 | u64 when = (u64)tcp_probe0_base(sk) << inet_csk(sk)->icsk_backoff; | |
1097 | ||
1098 | return (unsigned long)min_t(u64, when, max_when); | |
1099 | } | |
1100 | ||
1101 | static inline void tcp_check_probe_timer(struct sock *sk) | |
1102 | { | |
1103 | if (!tcp_sk(sk)->packets_out && !inet_csk(sk)->icsk_pending) | |
3f421baa | 1104 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, |
21c8fe99 | 1105 | tcp_probe0_base(sk), TCP_RTO_MAX); |
1da177e4 LT |
1106 | } |
1107 | ||
ee7537b6 | 1108 | static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq) |
1da177e4 LT |
1109 | { |
1110 | tp->snd_wl1 = seq; | |
1111 | } | |
1112 | ||
ee7537b6 | 1113 | static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq) |
1da177e4 LT |
1114 | { |
1115 | tp->snd_wl1 = seq; | |
1116 | } | |
1117 | ||
1da177e4 LT |
1118 | /* |
1119 | * Calculate(/check) TCP checksum | |
1120 | */ | |
ba7808ea FD |
1121 | static inline __sum16 tcp_v4_check(int len, __be32 saddr, |
1122 | __be32 daddr, __wsum base) | |
1da177e4 LT |
1123 | { |
1124 | return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base); | |
1125 | } | |
1126 | ||
b51655b9 | 1127 | static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 1128 | { |
fb286bb2 | 1129 | return __skb_checksum_complete(skb); |
1da177e4 LT |
1130 | } |
1131 | ||
a2a385d6 | 1132 | static inline bool tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 1133 | { |
60476372 | 1134 | return !skb_csum_unnecessary(skb) && |
1da177e4 LT |
1135 | __tcp_checksum_complete(skb); |
1136 | } | |
1137 | ||
1138 | /* Prequeue for VJ style copy to user, combined with checksumming. */ | |
1139 | ||
40efc6fa | 1140 | static inline void tcp_prequeue_init(struct tcp_sock *tp) |
1da177e4 LT |
1141 | { |
1142 | tp->ucopy.task = NULL; | |
1143 | tp->ucopy.len = 0; | |
1144 | tp->ucopy.memory = 0; | |
1145 | skb_queue_head_init(&tp->ucopy.prequeue); | |
1146 | } | |
1147 | ||
5c9f3023 | 1148 | bool tcp_prequeue(struct sock *sk, struct sk_buff *skb); |
1da177e4 LT |
1149 | |
1150 | #undef STATE_TRACE | |
1151 | ||
1152 | #ifdef STATE_TRACE | |
1153 | static const char *statename[]={ | |
1154 | "Unused","Established","Syn Sent","Syn Recv", | |
1155 | "Fin Wait 1","Fin Wait 2","Time Wait", "Close", | |
1156 | "Close Wait","Last ACK","Listen","Closing" | |
1157 | }; | |
1158 | #endif | |
5c9f3023 | 1159 | void tcp_set_state(struct sock *sk, int state); |
1da177e4 | 1160 | |
5c9f3023 | 1161 | void tcp_done(struct sock *sk); |
1da177e4 | 1162 | |
40efc6fa | 1163 | static inline void tcp_sack_reset(struct tcp_options_received *rx_opt) |
1da177e4 LT |
1164 | { |
1165 | rx_opt->dsack = 0; | |
1da177e4 LT |
1166 | rx_opt->num_sacks = 0; |
1167 | } | |
1168 | ||
5c9f3023 | 1169 | u32 tcp_default_init_rwnd(u32 mss); |
6f021c62 ED |
1170 | void tcp_cwnd_restart(struct sock *sk, s32 delta); |
1171 | ||
1172 | static inline void tcp_slow_start_after_idle_check(struct sock *sk) | |
1173 | { | |
1174 | struct tcp_sock *tp = tcp_sk(sk); | |
1175 | s32 delta; | |
1176 | ||
1177 | if (!sysctl_tcp_slow_start_after_idle || tp->packets_out) | |
1178 | return; | |
1179 | delta = tcp_time_stamp - tp->lsndtime; | |
1180 | if (delta > inet_csk(sk)->icsk_rto) | |
1181 | tcp_cwnd_restart(sk, delta); | |
1182 | } | |
85f16525 | 1183 | |
1da177e4 | 1184 | /* Determine a window scaling and initial window to offer. */ |
5c9f3023 JP |
1185 | void tcp_select_initial_window(int __space, __u32 mss, __u32 *rcv_wnd, |
1186 | __u32 *window_clamp, int wscale_ok, | |
1187 | __u8 *rcv_wscale, __u32 init_rcv_wnd); | |
1da177e4 LT |
1188 | |
1189 | static inline int tcp_win_from_space(int space) | |
1190 | { | |
1191 | return sysctl_tcp_adv_win_scale<=0 ? | |
1192 | (space>>(-sysctl_tcp_adv_win_scale)) : | |
1193 | space - (space>>sysctl_tcp_adv_win_scale); | |
1194 | } | |
1195 | ||
105970f6 | 1196 | /* Note: caller must be prepared to deal with negative returns */ |
1da177e4 LT |
1197 | static inline int tcp_space(const struct sock *sk) |
1198 | { | |
1199 | return tcp_win_from_space(sk->sk_rcvbuf - | |
1200 | atomic_read(&sk->sk_rmem_alloc)); | |
105970f6 | 1201 | } |
1da177e4 LT |
1202 | |
1203 | static inline int tcp_full_space(const struct sock *sk) | |
1204 | { | |
105970f6 | 1205 | return tcp_win_from_space(sk->sk_rcvbuf); |
1da177e4 LT |
1206 | } |
1207 | ||
843f4a55 YC |
1208 | extern void tcp_openreq_init_rwin(struct request_sock *req, |
1209 | struct sock *sk, struct dst_entry *dst); | |
1210 | ||
5c9f3023 | 1211 | void tcp_enter_memory_pressure(struct sock *sk); |
1da177e4 | 1212 | |
1da177e4 LT |
1213 | static inline int keepalive_intvl_when(const struct tcp_sock *tp) |
1214 | { | |
1215 | return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl; | |
1216 | } | |
1217 | ||
1218 | static inline int keepalive_time_when(const struct tcp_sock *tp) | |
1219 | { | |
1220 | return tp->keepalive_time ? : sysctl_tcp_keepalive_time; | |
1221 | } | |
1222 | ||
df19a626 ED |
1223 | static inline int keepalive_probes(const struct tcp_sock *tp) |
1224 | { | |
1225 | return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes; | |
1226 | } | |
1227 | ||
6c37e5de FL |
1228 | static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp) |
1229 | { | |
1230 | const struct inet_connection_sock *icsk = &tp->inet_conn; | |
1231 | ||
1232 | return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime, | |
1233 | tcp_time_stamp - tp->rcv_tstamp); | |
1234 | } | |
1235 | ||
463c84b9 | 1236 | static inline int tcp_fin_time(const struct sock *sk) |
1da177e4 | 1237 | { |
463c84b9 ACM |
1238 | int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout; |
1239 | const int rto = inet_csk(sk)->icsk_rto; | |
1da177e4 | 1240 | |
463c84b9 ACM |
1241 | if (fin_timeout < (rto << 2) - (rto >> 1)) |
1242 | fin_timeout = (rto << 2) - (rto >> 1); | |
1da177e4 LT |
1243 | |
1244 | return fin_timeout; | |
1245 | } | |
1246 | ||
a2a385d6 ED |
1247 | static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt, |
1248 | int paws_win) | |
1da177e4 | 1249 | { |
c887e6d2 | 1250 | if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win) |
a2a385d6 | 1251 | return true; |
c887e6d2 | 1252 | if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)) |
a2a385d6 | 1253 | return true; |
bc2ce894 ED |
1254 | /* |
1255 | * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0, | |
1256 | * then following tcp messages have valid values. Ignore 0 value, | |
1257 | * or else 'negative' tsval might forbid us to accept their packets. | |
1258 | */ | |
1259 | if (!rx_opt->ts_recent) | |
a2a385d6 ED |
1260 | return true; |
1261 | return false; | |
c887e6d2 IJ |
1262 | } |
1263 | ||
a2a385d6 ED |
1264 | static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt, |
1265 | int rst) | |
c887e6d2 IJ |
1266 | { |
1267 | if (tcp_paws_check(rx_opt, 0)) | |
a2a385d6 | 1268 | return false; |
1da177e4 LT |
1269 | |
1270 | /* RST segments are not recommended to carry timestamp, | |
1271 | and, if they do, it is recommended to ignore PAWS because | |
1272 | "their cleanup function should take precedence over timestamps." | |
1273 | Certainly, it is mistake. It is necessary to understand the reasons | |
1274 | of this constraint to relax it: if peer reboots, clock may go | |
1275 | out-of-sync and half-open connections will not be reset. | |
1276 | Actually, the problem would be not existing if all | |
1277 | the implementations followed draft about maintaining clock | |
1278 | via reboots. Linux-2.2 DOES NOT! | |
1279 | ||
1280 | However, we can relax time bounds for RST segments to MSL. | |
1281 | */ | |
9d729f72 | 1282 | if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL) |
a2a385d6 ED |
1283 | return false; |
1284 | return true; | |
1da177e4 LT |
1285 | } |
1286 | ||
7970ddc8 ED |
1287 | bool tcp_oow_rate_limited(struct net *net, const struct sk_buff *skb, |
1288 | int mib_idx, u32 *last_oow_ack_time); | |
032ee423 | 1289 | |
a9c19329 | 1290 | static inline void tcp_mib_init(struct net *net) |
1da177e4 LT |
1291 | { |
1292 | /* See RFC 2012 */ | |
cf1100a7 PE |
1293 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1); |
1294 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ); | |
1295 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ); | |
1296 | TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1); | |
1da177e4 LT |
1297 | } |
1298 | ||
5af4ec23 | 1299 | /* from STCP */ |
ef9da47c | 1300 | static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp) |
0800f170 | 1301 | { |
6a438bbe | 1302 | tp->lost_skb_hint = NULL; |
ef9da47c IJ |
1303 | } |
1304 | ||
1305 | static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp) | |
1306 | { | |
1307 | tcp_clear_retrans_hints_partial(tp); | |
6a438bbe | 1308 | tp->retransmit_skb_hint = NULL; |
b7689205 IJ |
1309 | } |
1310 | ||
cfb6eeb4 YH |
1311 | /* MD5 Signature */ |
1312 | struct crypto_hash; | |
1313 | ||
a915da9b ED |
1314 | union tcp_md5_addr { |
1315 | struct in_addr a4; | |
1316 | #if IS_ENABLED(CONFIG_IPV6) | |
1317 | struct in6_addr a6; | |
1318 | #endif | |
1319 | }; | |
1320 | ||
cfb6eeb4 YH |
1321 | /* - key database */ |
1322 | struct tcp_md5sig_key { | |
a915da9b | 1323 | struct hlist_node node; |
cfb6eeb4 | 1324 | u8 keylen; |
a915da9b ED |
1325 | u8 family; /* AF_INET or AF_INET6 */ |
1326 | union tcp_md5_addr addr; | |
1327 | u8 key[TCP_MD5SIG_MAXKEYLEN]; | |
1328 | struct rcu_head rcu; | |
cfb6eeb4 YH |
1329 | }; |
1330 | ||
1331 | /* - sock block */ | |
1332 | struct tcp_md5sig_info { | |
a915da9b | 1333 | struct hlist_head head; |
a8afca03 | 1334 | struct rcu_head rcu; |
cfb6eeb4 YH |
1335 | }; |
1336 | ||
1337 | /* - pseudo header */ | |
1338 | struct tcp4_pseudohdr { | |
1339 | __be32 saddr; | |
1340 | __be32 daddr; | |
1341 | __u8 pad; | |
1342 | __u8 protocol; | |
1343 | __be16 len; | |
1344 | }; | |
1345 | ||
1346 | struct tcp6_pseudohdr { | |
1347 | struct in6_addr saddr; | |
1348 | struct in6_addr daddr; | |
1349 | __be32 len; | |
1350 | __be32 protocol; /* including padding */ | |
1351 | }; | |
1352 | ||
1353 | union tcp_md5sum_block { | |
1354 | struct tcp4_pseudohdr ip4; | |
dfd56b8b | 1355 | #if IS_ENABLED(CONFIG_IPV6) |
cfb6eeb4 YH |
1356 | struct tcp6_pseudohdr ip6; |
1357 | #endif | |
1358 | }; | |
1359 | ||
1360 | /* - pool: digest algorithm, hash description and scratch buffer */ | |
1361 | struct tcp_md5sig_pool { | |
1362 | struct hash_desc md5_desc; | |
1363 | union tcp_md5sum_block md5_blk; | |
1364 | }; | |
1365 | ||
cfb6eeb4 | 1366 | /* - functions */ |
39f8e58e ED |
1367 | int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key, |
1368 | const struct sock *sk, const struct sk_buff *skb); | |
5c9f3023 JP |
1369 | int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr, |
1370 | int family, const u8 *newkey, u8 newkeylen, gfp_t gfp); | |
1371 | int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, | |
1372 | int family); | |
1373 | struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, | |
fd3a154a | 1374 | const struct sock *addr_sk); |
cfb6eeb4 | 1375 | |
9501f972 | 1376 | #ifdef CONFIG_TCP_MD5SIG |
5c9f3023 JP |
1377 | struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk, |
1378 | const union tcp_md5_addr *addr, | |
1379 | int family); | |
a915da9b | 1380 | #define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_key) |
9501f972 | 1381 | #else |
a915da9b ED |
1382 | static inline struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk, |
1383 | const union tcp_md5_addr *addr, | |
1384 | int family) | |
1385 | { | |
1386 | return NULL; | |
1387 | } | |
9501f972 YH |
1388 | #define tcp_twsk_md5_key(twsk) NULL |
1389 | #endif | |
1390 | ||
5c9f3023 | 1391 | bool tcp_alloc_md5sig_pool(void); |
cfb6eeb4 | 1392 | |
5c9f3023 | 1393 | struct tcp_md5sig_pool *tcp_get_md5sig_pool(void); |
71cea17e ED |
1394 | static inline void tcp_put_md5sig_pool(void) |
1395 | { | |
1396 | local_bh_enable(); | |
1397 | } | |
35790c04 | 1398 | |
5c9f3023 JP |
1399 | int tcp_md5_hash_header(struct tcp_md5sig_pool *, const struct tcphdr *); |
1400 | int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *, | |
1401 | unsigned int header_len); | |
1402 | int tcp_md5_hash_key(struct tcp_md5sig_pool *hp, | |
1403 | const struct tcp_md5sig_key *key); | |
cfb6eeb4 | 1404 | |
10467163 | 1405 | /* From tcp_fastopen.c */ |
5c9f3023 JP |
1406 | void tcp_fastopen_cache_get(struct sock *sk, u16 *mss, |
1407 | struct tcp_fastopen_cookie *cookie, int *syn_loss, | |
1408 | unsigned long *last_syn_loss); | |
1409 | void tcp_fastopen_cache_set(struct sock *sk, u16 mss, | |
2646c831 DL |
1410 | struct tcp_fastopen_cookie *cookie, bool syn_lost, |
1411 | u16 try_exp); | |
783237e8 YC |
1412 | struct tcp_fastopen_request { |
1413 | /* Fast Open cookie. Size 0 means a cookie request */ | |
1414 | struct tcp_fastopen_cookie cookie; | |
1415 | struct msghdr *data; /* data in MSG_FASTOPEN */ | |
f5ddcbbb ED |
1416 | size_t size; |
1417 | int copied; /* queued in tcp_connect() */ | |
783237e8 | 1418 | }; |
783237e8 YC |
1419 | void tcp_free_fastopen_req(struct tcp_sock *tp); |
1420 | ||
10467163 JC |
1421 | extern struct tcp_fastopen_context __rcu *tcp_fastopen_ctx; |
1422 | int tcp_fastopen_reset_cipher(void *key, unsigned int len); | |
843f4a55 YC |
1423 | bool tcp_try_fastopen(struct sock *sk, struct sk_buff *skb, |
1424 | struct request_sock *req, | |
1425 | struct tcp_fastopen_cookie *foc, | |
1426 | struct dst_entry *dst); | |
222e83d2 | 1427 | void tcp_fastopen_init_key_once(bool publish); |
10467163 JC |
1428 | #define TCP_FASTOPEN_KEY_LENGTH 16 |
1429 | ||
1430 | /* Fastopen key context */ | |
1431 | struct tcp_fastopen_context { | |
7ae8639c ED |
1432 | struct crypto_cipher *tfm; |
1433 | __u8 key[TCP_FASTOPEN_KEY_LENGTH]; | |
1434 | struct rcu_head rcu; | |
10467163 JC |
1435 | }; |
1436 | ||
fe067e8a DM |
1437 | /* write queue abstraction */ |
1438 | static inline void tcp_write_queue_purge(struct sock *sk) | |
1439 | { | |
1440 | struct sk_buff *skb; | |
1441 | ||
1442 | while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) | |
3ab224be HA |
1443 | sk_wmem_free_skb(sk, skb); |
1444 | sk_mem_reclaim(sk); | |
8818a9d8 | 1445 | tcp_clear_all_retrans_hints(tcp_sk(sk)); |
fe067e8a DM |
1446 | } |
1447 | ||
cf533ea5 | 1448 | static inline struct sk_buff *tcp_write_queue_head(const struct sock *sk) |
fe067e8a | 1449 | { |
cd07a8ea | 1450 | return skb_peek(&sk->sk_write_queue); |
fe067e8a DM |
1451 | } |
1452 | ||
cf533ea5 | 1453 | static inline struct sk_buff *tcp_write_queue_tail(const struct sock *sk) |
fe067e8a | 1454 | { |
cd07a8ea | 1455 | return skb_peek_tail(&sk->sk_write_queue); |
fe067e8a DM |
1456 | } |
1457 | ||
cf533ea5 ED |
1458 | static inline struct sk_buff *tcp_write_queue_next(const struct sock *sk, |
1459 | const struct sk_buff *skb) | |
fe067e8a | 1460 | { |
cd07a8ea | 1461 | return skb_queue_next(&sk->sk_write_queue, skb); |
fe067e8a DM |
1462 | } |
1463 | ||
cf533ea5 ED |
1464 | static inline struct sk_buff *tcp_write_queue_prev(const struct sock *sk, |
1465 | const struct sk_buff *skb) | |
832d11c5 IJ |
1466 | { |
1467 | return skb_queue_prev(&sk->sk_write_queue, skb); | |
1468 | } | |
1469 | ||
fe067e8a | 1470 | #define tcp_for_write_queue(skb, sk) \ |
cd07a8ea | 1471 | skb_queue_walk(&(sk)->sk_write_queue, skb) |
fe067e8a DM |
1472 | |
1473 | #define tcp_for_write_queue_from(skb, sk) \ | |
cd07a8ea | 1474 | skb_queue_walk_from(&(sk)->sk_write_queue, skb) |
fe067e8a | 1475 | |
234b6860 | 1476 | #define tcp_for_write_queue_from_safe(skb, tmp, sk) \ |
cd07a8ea | 1477 | skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp) |
234b6860 | 1478 | |
cf533ea5 | 1479 | static inline struct sk_buff *tcp_send_head(const struct sock *sk) |
fe067e8a DM |
1480 | { |
1481 | return sk->sk_send_head; | |
1482 | } | |
1483 | ||
cd07a8ea DM |
1484 | static inline bool tcp_skb_is_last(const struct sock *sk, |
1485 | const struct sk_buff *skb) | |
1486 | { | |
1487 | return skb_queue_is_last(&sk->sk_write_queue, skb); | |
1488 | } | |
1489 | ||
cf533ea5 | 1490 | static inline void tcp_advance_send_head(struct sock *sk, const struct sk_buff *skb) |
fe067e8a | 1491 | { |
cd07a8ea | 1492 | if (tcp_skb_is_last(sk, skb)) |
fe067e8a | 1493 | sk->sk_send_head = NULL; |
cd07a8ea DM |
1494 | else |
1495 | sk->sk_send_head = tcp_write_queue_next(sk, skb); | |
fe067e8a DM |
1496 | } |
1497 | ||
1498 | static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked) | |
1499 | { | |
1500 | if (sk->sk_send_head == skb_unlinked) | |
1501 | sk->sk_send_head = NULL; | |
1502 | } | |
1503 | ||
1504 | static inline void tcp_init_send_head(struct sock *sk) | |
1505 | { | |
1506 | sk->sk_send_head = NULL; | |
1507 | } | |
1508 | ||
1509 | static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1510 | { | |
1511 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1512 | } | |
1513 | ||
1514 | static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1515 | { | |
1516 | __tcp_add_write_queue_tail(sk, skb); | |
1517 | ||
1518 | /* Queue it, remembering where we must start sending. */ | |
6859d494 | 1519 | if (sk->sk_send_head == NULL) { |
fe067e8a | 1520 | sk->sk_send_head = skb; |
6859d494 IJ |
1521 | |
1522 | if (tcp_sk(sk)->highest_sack == NULL) | |
1523 | tcp_sk(sk)->highest_sack = skb; | |
1524 | } | |
fe067e8a DM |
1525 | } |
1526 | ||
1527 | static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb) | |
1528 | { | |
1529 | __skb_queue_head(&sk->sk_write_queue, skb); | |
1530 | } | |
1531 | ||
1532 | /* Insert buff after skb on the write queue of sk. */ | |
1533 | static inline void tcp_insert_write_queue_after(struct sk_buff *skb, | |
1534 | struct sk_buff *buff, | |
1535 | struct sock *sk) | |
1536 | { | |
7de6c033 | 1537 | __skb_queue_after(&sk->sk_write_queue, skb, buff); |
fe067e8a DM |
1538 | } |
1539 | ||
43f59c89 | 1540 | /* Insert new before skb on the write queue of sk. */ |
fe067e8a DM |
1541 | static inline void tcp_insert_write_queue_before(struct sk_buff *new, |
1542 | struct sk_buff *skb, | |
1543 | struct sock *sk) | |
1544 | { | |
43f59c89 | 1545 | __skb_queue_before(&sk->sk_write_queue, skb, new); |