]>
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 JH |
67 | /* The least MTU to use for probing */ |
68 | #define TCP_BASE_MSS 512 | |
69 | ||
1da177e4 LT |
70 | /* After receiving this amount of duplicate ACKs fast retransmit starts. */ |
71 | #define TCP_FASTRETRANS_THRESH 3 | |
72 | ||
1da177e4 LT |
73 | /* Maximal number of ACKs sent quickly to accelerate slow-start. */ |
74 | #define TCP_MAX_QUICKACKS 16U | |
75 | ||
76 | /* urg_data states */ | |
77 | #define TCP_URG_VALID 0x0100 | |
78 | #define TCP_URG_NOTYET 0x0200 | |
79 | #define TCP_URG_READ 0x0400 | |
80 | ||
81 | #define TCP_RETR1 3 /* | |
82 | * This is how many retries it does before it | |
83 | * tries to figure out if the gateway is | |
84 | * down. Minimal RFC value is 3; it corresponds | |
85 | * to ~3sec-8min depending on RTO. | |
86 | */ | |
87 | ||
88 | #define TCP_RETR2 15 /* | |
89 | * This should take at least | |
90 | * 90 minutes to time out. | |
91 | * RFC1122 says that the limit is 100 sec. | |
92 | * 15 is ~13-30min depending on RTO. | |
93 | */ | |
94 | ||
6c9ff979 AB |
95 | #define TCP_SYN_RETRIES 6 /* This is how many retries are done |
96 | * when active opening a connection. | |
97 | * RFC1122 says the minimum retry MUST | |
98 | * be at least 180secs. Nevertheless | |
99 | * this value is corresponding to | |
100 | * 63secs of retransmission with the | |
101 | * current initial RTO. | |
102 | */ | |
1da177e4 | 103 | |
6c9ff979 AB |
104 | #define TCP_SYNACK_RETRIES 5 /* This is how may retries are done |
105 | * when passive opening a connection. | |
106 | * This is corresponding to 31secs of | |
107 | * retransmission with the current | |
108 | * initial RTO. | |
109 | */ | |
1da177e4 | 110 | |
1da177e4 LT |
111 | #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT |
112 | * state, about 60 seconds */ | |
113 | #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN | |
114 | /* BSD style FIN_WAIT2 deadlock breaker. | |
115 | * It used to be 3min, new value is 60sec, | |
116 | * to combine FIN-WAIT-2 timeout with | |
117 | * TIME-WAIT timer. | |
118 | */ | |
119 | ||
120 | #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */ | |
121 | #if HZ >= 100 | |
122 | #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */ | |
123 | #define TCP_ATO_MIN ((unsigned)(HZ/25)) | |
124 | #else | |
125 | #define TCP_DELACK_MIN 4U | |
126 | #define TCP_ATO_MIN 4U | |
127 | #endif | |
128 | #define TCP_RTO_MAX ((unsigned)(120*HZ)) | |
129 | #define TCP_RTO_MIN ((unsigned)(HZ/5)) | |
fd4f2cea | 130 | #define TCP_TIMEOUT_INIT ((unsigned)(1*HZ)) /* RFC6298 2.1 initial RTO value */ |
9ad7c049 JC |
131 | #define TCP_TIMEOUT_FALLBACK ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value, now |
132 | * used as a fallback RTO for the | |
133 | * initial data transmission if no | |
134 | * valid RTT sample has been acquired, | |
135 | * most likely due to retrans in 3WHS. | |
136 | */ | |
1da177e4 LT |
137 | |
138 | #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes | |
139 | * for local resources. | |
140 | */ | |
141 | ||
142 | #define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */ | |
143 | #define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */ | |
144 | #define TCP_KEEPALIVE_INTVL (75*HZ) | |
145 | ||
146 | #define MAX_TCP_KEEPIDLE 32767 | |
147 | #define MAX_TCP_KEEPINTVL 32767 | |
148 | #define MAX_TCP_KEEPCNT 127 | |
149 | #define MAX_TCP_SYNCNT 127 | |
150 | ||
151 | #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */ | |
1da177e4 LT |
152 | |
153 | #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24) | |
154 | #define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated | |
155 | * after this time. It should be equal | |
156 | * (or greater than) TCP_TIMEWAIT_LEN | |
157 | * to provide reliability equal to one | |
158 | * provided by timewait state. | |
159 | */ | |
160 | #define TCP_PAWS_WINDOW 1 /* Replay window for per-host | |
161 | * timestamps. It must be less than | |
162 | * minimal timewait lifetime. | |
163 | */ | |
1da177e4 LT |
164 | /* |
165 | * TCP option | |
166 | */ | |
105970f6 | 167 | |
1da177e4 LT |
168 | #define TCPOPT_NOP 1 /* Padding */ |
169 | #define TCPOPT_EOL 0 /* End of options */ | |
170 | #define TCPOPT_MSS 2 /* Segment size negotiating */ | |
171 | #define TCPOPT_WINDOW 3 /* Window scaling */ | |
172 | #define TCPOPT_SACK_PERM 4 /* SACK Permitted */ | |
173 | #define TCPOPT_SACK 5 /* SACK Block */ | |
174 | #define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */ | |
cfb6eeb4 | 175 | #define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */ |
2100c8d2 YC |
176 | #define TCPOPT_EXP 254 /* Experimental */ |
177 | /* Magic number to be after the option value for sharing TCP | |
178 | * experimental options. See draft-ietf-tcpm-experimental-options-00.txt | |
179 | */ | |
180 | #define TCPOPT_FASTOPEN_MAGIC 0xF989 | |
1da177e4 LT |
181 | |
182 | /* | |
183 | * TCP option lengths | |
184 | */ | |
185 | ||
186 | #define TCPOLEN_MSS 4 | |
187 | #define TCPOLEN_WINDOW 3 | |
188 | #define TCPOLEN_SACK_PERM 2 | |
189 | #define TCPOLEN_TIMESTAMP 10 | |
cfb6eeb4 | 190 | #define TCPOLEN_MD5SIG 18 |
2100c8d2 | 191 | #define TCPOLEN_EXP_FASTOPEN_BASE 4 |
1da177e4 LT |
192 | |
193 | /* But this is what stacks really send out. */ | |
194 | #define TCPOLEN_TSTAMP_ALIGNED 12 | |
195 | #define TCPOLEN_WSCALE_ALIGNED 4 | |
196 | #define TCPOLEN_SACKPERM_ALIGNED 4 | |
197 | #define TCPOLEN_SACK_BASE 2 | |
198 | #define TCPOLEN_SACK_BASE_ALIGNED 4 | |
199 | #define TCPOLEN_SACK_PERBLOCK 8 | |
cfb6eeb4 | 200 | #define TCPOLEN_MD5SIG_ALIGNED 20 |
33ad798c | 201 | #define TCPOLEN_MSS_ALIGNED 4 |
1da177e4 | 202 | |
1da177e4 LT |
203 | /* Flags in tp->nonagle */ |
204 | #define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */ | |
205 | #define TCP_NAGLE_CORK 2 /* Socket is corked */ | |
caa20d9a | 206 | #define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */ |
1da177e4 | 207 | |
36e31b0a AP |
208 | /* TCP thin-stream limits */ |
209 | #define TCP_THIN_LINEAR_RETRIES 6 /* After 6 linear retries, do exp. backoff */ | |
210 | ||
7eb38527 | 211 | /* TCP initial congestion window as per draft-hkchu-tcpm-initcwnd-01 */ |
442b9635 DM |
212 | #define TCP_INIT_CWND 10 |
213 | ||
cf60af03 YC |
214 | /* Bit Flags for sysctl_tcp_fastopen */ |
215 | #define TFO_CLIENT_ENABLE 1 | |
10467163 | 216 | #define TFO_SERVER_ENABLE 2 |
67da22d2 | 217 | #define TFO_CLIENT_NO_COOKIE 4 /* Data in SYN w/o cookie option */ |
cf60af03 | 218 | |
10467163 JC |
219 | /* Accept SYN data w/o any cookie option */ |
220 | #define TFO_SERVER_COOKIE_NOT_REQD 0x200 | |
221 | ||
222 | /* Force enable TFO on all listeners, i.e., not requiring the | |
223 | * TCP_FASTOPEN socket option. SOCKOPT1/2 determine how to set max_qlen. | |
224 | */ | |
225 | #define TFO_SERVER_WO_SOCKOPT1 0x400 | |
226 | #define TFO_SERVER_WO_SOCKOPT2 0x800 | |
10467163 | 227 | |
295ff7ed ACM |
228 | extern struct inet_timewait_death_row tcp_death_row; |
229 | ||
1da177e4 | 230 | /* sysctl variables for tcp */ |
1da177e4 LT |
231 | extern int sysctl_tcp_timestamps; |
232 | extern int sysctl_tcp_window_scaling; | |
233 | extern int sysctl_tcp_sack; | |
234 | extern int sysctl_tcp_fin_timeout; | |
1da177e4 LT |
235 | extern int sysctl_tcp_keepalive_time; |
236 | extern int sysctl_tcp_keepalive_probes; | |
237 | extern int sysctl_tcp_keepalive_intvl; | |
238 | extern int sysctl_tcp_syn_retries; | |
239 | extern int sysctl_tcp_synack_retries; | |
240 | extern int sysctl_tcp_retries1; | |
241 | extern int sysctl_tcp_retries2; | |
242 | extern int sysctl_tcp_orphan_retries; | |
243 | extern int sysctl_tcp_syncookies; | |
2100c8d2 | 244 | extern int sysctl_tcp_fastopen; |
1da177e4 LT |
245 | extern int sysctl_tcp_retrans_collapse; |
246 | extern int sysctl_tcp_stdurg; | |
247 | extern int sysctl_tcp_rfc1337; | |
248 | extern int sysctl_tcp_abort_on_overflow; | |
249 | extern int sysctl_tcp_max_orphans; | |
1da177e4 LT |
250 | extern int sysctl_tcp_fack; |
251 | extern int sysctl_tcp_reordering; | |
dca145ff | 252 | extern int sysctl_tcp_max_reordering; |
1da177e4 | 253 | extern int sysctl_tcp_dsack; |
a4fe34bf | 254 | extern long sysctl_tcp_mem[3]; |
1da177e4 LT |
255 | extern int sysctl_tcp_wmem[3]; |
256 | extern int sysctl_tcp_rmem[3]; | |
257 | extern int sysctl_tcp_app_win; | |
258 | extern int sysctl_tcp_adv_win_scale; | |
259 | extern int sysctl_tcp_tw_reuse; | |
260 | extern int sysctl_tcp_frto; | |
261 | extern int sysctl_tcp_low_latency; | |
1da177e4 | 262 | extern int sysctl_tcp_nometrics_save; |
1da177e4 LT |
263 | extern int sysctl_tcp_moderate_rcvbuf; |
264 | extern int sysctl_tcp_tso_win_divisor; | |
5d424d5a JH |
265 | extern int sysctl_tcp_mtu_probing; |
266 | extern int sysctl_tcp_base_mss; | |
15d99e02 | 267 | extern int sysctl_tcp_workaround_signed_windows; |
35089bb2 | 268 | extern int sysctl_tcp_slow_start_after_idle; |
36e31b0a | 269 | extern int sysctl_tcp_thin_linear_timeouts; |
7e380175 | 270 | extern int sysctl_tcp_thin_dupack; |
eed530b6 | 271 | extern int sysctl_tcp_early_retrans; |
46d3ceab | 272 | extern int sysctl_tcp_limit_output_bytes; |
282f23c6 | 273 | extern int sysctl_tcp_challenge_ack_limit; |
c9bee3b7 | 274 | extern unsigned int sysctl_tcp_notsent_lowat; |
95bd09eb | 275 | extern int sysctl_tcp_min_tso_segs; |
f54b3111 | 276 | extern int sysctl_tcp_autocorking; |
1da177e4 | 277 | |
8d987e5c | 278 | extern atomic_long_t tcp_memory_allocated; |
1748376b | 279 | extern struct percpu_counter tcp_sockets_allocated; |
1da177e4 LT |
280 | extern int tcp_memory_pressure; |
281 | ||
1da177e4 LT |
282 | /* |
283 | * The next routines deal with comparing 32 bit unsigned ints | |
284 | * and worry about wraparound (automatic with unsigned arithmetic). | |
285 | */ | |
286 | ||
a2a385d6 | 287 | static inline bool before(__u32 seq1, __u32 seq2) |
1da177e4 | 288 | { |
0d630cc0 | 289 | return (__s32)(seq1-seq2) < 0; |
1da177e4 | 290 | } |
9a036b9c | 291 | #define after(seq2, seq1) before(seq1, seq2) |
1da177e4 LT |
292 | |
293 | /* is s2<=s1<=s3 ? */ | |
a2a385d6 | 294 | static inline bool between(__u32 seq1, __u32 seq2, __u32 seq3) |
1da177e4 LT |
295 | { |
296 | return seq3 - seq2 >= seq1 - seq2; | |
297 | } | |
298 | ||
efcdbf24 AS |
299 | static inline bool tcp_out_of_memory(struct sock *sk) |
300 | { | |
301 | if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF && | |
302 | sk_memory_allocated(sk) > sk_prot_mem_limits(sk, 2)) | |
303 | return true; | |
304 | return false; | |
305 | } | |
306 | ||
ad1af0fe | 307 | static inline bool tcp_too_many_orphans(struct sock *sk, int shift) |
e4fd5da3 | 308 | { |
ad1af0fe DM |
309 | struct percpu_counter *ocp = sk->sk_prot->orphan_count; |
310 | int orphans = percpu_counter_read_positive(ocp); | |
311 | ||
312 | if (orphans << shift > sysctl_tcp_max_orphans) { | |
313 | orphans = percpu_counter_sum_positive(ocp); | |
314 | if (orphans << shift > sysctl_tcp_max_orphans) | |
315 | return true; | |
316 | } | |
ad1af0fe | 317 | return false; |
e4fd5da3 | 318 | } |
1da177e4 | 319 | |
5c9f3023 | 320 | bool tcp_check_oom(struct sock *sk, int shift); |
efcdbf24 | 321 | |
a0f82f64 FW |
322 | /* syncookies: remember time of last synqueue overflow */ |
323 | static inline void tcp_synq_overflow(struct sock *sk) | |
324 | { | |
325 | tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies; | |
326 | } | |
327 | ||
328 | /* syncookies: no recent synqueue overflow on this listening socket? */ | |
a2a385d6 | 329 | static inline bool tcp_synq_no_recent_overflow(const struct sock *sk) |
a0f82f64 FW |
330 | { |
331 | unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp; | |
9ad7c049 | 332 | return time_after(jiffies, last_overflow + TCP_TIMEOUT_FALLBACK); |
a0f82f64 FW |
333 | } |
334 | ||
1da177e4 LT |
335 | extern struct proto tcp_prot; |
336 | ||
57ef42d5 PE |
337 | #define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field) |
338 | #define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field) | |
339 | #define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field) | |
340 | #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val) | |
aa2ea058 | 341 | #define TCP_ADD_STATS(net, field, val) SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val) |
1da177e4 | 342 | |
5c9f3023 JP |
343 | void tcp_tasklet_init(void); |
344 | ||
345 | void tcp_v4_err(struct sk_buff *skb, u32); | |
346 | ||
347 | void tcp_shutdown(struct sock *sk, int how); | |
348 | ||
349 | void tcp_v4_early_demux(struct sk_buff *skb); | |
350 | int tcp_v4_rcv(struct sk_buff *skb); | |
351 | ||
352 | int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw); | |
353 | int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
354 | size_t size); | |
355 | int tcp_sendpage(struct sock *sk, struct page *page, int offset, size_t size, | |
356 | int flags); | |
357 | void tcp_release_cb(struct sock *sk); | |
358 | void tcp_wfree(struct sk_buff *skb); | |
359 | void tcp_write_timer_handler(struct sock *sk); | |
360 | void tcp_delack_timer_handler(struct sock *sk); | |
361 | int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg); | |
362 | int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, | |
363 | const struct tcphdr *th, unsigned int len); | |
364 | void tcp_rcv_established(struct sock *sk, struct sk_buff *skb, | |
365 | const struct tcphdr *th, unsigned int len); | |
366 | void tcp_rcv_space_adjust(struct sock *sk); | |
5c9f3023 JP |
367 | int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp); |
368 | void tcp_twsk_destructor(struct sock *sk); | |
369 | ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos, | |
370 | struct pipe_inode_info *pipe, size_t len, | |
371 | unsigned int flags); | |
9c55e01c | 372 | |
463c84b9 ACM |
373 | static inline void tcp_dec_quickack_mode(struct sock *sk, |
374 | const unsigned int pkts) | |
1da177e4 | 375 | { |
463c84b9 | 376 | struct inet_connection_sock *icsk = inet_csk(sk); |
fc6415bc | 377 | |
463c84b9 ACM |
378 | if (icsk->icsk_ack.quick) { |
379 | if (pkts >= icsk->icsk_ack.quick) { | |
380 | icsk->icsk_ack.quick = 0; | |
fc6415bc | 381 | /* Leaving quickack mode we deflate ATO. */ |
463c84b9 | 382 | icsk->icsk_ack.ato = TCP_ATO_MIN; |
fc6415bc | 383 | } else |
463c84b9 | 384 | icsk->icsk_ack.quick -= pkts; |
1da177e4 LT |
385 | } |
386 | } | |
387 | ||
bdf1ee5d IJ |
388 | #define TCP_ECN_OK 1 |
389 | #define TCP_ECN_QUEUE_CWR 2 | |
390 | #define TCP_ECN_DEMAND_CWR 4 | |
7a269ffa | 391 | #define TCP_ECN_SEEN 8 |
bdf1ee5d | 392 | |
fd2c3ef7 | 393 | enum tcp_tw_status { |
1da177e4 LT |
394 | TCP_TW_SUCCESS = 0, |
395 | TCP_TW_RST = 1, | |
396 | TCP_TW_ACK = 2, | |
397 | TCP_TW_SYN = 3 | |
398 | }; | |
399 | ||
400 | ||
5c9f3023 JP |
401 | enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw, |
402 | struct sk_buff *skb, | |
403 | const struct tcphdr *th); | |
404 | struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb, | |
405 | struct request_sock *req, struct request_sock **prev, | |
406 | bool fastopen); | |
407 | int tcp_child_process(struct sock *parent, struct sock *child, | |
408 | struct sk_buff *skb); | |
5ae344c9 | 409 | void tcp_enter_loss(struct sock *sk); |
5c9f3023 JP |
410 | void tcp_clear_retrans(struct tcp_sock *tp); |
411 | void tcp_update_metrics(struct sock *sk); | |
412 | void tcp_init_metrics(struct sock *sk); | |
413 | void tcp_metrics_init(void); | |
414 | bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst, | |
a26552af | 415 | bool paws_check, bool timestamps); |
5c9f3023 JP |
416 | bool tcp_remember_stamp(struct sock *sk); |
417 | bool tcp_tw_remember_stamp(struct inet_timewait_sock *tw); | |
418 | void tcp_fetch_timewait_stamp(struct sock *sk, struct dst_entry *dst); | |
419 | void tcp_disable_fack(struct tcp_sock *tp); | |
420 | void tcp_close(struct sock *sk, long timeout); | |
421 | void tcp_init_sock(struct sock *sk); | |
422 | unsigned int tcp_poll(struct file *file, struct socket *sock, | |
423 | struct poll_table_struct *wait); | |
424 | int tcp_getsockopt(struct sock *sk, int level, int optname, | |
425 | char __user *optval, int __user *optlen); | |
426 | int tcp_setsockopt(struct sock *sk, int level, int optname, | |
427 | char __user *optval, unsigned int optlen); | |
428 | int compat_tcp_getsockopt(struct sock *sk, int level, int optname, | |
53d3176b | 429 | char __user *optval, int __user *optlen); |
5c9f3023 | 430 | int compat_tcp_setsockopt(struct sock *sk, int level, int optname, |
53d3176b | 431 | char __user *optval, unsigned int optlen); |
5c9f3023 JP |
432 | void tcp_set_keepalive(struct sock *sk, int val); |
433 | void tcp_syn_ack_timeout(struct sock *sk, struct request_sock *req); | |
434 | int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
435 | size_t len, int nonblock, int flags, int *addr_len); | |
436 | void tcp_parse_options(const struct sk_buff *skb, | |
437 | struct tcp_options_received *opt_rx, | |
438 | int estab, struct tcp_fastopen_cookie *foc); | |
439 | const u8 *tcp_parse_md5sig_option(const struct tcphdr *th); | |
7d5d5525 | 440 | |
1da177e4 LT |
441 | /* |
442 | * TCP v4 functions exported for the inet6 API | |
443 | */ | |
444 | ||
5c9f3023 | 445 | void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb); |
4fab9071 | 446 | void tcp_v4_mtu_reduced(struct sock *sk); |
5c9f3023 JP |
447 | int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb); |
448 | struct sock *tcp_create_openreq_child(struct sock *sk, | |
449 | struct request_sock *req, | |
450 | struct sk_buff *skb); | |
451 | struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb, | |
452 | struct request_sock *req, | |
453 | struct dst_entry *dst); | |
454 | int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb); | |
455 | int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len); | |
456 | int tcp_connect(struct sock *sk); | |
457 | struct sk_buff *tcp_make_synack(struct sock *sk, struct dst_entry *dst, | |
458 | struct request_sock *req, | |
459 | struct tcp_fastopen_cookie *foc); | |
460 | int tcp_disconnect(struct sock *sk, int flags); | |
1da177e4 | 461 | |
370816ae | 462 | void tcp_finish_connect(struct sock *sk, struct sk_buff *skb); |
292e8d8c | 463 | int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size); |
63d02d15 | 464 | void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb); |
1da177e4 | 465 | |
1da177e4 | 466 | /* From syncookies.c */ |
5c9f3023 JP |
467 | int __cookie_v4_check(const struct iphdr *iph, const struct tcphdr *th, |
468 | u32 cookie); | |
461b74c3 | 469 | struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb); |
e05c82d3 | 470 | #ifdef CONFIG_SYN_COOKIES |
8c27bd75 | 471 | |
63262315 | 472 | /* Syncookies use a monotonic timer which increments every 60 seconds. |
8c27bd75 FW |
473 | * This counter is used both as a hash input and partially encoded into |
474 | * the cookie value. A cookie is only validated further if the delta | |
475 | * between the current counter value and the encoded one is less than this, | |
63262315 | 476 | * i.e. a sent cookie is valid only at most for 2*60 seconds (or less if |
8c27bd75 FW |
477 | * the counter advances immediately after a cookie is generated). |
478 | */ | |
479 | #define MAX_SYNCOOKIE_AGE 2 | |
480 | ||
481 | static inline u32 tcp_cookie_time(void) | |
482 | { | |
63262315 ED |
483 | u64 val = get_jiffies_64(); |
484 | ||
485 | do_div(val, 60 * HZ); | |
486 | return val; | |
8c27bd75 FW |
487 | } |
488 | ||
5c9f3023 JP |
489 | u32 __cookie_v4_init_sequence(const struct iphdr *iph, const struct tcphdr *th, |
490 | u16 *mssp); | |
57b47553 OP |
491 | __u32 cookie_v4_init_sequence(struct sock *sk, const struct sk_buff *skb, |
492 | __u16 *mss); | |
5c9f3023 | 493 | __u32 cookie_init_timestamp(struct request_sock *req); |
f1673381 FW |
494 | bool cookie_timestamp_decode(struct tcp_options_received *opt); |
495 | bool cookie_ecn_ok(const struct tcp_options_received *opt, | |
f7b3bec6 | 496 | const struct net *net, const struct dst_entry *dst); |
4dfc2817 | 497 | |
c6aefafb | 498 | /* From net/ipv6/syncookies.c */ |
5c9f3023 JP |
499 | int __cookie_v6_check(const struct ipv6hdr *iph, const struct tcphdr *th, |
500 | u32 cookie); | |
501 | struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb); | |
f1673381 | 502 | |
5c9f3023 JP |
503 | u32 __cookie_v6_init_sequence(const struct ipv6hdr *iph, |
504 | const struct tcphdr *th, u16 *mssp); | |
505 | __u32 cookie_v6_init_sequence(struct sock *sk, const struct sk_buff *skb, | |
506 | __u16 *mss); | |
e05c82d3 | 507 | #endif |
1da177e4 LT |
508 | /* tcp_output.c */ |
509 | ||
5c9f3023 JP |
510 | void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss, |
511 | int nonagle); | |
512 | bool tcp_may_send_now(struct sock *sk); | |
513 | int __tcp_retransmit_skb(struct sock *, struct sk_buff *); | |
514 | int tcp_retransmit_skb(struct sock *, struct sk_buff *); | |
515 | void tcp_retransmit_timer(struct sock *sk); | |
516 | void tcp_xmit_retransmit_queue(struct sock *); | |
517 | void tcp_simple_retransmit(struct sock *); | |
518 | int tcp_trim_head(struct sock *, struct sk_buff *, u32); | |
6cc55e09 | 519 | int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int, gfp_t); |
5c9f3023 JP |
520 | |
521 | void tcp_send_probe0(struct sock *); | |
522 | void tcp_send_partial(struct sock *); | |
523 | int tcp_write_wakeup(struct sock *); | |
524 | void tcp_send_fin(struct sock *sk); | |
525 | void tcp_send_active_reset(struct sock *sk, gfp_t priority); | |
526 | int tcp_send_synack(struct sock *); | |
527 | bool tcp_syn_flood_action(struct sock *sk, const struct sk_buff *skb, | |
528 | const char *proto); | |
529 | void tcp_push_one(struct sock *, unsigned int mss_now); | |
530 | void tcp_send_ack(struct sock *sk); | |
531 | void tcp_send_delayed_ack(struct sock *sk); | |
532 | void tcp_send_loss_probe(struct sock *sk); | |
533 | bool tcp_schedule_loss_probe(struct sock *sk); | |
1da177e4 | 534 | |
a762a980 | 535 | /* tcp_input.c */ |
5c9f3023 JP |
536 | void tcp_resume_early_retransmit(struct sock *sk); |
537 | void tcp_rearm_rto(struct sock *sk); | |
538 | void tcp_reset(struct sock *sk); | |
a762a980 | 539 | |
1da177e4 | 540 | /* tcp_timer.c */ |
5c9f3023 | 541 | void tcp_init_xmit_timers(struct sock *); |
463c84b9 ACM |
542 | static inline void tcp_clear_xmit_timers(struct sock *sk) |
543 | { | |
544 | inet_csk_clear_xmit_timers(sk); | |
545 | } | |
1da177e4 | 546 | |
5c9f3023 JP |
547 | unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu); |
548 | unsigned int tcp_current_mss(struct sock *sk); | |
0c54b85f IJ |
549 | |
550 | /* Bound MSS / TSO packet size with the half of the window */ | |
551 | static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize) | |
552 | { | |
01f83d69 AK |
553 | int cutoff; |
554 | ||
555 | /* When peer uses tiny windows, there is no use in packetizing | |
556 | * to sub-MSS pieces for the sake of SWS or making sure there | |
557 | * are enough packets in the pipe for fast recovery. | |
558 | * | |
559 | * On the other hand, for extremely large MSS devices, handling | |
560 | * smaller than MSS windows in this way does make sense. | |
561 | */ | |
562 | if (tp->max_window >= 512) | |
563 | cutoff = (tp->max_window >> 1); | |
564 | else | |
565 | cutoff = tp->max_window; | |
566 | ||
567 | if (cutoff && pktsize > cutoff) | |
568 | return max_t(int, cutoff, 68U - tp->tcp_header_len); | |
0c54b85f IJ |
569 | else |
570 | return pktsize; | |
571 | } | |
1da177e4 | 572 | |
17b085ea | 573 | /* tcp.c */ |
5c9f3023 | 574 | void tcp_get_info(const struct sock *, struct tcp_info *); |
1da177e4 LT |
575 | |
576 | /* Read 'sendfile()'-style from a TCP socket */ | |
577 | typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *, | |
578 | unsigned int, size_t); | |
5c9f3023 JP |
579 | int tcp_read_sock(struct sock *sk, read_descriptor_t *desc, |
580 | sk_read_actor_t recv_actor); | |
1da177e4 | 581 | |
5c9f3023 | 582 | void tcp_initialize_rcv_mss(struct sock *sk); |
1da177e4 | 583 | |
5c9f3023 JP |
584 | int tcp_mtu_to_mss(struct sock *sk, int pmtu); |
585 | int tcp_mss_to_mtu(struct sock *sk, int mss); | |
586 | void tcp_mtup_init(struct sock *sk); | |
587 | void tcp_init_buffer_space(struct sock *sk); | |
5d424d5a | 588 | |
f1ecd5d9 DL |
589 | static inline void tcp_bound_rto(const struct sock *sk) |
590 | { | |
591 | if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX) | |
592 | inet_csk(sk)->icsk_rto = TCP_RTO_MAX; | |
593 | } | |
594 | ||
595 | static inline u32 __tcp_set_rto(const struct tcp_sock *tp) | |
596 | { | |
740b0f18 | 597 | return usecs_to_jiffies((tp->srtt_us >> 3) + tp->rttvar_us); |
f1ecd5d9 DL |
598 | } |
599 | ||
40efc6fa | 600 | static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd) |
1da177e4 LT |
601 | { |
602 | tp->pred_flags = htonl((tp->tcp_header_len << 26) | | |
603 | ntohl(TCP_FLAG_ACK) | | |
604 | snd_wnd); | |
605 | } | |
606 | ||
40efc6fa | 607 | static inline void tcp_fast_path_on(struct tcp_sock *tp) |
1da177e4 LT |
608 | { |
609 | __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale); | |
610 | } | |
611 | ||
9e412ba7 | 612 | static inline void tcp_fast_path_check(struct sock *sk) |
1da177e4 | 613 | { |
9e412ba7 IJ |
614 | struct tcp_sock *tp = tcp_sk(sk); |
615 | ||
b03efcfb | 616 | if (skb_queue_empty(&tp->out_of_order_queue) && |
1da177e4 LT |
617 | tp->rcv_wnd && |
618 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf && | |
619 | !tp->urg_data) | |
620 | tcp_fast_path_on(tp); | |
621 | } | |
622 | ||
0c266898 SS |
623 | /* Compute the actual rto_min value */ |
624 | static inline u32 tcp_rto_min(struct sock *sk) | |
625 | { | |
cf533ea5 | 626 | const struct dst_entry *dst = __sk_dst_get(sk); |
0c266898 SS |
627 | u32 rto_min = TCP_RTO_MIN; |
628 | ||
629 | if (dst && dst_metric_locked(dst, RTAX_RTO_MIN)) | |
630 | rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN); | |
631 | return rto_min; | |
632 | } | |
633 | ||
740b0f18 ED |
634 | static inline u32 tcp_rto_min_us(struct sock *sk) |
635 | { | |
636 | return jiffies_to_usecs(tcp_rto_min(sk)); | |
637 | } | |
638 | ||
1da177e4 LT |
639 | /* Compute the actual receive window we are currently advertising. |
640 | * Rcv_nxt can be after the window if our peer push more data | |
641 | * than the offered window. | |
642 | */ | |
40efc6fa | 643 | static inline u32 tcp_receive_window(const struct tcp_sock *tp) |
1da177e4 LT |
644 | { |
645 | s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt; | |
646 | ||
647 | if (win < 0) | |
648 | win = 0; | |
649 | return (u32) win; | |
650 | } | |
651 | ||
652 | /* Choose a new window, without checks for shrinking, and without | |
653 | * scaling applied to the result. The caller does these things | |
654 | * if necessary. This is a "raw" window selection. | |
655 | */ | |
5c9f3023 | 656 | u32 __tcp_select_window(struct sock *sk); |
1da177e4 | 657 | |
ee995283 PE |
658 | void tcp_send_window_probe(struct sock *sk); |
659 | ||
1da177e4 LT |
660 | /* TCP timestamps are only 32-bits, this causes a slight |
661 | * complication on 64-bit systems since we store a snapshot | |
31f34269 SH |
662 | * of jiffies in the buffer control blocks below. We decided |
663 | * to use only the low 32-bits of jiffies and hide the ugly | |
1da177e4 LT |
664 | * casts with the following macro. |
665 | */ | |
666 | #define tcp_time_stamp ((__u32)(jiffies)) | |
667 | ||
7faee5c0 ED |
668 | static inline u32 tcp_skb_timestamp(const struct sk_buff *skb) |
669 | { | |
670 | return skb->skb_mstamp.stamp_jiffies; | |
671 | } | |
672 | ||
673 | ||
a3433f35 CG |
674 | #define tcp_flag_byte(th) (((u_int8_t *)th)[13]) |
675 | ||
676 | #define TCPHDR_FIN 0x01 | |
677 | #define TCPHDR_SYN 0x02 | |
678 | #define TCPHDR_RST 0x04 | |
679 | #define TCPHDR_PSH 0x08 | |
680 | #define TCPHDR_ACK 0x10 | |
681 | #define TCPHDR_URG 0x20 | |
682 | #define TCPHDR_ECE 0x40 | |
683 | #define TCPHDR_CWR 0x80 | |
684 | ||
caa20d9a | 685 | /* This is what the send packet queuing engine uses to pass |
f86586fa ED |
686 | * TCP per-packet control information to the transmission code. |
687 | * We also store the host-order sequence numbers in here too. | |
688 | * This is 44 bytes if IPV6 is enabled. | |
689 | * If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately. | |
1da177e4 LT |
690 | */ |
691 | struct tcp_skb_cb { | |
1da177e4 LT |
692 | __u32 seq; /* Starting sequence number */ |
693 | __u32 end_seq; /* SEQ + FIN + SYN + datalen */ | |
cd7d8498 ED |
694 | union { |
695 | /* Note : tcp_tw_isn is used in input path only | |
696 | * (isn chosen by tcp_timewait_state_process()) | |
697 | * | |
698 | * tcp_gso_segs is used in write queue only, | |
699 | * cf tcp_skb_pcount() | |
700 | */ | |
701 | __u32 tcp_tw_isn; | |
702 | __u32 tcp_gso_segs; | |
703 | }; | |
4de075e0 | 704 | __u8 tcp_flags; /* TCP header flags. (tcp[13]) */ |
f4f9f6e7 | 705 | |
1da177e4 LT |
706 | __u8 sacked; /* State flags for SACK/FACK. */ |
707 | #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */ | |
708 | #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */ | |
709 | #define TCPCB_LOST 0x04 /* SKB is lost */ | |
710 | #define TCPCB_TAGBITS 0x07 /* All tag bits */ | |
9d186cac | 711 | #define TCPCB_REPAIRED 0x10 /* SKB repaired (no skb_mstamp) */ |
1da177e4 | 712 | #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */ |
9d186cac AV |
713 | #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS| \ |
714 | TCPCB_REPAIRED) | |
1da177e4 | 715 | |
f4f9f6e7 NC |
716 | __u8 ip_dsfield; /* IPv4 tos or IPv6 dsfield */ |
717 | /* 1 byte hole */ | |
1da177e4 | 718 | __u32 ack_seq; /* Sequence number ACK'd */ |
971f10ec ED |
719 | union { |
720 | struct inet_skb_parm h4; | |
721 | #if IS_ENABLED(CONFIG_IPV6) | |
722 | struct inet6_skb_parm h6; | |
723 | #endif | |
724 | } header; /* For incoming frames */ | |
1da177e4 LT |
725 | }; |
726 | ||
727 | #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0])) | |
728 | ||
870c3151 | 729 | |
815afe17 | 730 | #if IS_ENABLED(CONFIG_IPV6) |
870c3151 ED |
731 | /* This is the variant of inet6_iif() that must be used by TCP, |
732 | * as TCP moves IP6CB into a different location in skb->cb[] | |
733 | */ | |
734 | static inline int tcp_v6_iif(const struct sk_buff *skb) | |
735 | { | |
736 | return TCP_SKB_CB(skb)->header.h6.iif; | |
737 | } | |
815afe17 | 738 | #endif |
870c3151 | 739 | |
1da177e4 LT |
740 | /* Due to TSO, an SKB can be composed of multiple actual |
741 | * packets. To keep these tracked properly, we use this. | |
bd14b1b2 | 742 | */ |
1da177e4 | 743 | static inline int tcp_skb_pcount(const struct sk_buff *skb) |
bd14b1b2 | 744 | { |
cd7d8498 ED |
745 | return TCP_SKB_CB(skb)->tcp_gso_segs; |
746 | } | |
bd14b1b2 | 747 | |
cd7d8498 ED |
748 | static inline void tcp_skb_pcount_set(struct sk_buff *skb, int segs) |
749 | { | |
750 | TCP_SKB_CB(skb)->tcp_gso_segs = segs; | |
bd14b1b2 ED |
751 | } |
752 | ||
cd7d8498 | 753 | static inline void tcp_skb_pcount_add(struct sk_buff *skb, int segs) |
1da177e4 | 754 | { |
cd7d8498 | 755 | TCP_SKB_CB(skb)->tcp_gso_segs += segs; |
1da177e4 LT |
756 | } |
757 | ||
758 | /* This is valid iff tcp_skb_pcount() > 1. */ | |
759 | static inline int tcp_skb_mss(const struct sk_buff *skb) | |
760 | { | |
7967168c | 761 | return skb_shinfo(skb)->gso_size; |
1da177e4 LT |
762 | } |
763 | ||
317a76f9 SH |
764 | /* Events passed to congestion control interface */ |
765 | enum tcp_ca_event { | |
766 | CA_EVENT_TX_START, /* first transmit when no packets in flight */ | |
767 | CA_EVENT_CWND_RESTART, /* congestion window restart */ | |
768 | CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */ | |
317a76f9 | 769 | CA_EVENT_LOSS, /* loss timeout */ |
9890092e FW |
770 | CA_EVENT_ECN_NO_CE, /* ECT set, but not CE marked */ |
771 | CA_EVENT_ECN_IS_CE, /* received CE marked IP packet */ | |
772 | CA_EVENT_DELAYED_ACK, /* Delayed ack is sent */ | |
773 | CA_EVENT_NON_DELAYED_ACK, | |
7354c8c3 FW |
774 | }; |
775 | ||
9890092e | 776 | /* Information about inbound ACK, passed to cong_ops->in_ack_event() */ |
7354c8c3 | 777 | enum tcp_ca_ack_event_flags { |
9890092e FW |
778 | CA_ACK_SLOWPATH = (1 << 0), /* In slow path processing */ |
779 | CA_ACK_WIN_UPDATE = (1 << 1), /* ACK updated window */ | |
780 | CA_ACK_ECE = (1 << 2), /* ECE bit is set on ack */ | |
317a76f9 SH |
781 | }; |
782 | ||
783 | /* | |
784 | * Interface for adding new TCP congestion control handlers | |
785 | */ | |
786 | #define TCP_CA_NAME_MAX 16 | |
3ff825b2 SH |
787 | #define TCP_CA_MAX 128 |
788 | #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX) | |
789 | ||
c5c6a8ab DB |
790 | #define TCP_CA_UNSPEC 0 |
791 | ||
30e502a3 | 792 | /* Algorithm can be set on socket without CAP_NET_ADMIN privileges */ |
164891aa | 793 | #define TCP_CONG_NON_RESTRICTED 0x1 |
30e502a3 DB |
794 | /* Requires ECN/ECT set on all packets */ |
795 | #define TCP_CONG_NEEDS_ECN 0x2 | |
164891aa | 796 | |
317a76f9 SH |
797 | struct tcp_congestion_ops { |
798 | struct list_head list; | |
c5c6a8ab DB |
799 | u32 key; |
800 | u32 flags; | |
317a76f9 SH |
801 | |
802 | /* initialize private data (optional) */ | |
6687e988 | 803 | void (*init)(struct sock *sk); |
317a76f9 | 804 | /* cleanup private data (optional) */ |
6687e988 | 805 | void (*release)(struct sock *sk); |
317a76f9 SH |
806 | |
807 | /* return slow start threshold (required) */ | |
6687e988 | 808 | u32 (*ssthresh)(struct sock *sk); |
317a76f9 | 809 | /* do new cwnd calculation (required) */ |
24901551 | 810 | void (*cong_avoid)(struct sock *sk, u32 ack, u32 acked); |
317a76f9 | 811 | /* call before changing ca_state (optional) */ |
6687e988 | 812 | void (*set_state)(struct sock *sk, u8 new_state); |
317a76f9 | 813 | /* call when cwnd event occurs (optional) */ |
6687e988 | 814 | void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev); |
7354c8c3 FW |
815 | /* call when ack arrives (optional) */ |
816 | void (*in_ack_event)(struct sock *sk, u32 flags); | |
317a76f9 | 817 | /* new value of cwnd after loss (optional) */ |
6687e988 | 818 | u32 (*undo_cwnd)(struct sock *sk); |
317a76f9 | 819 | /* hook for packet ack accounting (optional) */ |
30cfd0ba | 820 | void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us); |
73c1f4a0 | 821 | /* get info for inet_diag (optional) */ |
6687e988 | 822 | void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb); |
317a76f9 SH |
823 | |
824 | char name[TCP_CA_NAME_MAX]; | |
825 | struct module *owner; | |
826 | }; | |
827 | ||
5c9f3023 JP |
828 | int tcp_register_congestion_control(struct tcp_congestion_ops *type); |
829 | void tcp_unregister_congestion_control(struct tcp_congestion_ops *type); | |
317a76f9 | 830 | |
55d8694f | 831 | void tcp_assign_congestion_control(struct sock *sk); |
5c9f3023 JP |
832 | void tcp_init_congestion_control(struct sock *sk); |
833 | void tcp_cleanup_congestion_control(struct sock *sk); | |
834 | int tcp_set_default_congestion_control(const char *name); | |
835 | void tcp_get_default_congestion_control(char *name); | |
836 | void tcp_get_available_congestion_control(char *buf, size_t len); | |
837 | void tcp_get_allowed_congestion_control(char *buf, size_t len); | |
838 | int tcp_set_allowed_congestion_control(char *allowed); | |
839 | int tcp_set_congestion_control(struct sock *sk, const char *name); | |
a12a601e | 840 | void tcp_slow_start(struct tcp_sock *tp, u32 acked); |
5c9f3023 | 841 | void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w); |
317a76f9 | 842 | |
5c9f3023 | 843 | u32 tcp_reno_ssthresh(struct sock *sk); |
24901551 | 844 | void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 acked); |
a8acfbac | 845 | extern struct tcp_congestion_ops tcp_reno; |
317a76f9 | 846 | |
c5c6a8ab DB |
847 | struct tcp_congestion_ops *tcp_ca_find_key(u32 key); |
848 | u32 tcp_ca_get_key_by_name(const char *name); | |
ea697639 | 849 | #ifdef CONFIG_INET |
c5c6a8ab | 850 | char *tcp_ca_get_name_by_key(u32 key, char *buffer); |
ea697639 DB |
851 | #else |
852 | static inline char *tcp_ca_get_name_by_key(u32 key, char *buffer) | |
853 | { | |
854 | return NULL; | |
855 | } | |
856 | #endif | |
c5c6a8ab | 857 | |
30e502a3 DB |
858 | static inline bool tcp_ca_needs_ecn(const struct sock *sk) |
859 | { | |
860 | const struct inet_connection_sock *icsk = inet_csk(sk); | |
861 | ||
862 | return icsk->icsk_ca_ops->flags & TCP_CONG_NEEDS_ECN; | |
863 | } | |
864 | ||
6687e988 | 865 | static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state) |
317a76f9 | 866 | { |
6687e988 ACM |
867 | struct inet_connection_sock *icsk = inet_csk(sk); |
868 | ||
869 | if (icsk->icsk_ca_ops->set_state) | |
870 | icsk->icsk_ca_ops->set_state(sk, ca_state); | |
871 | icsk->icsk_ca_state = ca_state; | |
317a76f9 SH |
872 | } |
873 | ||
6687e988 | 874 | static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event) |
317a76f9 | 875 | { |
6687e988 ACM |
876 | const struct inet_connection_sock *icsk = inet_csk(sk); |
877 | ||
878 | if (icsk->icsk_ca_ops->cwnd_event) | |
879 | icsk->icsk_ca_ops->cwnd_event(sk, event); | |
317a76f9 SH |
880 | } |
881 | ||
e60402d0 IJ |
882 | /* These functions determine how the current flow behaves in respect of SACK |
883 | * handling. SACK is negotiated with the peer, and therefore it can vary | |
884 | * between different flows. | |
885 | * | |
886 | * tcp_is_sack - SACK enabled | |
887 | * tcp_is_reno - No SACK | |
888 | * tcp_is_fack - FACK enabled, implies SACK enabled | |
889 | */ | |
890 | static inline int tcp_is_sack(const struct tcp_sock *tp) | |
891 | { | |
892 | return tp->rx_opt.sack_ok; | |
893 | } | |
894 | ||
a2a385d6 | 895 | static inline bool tcp_is_reno(const struct tcp_sock *tp) |
e60402d0 IJ |
896 | { |
897 | return !tcp_is_sack(tp); | |
898 | } | |
899 | ||
a2a385d6 | 900 | static inline bool tcp_is_fack(const struct tcp_sock *tp) |
e60402d0 | 901 | { |
ab56222a | 902 | return tp->rx_opt.sack_ok & TCP_FACK_ENABLED; |
e60402d0 IJ |
903 | } |
904 | ||
905 | static inline void tcp_enable_fack(struct tcp_sock *tp) | |
906 | { | |
ab56222a | 907 | tp->rx_opt.sack_ok |= TCP_FACK_ENABLED; |
e60402d0 IJ |
908 | } |
909 | ||
eed530b6 YC |
910 | /* TCP early-retransmit (ER) is similar to but more conservative than |
911 | * the thin-dupack feature. Enable ER only if thin-dupack is disabled. | |
912 | */ | |
913 | static inline void tcp_enable_early_retrans(struct tcp_sock *tp) | |
914 | { | |
915 | tp->do_early_retrans = sysctl_tcp_early_retrans && | |
6ba8a3b1 ND |
916 | sysctl_tcp_early_retrans < 4 && !sysctl_tcp_thin_dupack && |
917 | sysctl_tcp_reordering == 3; | |
eed530b6 YC |
918 | } |
919 | ||
920 | static inline void tcp_disable_early_retrans(struct tcp_sock *tp) | |
921 | { | |
922 | tp->do_early_retrans = 0; | |
923 | } | |
924 | ||
83ae4088 IJ |
925 | static inline unsigned int tcp_left_out(const struct tcp_sock *tp) |
926 | { | |
927 | return tp->sacked_out + tp->lost_out; | |
928 | } | |
929 | ||
1da177e4 LT |
930 | /* This determines how many packets are "in the network" to the best |
931 | * of our knowledge. In many cases it is conservative, but where | |
932 | * detailed information is available from the receiver (via SACK | |
933 | * blocks etc.) we can make more aggressive calculations. | |
934 | * | |
935 | * Use this for decisions involving congestion control, use just | |
936 | * tp->packets_out to determine if the send queue is empty or not. | |
937 | * | |
938 | * Read this equation as: | |
939 | * | |
940 | * "Packets sent once on transmission queue" MINUS | |
941 | * "Packets left network, but not honestly ACKed yet" PLUS | |
942 | * "Packets fast retransmitted" | |
943 | */ | |
40efc6fa | 944 | static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp) |
1da177e4 | 945 | { |
83ae4088 | 946 | return tp->packets_out - tcp_left_out(tp) + tp->retrans_out; |
1da177e4 LT |
947 | } |
948 | ||
0b6a05c1 IJ |
949 | #define TCP_INFINITE_SSTHRESH 0x7fffffff |
950 | ||
951 | static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp) | |
952 | { | |
953 | return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH; | |
954 | } | |
955 | ||
684bad11 YC |
956 | static inline bool tcp_in_cwnd_reduction(const struct sock *sk) |
957 | { | |
958 | return (TCPF_CA_CWR | TCPF_CA_Recovery) & | |
959 | (1 << inet_csk(sk)->icsk_ca_state); | |
960 | } | |
961 | ||
1da177e4 | 962 | /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd. |
684bad11 | 963 | * The exception is cwnd reduction phase, when cwnd is decreasing towards |
1da177e4 LT |
964 | * ssthresh. |
965 | */ | |
6687e988 | 966 | static inline __u32 tcp_current_ssthresh(const struct sock *sk) |
1da177e4 | 967 | { |
6687e988 | 968 | const struct tcp_sock *tp = tcp_sk(sk); |
cf533ea5 | 969 | |
684bad11 | 970 | if (tcp_in_cwnd_reduction(sk)) |
1da177e4 LT |
971 | return tp->snd_ssthresh; |
972 | else | |
973 | return max(tp->snd_ssthresh, | |
974 | ((tp->snd_cwnd >> 1) + | |
975 | (tp->snd_cwnd >> 2))); | |
976 | } | |
977 | ||
b9c4595b IJ |
978 | /* Use define here intentionally to get WARN_ON location shown at the caller */ |
979 | #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out) | |
1da177e4 | 980 | |
5ee2c941 | 981 | void tcp_enter_cwr(struct sock *sk); |
5c9f3023 | 982 | __u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst); |
1da177e4 | 983 | |
6b5a5c0d NC |
984 | /* The maximum number of MSS of available cwnd for which TSO defers |
985 | * sending if not using sysctl_tcp_tso_win_divisor. | |
986 | */ | |
987 | static inline __u32 tcp_max_tso_deferred_mss(const struct tcp_sock *tp) | |
988 | { | |
989 | return 3; | |
990 | } | |
991 | ||
1da177e4 | 992 | /* Slow start with delack produces 3 packets of burst, so that |
dd9e0dda JH |
993 | * it is safe "de facto". This will be the default - same as |
994 | * the default reordering threshold - but if reordering increases, | |
995 | * we must be able to allow cwnd to burst at least this much in order | |
996 | * to not pull it back when holes are filled. | |
1da177e4 LT |
997 | */ |
998 | static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp) | |
999 | { | |
dd9e0dda | 1000 | return tp->reordering; |
1da177e4 LT |
1001 | } |
1002 | ||
90840def IJ |
1003 | /* Returns end sequence number of the receiver's advertised window */ |
1004 | static inline u32 tcp_wnd_end(const struct tcp_sock *tp) | |
1005 | { | |
1006 | return tp->snd_una + tp->snd_wnd; | |
1007 | } | |
e114a710 ED |
1008 | |
1009 | /* We follow the spirit of RFC2861 to validate cwnd but implement a more | |
1010 | * flexible approach. The RFC suggests cwnd should not be raised unless | |
ca8a2263 NC |
1011 | * it was fully used previously. And that's exactly what we do in |
1012 | * congestion avoidance mode. But in slow start we allow cwnd to grow | |
1013 | * as long as the application has used half the cwnd. | |
e114a710 ED |
1014 | * Example : |
1015 | * cwnd is 10 (IW10), but application sends 9 frames. | |
1016 | * We allow cwnd to reach 18 when all frames are ACKed. | |
1017 | * This check is safe because it's as aggressive as slow start which already | |
1018 | * risks 100% overshoot. The advantage is that we discourage application to | |
1019 | * either send more filler packets or data to artificially blow up the cwnd | |
1020 | * usage, and allow application-limited process to probe bw more aggressively. | |
e114a710 | 1021 | */ |
24901551 | 1022 | static inline bool tcp_is_cwnd_limited(const struct sock *sk) |
e114a710 ED |
1023 | { |
1024 | const struct tcp_sock *tp = tcp_sk(sk); | |
1025 | ||
ca8a2263 NC |
1026 | /* If in slow start, ensure cwnd grows to twice what was ACKed. */ |
1027 | if (tp->snd_cwnd <= tp->snd_ssthresh) | |
1028 | return tp->snd_cwnd < 2 * tp->max_packets_out; | |
1029 | ||
1030 | return tp->is_cwnd_limited; | |
e114a710 | 1031 | } |
f4805ede | 1032 | |
9e412ba7 | 1033 | static inline void tcp_check_probe_timer(struct sock *sk) |
1da177e4 | 1034 | { |
cf533ea5 | 1035 | const struct tcp_sock *tp = tcp_sk(sk); |
463c84b9 | 1036 | const struct inet_connection_sock *icsk = inet_csk(sk); |
9e412ba7 | 1037 | |
463c84b9 | 1038 | if (!tp->packets_out && !icsk->icsk_pending) |
3f421baa ACM |
1039 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, |
1040 | icsk->icsk_rto, TCP_RTO_MAX); | |
1da177e4 LT |
1041 | } |
1042 | ||
ee7537b6 | 1043 | static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq) |
1da177e4 LT |
1044 | { |
1045 | tp->snd_wl1 = seq; | |
1046 | } | |
1047 | ||
ee7537b6 | 1048 | static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq) |
1da177e4 LT |
1049 | { |
1050 | tp->snd_wl1 = seq; | |
1051 | } | |
1052 | ||
1da177e4 LT |
1053 | /* |
1054 | * Calculate(/check) TCP checksum | |
1055 | */ | |
ba7808ea FD |
1056 | static inline __sum16 tcp_v4_check(int len, __be32 saddr, |
1057 | __be32 daddr, __wsum base) | |
1da177e4 LT |
1058 | { |
1059 | return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base); | |
1060 | } | |
1061 | ||
b51655b9 | 1062 | static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 1063 | { |
fb286bb2 | 1064 | return __skb_checksum_complete(skb); |
1da177e4 LT |
1065 | } |
1066 | ||
a2a385d6 | 1067 | static inline bool tcp_checksum_complete(struct sk_buff *skb) |
1da177e4 | 1068 | { |
60476372 | 1069 | return !skb_csum_unnecessary(skb) && |
1da177e4 LT |
1070 | __tcp_checksum_complete(skb); |
1071 | } | |
1072 | ||
1073 | /* Prequeue for VJ style copy to user, combined with checksumming. */ | |
1074 | ||
40efc6fa | 1075 | static inline void tcp_prequeue_init(struct tcp_sock *tp) |
1da177e4 LT |
1076 | { |
1077 | tp->ucopy.task = NULL; | |
1078 | tp->ucopy.len = 0; | |
1079 | tp->ucopy.memory = 0; | |
1080 | skb_queue_head_init(&tp->ucopy.prequeue); | |
1081 | } | |
1082 | ||
5c9f3023 | 1083 | bool tcp_prequeue(struct sock *sk, struct sk_buff *skb); |
1da177e4 LT |
1084 | |
1085 | #undef STATE_TRACE | |
1086 | ||
1087 | #ifdef STATE_TRACE | |
1088 | static const char *statename[]={ | |
1089 | "Unused","Established","Syn Sent","Syn Recv", | |
1090 | "Fin Wait 1","Fin Wait 2","Time Wait", "Close", | |
1091 | "Close Wait","Last ACK","Listen","Closing" | |
1092 | }; | |
1093 | #endif | |
5c9f3023 | 1094 | void tcp_set_state(struct sock *sk, int state); |
1da177e4 | 1095 | |
5c9f3023 | 1096 | void tcp_done(struct sock *sk); |
1da177e4 | 1097 | |
40efc6fa | 1098 | static inline void tcp_sack_reset(struct tcp_options_received *rx_opt) |
1da177e4 LT |
1099 | { |
1100 | rx_opt->dsack = 0; | |
1da177e4 LT |
1101 | rx_opt->num_sacks = 0; |
1102 | } | |
1103 | ||
5c9f3023 | 1104 | u32 tcp_default_init_rwnd(u32 mss); |
85f16525 | 1105 | |
1da177e4 | 1106 | /* Determine a window scaling and initial window to offer. */ |
5c9f3023 JP |
1107 | void tcp_select_initial_window(int __space, __u32 mss, __u32 *rcv_wnd, |
1108 | __u32 *window_clamp, int wscale_ok, | |
1109 | __u8 *rcv_wscale, __u32 init_rcv_wnd); | |
1da177e4 LT |
1110 | |
1111 | static inline int tcp_win_from_space(int space) | |
1112 | { | |
1113 | return sysctl_tcp_adv_win_scale<=0 ? | |
1114 | (space>>(-sysctl_tcp_adv_win_scale)) : | |
1115 | space - (space>>sysctl_tcp_adv_win_scale); | |
1116 | } | |
1117 | ||
105970f6 | 1118 | /* Note: caller must be prepared to deal with negative returns */ |
1da177e4 LT |
1119 | static inline int tcp_space(const struct sock *sk) |
1120 | { | |
1121 | return tcp_win_from_space(sk->sk_rcvbuf - | |
1122 | atomic_read(&sk->sk_rmem_alloc)); | |
105970f6 | 1123 | } |
1da177e4 LT |
1124 | |
1125 | static inline int tcp_full_space(const struct sock *sk) | |
1126 | { | |
105970f6 | 1127 | return tcp_win_from_space(sk->sk_rcvbuf); |
1da177e4 LT |
1128 | } |
1129 | ||
40efc6fa SH |
1130 | static inline void tcp_openreq_init(struct request_sock *req, |
1131 | struct tcp_options_received *rx_opt, | |
e0f802fb | 1132 | struct sk_buff *skb, struct sock *sk) |
1da177e4 | 1133 | { |
2e6599cb ACM |
1134 | struct inet_request_sock *ireq = inet_rsk(req); |
1135 | ||
1da177e4 | 1136 | req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */ |
4dfc2817 | 1137 | req->cookie_ts = 0; |
2e6599cb | 1138 | tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq; |
10467163 | 1139 | tcp_rsk(req)->rcv_nxt = TCP_SKB_CB(skb)->seq + 1; |
86c6a2c7 | 1140 | tcp_rsk(req)->snt_synack = tcp_time_stamp; |
1da177e4 LT |
1141 | req->mss = rx_opt->mss_clamp; |
1142 | req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0; | |
2e6599cb ACM |
1143 | ireq->tstamp_ok = rx_opt->tstamp_ok; |
1144 | ireq->sack_ok = rx_opt->sack_ok; | |
1145 | ireq->snd_wscale = rx_opt->snd_wscale; | |
1146 | ireq->wscale_ok = rx_opt->wscale_ok; | |
1147 | ireq->acked = 0; | |
1148 | ireq->ecn_ok = 0; | |
634fb979 | 1149 | ireq->ir_rmt_port = tcp_hdr(skb)->source; |
b44084c2 | 1150 | ireq->ir_num = ntohs(tcp_hdr(skb)->dest); |
e0f802fb | 1151 | ireq->ir_mark = inet_request_mark(sk, skb); |
1da177e4 LT |
1152 | } |
1153 | ||
843f4a55 YC |
1154 | extern void tcp_openreq_init_rwin(struct request_sock *req, |
1155 | struct sock *sk, struct dst_entry *dst); | |
1156 | ||
5c9f3023 | 1157 | void tcp_enter_memory_pressure(struct sock *sk); |
1da177e4 | 1158 | |
1da177e4 LT |
1159 | static inline int keepalive_intvl_when(const struct tcp_sock *tp) |
1160 | { | |
1161 | return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl; | |
1162 | } | |
1163 | ||
1164 | static inline int keepalive_time_when(const struct tcp_sock *tp) | |
1165 | { | |
1166 | return tp->keepalive_time ? : sysctl_tcp_keepalive_time; | |
1167 | } | |
1168 | ||
df19a626 ED |
1169 | static inline int keepalive_probes(const struct tcp_sock *tp) |
1170 | { | |
1171 | return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes; | |
1172 | } | |
1173 | ||
6c37e5de FL |
1174 | static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp) |
1175 | { | |
1176 | const struct inet_connection_sock *icsk = &tp->inet_conn; | |
1177 | ||
1178 | return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime, | |
1179 | tcp_time_stamp - tp->rcv_tstamp); | |
1180 | } | |
1181 | ||
463c84b9 | 1182 | static inline int tcp_fin_time(const struct sock *sk) |
1da177e4 | 1183 | { |
463c84b9 ACM |
1184 | int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout; |
1185 | const int rto = inet_csk(sk)->icsk_rto; | |
1da177e4 | 1186 | |
463c84b9 ACM |
1187 | if (fin_timeout < (rto << 2) - (rto >> 1)) |
1188 | fin_timeout = (rto << 2) - (rto >> 1); | |
1da177e4 LT |
1189 | |
1190 | return fin_timeout; | |
1191 | } | |
1192 | ||
a2a385d6 ED |
1193 | static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt, |
1194 | int paws_win) | |
1da177e4 | 1195 | { |
c887e6d2 | 1196 | if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win) |
a2a385d6 | 1197 | return true; |
c887e6d2 | 1198 | if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)) |
a2a385d6 | 1199 | return true; |
bc2ce894 ED |
1200 | /* |
1201 | * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0, | |
1202 | * then following tcp messages have valid values. Ignore 0 value, | |
1203 | * or else 'negative' tsval might forbid us to accept their packets. | |
1204 | */ | |
1205 | if (!rx_opt->ts_recent) | |
a2a385d6 ED |
1206 | return true; |
1207 | return false; | |
c887e6d2 IJ |
1208 | } |
1209 | ||
a2a385d6 ED |
1210 | static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt, |
1211 | int rst) | |
c887e6d2 IJ |
1212 | { |
1213 | if (tcp_paws_check(rx_opt, 0)) | |
a2a385d6 | 1214 | return false; |
1da177e4 LT |
1215 | |
1216 | /* RST segments are not recommended to carry timestamp, | |
1217 | and, if they do, it is recommended to ignore PAWS because | |
1218 | "their cleanup function should take precedence over timestamps." | |
1219 | Certainly, it is mistake. It is necessary to understand the reasons | |
1220 | of this constraint to relax it: if peer reboots, clock may go | |
1221 | out-of-sync and half-open connections will not be reset. | |
1222 | Actually, the problem would be not existing if all | |
1223 | the implementations followed draft about maintaining clock | |
1224 | via reboots. Linux-2.2 DOES NOT! | |
1225 | ||
1226 | However, we can relax time bounds for RST segments to MSL. | |
1227 | */ | |
9d729f72 | 1228 | if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL) |
a2a385d6 ED |
1229 | return false; |
1230 | return true; | |
1da177e4 LT |
1231 | } |
1232 | ||
a9c19329 | 1233 | static inline void tcp_mib_init(struct net *net) |
1da177e4 LT |
1234 | { |
1235 | /* See RFC 2012 */ | |
cf1100a7 PE |
1236 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1); |
1237 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ); | |
1238 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ); | |
1239 | TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1); | |
1da177e4 LT |
1240 | } |
1241 | ||
5af4ec23 | 1242 | /* from STCP */ |
ef9da47c | 1243 | static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp) |
0800f170 | 1244 | { |
6a438bbe | 1245 | tp->lost_skb_hint = NULL; |
ef9da47c IJ |
1246 | } |
1247 | ||
1248 | static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp) | |
1249 | { | |
1250 | tcp_clear_retrans_hints_partial(tp); | |
6a438bbe | 1251 | tp->retransmit_skb_hint = NULL; |
b7689205 IJ |
1252 | } |
1253 | ||
cfb6eeb4 YH |
1254 | /* MD5 Signature */ |
1255 | struct crypto_hash; | |
1256 | ||
a915da9b ED |
1257 | union tcp_md5_addr { |
1258 | struct in_addr a4; | |
1259 | #if IS_ENABLED(CONFIG_IPV6) | |
1260 | struct in6_addr a6; | |
1261 | #endif | |
1262 | }; | |
1263 | ||
cfb6eeb4 YH |
1264 | /* - key database */ |
1265 | struct tcp_md5sig_key { | |
a915da9b | 1266 | struct hlist_node node; |
cfb6eeb4 | 1267 | u8 keylen; |
a915da9b ED |
1268 | u8 family; /* AF_INET or AF_INET6 */ |
1269 | union tcp_md5_addr addr; | |
1270 | u8 key[TCP_MD5SIG_MAXKEYLEN]; | |
1271 | struct rcu_head rcu; | |
cfb6eeb4 YH |
1272 | }; |
1273 | ||
1274 | /* - sock block */ | |
1275 | struct tcp_md5sig_info { | |
a915da9b | 1276 | struct hlist_head head; |
a8afca03 | 1277 | struct rcu_head rcu; |
cfb6eeb4 YH |
1278 | }; |
1279 | ||
1280 | /* - pseudo header */ | |
1281 | struct tcp4_pseudohdr { | |
1282 | __be32 saddr; | |
1283 | __be32 daddr; | |
1284 | __u8 pad; | |
1285 | __u8 protocol; | |
1286 | __be16 len; | |
1287 | }; | |
1288 | ||
1289 | struct tcp6_pseudohdr { | |
1290 | struct in6_addr saddr; | |
1291 | struct in6_addr daddr; | |
1292 | __be32 len; | |
1293 | __be32 protocol; /* including padding */ | |
1294 | }; | |
1295 | ||
1296 | union tcp_md5sum_block { | |
1297 | struct tcp4_pseudohdr ip4; | |
dfd56b8b | 1298 | #if IS_ENABLED(CONFIG_IPV6) |
cfb6eeb4 YH |
1299 | struct tcp6_pseudohdr ip6; |
1300 | #endif | |
1301 | }; | |
1302 | ||
1303 | /* - pool: digest algorithm, hash description and scratch buffer */ | |
1304 | struct tcp_md5sig_pool { | |
1305 | struct hash_desc md5_desc; | |
1306 | union tcp_md5sum_block md5_blk; | |
1307 | }; | |
1308 | ||
cfb6eeb4 | 1309 | /* - functions */ |
5c9f3023 JP |
1310 | int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key, |
1311 | const struct sock *sk, const struct request_sock *req, | |
1312 | const struct sk_buff *skb); | |
1313 | int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr, | |
1314 | int family, const u8 *newkey, u8 newkeylen, gfp_t gfp); | |
1315 | int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, | |
1316 | int family); | |
1317 | struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, | |
a915da9b | 1318 | struct sock *addr_sk); |
cfb6eeb4 | 1319 | |
9501f972 | 1320 | #ifdef CONFIG_TCP_MD5SIG |
5c9f3023 JP |
1321 | struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk, |
1322 | const union tcp_md5_addr *addr, | |
1323 | int family); | |
a915da9b | 1324 | #define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_key) |
9501f972 | 1325 | #else |
a915da9b ED |
1326 | static inline struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk, |
1327 | const union tcp_md5_addr *addr, | |
1328 | int family) | |
1329 | { | |
1330 | return NULL; | |
1331 | } | |
9501f972 YH |
1332 | #define tcp_twsk_md5_key(twsk) NULL |
1333 | #endif | |
1334 | ||
5c9f3023 | 1335 | bool tcp_alloc_md5sig_pool(void); |
cfb6eeb4 | 1336 | |
5c9f3023 | 1337 | struct tcp_md5sig_pool *tcp_get_md5sig_pool(void); |
71cea17e ED |
1338 | static inline void tcp_put_md5sig_pool(void) |
1339 | { | |
1340 | local_bh_enable(); | |
1341 | } | |
35790c04 | 1342 | |
5c9f3023 JP |
1343 | int tcp_md5_hash_header(struct tcp_md5sig_pool *, const struct tcphdr *); |
1344 | int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *, | |
1345 | unsigned int header_len); | |
1346 | int tcp_md5_hash_key(struct tcp_md5sig_pool *hp, | |
1347 | const struct tcp_md5sig_key *key); | |
cfb6eeb4 | 1348 | |
10467163 | 1349 | /* From tcp_fastopen.c */ |
5c9f3023 JP |
1350 | void tcp_fastopen_cache_get(struct sock *sk, u16 *mss, |
1351 | struct tcp_fastopen_cookie *cookie, int *syn_loss, | |
1352 | unsigned long *last_syn_loss); | |
1353 | void tcp_fastopen_cache_set(struct sock *sk, u16 mss, | |
1354 | struct tcp_fastopen_cookie *cookie, bool syn_lost); | |
783237e8 YC |
1355 | struct tcp_fastopen_request { |
1356 | /* Fast Open cookie. Size 0 means a cookie request */ | |
1357 | struct tcp_fastopen_cookie cookie; | |
1358 | struct msghdr *data; /* data in MSG_FASTOPEN */ | |
f5ddcbbb ED |
1359 | size_t size; |
1360 | int copied; /* queued in tcp_connect() */ | |
783237e8 | 1361 | }; |
783237e8 YC |
1362 | void tcp_free_fastopen_req(struct tcp_sock *tp); |
1363 | ||
10467163 JC |
1364 | extern struct tcp_fastopen_context __rcu *tcp_fastopen_ctx; |
1365 | int tcp_fastopen_reset_cipher(void *key, unsigned int len); | |
843f4a55 YC |
1366 | bool tcp_try_fastopen(struct sock *sk, struct sk_buff *skb, |
1367 | struct request_sock *req, | |
1368 | struct tcp_fastopen_cookie *foc, | |
1369 | struct dst_entry *dst); | |
222e83d2 | 1370 | void tcp_fastopen_init_key_once(bool publish); |
10467163 JC |
1371 | #define TCP_FASTOPEN_KEY_LENGTH 16 |
1372 | ||
1373 | /* Fastopen key context */ | |
1374 | struct tcp_fastopen_context { | |
7ae8639c ED |
1375 | struct crypto_cipher *tfm; |
1376 | __u8 key[TCP_FASTOPEN_KEY_LENGTH]; | |
1377 | struct rcu_head rcu; | |
10467163 JC |
1378 | }; |
1379 | ||
fe067e8a DM |
1380 | /* write queue abstraction */ |
1381 | static inline void tcp_write_queue_purge(struct sock *sk) | |
1382 | { | |
1383 | struct sk_buff *skb; | |
1384 | ||
1385 | while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) | |
3ab224be HA |
1386 | sk_wmem_free_skb(sk, skb); |
1387 | sk_mem_reclaim(sk); | |
8818a9d8 | 1388 | tcp_clear_all_retrans_hints(tcp_sk(sk)); |
fe067e8a DM |
1389 | } |
1390 | ||
cf533ea5 | 1391 | static inline struct sk_buff *tcp_write_queue_head(const struct sock *sk) |
fe067e8a | 1392 | { |
cd07a8ea | 1393 | return skb_peek(&sk->sk_write_queue); |
fe067e8a DM |
1394 | } |
1395 | ||
cf533ea5 | 1396 | static inline struct sk_buff *tcp_write_queue_tail(const struct sock *sk) |
fe067e8a | 1397 | { |
cd07a8ea | 1398 | return skb_peek_tail(&sk->sk_write_queue); |
fe067e8a DM |
1399 | } |
1400 | ||
cf533ea5 ED |
1401 | static inline struct sk_buff *tcp_write_queue_next(const struct sock *sk, |
1402 | const struct sk_buff *skb) | |
fe067e8a | 1403 | { |
cd07a8ea | 1404 | return skb_queue_next(&sk->sk_write_queue, skb); |
fe067e8a DM |
1405 | } |
1406 | ||
cf533ea5 ED |
1407 | static inline struct sk_buff *tcp_write_queue_prev(const struct sock *sk, |
1408 | const struct sk_buff *skb) | |
832d11c5 IJ |
1409 | { |
1410 | return skb_queue_prev(&sk->sk_write_queue, skb); | |
1411 | } | |
1412 | ||
fe067e8a | 1413 | #define tcp_for_write_queue(skb, sk) \ |
cd07a8ea | 1414 | skb_queue_walk(&(sk)->sk_write_queue, skb) |
fe067e8a DM |
1415 | |
1416 | #define tcp_for_write_queue_from(skb, sk) \ | |
cd07a8ea | 1417 | skb_queue_walk_from(&(sk)->sk_write_queue, skb) |
fe067e8a | 1418 | |
234b6860 | 1419 | #define tcp_for_write_queue_from_safe(skb, tmp, sk) \ |
cd07a8ea | 1420 | skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp) |
234b6860 | 1421 | |
cf533ea5 | 1422 | static inline struct sk_buff *tcp_send_head(const struct sock *sk) |
fe067e8a DM |
1423 | { |
1424 | return sk->sk_send_head; | |
1425 | } | |
1426 | ||
cd07a8ea DM |
1427 | static inline bool tcp_skb_is_last(const struct sock *sk, |
1428 | const struct sk_buff *skb) | |
1429 | { | |
1430 | return skb_queue_is_last(&sk->sk_write_queue, skb); | |
1431 | } | |
1432 | ||
cf533ea5 | 1433 | static inline void tcp_advance_send_head(struct sock *sk, const struct sk_buff *skb) |
fe067e8a | 1434 | { |
cd07a8ea | 1435 | if (tcp_skb_is_last(sk, skb)) |
fe067e8a | 1436 | sk->sk_send_head = NULL; |
cd07a8ea DM |
1437 | else |
1438 | sk->sk_send_head = tcp_write_queue_next(sk, skb); | |
fe067e8a DM |
1439 | } |
1440 | ||
1441 | static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked) | |
1442 | { | |
1443 | if (sk->sk_send_head == skb_unlinked) | |
1444 | sk->sk_send_head = NULL; | |
1445 | } | |
1446 | ||
1447 | static inline void tcp_init_send_head(struct sock *sk) | |
1448 | { | |
1449 | sk->sk_send_head = NULL; | |
1450 | } | |
1451 | ||
1452 | static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1453 | { | |
1454 | __skb_queue_tail(&sk->sk_write_queue, skb); | |
1455 | } | |
1456 | ||
1457 | static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) | |
1458 | { | |
1459 | __tcp_add_write_queue_tail(sk, skb); | |
1460 | ||
1461 | /* Queue it, remembering where we must start sending. */ | |
6859d494 | 1462 | if (sk->sk_send_head == NULL) { |
fe067e8a | 1463 | sk->sk_send_head = skb; |
6859d494 IJ |
1464 | |
1465 | if (tcp_sk(sk)->highest_sack == NULL) | |
1466 | tcp_sk(sk)->highest_sack = skb; | |
1467 | } | |
fe067e8a DM |
1468 | } |
1469 | ||
1470 | static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb) | |
1471 | { | |
1472 | __skb_queue_head(&sk->sk_write_queue, skb); | |
1473 | } | |
1474 | ||
1475 | /* Insert buff after skb on the write queue of sk. */ | |
1476 | static inline void tcp_insert_write_queue_after(struct sk_buff *skb, | |
1477 | struct sk_buff *buff, | |
1478 | struct sock *sk) | |
1479 | { | |
7de6c033 | 1480 | __skb_queue_after(&sk->sk_write_queue, skb, buff); |
fe067e8a DM |
1481 | } |
1482 | ||
43f59c89 | 1483 | /* Insert new before skb on the write queue of sk. */ |
fe067e8a DM |
1484 | static inline void tcp_insert_write_queue_before(struct sk_buff *new, |
1485 | struct sk_buff *skb, | |
1486 | struct sock *sk) | |
1487 | { | |
43f59c89 | 1488 | __skb_queue_before(&sk->sk_write_queue, skb, new); |