]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Copyright (c) 2000-2014 Apple Computer, Inc. All rights reserved. | |
3 | * | |
4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ | |
5 | * | |
6 | * This file contains Original Code and/or Modifications of Original Code | |
7 | * as defined in and that are subject to the Apple Public Source License | |
8 | * Version 2.0 (the 'License'). You may not use this file except in | |
9 | * compliance with the License. The rights granted to you under the License | |
10 | * may not be used to create, or enable the creation or redistribution of, | |
11 | * unlawful or unlicensed copies of an Apple operating system, or to | |
12 | * circumvent, violate, or enable the circumvention or violation of, any | |
13 | * terms of an Apple operating system software license agreement. | |
14 | * | |
15 | * Please obtain a copy of the License at | |
16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. | |
17 | * | |
18 | * The Original Code and all software distributed under the License are | |
19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER | |
20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, | |
21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, | |
22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. | |
23 | * Please see the License for the specific language governing rights and | |
24 | * limitations under the License. | |
25 | * | |
26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ | |
27 | */ | |
28 | /* | |
29 | * Copyright (c) 1982, 1986, 1993 | |
30 | * The Regents of the University of California. All rights reserved. | |
31 | * | |
32 | * Redistribution and use in source and binary forms, with or without | |
33 | * modification, are permitted provided that the following conditions | |
34 | * are met: | |
35 | * 1. Redistributions of source code must retain the above copyright | |
36 | * notice, this list of conditions and the following disclaimer. | |
37 | * 2. Redistributions in binary form must reproduce the above copyright | |
38 | * notice, this list of conditions and the following disclaimer in the | |
39 | * documentation and/or other materials provided with the distribution. | |
40 | * 3. All advertising materials mentioning features or use of this software | |
41 | * must display the following acknowledgement: | |
42 | * This product includes software developed by the University of | |
43 | * California, Berkeley and its contributors. | |
44 | * 4. Neither the name of the University nor the names of its contributors | |
45 | * may be used to endorse or promote products derived from this software | |
46 | * without specific prior written permission. | |
47 | * | |
48 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND | |
49 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | |
50 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | |
51 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | |
52 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | |
53 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | |
54 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | |
55 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | |
56 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | |
57 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | |
58 | * SUCH DAMAGE. | |
59 | * | |
60 | * @(#)tcp_timer.h 8.1 (Berkeley) 6/10/93 | |
61 | * $FreeBSD: src/sys/netinet/tcp_timer.h,v 1.18 1999/12/29 04:41:03 peter Exp $ | |
62 | */ | |
63 | ||
64 | #ifndef _NETINET_TCP_TIMER_H_ | |
65 | #define _NETINET_TCP_TIMER_H_ | |
66 | #include <sys/appleapiopts.h> | |
67 | ||
68 | #ifdef BSD_KERNEL_PRIVATE | |
69 | #include <kern/thread_call.h> | |
70 | #endif /* BSD_KERNEL_PRIVATE */ | |
71 | ||
72 | /* Keep the external definition the same for binary compatibility */ | |
73 | #define TCPT_NTIMERS_EXT 4 | |
74 | ||
75 | /* | |
76 | * Definitions of the TCP timers. | |
77 | * | |
78 | * The TCPT_PTO timer is used for probing for a tail loss in a send window. | |
79 | * If this probe gets acknowledged using SACK, it will allow the connection | |
80 | * to enter fast-recovery instead of hitting a retransmit timeout. A probe | |
81 | * timeout will send the last unacknowledged segment to generate more acks | |
82 | * with SACK information which can be used for fast-retransmiting the lost | |
83 | * packets. This will fire in the order of 10ms. | |
84 | * | |
85 | * The TCPT_REXMT timer is used to force retransmissions. | |
86 | * The TCP has the TCPT_REXMT timer set whenever segments | |
87 | * have been sent for which ACKs are expected but not yet | |
88 | * received. If an ACK is received which advances tp->snd_una, | |
89 | * then the retransmit timer is cleared (if there are no more | |
90 | * outstanding segments) or reset to the base value (if there | |
91 | * are more ACKs expected). Whenever the retransmit timer goes off, | |
92 | * we retransmit one unacknowledged segment, and do a backoff | |
93 | * on the retransmit timer. | |
94 | * | |
95 | * The TCPT_DELACK timer is used for transmitting delayed acknowledgements | |
96 | * if an acknowledgement was delayed in anticipation of a new segment. | |
97 | * | |
98 | * The TCPT_PERSIST timer is used to keep window size information | |
99 | * flowing even if the window goes shut. If all previous transmissions | |
100 | * have been acknowledged(so that there are no retransmissions in progress), | |
101 | * and the window is too small to bother sending anything, then we start | |
102 | * the TCPT_PERSIST timer. When it expires, if the window is nonzero, | |
103 | * we go to transmit state. Otherwise, at intervals send a single byte | |
104 | * into the peer's window to force him to update our window information. | |
105 | * We do this at most as often as TCPT_PERSMIN time intervals, | |
106 | * but no more frequently than the current estimate of round-trip | |
107 | * packet time. The TCPT_PERSIST timer is cleared whenever we receive | |
108 | * a window update from the peer. | |
109 | * | |
110 | * The TCPT_KEEP timer is used to keep connections alive. If an | |
111 | * connection is idle (no segments received) for TCPTV_KEEP_INIT amount | |
112 | * of time, but not yet established, then we drop the connection. | |
113 | * Once the connection is established, if the connection is idle for | |
114 | * TCPTV_KEEP_IDLE time (and keepalives have been enabled on the socket), | |
115 | * we begin to probe the connection. We force the peer to send us a | |
116 | * segment by sending: | |
117 | * <SEQ=SND.UNA-1><ACK=RCV.NXT><CTL=ACK> | |
118 | * This segment is (deliberately) outside the window, and should elicit | |
119 | * an ack segment in response from the peer. If, despite the TCPT_KEEP | |
120 | * initiated segments we cannot elicit a response from a peer in | |
121 | * TCPT_MAXIDLE amount of time probing, then we drop the connection. | |
122 | * | |
123 | * The TCPT_2MSL timer is used for keeping the conenction in Time-wait state | |
124 | * before fully closing it so that the connection 4-tuple can be reused. | |
125 | */ | |
126 | #ifdef BSD_KERNEL_PRIVATE | |
127 | ||
128 | #define TCPT_PTO 0 /* Probe timeout */ | |
129 | #define TCPT_DELAYFR 1 /* Delay recovery if there is reordering */ | |
130 | #define TCPT_REXMT 2 /* retransmit */ | |
131 | #define TCPT_DELACK 3 /* delayed ack */ | |
132 | #define TCPT_PERSIST 4 /* retransmit persistence */ | |
133 | #define TCPT_KEEP 5 /* keep alive */ | |
134 | #define TCPT_2MSL 6 /* 2*msl quiet time timer */ | |
135 | #if MPTCP | |
136 | #define TCPT_JACK_RXMT 7 /* retransmit timer for join ack */ | |
137 | #define TCPT_CELLICON 8 /* Timer to check for cell-activity */ | |
138 | #define TCPT_MAX 8 | |
139 | #else /* MPTCP */ | |
140 | #define TCPT_MAX 6 | |
141 | #endif /* !MPTCP */ | |
142 | ||
143 | #define TCPT_NONE (TCPT_MAX + 1) | |
144 | #define TCPT_NTIMERS (TCPT_MAX + 1) | |
145 | ||
146 | /* External definitions */ | |
147 | #define TCPT_REXMT_EXT 0 | |
148 | #define TCPT_PERSIST_EXT 1 | |
149 | #define TCPT_KEEP_EXT 2 | |
150 | #define TCPT_2MSL_EXT 3 | |
151 | #define TCPT_DELACK_EXT 4 | |
152 | ||
153 | #else /* !BSD_KERNEL_PRIVATE */ | |
154 | #define TCPT_REXMT 0 /* retransmit */ | |
155 | #define TCPT_PERSIST 1 /* retransmit persistence */ | |
156 | #define TCPT_KEEP 2 /* keep alive */ | |
157 | #define TCPT_2MSL 3 /* 2*msl quiet time timer */ | |
158 | #define TCPT_DELACK 4 /* delayed ack timer */ | |
159 | #if MPTCP | |
160 | #define TCPT_JACK_RXMT 5 /* retransmit timer for join ack */ | |
161 | #define TCPT_MAX 5 | |
162 | #else /* MPTCP */ | |
163 | #define TCPT_MAX 4 | |
164 | #endif /* !MPTCP */ | |
165 | #define TCPT_NONE (TCPT_MAX + 1) | |
166 | #define TCPT_NTIMERS (TCPT_MAX + 1) | |
167 | ||
168 | #endif /* BSD_KERNEL_PRIVATE */ | |
169 | ||
170 | #ifdef BSD_KERNEL_PRIVATE | |
171 | /* | |
172 | * Time constants. | |
173 | */ | |
174 | #define TCPTV_MSL ( 15*TCP_RETRANSHZ) /* max seg lifetime */ | |
175 | #define TCPTV_SRTTBASE 0 /* base roundtrip time; if 0, no idea yet */ | |
176 | #define TCPTV_RTOBASE ( 1*TCP_RETRANSHZ) /* assumed RTO if no info */ | |
177 | #define TCPTV_SRTTDFLT ( 1*TCP_RETRANSHZ) /* assumed RTT if no info */ | |
178 | #define TCPTV_PERSMIN ( 5*TCP_RETRANSHZ) /* retransmit persistence */ | |
179 | #define TCPTV_PERSMAX ( 60*TCP_RETRANSHZ) /* maximum persist interval */ | |
180 | ||
181 | extern int tcptv_persmin_val; | |
182 | ||
183 | #define TCPTV_KEEP_INIT ( 75*TCP_RETRANSHZ) /* connect keep alive */ | |
184 | #define TCPTV_KEEP_IDLE (120*60*TCP_RETRANSHZ) /* time before probing */ | |
185 | #define TCPTV_KEEPINTVL ( 75*TCP_RETRANSHZ) /* default probe interval */ | |
186 | #define TCPTV_KEEPCNT 8 /* max probes before drop */ | |
187 | ||
188 | #define TCPTV_REXMTMAX ( 64*TCP_RETRANSHZ ) /* max REXMT value */ | |
189 | #define TCPTV_REXMTMIN ( TCP_RETRANSHZ/33 ) /* min REXMT for non-local connections */ | |
190 | ||
191 | /* | |
192 | * Window for counting received bytes to see if ack-stretching | |
193 | * can start (default 100 ms) | |
194 | */ | |
195 | #define TCPTV_UNACKWIN ( TCP_RETRANSHZ/10 ) | |
196 | ||
197 | /* Receiver idle time, avoid ack-stretching after this idle time */ | |
198 | #define TCPTV_MAXRCVIDLE (TCP_RETRANSHZ/5 ) | |
199 | ||
200 | /* | |
201 | * No ack stretching during slow-start, until we see some packets. | |
202 | * By the time the receiver gets 512 packets, the senders cwnd | |
203 | * should open by a few hundred packets consdering the | |
204 | * slow-start progression. | |
205 | */ | |
206 | #define TCP_RCV_SS_PKTCOUNT 512 | |
207 | ||
208 | #define TCPTV_TWTRUNC 8 /* RTO factor to truncate TW */ | |
209 | ||
210 | #define TCP_LINGERTIME 120 /* linger at most 2 minutes */ | |
211 | ||
212 | #define TCP_MAXRXTSHIFT 12 /* maximum retransmits */ | |
213 | ||
214 | #ifdef TCPTIMERS | |
215 | static char *tcptimers[] = | |
216 | { "REXMT", "PERSIST", "KEEP", "2MSL", "DELACK"}; | |
217 | #endif /* TCPTIMERS */ | |
218 | ||
219 | /* | |
220 | * Persist, keep, 2msl and MPTCP's join-ack timer as slow timers which can | |
221 | * be coalesced at a higher granularity (500 ms). | |
222 | * | |
223 | * Rexmt and delayed ack timers are considered as fast timers which run | |
224 | * in the order of 100ms. | |
225 | * | |
226 | * Probe timeout is a quick timer which will run in the order of 10ms. | |
227 | */ | |
228 | #define IS_TIMER_HZ_500MS(i) ((i) >= TCPT_PERSIST) | |
229 | #define IS_TIMER_HZ_100MS(i) ((i) >= TCPT_REXMT && (i) < TCPT_PERSIST) | |
230 | #define IS_TIMER_HZ_10MS(i) ((i) < TCPT_REXMT) | |
231 | ||
232 | struct tcptimerlist; | |
233 | ||
234 | struct tcptimerentry { | |
235 | LIST_ENTRY(tcptimerentry) le; /* links for timer list */ | |
236 | uint32_t timer_start; /* tcp clock when the timer was started */ | |
237 | uint16_t index; /* index of lowest timer that needs to run first */ | |
238 | uint16_t mode; /* Bit-wise OR of timers that are active */ | |
239 | uint32_t runtime; /* deadline at which the first timer has to fire */ | |
240 | }; | |
241 | ||
242 | LIST_HEAD(timerlisthead, tcptimerentry); | |
243 | ||
244 | struct tcptimerlist { | |
245 | struct timerlisthead lhead; /* head of the list */ | |
246 | lck_mtx_t *mtx; /* lock to protect the list */ | |
247 | lck_attr_t *mtx_attr; /* mutex attributes */ | |
248 | lck_grp_t *mtx_grp; /* mutex group definition */ | |
249 | lck_grp_attr_t *mtx_grp_attr; /* mutex group attributes */ | |
250 | thread_call_t call; /* call entry */ | |
251 | uint32_t runtime; /* time at which this list is going to run */ | |
252 | uint32_t schedtime; /* time at which this list was scheduled */ | |
253 | uint32_t entries; /* Number of entries on the list */ | |
254 | uint32_t maxentries; /* Max number of entries at any time */ | |
255 | ||
256 | /* Set desired mode when timer list running */ | |
257 | boolean_t running; /* Set when timer list is being processed */ | |
258 | boolean_t scheduled; /* set when the timer is scheduled */ | |
259 | #define TCP_TIMERLIST_10MS_MODE 0x1 | |
260 | #define TCP_TIMERLIST_100MS_MODE 0x2 | |
261 | #define TCP_TIMERLIST_500MS_MODE 0x4 | |
262 | uint32_t mode; /* Current mode of the timer */ | |
263 | uint32_t pref_mode; /* Preferred mode set by a connection */ | |
264 | uint32_t pref_offset; /* Preferred offset set by a connection */ | |
265 | uint32_t idleruns; /* Number of times the list has been idle in fast mode */ | |
266 | struct tcptimerentry *next_te; /* next timer entry pointer to process */ | |
267 | u_int16_t probe_if_index; /* Interface index that needs to send probes */ | |
268 | }; | |
269 | ||
270 | /* number of idle runs allowed for TCP timer list in fast or quick modes */ | |
271 | #define TCP_FASTMODE_IDLERUN_MAX 10 | |
272 | ||
273 | /* | |
274 | * Minimum retransmit timeout is set to 30ms. We add a slop of | |
275 | * 200 ms to the retransmit value to account for processing | |
276 | * variance and delayed ack. This extra 200ms will help to avoid | |
277 | * spurious retransmits by taking into consideration the receivers | |
278 | * that wait for delayed ack timer instead of generating an ack | |
279 | * for every two packets. | |
280 | * | |
281 | * On a local link, the minimum retransmit timeout is 100ms and | |
282 | * variance is set to 0. This will make the sender a little bit more | |
283 | * aggressive on local link. When the connection is not established yet, | |
284 | * there is no need to add an extra 200ms to retransmit timeout because | |
285 | * the initial value is high (1s) and delayed ack is not a problem in | |
286 | * that case. | |
287 | */ | |
288 | #define TCPTV_REXMTSLOP ( TCP_RETRANSHZ/5 ) /* extra 200 ms slop */ | |
289 | ||
290 | /* macro to decide when retransmit slop (described above) should be added */ | |
291 | #define TCP_ADD_REXMTSLOP(tp) (tp->t_state >= TCPS_ESTABLISHED) | |
292 | ||
293 | #define TCPT_RANGESET(tv, value, tvmin, tvmax, addslop) do { \ | |
294 | (tv) = ((addslop) ? tcp_rexmt_slop : 0) + (value); \ | |
295 | if ((uint32_t)(tv) < (uint32_t)(tvmin)) \ | |
296 | (tv) = (tvmin); \ | |
297 | else if ((uint32_t)(tv) > (uint32_t)(tvmax)) \ | |
298 | (tv) = (tvmax); \ | |
299 | } while(0) | |
300 | ||
301 | #define TCP_CONN_KEEPIDLE(tp) \ | |
302 | ((tp)->t_keepidle && \ | |
303 | ((tp)->t_inpcb->inp_socket->so_options & SO_KEEPALIVE) ? \ | |
304 | (tp)->t_keepidle : (uint32_t)tcp_keepidle) | |
305 | #define TCP_CONN_KEEPINIT(tp) \ | |
306 | (((tp)->t_keepinit > 0) ? (tp)->t_keepinit : (uint32_t)tcp_keepinit) | |
307 | #define TCP_CONN_KEEPCNT(tp) \ | |
308 | (((tp)->t_keepcnt > 0) ? (tp)->t_keepcnt : (uint32_t)tcp_keepcnt) | |
309 | #define TCP_CONN_KEEPINTVL(tp) \ | |
310 | (((tp)->t_keepintvl > 0) ? (tp)->t_keepintvl : (uint32_t)tcp_keepintvl) | |
311 | #define TCP_CONN_MAXIDLE(tp) \ | |
312 | (TCP_CONN_KEEPCNT(tp) * TCP_CONN_KEEPINTVL(tp)) | |
313 | ||
314 | #define TCP_IDLETIMEOUT(tp) \ | |
315 | (((TCP_ADD_REXMTSLOP(tp)) ? 0 : tcp_rexmt_slop) + tp->t_rxtcur) | |
316 | ||
317 | TAILQ_HEAD(tcptailq, tcpcb); | |
318 | ||
319 | extern int tcp_keepinit; /* time to establish connection */ | |
320 | extern int tcp_keepidle; /* time before keepalive probes begin */ | |
321 | extern int tcp_keepintvl; /* time between keepalive probes */ | |
322 | extern int tcp_keepcnt; /* number of keepalives */ | |
323 | extern int tcp_delack; /* delayed ack timer */ | |
324 | extern int tcp_maxpersistidle; | |
325 | extern int tcp_msl; | |
326 | extern int tcp_ttl; /* time to live for TCP segs */ | |
327 | extern int tcp_backoff[]; | |
328 | extern int tcp_rexmt_slop; | |
329 | extern u_int32_t tcp_max_persist_timeout; /* Maximum persistence for Zero Window Probes */ | |
330 | ||
331 | #define OFFSET_FROM_START(tp, off) ((tcp_now + (off)) - (tp)->tentry.timer_start) | |
332 | ||
333 | #endif /* BSD_KERNEL_PRIVATE */ | |
334 | #endif /* !_NETINET_TCP_TIMER_H_ */ |