]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet/tcp_var.h
xnu-201.42.3.tar.gz
[apple/xnu.git] / bsd / netinet / tcp_var.h
1 /*
2 * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
3 *
4 * @APPLE_LICENSE_HEADER_START@
5 *
6 * The contents of this file constitute Original Code as defined in and
7 * are subject to the Apple Public Source License Version 1.1 (the
8 * "License"). You may not use this file except in compliance with the
9 * License. Please obtain a copy of the License at
10 * http://www.apple.com/publicsource and read it before using this file.
11 *
12 * This Original Code and all software distributed under the License are
13 * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
14 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
15 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the
17 * License for the specific language governing rights and limitations
18 * under the License.
19 *
20 * @APPLE_LICENSE_HEADER_END@
21 */
22 /*
23 * Copyright (c) 1982, 1986, 1993, 1994, 1995
24 * The Regents of the University of California. All rights reserved.
25 *
26 * Redistribution and use in source and binary forms, with or without
27 * modification, are permitted provided that the following conditions
28 * are met:
29 * 1. Redistributions of source code must retain the above copyright
30 * notice, this list of conditions and the following disclaimer.
31 * 2. Redistributions in binary form must reproduce the above copyright
32 * notice, this list of conditions and the following disclaimer in the
33 * documentation and/or other materials provided with the distribution.
34 * 3. All advertising materials mentioning features or use of this software
35 * must display the following acknowledgement:
36 * This product includes software developed by the University of
37 * California, Berkeley and its contributors.
38 * 4. Neither the name of the University nor the names of its contributors
39 * may be used to endorse or promote products derived from this software
40 * without specific prior written permission.
41 *
42 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
43 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
44 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
45 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
46 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
47 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
48 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
49 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
50 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
51 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
52 * SUCH DAMAGE.
53 *
54 * @(#)tcp_var.h 8.4 (Berkeley) 5/24/95
55 */
56
57 #ifndef _NETINET_TCP_VAR_H_
58 #define _NETINET_TCP_VAR_H_
59 #define N_TIME_WAIT_SLOTS 128 /* must be power of 2 */
60
61 /*
62 * Ip (reassembly or sequence) queue structures.
63 *
64 * XXX -- The following explains why the ipqe_m field is here, for TCP's use:
65 * We want to avoid doing m_pullup on incoming packets but that
66 * means avoiding dtom on the tcp reassembly code. That in turn means
67 * keeping an mbuf pointer in the reassembly queue (since we might
68 * have a cluster). As a quick hack, the source & destination
69 * port numbers (which are no longer needed once we've located the
70 * tcpcb) are overlayed with an mbuf pointer.
71 */
72 LIST_HEAD(ipqehead, ipqent);
73 struct ipqent {
74 LIST_ENTRY(ipqent) ipqe_q;
75 union {
76 struct ip *_ip;
77 #if INET6
78 struct ipv6 *_ip6;
79 #endif
80 struct tcpiphdr *_tcp;
81 } _ipqe_u1;
82 struct mbuf *ipqe_m; /* mbuf contains packet */
83 u_int8_t ipqe_mff; /* for IP fragmentation */
84 };
85 #define ipqe_ip _ipqe_u1._ip
86 #if INET6
87 #define ipqe_ip6 _ipqe_u1._ip6
88 #endif
89 #define ipqe_tcp _ipqe_u1._tcp
90 #define tcp6cb tcpcb /* for KAME src sync over BSD*'s */
91
92 #define TCP_DELACK_BITSET(hash_elem)\
93 delack_bitmask[((hash_elem) >> 5)] |= 1 << ((hash_elem) & 0x1F)
94
95 #define DELACK_BITMASK_ON 1
96 #define DELACK_BITMASK_THRESH 300
97
98 /*
99 * Kernel variables for tcp.
100 */
101
102 /*
103 * Tcp control block, one per tcp; fields:
104 * Organized for 16 byte cacheline efficiency.
105 */
106 struct tcpcb {
107 struct ipqehead segq; /* sequencing queue */
108 int t_dupacks; /* consecutive dup acks recd */
109 struct tcptemp *t_template; /* skeletal packet for transmit */
110
111 int t_timer[TCPT_NTIMERS]; /* tcp timers */
112
113 struct inpcb *t_inpcb; /* back pointer to internet pcb */
114 int t_state; /* state of this connection */
115 u_int t_flags;
116 #define TF_ACKNOW 0x00001 /* ack peer immediately */
117 #define TF_DELACK 0x00002 /* ack, but try to delay it */
118 #define TF_NODELAY 0x00004 /* don't delay packets to coalesce */
119 #define TF_NOOPT 0x00008 /* don't use tcp options */
120 #define TF_SENTFIN 0x00010 /* have sent FIN */
121 #define TF_REQ_SCALE 0x00020 /* have/will request window scaling */
122 #define TF_RCVD_SCALE 0x00040 /* other side has requested scaling */
123 #define TF_REQ_TSTMP 0x00080 /* have/will request timestamps */
124 #define TF_RCVD_TSTMP 0x00100 /* a timestamp was received in SYN */
125 #define TF_SACK_PERMIT 0x00200 /* other side said I could SACK */
126 #define TF_NEEDSYN 0x00400 /* send SYN (implicit state) */
127 #define TF_NEEDFIN 0x00800 /* send FIN (implicit state) */
128 #define TF_NOPUSH 0x01000 /* don't push */
129 #define TF_REQ_CC 0x02000 /* have/will request CC */
130 #define TF_RCVD_CC 0x04000 /* a CC was received in SYN */
131 #define TF_SENDCCNEW 0x08000 /* send CCnew instead of CC in SYN */
132 #define TF_MORETOCOME 0x10000 /* More data to be appended to sock */
133 int t_force; /* 1 if forcing out a byte */
134
135 tcp_seq snd_una; /* send unacknowledged */
136 tcp_seq snd_max; /* highest sequence number sent;
137 * used to recognize retransmits
138 */
139 tcp_seq snd_nxt; /* send next */
140 tcp_seq snd_up; /* send urgent pointer */
141
142 tcp_seq snd_wl1; /* window update seg seq number */
143 tcp_seq snd_wl2; /* window update seg ack number */
144 tcp_seq iss; /* initial send sequence number */
145 tcp_seq irs; /* initial receive sequence number */
146
147 tcp_seq rcv_nxt; /* receive next */
148 tcp_seq rcv_adv; /* advertised window */
149 u_long rcv_wnd; /* receive window */
150 tcp_seq rcv_up; /* receive urgent pointer */
151
152 u_long snd_wnd; /* send window */
153 u_long snd_cwnd; /* congestion-controlled window */
154 u_long snd_ssthresh; /* snd_cwnd size threshold for
155 * for slow start exponential to
156 * linear switch
157 */
158 u_int t_maxopd; /* mss plus options */
159
160 u_int t_idle; /* inactivity time */
161 u_long t_duration; /* connection duration */
162 int t_rtt; /* round trip time */
163 tcp_seq t_rtseq; /* sequence number being timed */
164
165 int t_rxtcur; /* current retransmit value */
166 u_int t_maxseg; /* maximum segment size */
167 int t_srtt; /* smoothed round-trip time */
168 int t_rttvar; /* variance in round-trip time */
169
170 int t_rxtshift; /* log(2) of rexmt exp. backoff */
171 u_int t_rttmin; /* minimum rtt allowed */
172 u_long t_rttupdated; /* number of times rtt sampled */
173 u_long max_sndwnd; /* largest window peer has offered */
174
175 int t_softerror; /* possible error not yet reported */
176 /* out-of-band data */
177 char t_oobflags; /* have some */
178 char t_iobc; /* input character */
179 #define TCPOOB_HAVEDATA 0x01
180 #define TCPOOB_HADDATA 0x02
181 /* RFC 1323 variables */
182 u_char snd_scale; /* window scaling for send window */
183 u_char rcv_scale; /* window scaling for recv window */
184 u_char request_r_scale; /* pending window scaling */
185 u_char requested_s_scale;
186 u_long ts_recent; /* timestamp echo data */
187
188 u_long ts_recent_age; /* when last updated */
189 tcp_seq last_ack_sent;
190 /* RFC 1644 variables */
191 tcp_cc cc_send; /* send connection count */
192 tcp_cc cc_recv; /* receive connection count */
193 u_long reserved[4];
194 };
195
196 /*
197 * Structure to hold TCP options that are only used during segment
198 * processing (in tcp_input), but not held in the tcpcb.
199 * It's basically used to reduce the number of parameters
200 * to tcp_dooptions.
201 */
202 struct tcpopt {
203 u_long to_flag; /* which options are present */
204 #define TOF_TS 0x0001 /* timestamp */
205 #define TOF_CC 0x0002 /* CC and CCnew are exclusive */
206 #define TOF_CCNEW 0x0004
207 #define TOF_CCECHO 0x0008
208 u_long to_tsval;
209 u_long to_tsecr;
210 tcp_cc to_cc; /* holds CC or CCnew */
211 tcp_cc to_ccecho;
212 u_short to_maxseg;
213 };
214
215 /*
216 * The TAO cache entry which is stored in the protocol family specific
217 * portion of the route metrics.
218 */
219 struct rmxp_tao {
220 tcp_cc tao_cc; /* latest CC in valid SYN */
221 tcp_cc tao_ccsent; /* latest CC sent to peer */
222 u_short tao_mssopt; /* peer's cached MSS */
223 #ifdef notyet
224 u_short tao_flags; /* cache status flags */
225 #define TAOF_DONT 0x0001 /* peer doesn't understand rfc1644 */
226 #define TAOF_OK 0x0002 /* peer does understand rfc1644 */
227 #define TAOF_UNDEF 0 /* we don't know yet */
228 #endif /* notyet */
229 };
230 #define rmx_taop(r) ((struct rmxp_tao *)(r).rmx_filler)
231
232 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb)
233 #define sototcpcb(so) (intotcpcb(sotoinpcb(so)))
234
235 /*
236 * The smoothed round-trip time and estimated variance
237 * are stored as fixed point numbers scaled by the values below.
238 * For convenience, these scales are also used in smoothing the average
239 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
240 * With these scales, srtt has 3 bits to the right of the binary point,
241 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the
242 * binary point, and is smoothed with an ALPHA of 0.75.
243 */
244 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */
245 #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */
246 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */
247 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */
248 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
249
250 /*
251 * The initial retransmission should happen at rtt + 4 * rttvar.
252 * Because of the way we do the smoothing, srtt and rttvar
253 * will each average +1/2 tick of bias. When we compute
254 * the retransmit timer, we want 1/2 tick of rounding and
255 * 1 extra tick because of +-1/2 tick uncertainty in the
256 * firing of the timer. The bias will give us exactly the
257 * 1.5 tick we need. But, because the bias is
258 * statistical, we have to test that we don't drop below
259 * the minimum feasible timer (which is 2 ticks).
260 * This version of the macro adapted from a paper by Lawrence
261 * Brakmo and Larry Peterson which outlines a problem caused
262 * by insufficient precision in the original implementation,
263 * which results in inappropriately large RTO values for very
264 * fast networks.
265 */
266 #define TCP_REXMTVAL(tp) \
267 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
268 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
269
270 /*
271 * TCP statistics.
272 * Many of these should be kept per connection,
273 * but that's inconvenient at the moment.
274 */
275 struct tcpstat {
276 u_long tcps_connattempt; /* connections initiated */
277 u_long tcps_accepts; /* connections accepted */
278 u_long tcps_connects; /* connections established */
279 u_long tcps_drops; /* connections dropped */
280 u_long tcps_conndrops; /* embryonic connections dropped */
281 u_long tcps_closed; /* conn. closed (includes drops) */
282 u_long tcps_segstimed; /* segs where we tried to get rtt */
283 u_long tcps_rttupdated; /* times we succeeded */
284 u_long tcps_delack; /* delayed acks sent */
285 u_long tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
286 u_long tcps_rexmttimeo; /* retransmit timeouts */
287 u_long tcps_persisttimeo; /* persist timeouts */
288 u_long tcps_keeptimeo; /* keepalive timeouts */
289 u_long tcps_keepprobe; /* keepalive probes sent */
290 u_long tcps_keepdrops; /* connections dropped in keepalive */
291
292 u_long tcps_sndtotal; /* total packets sent */
293 u_long tcps_sndpack; /* data packets sent */
294 u_long tcps_sndbyte; /* data bytes sent */
295 u_long tcps_sndrexmitpack; /* data packets retransmitted */
296 u_long tcps_sndrexmitbyte; /* data bytes retransmitted */
297 u_long tcps_sndacks; /* ack-only packets sent */
298 u_long tcps_sndprobe; /* window probes sent */
299 u_long tcps_sndurg; /* packets sent with URG only */
300 u_long tcps_sndwinup; /* window update-only packets sent */
301 u_long tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
302
303 u_long tcps_rcvtotal; /* total packets received */
304 u_long tcps_rcvpack; /* packets received in sequence */
305 u_long tcps_rcvbyte; /* bytes received in sequence */
306 u_long tcps_rcvbadsum; /* packets received with ccksum errs */
307 u_long tcps_rcvbadoff; /* packets received with bad offset */
308 u_long tcps_rcvmemdrop; /* packets dropped for lack of memory */
309 u_long tcps_rcvshort; /* packets received too short */
310 u_long tcps_rcvduppack; /* duplicate-only packets received */
311 u_long tcps_rcvdupbyte; /* duplicate-only bytes received */
312 u_long tcps_rcvpartduppack; /* packets with some duplicate data */
313 u_long tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
314 u_long tcps_rcvoopack; /* out-of-order packets received */
315 u_long tcps_rcvoobyte; /* out-of-order bytes received */
316 u_long tcps_rcvpackafterwin; /* packets with data after window */
317 u_long tcps_rcvbyteafterwin; /* bytes rcvd after window */
318 u_long tcps_rcvafterclose; /* packets rcvd after "close" */
319 u_long tcps_rcvwinprobe; /* rcvd window probe packets */
320 u_long tcps_rcvdupack; /* rcvd duplicate acks */
321 u_long tcps_rcvacktoomuch; /* rcvd acks for unsent data */
322 u_long tcps_rcvackpack; /* rcvd ack packets */
323 u_long tcps_rcvackbyte; /* bytes acked by rcvd acks */
324 u_long tcps_rcvwinupd; /* rcvd window update packets */
325 u_long tcps_pawsdrop; /* segments dropped due to PAWS */
326 u_long tcps_predack; /* times hdr predict ok for acks */
327 u_long tcps_preddat; /* times hdr predict ok for data pkts */
328 u_long tcps_pcbcachemiss;
329 u_long tcps_cachedrtt; /* times cached RTT in route updated */
330 u_long tcps_cachedrttvar; /* times cached rttvar updated */
331 u_long tcps_cachedssthresh; /* times cached ssthresh updated */
332 u_long tcps_usedrtt; /* times RTT initialized from route */
333 u_long tcps_usedrttvar; /* times RTTVAR initialized from rt */
334 u_long tcps_usedssthresh; /* times ssthresh initialized from rt*/
335 u_long tcps_persistdrop; /* timeout in persist state */
336 u_long tcps_badsyn; /* bogus SYN, e.g. premature ACK */
337 u_long tcps_mturesent; /* resends due to MTU discovery */
338 u_long tcps_listendrop; /* listen queue overflows */
339 };
340
341 /*
342 * TCB structure exported to user-land via sysctl(3).
343 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
344 * included. Not all of our clients do.
345 */
346 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
347 struct xtcpcb {
348 size_t xt_len;
349 struct inpcb xt_inp;
350 struct tcpcb xt_tp;
351 struct xsocket xt_socket;
352 u_quad_t xt_alignment_hack;
353 };
354 #endif
355
356 /*
357 * Names for TCP sysctl objects
358 */
359 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
360 #define TCPCTL_DO_RFC1644 2 /* use RFC-1644 extensions */
361 #define TCPCTL_MSSDFLT 3 /* MSS default */
362 #define TCPCTL_STATS 4 /* statistics (read-only) */
363 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */
364 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
365 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
366 #define TCPCTL_SENDSPACE 8 /* send buffer space */
367 #define TCPCTL_RECVSPACE 9 /* receive buffer space */
368 #define TCPCTL_KEEPINIT 10 /* receive buffer space */
369 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */
370 #define TCPCTL_V6MSSDFLT 12 /* MSS default for IPv6 */
371 #define TCPCTL_MAXID 13
372
373 #define TCPCTL_NAMES { \
374 { 0, 0 }, \
375 { "rfc1323", CTLTYPE_INT }, \
376 { "rfc1644", CTLTYPE_INT }, \
377 { "mssdflt", CTLTYPE_INT }, \
378 { "stats", CTLTYPE_STRUCT }, \
379 { "rttdflt", CTLTYPE_INT }, \
380 { "keepidle", CTLTYPE_INT }, \
381 { "keepintvl", CTLTYPE_INT }, \
382 { "sendspace", CTLTYPE_INT }, \
383 { "recvspace", CTLTYPE_INT }, \
384 { "keepinit", CTLTYPE_INT }, \
385 { "pcblist", CTLTYPE_STRUCT }, \
386 { "v6mssdflt", CTLTYPE_INT }, \
387 }
388
389 #ifdef KERNEL
390 #ifdef SYSCTL_DECL
391 SYSCTL_DECL(_net_inet_tcp);
392 #endif
393
394 extern struct inpcbhead tcb; /* head of queue of active tcpcb's */
395 extern struct inpcbinfo tcbinfo;
396 extern struct tcpstat tcpstat; /* tcp statistics */
397 extern int tcp_mssdflt; /* XXX */
398 extern int tcp_v6mssdflt; /* XXX */
399 extern u_long tcp_now; /* for RFC 1323 timestamps */
400 extern int tcp_delack_enabled;
401
402 void tcp_canceltimers __P((struct tcpcb *));
403 struct tcpcb *
404 tcp_close __P((struct tcpcb *));
405 void tcp_ctlinput __P((int, struct sockaddr *, void *));
406 #if INET6
407 struct ip6_hdr;
408 void tcp6_ctlinput __P((int, struct sockaddr *,void *));
409 #endif
410 int tcp_ctloutput __P((struct socket *, struct sockopt *));
411 struct tcpcb *
412 tcp_drop __P((struct tcpcb *, int));
413 void tcp_drain __P((void));
414 void tcp_fasttimo __P((void));
415 struct rmxp_tao *
416 tcp_gettaocache __P((struct inpcb *));
417 void tcp_init __P((void));
418 #if INET6
419 void tcp6_init __P((void));
420 int tcp6_input __P((struct mbuf **, int *, int));
421 #endif /* INET6 */
422 void tcp_input __P((struct mbuf *, int));
423 #if INET6
424 void tcp_mss __P((struct tcpcb *, int, int));
425 int tcp_mssopt __P((struct tcpcb *, int));
426 #else /* INET6 */
427 void tcp_mss __P((struct tcpcb *, int));
428 int tcp_mssopt __P((struct tcpcb *));
429 #endif /* INET6 */
430
431 void tcp_mtudisc __P((struct inpcb *, int));
432 struct tcpcb *
433 tcp_newtcpcb __P((struct inpcb *));
434 int tcp_output __P((struct tcpcb *));
435 void tcp_quench __P((struct inpcb *, int));
436 #if INET6
437 void tcp_respond __P((struct tcpcb *, void *, struct tcphdr *,
438 struct mbuf *, tcp_seq, tcp_seq, int, int));
439 #else /* INET6 */
440 void tcp_respond __P((struct tcpcb *, void *, struct tcphdr *,
441 struct mbuf *, tcp_seq, tcp_seq, int));
442 #endif /* INET6 */
443
444 struct rtentry *
445 tcp_rtlookup __P((struct inpcb *));
446 #if INET6
447 struct rtentry *
448 tcp_rtlookup6 __P((struct inpcb *));
449 #endif /* INET6 */
450 void tcp_setpersist __P((struct tcpcb *));
451 void tcp_slowtimo __P((void));
452 struct tcptemp *
453 tcp_template __P((struct tcpcb *));
454 struct tcpcb *
455 tcp_timers __P((struct tcpcb *, int));
456 #if INET6
457 void tcp_trace __P((int, int, struct tcpcb *, void *, struct tcphdr *,
458 int));
459 #else
460 void tcp_trace __P((int, int, struct tcpcb *, struct ip *,
461 struct tcphdr *, int));
462 #endif
463
464 #if INET6
465 int tcp_reass __P((struct tcpcb *, struct tcphdr *, int,
466 struct mbuf *, int));
467 #else /* INET6 */
468 int tcp_reass __P((struct tcpcb *, struct tcphdr *, int, struct mbuf *));
469 /* suppress INET6 only args */
470 #define tcp_reass(x, y, z, t, i) tcp_reass(x, y, z, t)
471 #define tcp_mss(x, y, i) tcp_mss(x, y)
472 #define tcp_mssopt(x, i) tcp_mssopt(x)
473 #define tcp_respond(x, y, z, m, s1, s2, f, i) tcp_respond(x, y, z, m, s1, \
474 s2, f)
475
476 #endif /* INET6 */
477
478 extern struct pr_usrreqs tcp_usrreqs;
479 #if INET6
480 extern struct pr_usrreqs tcp6_usrreqs;
481 #endif /* INET6 */
482 extern u_long tcp_sendspace;
483 extern u_long tcp_recvspace;
484 void tcp_rndiss_init __P((void));
485 tcp_seq tcp_rndiss_next __P((void));
486 u_int16_t tcp_rndiss_encrypt __P((u_int16_t));
487
488
489
490 #endif /* KERNEL */
491
492 #endif /* _NETINET_TCP_VAR_H_ */