]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet/tcp_var.h
projects / apple / xnu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xnu-2782.30.5.tar.gz
[apple/xnu.git] / bsd / netinet / tcp_var.h
1 /*
2 * Copyright (c) 2000-2014 Apple 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, 1994, 1995
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_var.h 8.4 (Berkeley) 5/24/95
61 * $FreeBSD: src/sys/netinet/tcp_var.h,v 1.56.2.8 2001/08/22 00:59:13 silby Exp $
62 */
63
64 #ifndef _NETINET_TCP_VAR_H_
65 #define _NETINET_TCP_VAR_H_
66 #include <sys/appleapiopts.h>
67 #include <sys/queue.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/tcp_timer.h>
70
71 #if defined(__LP64__)
72 #define _TCPCB_PTR(x) u_int32_t
73 #define _TCPCB_LIST_HEAD(name, type) \
74 struct name { \
75 u_int32_t lh_first; \
76 };
77 #else
78 #define _TCPCB_PTR(x) x
79 #define _TCPCB_LIST_HEAD(name, type) LIST_HEAD(name, type)
80 #endif
81
82 #ifdef KERNEL_PRIVATE
83
84 #define TCP_RETRANSHZ 1000 /* granularity of TCP timestamps, 1ms */
85 /* Minimum time quantum within which the timers are coalesced */
86 #define TCP_TIMER_10MS_QUANTUM (TCP_RETRANSHZ/100) /* every 10ms */
87 #define TCP_TIMER_100MS_QUANTUM (TCP_RETRANSHZ/10) /* every 100ms */
88 #define TCP_TIMER_500MS_QUANTUM (TCP_RETRANSHZ/2) /* every 500ms */
89
90 #define TCP_RETRANSHZ_TO_USEC 1000
91
92 #define N_TIME_WAIT_SLOTS 128 /* must be power of 2 */
93
94 /* Base RTT is stored for N_MIN_RTT_HISTORY slots. This is used to
95 * estimate expected minimum RTT for delay based congestion control
96 * algorithms.
97 */
98 #define N_RTT_BASE 5
99
100 /* Always allow at least 4 packets worth of recv window when adjusting
101 * recv window using inter-packet arrival jitter.
102 */
103 #define MIN_IAJ_WIN 4
104
105 /* A variation in delay of this many milliseconds is tolerable. This limit has to
106 * be low but greater than zero. We also use standard deviation on jitter to adjust
107 * this limit for different link and connection types.
108 */
109 #define ALLOWED_IAJ 5
110
111 /* Ignore the first few packets on a connection until the ACK clock gets going
112 */
113 #define IAJ_IGNORE_PKTCNT 40
114
115 /* Let the accumulated IAJ value increase by this threshold at most. This limit
116 * will control how many ALLOWED_IAJ measurements a receiver will have to see
117 * before opening the receive window
118 */
119 #define ACC_IAJ_HIGH_THRESH 100
120
121 /* When accumulated IAJ reaches this value, the receiver starts to react by
122 * closing the window
123 */
124 #define ACC_IAJ_REACT_LIMIT 200
125
126 /* If the number of small packets (smaller than IAJ packet size) seen on a
127 * connection is more than this threshold, reset the size and learn it again.
128 * This is needed because the sender might send smaller segments after PMTU
129 * discovery and the receiver has to learn the new size.
130 */
131 #define RESET_IAJ_SIZE_THRESH 20
132
133 /*
134 * Adaptive timeout is a read/write timeout specified by the application to
135 * get a socket event when the transport layer detects a stall in data
136 * transfer. The value specified is the number of probes that can be sent
137 * to the peer before generating an event. Since it is not specified as
138 * a time value, the timeout will adjust based on the RTT seen on the link.
139 * The timeout will start only when there is an indication that the read/write
140 * operation is not making progress.
141 *
142 * If a write operation stalls, the probe will be retransmission of data.
143 * If a read operation stalls, the probe will be a keep-alive packet.
144 *
145 * The maximum value of adaptive timeout is set to 10 which will allow
146 * transmission of enough number of probes to the peer.
147 */
148 #define TCP_ADAPTIVE_TIMEOUT_MAX 10
149
150 /*
151 * Kernel variables for tcp.
152 */
153
154 /* TCP segment queue entry */
155 struct tseg_qent {
156 LIST_ENTRY(tseg_qent) tqe_q;
157 int tqe_len; /* TCP segment data length */
158 struct tcphdr *tqe_th; /* a pointer to tcp header */
159 struct mbuf *tqe_m; /* mbuf contains packet */
160 };
161 LIST_HEAD(tsegqe_head, tseg_qent);
162
163 struct sackblk {
164 tcp_seq start; /* start seq no. of sack block */
165 tcp_seq end; /* end seq no. */
166 };
167
168 struct sackhole {
169 tcp_seq start; /* start seq no. of hole */
170 tcp_seq end; /* end seq no. */
171 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */
172 u_int32_t rxmit_start; /* timestamp of first retransmission */
173 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */
174 };
175
176 struct sackhint {
177 struct sackhole *nexthole;
178 int sack_bytes_rexmit;
179 };
180
181 struct tcptemp {
182 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
183 struct tcphdr tt_t;
184 };
185
186 struct bwmeas {
187 tcp_seq bw_start; /* start of bw measurement */
188 uint32_t bw_ts; /* timestamp when bw measurement started */
189 uint32_t bw_size; /* burst size in bytes for this bw measurement */
190 uint32_t bw_minsizepkts; /* Min burst size as segments */
191 uint32_t bw_maxsizepkts; /* Max burst size as segments */
192 uint32_t bw_minsize; /* Min size in bytes */
193 uint32_t bw_maxsize; /* Max size in bytes */
194 uint32_t bw_sndbw; /* Measured send bw */
195 };
196
197 /* MPTCP Data sequence map entry */
198 struct mpt_dsn_map {
199 uint64_t mpt_dsn; /* data seq num recvd */
200 uint32_t mpt_sseq; /* relative subflow # */
201 uint16_t mpt_len; /* length of mapping */
202 uint16_t mpt_csum; /* checksum value if on */
203 };
204 #define tcp6cb tcpcb /* for KAME src sync over BSD*'s */
205
206 struct tcp_ccstate {
207 union {
208 struct tcp_cubic_state {
209 u_int32_t tc_last_max; /* cwnd at last loss */
210 u_int32_t tc_epoch_start; /* TS of last loss */
211 u_int32_t tc_origin_point; /* window at the start of an epoch */
212 u_int32_t tc_tcp_win; /* computed tcp win */
213 u_int32_t tc_tcp_bytes_acked; /* bytes acked */
214 u_int32_t tc_target_win; /* cubic target win */
215 u_int32_t tc_avg_lastmax; /* Average of last max */
216 u_int32_t tc_mean_deviation; /* Mean absolute deviation */
217 float tc_epoch_period; /* K parameter */
218 } _cubic_state_;
219 #define cub_last_max __u__._cubic_state_.tc_last_max
220 #define cub_epoch_start __u__._cubic_state_.tc_epoch_start
221 #define cub_origin_point __u__._cubic_state_.tc_origin_point
222 #define cub_tcp_win __u__._cubic_state_.tc_tcp_win
223 #define cub_tcp_bytes_acked __u__._cubic_state_.tc_tcp_bytes_acked
224 #define cub_epoch_period __u__._cubic_state_.tc_epoch_period
225 #define cub_target_win __u__._cubic_state_.tc_target_win
226 #define cub_avg_lastmax __u__._cubic_state_.tc_avg_lastmax
227 #define cub_mean_dev __u__._cubic_state_.tc_mean_deviation
228 } __u__;
229 };
230
231 /*
232 * Tcp control block, one per tcp; fields:
233 * Organized for 16 byte cacheline efficiency.
234 */
235 struct tcpcb {
236 struct tsegqe_head t_segq;
237 int t_dupacks; /* consecutive dup acks recd */
238 uint32_t t_timer[TCPT_NTIMERS]; /* tcp timers */
239 struct tcptimerentry tentry; /* entry in timer list */
240
241 struct inpcb *t_inpcb; /* back pointer to internet pcb */
242 int t_state; /* state of this connection */
243 uint32_t t_flags;
244 #define TF_ACKNOW 0x00001 /* ack peer immediately */
245 #define TF_DELACK 0x00002 /* ack, but try to delay it */
246 #define TF_NODELAY 0x00004 /* don't delay packets to coalesce */
247 #define TF_NOOPT 0x00008 /* don't use tcp options */
248 #define TF_SENTFIN 0x00010 /* have sent FIN */
249 #define TF_REQ_SCALE 0x00020 /* have/will request window scaling */
250 #define TF_RCVD_SCALE 0x00040 /* other side has requested scaling */
251 #define TF_REQ_TSTMP 0x00080 /* have/will request timestamps */
252 #define TF_RCVD_TSTMP 0x00100 /* a timestamp was received in SYN */
253 #define TF_SACK_PERMIT 0x00200 /* other side said I could SACK */
254 #define TF_NEEDSYN 0x00400 /* send SYN (implicit state) */
255 #define TF_NEEDFIN 0x00800 /* send FIN (implicit state) */
256 #define TF_NOPUSH 0x01000 /* don't push */
257 #define TF_REQ_CC 0x02000 /* have/will request CC */
258 #define TF_RCVD_CC 0x04000 /* a CC was received in SYN */
259 #define TF_SENDCCNEW 0x08000 /* send CCnew instead of CC in SYN */
260 #define TF_MORETOCOME 0x10000 /* More data to be appended to sock */
261 #define TF_LOCAL 0x20000 /* connection to a host on local link */
262 #define TF_RXWIN0SENT 0x40000 /* sent a receiver win 0 in response */
263 #define TF_SLOWLINK 0x80000 /* route is a on a modem speed link */
264 #define TF_LASTIDLE 0x100000 /* connection was previously idle */
265 #define TF_FASTRECOVERY 0x200000 /* in NewReno Fast Recovery */
266 #define TF_WASFRECOVERY 0x400000 /* was in NewReno Fast Recovery */
267 #define TF_SIGNATURE 0x800000 /* require MD5 digests (RFC2385) */
268 #define TF_MAXSEGSNT 0x1000000 /* last segment sent was a full segment */
269 #define TF_ENABLE_ECN 0x2000000 /* Enable ECN */
270 #define TF_PMTUD 0x4000000 /* Perform Path MTU Discovery for this connection */
271 #define TF_CLOSING 0x8000000 /* pending tcp close */
272 #define TF_TSO 0x10000000 /* TCP Segment Offloading is enable on this connection */
273 #define TF_BLACKHOLE 0x20000000 /* Path MTU Discovery Black Hole detection */
274 #define TF_TIMER_ONLIST 0x40000000 /* pcb is on tcp_timer_list */
275 #define TF_STRETCHACK 0x80000000 /* receiver is going to delay acks */
276
277 tcp_seq snd_una; /* send unacknowledged */
278 tcp_seq snd_max; /* highest sequence number sent;
279 * used to recognize retransmits
280 */
281 tcp_seq snd_nxt; /* send next */
282 tcp_seq snd_up; /* send urgent pointer */
283
284 tcp_seq snd_wl1; /* window update seg seq number */
285 tcp_seq snd_wl2; /* window update seg ack number */
286 tcp_seq iss; /* initial send sequence number */
287 tcp_seq irs; /* initial receive sequence number */
288
289 tcp_seq rcv_nxt; /* receive next */
290 tcp_seq rcv_adv; /* advertised window */
291 u_int32_t rcv_wnd; /* receive window */
292 tcp_seq rcv_up; /* receive urgent pointer */
293
294 u_int32_t snd_wnd; /* send window */
295 u_int32_t snd_cwnd; /* congestion-controlled window */
296 u_int32_t snd_ssthresh; /* snd_cwnd size threshold for
297 * for slow start exponential to
298 * linear switch
299 */
300 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */
301
302 u_int t_maxopd; /* mss plus options */
303
304 u_int32_t t_rcvtime; /* time at which a packet was received */
305 u_int32_t t_starttime; /* time connection was established */
306 int t_rtttime; /* tcp clock when rtt calculation was started */
307 tcp_seq t_rtseq; /* sequence number being timed */
308
309 u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */
310 u_int32_t rfbuf_cnt; /* recv buffer autoscaling byte count */
311
312 int t_rxtcur; /* current retransmit value (ticks) */
313 u_int t_maxseg; /* maximum segment size */
314 int t_srtt; /* smoothed round-trip time */
315 int t_rttvar; /* variance in round-trip time */
316
317 u_int16_t t_reassqlen; /* length of reassembly queue */
318 u_int16_t t_rxtshift; /* log(2) of rexmt exp. backoff */
319 u_int t_rttmin; /* minimum rtt allowed */
320 u_int t_rttbest; /* best rtt we've seen */
321 u_int t_rttcur; /* most recent value of rtt */
322 u_int32_t t_rttupdated; /* number of times rtt sampled */
323 u_int32_t t_rxt_conndroptime; /* retxmt conn gets dropped after this time, when set */
324 u_int32_t t_rxtstart; /* time at which retransmission started */
325 u_int32_t max_sndwnd; /* largest window peer has offered */
326
327 int t_softerror; /* possible error not yet reported */
328 /* out-of-band data */
329 char t_oobflags; /* have some */
330 char t_iobc; /* input character */
331 #define TCPOOB_HAVEDATA 0x01
332 #define TCPOOB_HADDATA 0x02
333 /* RFC 1323 variables */
334 u_int8_t snd_scale; /* window scaling for send window */
335 u_int8_t rcv_scale; /* window scaling for recv window */
336 u_int8_t request_r_scale; /* pending window scaling */
337 u_int8_t requested_s_scale;
338 u_int8_t tcp_cc_index; /* index of congestion control algorithm */
339 u_int8_t t_adaptive_rtimo; /* Read timeout used as a multiple of RTT */
340 u_int8_t t_adaptive_wtimo; /* Write timeout used as a multiple of RTT */
341 u_int8_t t_stretchack_delayed; /* stretch ack delayed */
342
343 /* State for limiting early retransmits when SACK is not enabled */
344 u_int16_t t_early_rexmt_count; /* count of early rexmts */
345 u_int32_t t_early_rexmt_win; /* window for limiting early rexmts */
346
347 u_int32_t ts_recent; /* timestamp echo data */
348
349 u_int32_t ts_recent_age; /* when last updated */
350 tcp_seq last_ack_sent;
351 /* RFC 1644 variables */
352 tcp_cc cc_send; /* send connection count */
353 tcp_cc cc_recv; /* receive connection count */
354 /* RFC 3465 variables */
355 u_int32_t t_bytes_acked; /* ABC "bytes_acked" parameter */
356
357 int t_lastchain; /* amount of packets chained last time around */
358 u_int16_t t_unacksegs; /* received but unacked segments for delaying acks */
359 u_int8_t t_rexmtthresh; /* duplicate ack threshold for entering fast recovery */
360 u_int8_t t_rtimo_probes; /* number of adaptive rtimo probes sent */
361 u_int32_t t_persist_timeout; /* ZWP persistence limit as set by PERSIST_TIMEOUT */
362 u_int32_t t_persist_stop; /* persistence limit deadline if triggered by ZWP */
363 u_int32_t t_notsent_lowat; /* Low water for not sent data */
364
365 /* Receiver state for stretch-ack algorithm */
366 u_int32_t rcv_unackwin; /* to measure win for stretching acks */
367 u_int32_t rcv_by_unackwin; /* bytes seen during the last ack-stretching win */
368 u_int32_t rcv_nostrack_ts; /* timestamp when stretch ack was disabled automatically */
369 u_int16_t rcv_waitforss; /* wait for packets during slow-start */
370 u_int16_t ecn_flags;
371 #define TE_SETUPSENT 0x01 /* Indicate we have sent ECN-SETUP SYN or SYN-ACK */
372 #define TE_SETUPRECEIVED 0x02 /* Indicate we have received ECN-SETUP SYN or SYN-ACK */
373 #define TE_SENDIPECT 0x04 /* Indicate we haven't sent or received non-ECN-setup SYN or SYN-ACK */
374 #define TE_SENDCWR 0x08 /* Indicate that the next non-retransmit should have the TCP CWR flag set */
375 #define TE_SENDECE 0x10 /* Indicate that the next packet should have the TCP ECE flag set */
376 #define TE_INRECOVERY 0x20 /* connection entered recovery after receiving ECE */
377 #define TE_ECN_ON (TE_SETUPSENT | TE_SETUPRECEIVED) /* Indicate ECN was successfully negotiated on a connection) */
378
379 /* state for bad retransmit recovery */
380 u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */
381 u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
382 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */
383 int t_srtt_prev; /* srtt prior to retransmit */
384 int t_rttvar_prev; /* rttvar prior to retransmit */
385 u_int32_t t_badrexmt_time; /* bad rexmt detection time */
386
387 /* Packet reordering metric */
388 u_int16_t t_reorderwin; /* Reordering late time offset */
389
390 /* SACK related state */
391 int16_t snd_numholes; /* number of holes seen by sender */
392 TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
393 /* SACK scoreboard (sorted) */
394 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/
395 int rcv_numsacks; /* # distinct sack blks present */
396 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
397 tcp_seq sack_newdata; /* New data xmitted in this recovery
398 episode starts at this seq number */
399 struct sackhint sackhint; /* SACK scoreboard hint */
400
401 u_int32_t t_pktlist_sentlen; /* total bytes in transmit chain */
402 struct mbuf *t_pktlist_head; /* First packet in transmit chain */
403 struct mbuf *t_pktlist_tail; /* Last packet in transmit chain */
404
405 u_int32_t t_keepidle; /* keepalive idle timer (override global if > 0) */
406 u_int32_t t_keepinit; /* connection timeout, i.e. idle time
407 in SYN_SENT or SYN_RECV state */
408 u_int32_t t_keepintvl; /* interval between keepalives */
409 u_int32_t t_keepcnt; /* number of keepalives before close */
410
411 u_int32_t tso_max_segment_size; /* TSO maximum segment unit for NIC */
412 u_int32_t t_pmtud_saved_maxopd; /* MSS saved before performing PMTU-D BlackHole detection */
413 u_int32_t t_pmtud_start_ts; /* Time of PMTUD blackhole detection */
414
415 struct
416 {
417 u_int32_t rxduplicatebytes;
418 u_int32_t rxoutoforderbytes;
419 u_int32_t txretransmitbytes;
420 u_int8_t synrxtshift;
421 u_int8_t unused;
422 u_int16_t unused_pad_to_8;
423 } t_stat;
424
425 /* Background congestion related state */
426 uint32_t rtt_hist[N_RTT_BASE]; /* history of minimum RTT */
427 uint32_t rtt_count; /* Number of RTT samples in recent base history */
428 uint32_t bg_ssthresh; /* Slow start threshold until delay increases */
429 uint32_t t_flagsext; /* Another field to accommodate more flags */
430 #define TF_RXTFINDROP 0x1 /* Drop conn after retransmitting FIN 3 times */
431 #define TF_RCVUNACK_WAITSS 0x2 /* set when the receiver should not stretch acks */
432 #define TF_BWMEAS_INPROGRESS 0x4 /* Indicate BW meas is happening */
433 #define TF_MEASURESNDBW 0x8 /* Measure send bw on this connection */
434 #define TF_LRO_OFFLOADED 0x10 /* Connection LRO offloaded */
435 #define TF_SACK_ENABLE 0x20 /* SACK is enabled */
436 #define TF_RECOMPUTE_RTT 0x40 /* recompute RTT after spurious retransmit */
437 #define TF_DETECT_READSTALL 0x80 /* Used to detect a stall during read operation */
438 #define TF_RECV_THROTTLE 0x100 /* Input throttling active */
439 #define TF_NOSTRETCHACK 0x200 /* ack every other packet */
440 #define TF_STREAMEOW 0x400 /* Last packet was small indicating end of write */
441 #define TF_NOTIMEWAIT 0x800 /* Avoid going into time-wait */
442 #define TF_SENT_TLPROBE 0x1000 /* Sent data in PTO */
443 #define TF_PKTS_REORDERED 0x2000 /* Detected reordering */
444 #define TF_DELAY_RECOVERY 0x4000 /* delay fast recovery */
445 #define TF_FORCE 0x8000 /* force 1 byte out */
446 #define TF_DISABLE_STRETCHACK 0x10000 /* auto-disable stretch ack */
447 #define TF_NOBLACKHOLE_DETECTION 0x20000 /* Disable PMTU blackhole detection */
448
449 #if TRAFFIC_MGT
450 /* Inter-arrival jitter related state */
451 uint32_t iaj_rcv_ts; /* tcp clock when the first packet was received */
452 uint16_t iaj_size; /* Size of packet for iaj measurement */
453 uint16_t iaj_small_pkt; /* Count of packets smaller than iaj_size */
454 uint16_t iaj_pktcnt; /* packet count, to avoid throttling initially */
455 uint16_t acc_iaj; /* Accumulated iaj */
456 tcp_seq iaj_rwintop; /* recent max advertised window */
457 uint32_t avg_iaj; /* Mean */
458 uint32_t std_dev_iaj; /* Standard deviation */
459 #endif /* TRAFFIC_MGT */
460 struct bwmeas *t_bwmeas; /* State for bandwidth measurement */
461 uint32_t t_lropktlen; /* Bytes in a LRO frame */
462 tcp_seq t_idleat; /* rcv_nxt at idle time */
463 TAILQ_ENTRY(tcpcb) t_twentry; /* link for time wait queue */
464 struct tcp_ccstate *t_ccstate; /* congestion control related state */
465 /* Tail loss probe related state */
466 tcp_seq t_tlphighrxt; /* snd_nxt after PTO */
467 u_int32_t t_tlpstart; /* timestamp at PTO */
468 #if MPTCP
469 u_int32_t t_mpflags; /* flags for multipath TCP */
470
471 #define TMPF_PREESTABLISHED 0x00000001 /* conn in pre-established state */
472 #define TMPF_SENT_KEYS 0x00000002 /* indicates that keys were sent */
473 #define TMPF_MPTCP_TRUE 0x00000004 /* negotiated MPTCP successfully */
474 #define TMPF_MPTCP_RCVD_KEY 0x00000008 /* state for 3-way handshake */
475 #define TMPF_SND_MPPRIO 0x00000010 /* send priority of subflow */
476 #define TMPF_SND_REM_ADDR 0x00000020 /* initiate address removal */
477 #define TMPF_UNUSED 0x00000040 /* address addition acked by peer */
478 #define TMPF_JOINED_FLOW 0x00000080 /* Indicates additional flow */
479 #define TMPF_BACKUP_PATH 0x00000100 /* Indicates backup path */
480 #define TMPF_MPTCP_ACKNOW 0x00000200 /* Send Data ACK */
481 #define TMPF_SEND_DSN 0x00000400 /* Send DSN mapping */
482 #define TMPF_SEND_DFIN 0x00000800 /* Send Data FIN */
483 #define TMPF_RECV_DFIN 0x00001000 /* Recv Data FIN */
484 #define TMPF_SENT_JOIN 0x00002000 /* Sent Join */
485 #define TMPF_RECVD_JOIN 0x00004000 /* Received Join */
486 #define TMPF_RESET 0x00008000 /* Send RST */
487 #define TMPF_TCP_FALLBACK 0x00010000 /* Fallback to TCP */
488 #define TMPF_FASTCLOSE 0x00020000 /* Send Fastclose option */
489 #define TMPF_EMBED_DSN 0x00040000 /* tp has DSN mapping */
490 #define TMPF_MPTCP_READY 0x00080000 /* Can send DSS options on data */
491 #define TMPF_INFIN_SENT 0x00100000 /* Sent infinite mapping */
492 #define TMPF_SND_MPFAIL 0x00200000 /* Received mapping csum failure */
493 #define TMPF_FASTJOIN_SEND 0x00400000 /* Fast join early data send */
494 #define TMPF_FASTJOINBY2_SEND 0x00800000 /* Fast join send after 3 WHS */
495
496 void *t_mptcb; /* pointer to MPTCP TCB */
497 tcp_seq t_mpuna; /* unacknowledged sequence */
498 struct mpt_dsn_map t_rcv_map; /* Receive mapping list */
499 u_int8_t t_local_aid; /* Addr Id for authentication */
500 u_int8_t t_rem_aid; /* Addr ID of another subflow */
501 u_int8_t t_mprxtshift; /* join retransmission */
502 #endif /* MPTCP */
503 };
504
505 #define IN_FASTRECOVERY(tp) (tp->t_flags & TF_FASTRECOVERY)
506 #define SACK_ENABLED(tp) (tp->t_flagsext & TF_SACK_ENABLE)
507
508 /*
509 * If the connection is in a throttled state due to advisory feedback from
510 * the interface output queue, reset that state. We do this in favor
511 * of entering recovery because the data transfer during recovery
512 * should be just a trickle and it will help to improve performance.
513 * We also do not want to back off twice in the same RTT.
514 */
515 #define ENTER_FASTRECOVERY(_tp_) do { \
516 (_tp_)->t_flags |= TF_FASTRECOVERY; \
517 if (INP_IS_FLOW_CONTROLLED((_tp_)->t_inpcb)) \
518 inp_reset_fc_state((_tp_)->t_inpcb); \
519 } while(0)
520
521 #define EXIT_FASTRECOVERY(_tp_) do { \
522 (_tp_)->t_flags &= ~TF_FASTRECOVERY; \
523 (_tp_)->t_dupacks = 0; \
524 (_tp_)->t_rexmtthresh = tcprexmtthresh; \
525 (_tp_)->t_bytes_acked = 0; \
526 (_tp_)->ecn_flags &= ~TE_INRECOVERY; \
527 (_tp_)->t_timer[TCPT_PTO] = 0; \
528 } while(0)
529
530 /*
531 * When the number of duplicate acks received is less than
532 * the retransmit threshold, use Limited Transmit algorithm
533 */
534 extern int tcprexmtthresh;
535 #define ALLOW_LIMITED_TRANSMIT(_tp_) \
536 ((_tp_)->t_dupacks > 0 && \
537 (_tp_)->t_dupacks < (_tp_)->t_rexmtthresh && \
538 ((_tp_)->t_flagsext & (TF_PKTS_REORDERED|TF_DELAY_RECOVERY)) \
539 != (TF_PKTS_REORDERED|TF_DELAY_RECOVERY))
540
541 /*
542 * This condition is true is timestamp option is supported
543 * on a connection.
544 */
545 #define TSTMP_SUPPORTED(_tp_) \
546 (((_tp_)->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP)) == \
547 (TF_REQ_TSTMP|TF_RCVD_TSTMP))
548
549 /*
550 * Gives number of bytes acked by this ack
551 */
552 #define BYTES_ACKED(_th_, _tp_) \
553 ((_th_)->th_ack - (_tp_)->snd_una)
554
555 enum tcp_cc_event {
556 TCP_CC_CWND_INIT,
557 TCP_CC_INSEQ_ACK_RCVD,
558 TCP_CC_ACK_RCVD,
559 TCP_CC_ENTER_FASTRECOVERY,
560 TCP_CC_IN_FASTRECOVERY,
561 TCP_CC_EXIT_FASTRECOVERY,
562 TCP_CC_PARTIAL_ACK,
563 TCP_CC_IDLE_TIMEOUT,
564 TCP_CC_REXMT_TIMEOUT,
565 TCP_CC_ECN_RCVD,
566 TCP_CC_BAD_REXMT_RECOVERY,
567 TCP_CC_OUTPUT_ERROR,
568 TCP_CC_CHANGE_ALGO,
569 TCP_CC_FLOW_CONTROL,
570 TCP_CC_SUSPEND,
571 TCP_CC_LIMITED_TRANSMIT,
572 TCP_CC_EARLY_RETRANSMIT,
573 TCP_CC_TLP_RECOVERY,
574 TCP_CC_TLP_RECOVER_LASTPACKET,
575 TCP_CC_DELAY_FASTRECOVERY,
576 TCP_CC_TLP_IN_FASTRECOVERY
577 };
578
579 /*
580 * Structure to hold TCP options that are only used during segment
581 * processing (in tcp_input), but not held in the tcpcb.
582 * It's basically used to reduce the number of parameters
583 * to tcp_dooptions.
584 */
585 struct tcpopt {
586 u_int32_t to_flags; /* which options are present */
587 #define TOF_TS 0x0001 /* timestamp */
588 #define TOF_MSS 0x0010
589 #define TOF_SCALE 0x0020
590 #define TOF_SIGNATURE 0x0040 /* signature option present */
591 #define TOF_SIGLEN 0x0080 /* signature length valid (RFC2385) */
592 #define TOF_SACK 0x0100 /* Peer sent SACK option */
593 #define TOF_MPTCP 0x0200 /* MPTCP options to be dropped */
594 u_int32_t to_tsval;
595 u_int32_t to_tsecr;
596 u_int16_t to_mss;
597 u_int8_t to_requested_s_scale;
598 u_int8_t to_nsacks; /* number of SACK blocks */
599 u_char *to_sacks; /* pointer to the first SACK blocks */
600 };
601
602 /*
603 * The TAO cache entry which is stored in the protocol family specific
604 * portion of the route metrics.
605 */
606 struct rmxp_tao {
607 tcp_cc tao_cc; /* latest CC in valid SYN */
608 tcp_cc tao_ccsent; /* latest CC sent to peer */
609 u_short tao_mssopt; /* peer's cached MSS */
610 #ifdef notyet
611 u_short tao_flags; /* cache status flags */
612 #define TAOF_DONT 0x0001 /* peer doesn't understand rfc1644 */
613 #define TAOF_OK 0x0002 /* peer does understand rfc1644 */
614 #define TAOF_UNDEF 0 /* we don't know yet */
615 #endif /* notyet */
616 };
617 #define rmx_taop(r) ((struct rmxp_tao *)(r).rmx_filler)
618
619 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb)
620 #define sototcpcb(so) (intotcpcb(sotoinpcb(so)))
621
622 /*
623 * The rtt measured is in milliseconds as the timestamp granularity is
624 * a millisecond. The smoothed round-trip time and estimated variance
625 * are stored as fixed point numbers scaled by the values below.
626 * For convenience, these scales are also used in smoothing the average
627 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
628 * With these scales, srtt has 5 bits to the right of the binary point,
629 * and thus an "ALPHA" of 0.875. rttvar has 4 bits to the right of the
630 * binary point, and is smoothed with an ALPHA of 0.75.
631 */
632 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */
633 #define TCP_RTT_SHIFT 5 /* shift for srtt; 5 bits frac. */
634 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 4 bits */
635 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 4 bits */
636 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
637
638 /*
639 * The initial retransmission should happen at rtt + 4 * rttvar.
640 * Because of the way we do the smoothing, srtt and rttvar
641 * will each average +1/2 tick of bias. When we compute
642 * the retransmit timer, we want 1/2 tick of rounding and
643 * 1 extra tick because of +-1/2 tick uncertainty in the
644 * firing of the timer. The bias will give us exactly the
645 * 1.5 tick we need. But, because the bias is
646 * statistical, we have to test that we don't drop below
647 * the minimum feasible timer (which is 2 ticks).
648 * This version of the macro adapted from a paper by Lawrence
649 * Brakmo and Larry Peterson which outlines a problem caused
650 * by insufficient precision in the original implementation,
651 * which results in inappropriately large RTO values for very
652 * fast networks.
653 */
654 #define TCP_REXMTVAL(tp) \
655 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
656 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
657
658 /*
659 * Jaguar compatible TCP control block, for xtcpcb
660 * Does not have the old fields
661 */
662 struct otcpcb {
663 #else
664 struct tseg_qent;
665 _TCPCB_LIST_HEAD(tsegqe_head, tseg_qent);
666
667 struct tcpcb {
668 #endif /* KERNEL_PRIVATE */
669 #if defined(KERNEL_PRIVATE)
670 u_int32_t t_segq;
671 #else
672 struct tsegqe_head t_segq;
673 #endif /* KERNEL_PRIVATE */
674 int t_dupacks; /* consecutive dup acks recd */
675 u_int32_t unused; /* unused now: was t_template */
676
677 int t_timer[TCPT_NTIMERS_EXT]; /* tcp timers */
678
679 _TCPCB_PTR(struct inpcb *) t_inpcb; /* back pointer to internet pcb */
680 int t_state; /* state of this connection */
681 u_int t_flags;
682 #define TF_ACKNOW 0x00001 /* ack peer immediately */
683 #define TF_DELACK 0x00002 /* ack, but try to delay it */
684 #define TF_NODELAY 0x00004 /* don't delay packets to coalesce */
685 #define TF_NOOPT 0x00008 /* don't use tcp options */
686 #define TF_SENTFIN 0x00010 /* have sent FIN */
687 #define TF_REQ_SCALE 0x00020 /* have/will request window scaling */
688 #define TF_RCVD_SCALE 0x00040 /* other side has requested scaling */
689 #define TF_REQ_TSTMP 0x00080 /* have/will request timestamps */
690 #define TF_RCVD_TSTMP 0x00100 /* a timestamp was received in SYN */
691 #define TF_SACK_PERMIT 0x00200 /* other side said I could SACK */
692 #define TF_NEEDSYN 0x00400 /* send SYN (implicit state) */
693 #define TF_NEEDFIN 0x00800 /* send FIN (implicit state) */
694 #define TF_NOPUSH 0x01000 /* don't push */
695 #define TF_REQ_CC 0x02000 /* have/will request CC */
696 #define TF_RCVD_CC 0x04000 /* a CC was received in SYN */
697 #define TF_SENDCCNEW 0x08000 /* send CCnew instead of CC in SYN */
698 #define TF_MORETOCOME 0x10000 /* More data to be appended to sock */
699 #define TF_LQ_OVERFLOW 0x20000 /* listen queue overflow */
700 #define TF_RXWIN0SENT 0x40000 /* sent a receiver win 0 in response */
701 #define TF_SLOWLINK 0x80000 /* route is a on a modem speed link */
702
703 int t_force; /* 1 if forcing out a byte */
704
705 tcp_seq snd_una; /* send unacknowledged */
706 tcp_seq snd_max; /* highest sequence number sent;
707 * used to recognize retransmits
708 */
709 tcp_seq snd_nxt; /* send next */
710 tcp_seq snd_up; /* send urgent pointer */
711
712 tcp_seq snd_wl1; /* window update seg seq number */
713 tcp_seq snd_wl2; /* window update seg ack number */
714 tcp_seq iss; /* initial send sequence number */
715 tcp_seq irs; /* initial receive sequence number */
716
717 tcp_seq rcv_nxt; /* receive next */
718 tcp_seq rcv_adv; /* advertised window */
719 u_int32_t rcv_wnd; /* receive window */
720 tcp_seq rcv_up; /* receive urgent pointer */
721
722 u_int32_t snd_wnd; /* send window */
723 u_int32_t snd_cwnd; /* congestion-controlled window */
724 u_int32_t snd_ssthresh; /* snd_cwnd size threshold for
725 * for slow start exponential to
726 * linear switch
727 */
728 u_int t_maxopd; /* mss plus options */
729
730 u_int32_t t_rcvtime; /* time at which a packet was received */
731 u_int32_t t_starttime; /* time connection was established */
732 int t_rtttime; /* round trip time */
733 tcp_seq t_rtseq; /* sequence number being timed */
734
735 int t_rxtcur; /* current retransmit value (ticks) */
736 u_int t_maxseg; /* maximum segment size */
737 int t_srtt; /* smoothed round-trip time */
738 int t_rttvar; /* variance in round-trip time */
739
740 int t_rxtshift; /* log(2) of rexmt exp. backoff */
741 u_int t_rttmin; /* minimum rtt allowed */
742 u_int32_t t_rttupdated; /* number of times rtt sampled */
743 u_int32_t max_sndwnd; /* largest window peer has offered */
744
745 int t_softerror; /* possible error not yet reported */
746 /* out-of-band data */
747 char t_oobflags; /* have some */
748 char t_iobc; /* input character */
749 #define TCPOOB_HAVEDATA 0x01
750 #define TCPOOB_HADDATA 0x02
751 /* RFC 1323 variables */
752 u_char snd_scale; /* window scaling for send window */
753 u_char rcv_scale; /* window scaling for recv window */
754 u_char request_r_scale; /* pending window scaling */
755 u_char requested_s_scale;
756 u_int32_t ts_recent; /* timestamp echo data */
757
758 u_int32_t ts_recent_age; /* when last updated */
759 tcp_seq last_ack_sent;
760 /* RFC 1644 variables */
761 tcp_cc cc_send; /* send connection count */
762 tcp_cc cc_recv; /* receive connection count */
763 tcp_seq snd_recover; /* for use in fast recovery */
764 /* experimental */
765 u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */
766 u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
767 u_int32_t t_badrxtwin; /* window for retransmit recovery */
768 };
769
770
771 /*
772 * TCP statistics.
773 * Many of these should be kept per connection,
774 * but that's inconvenient at the moment.
775 */
776 struct tcpstat {
777 u_int32_t tcps_connattempt; /* connections initiated */
778 u_int32_t tcps_accepts; /* connections accepted */
779 u_int32_t tcps_connects; /* connections established */
780 u_int32_t tcps_drops; /* connections dropped */
781 u_int32_t tcps_conndrops; /* embryonic connections dropped */
782 u_int32_t tcps_closed; /* conn. closed (includes drops) */
783 u_int32_t tcps_segstimed; /* segs where we tried to get rtt */
784 u_int32_t tcps_rttupdated; /* times we succeeded */
785 u_int32_t tcps_delack; /* delayed acks sent */
786 u_int32_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
787 u_int32_t tcps_rexmttimeo; /* retransmit timeouts */
788 u_int32_t tcps_persisttimeo; /* persist timeouts */
789 u_int32_t tcps_keeptimeo; /* keepalive timeouts */
790 u_int32_t tcps_keepprobe; /* keepalive probes sent */
791 u_int32_t tcps_keepdrops; /* connections dropped in keepalive */
792
793 u_int32_t tcps_sndtotal; /* total packets sent */
794 u_int32_t tcps_sndpack; /* data packets sent */
795 u_int32_t tcps_sndbyte; /* data bytes sent */
796 u_int32_t tcps_sndrexmitpack; /* data packets retransmitted */
797 u_int32_t tcps_sndrexmitbyte; /* data bytes retransmitted */
798 u_int32_t tcps_sndacks; /* ack-only packets sent */
799 u_int32_t tcps_sndprobe; /* window probes sent */
800 u_int32_t tcps_sndurg; /* packets sent with URG only */
801 u_int32_t tcps_sndwinup; /* window update-only packets sent */
802 u_int32_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
803
804 u_int32_t tcps_rcvtotal; /* total packets received */
805 u_int32_t tcps_rcvpack; /* packets received in sequence */
806 u_int32_t tcps_rcvbyte; /* bytes received in sequence */
807 u_int32_t tcps_rcvbadsum; /* packets received with ccksum errs */
808 u_int32_t tcps_rcvbadoff; /* packets received with bad offset */
809 u_int32_t tcps_rcvmemdrop; /* packets dropped for lack of memory */
810 u_int32_t tcps_rcvshort; /* packets received too short */
811 u_int32_t tcps_rcvduppack; /* duplicate-only packets received */
812 u_int32_t tcps_rcvdupbyte; /* duplicate-only bytes received */
813 u_int32_t tcps_rcvpartduppack; /* packets with some duplicate data */
814 u_int32_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
815 u_int32_t tcps_rcvoopack; /* out-of-order packets received */
816 u_int32_t tcps_rcvoobyte; /* out-of-order bytes received */
817 u_int32_t tcps_rcvpackafterwin; /* packets with data after window */
818 u_int32_t tcps_rcvbyteafterwin; /* bytes rcvd after window */
819 u_int32_t tcps_rcvafterclose; /* packets rcvd after "close" */
820 u_int32_t tcps_rcvwinprobe; /* rcvd window probe packets */
821 u_int32_t tcps_rcvdupack; /* rcvd duplicate acks */
822 u_int32_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */
823 u_int32_t tcps_rcvackpack; /* rcvd ack packets */
824 u_int32_t tcps_rcvackbyte; /* bytes acked by rcvd acks */
825 u_int32_t tcps_rcvwinupd; /* rcvd window update packets */
826 u_int32_t tcps_pawsdrop; /* segments dropped due to PAWS */
827 u_int32_t tcps_predack; /* times hdr predict ok for acks */
828 u_int32_t tcps_preddat; /* times hdr predict ok for data pkts */
829 u_int32_t tcps_pcbcachemiss;
830 u_int32_t tcps_cachedrtt; /* times cached RTT in route updated */
831 u_int32_t tcps_cachedrttvar; /* times cached rttvar updated */
832 u_int32_t tcps_cachedssthresh; /* times cached ssthresh updated */
833 u_int32_t tcps_usedrtt; /* times RTT initialized from route */
834 u_int32_t tcps_usedrttvar; /* times RTTVAR initialized from rt */
835 u_int32_t tcps_usedssthresh; /* times ssthresh initialized from rt*/
836 u_int32_t tcps_persistdrop; /* timeout in persist state */
837 u_int32_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */
838 u_int32_t tcps_mturesent; /* resends due to MTU discovery */
839 u_int32_t tcps_listendrop; /* listen queue overflows */
840
841 /* new stats from FreeBSD 5.4 sync up */
842 u_int32_t tcps_minmssdrops; /* average minmss too low drops */
843 u_int32_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */
844 u_int32_t tcps_badrst; /* ignored RSTs in the window */
845
846 u_int32_t tcps_sc_added; /* entry added to syncache */
847 u_int32_t tcps_sc_retransmitted; /* syncache entry was retransmitted */
848 u_int32_t tcps_sc_dupsyn; /* duplicate SYN packet */
849 u_int32_t tcps_sc_dropped; /* could not reply to packet */
850 u_int32_t tcps_sc_completed; /* successful extraction of entry */
851 u_int32_t tcps_sc_bucketoverflow; /* syncache per-bucket limit hit */
852 u_int32_t tcps_sc_cacheoverflow; /* syncache cache limit hit */
853 u_int32_t tcps_sc_reset; /* RST removed entry from syncache */
854 u_int32_t tcps_sc_stale; /* timed out or listen socket gone */
855 u_int32_t tcps_sc_aborted; /* syncache entry aborted */
856 u_int32_t tcps_sc_badack; /* removed due to bad ACK */
857 u_int32_t tcps_sc_unreach; /* ICMP unreachable received */
858 u_int32_t tcps_sc_zonefail; /* zalloc() failed */
859 u_int32_t tcps_sc_sendcookie; /* SYN cookie sent */
860 u_int32_t tcps_sc_recvcookie; /* SYN cookie received */
861
862 u_int32_t tcps_hc_added; /* entry added to hostcache */
863 u_int32_t tcps_hc_bucketoverflow; /* hostcache per bucket limit hit */
864
865 /* SACK related stats */
866 u_int32_t tcps_sack_recovery_episode; /* SACK recovery episodes */
867 u_int32_t tcps_sack_rexmits; /* SACK rexmit segments */
868 u_int32_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */
869 u_int32_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */
870 u_int32_t tcps_sack_send_blocks; /* SACK blocks (options) sent */
871 u_int32_t tcps_sack_sboverflow; /* SACK sendblock overflow */
872
873 u_int32_t tcps_bg_rcvtotal; /* total background packets received */
874 u_int32_t tcps_rxtfindrop; /* drop conn after retransmitting FIN */
875 u_int32_t tcps_fcholdpacket; /* packets withheld because of flow control */
876
877 /* LRO related stats */
878 u_int32_t tcps_coalesced_pack; /* number of coalesced packets */
879 u_int32_t tcps_flowtbl_full; /* times flow table was full */
880 u_int32_t tcps_flowtbl_collision; /* collisions in flow tbl */
881 u_int32_t tcps_lro_twopack; /* 2 packets coalesced */
882 u_int32_t tcps_lro_multpack; /* 3 or 4 pkts coalesced */
883 u_int32_t tcps_lro_largepack; /* 5 or more pkts coalesced */
884
885 u_int32_t tcps_limited_txt; /* Limited transmit used */
886 u_int32_t tcps_early_rexmt; /* Early retransmit used */
887 u_int32_t tcps_sack_ackadv; /* Cumulative ack advanced along with sack */
888
889 /* Checksum related stats */
890 u_int32_t tcps_rcv_swcsum; /* tcp swcksum (inbound), packets */
891 u_int32_t tcps_rcv_swcsum_bytes; /* tcp swcksum (inbound), bytes */
892 u_int32_t tcps_rcv6_swcsum; /* tcp6 swcksum (inbound), packets */
893 u_int32_t tcps_rcv6_swcsum_bytes; /* tcp6 swcksum (inbound), bytes */
894 u_int32_t tcps_snd_swcsum; /* tcp swcksum (outbound), packets */
895 u_int32_t tcps_snd_swcsum_bytes; /* tcp swcksum (outbound), bytes */
896 u_int32_t tcps_snd6_swcsum; /* tcp6 swcksum (outbound), packets */
897 u_int32_t tcps_snd6_swcsum_bytes; /* tcp6 swcksum (outbound), bytes */
898 u_int32_t tcps_msg_unopkts; /* unordered packet on TCP msg stream */
899 u_int32_t tcps_msg_unoappendfail; /* failed to append unordered pkt */
900 u_int32_t tcps_msg_sndwaithipri; /* send is waiting for high priority data */
901
902 /* MPTCP Related stats */
903 u_int32_t tcps_invalid_mpcap; /* Invalid MPTCP capable opts */
904 u_int32_t tcps_invalid_joins; /* Invalid MPTCP joins */
905 u_int32_t tcps_mpcap_fallback; /* TCP fallback in primary */
906 u_int32_t tcps_join_fallback; /* No MPTCP in secondary */
907 u_int32_t tcps_estab_fallback; /* DSS option dropped */
908 u_int32_t tcps_invalid_opt; /* Catchall error stat */
909 u_int32_t tcps_mp_outofwin; /* Packet lies outside the
910 shared recv window */
911 u_int32_t tcps_mp_reducedwin; /* Reduced subflow window */
912 u_int32_t tcps_mp_badcsum; /* Bad DSS csum */
913 u_int32_t tcps_mp_oodata; /* Out of order data */
914 u_int32_t tcps_mp_switches; /* number of subflow switch */
915 u_int32_t tcps_mp_rcvtotal; /* number of rcvd packets */
916 u_int32_t tcps_mp_rcvbytes; /* number of bytes received */
917 u_int32_t tcps_mp_sndpacks; /* number of data packs sent */
918 u_int32_t tcps_mp_sndbytes; /* number of bytes sent */
919 u_int32_t tcps_join_rxmts; /* join ack retransmits */
920 u_int32_t tcps_tailloss_rto; /* RTO due to tail loss */
921 u_int32_t tcps_reordered_pkts; /* packets reorderd */
922 u_int32_t tcps_recovered_pkts; /* recovered after loss */
923 u_int32_t tcps_pto; /* probe timeout */
924 u_int32_t tcps_rto_after_pto; /* RTO after a probe */
925 u_int32_t tcps_tlp_recovery; /* TLP induced fast recovery */
926 u_int32_t tcps_tlp_recoverlastpkt; /* TLP recoverd last pkt */
927 u_int32_t tcps_ecn_setup; /* connection negotiated ECN */
928 u_int32_t tcps_sent_cwr; /* Sent CWR, ECE received */
929 u_int32_t tcps_sent_ece; /* Sent ECE notification */
930 u_int32_t tcps_detect_reordering; /* Detect pkt reordering */
931 u_int32_t tcps_delay_recovery; /* Delay fast recovery */
932 u_int32_t tcps_avoid_rxmt; /* Retransmission was avoided */
933 u_int32_t tcps_unnecessary_rxmt; /* Retransmission was not needed */
934 u_int32_t tcps_nostretchack; /* disabled stretch ack algorithm on a connection */
935 u_int32_t tcps_rescue_rxmt; /* SACK rescue retransmit */
936 u_int32_t tcps_pto_in_recovery; /* PTO during fast recovery */
937 u_int32_t tcps_pmtudbh_reverted; /* PMTU Blackhole detection, segement size reverted */
938 };
939
940 struct tcpstat_local {
941 u_int64_t badformat;
942 u_int64_t unspecv6;
943 u_int64_t synfin;
944 u_int64_t badformatipsec;
945 u_int64_t noconnnolist;
946 u_int64_t noconnlist;
947 u_int64_t listbadsyn;
948 u_int64_t icmp6unreach;
949 u_int64_t deprecate6;
950 u_int64_t ooopacket;
951 u_int64_t rstinsynrcv;
952 u_int64_t dospacket;
953 u_int64_t cleanup;
954 u_int64_t synwindow;
955 };
956
957 #pragma pack(4)
958
959 /*
960 * TCB structure exported to user-land via sysctl(3).
961 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
962 * included. Not all of our clients do.
963 */
964
965 struct xtcpcb {
966 u_int32_t xt_len;
967 #ifdef KERNEL_PRIVATE
968 struct inpcb_compat xt_inp;
969 #else
970 struct inpcb xt_inp;
971 #endif
972 #ifdef KERNEL_PRIVATE
973 struct otcpcb xt_tp;
974 #else
975 struct tcpcb xt_tp;
976 #endif
977 struct xsocket xt_socket;
978 u_quad_t xt_alignment_hack;
979 };
980
981
982 struct xtcpcb64 {
983 u_int32_t xt_len;
984 struct xinpcb64 xt_inpcb;
985
986 u_int64_t t_segq;
987 int t_dupacks; /* consecutive dup acks recd */
988
989 int t_timer[TCPT_NTIMERS_EXT]; /* tcp timers */
990
991 int t_state; /* state of this connection */
992 u_int t_flags;
993
994 int t_force; /* 1 if forcing out a byte */
995
996 tcp_seq snd_una; /* send unacknowledged */
997 tcp_seq snd_max; /* highest sequence number sent;
998 * used to recognize retransmits
999 */
1000 tcp_seq snd_nxt; /* send next */
1001 tcp_seq snd_up; /* send urgent pointer */
1002
1003 tcp_seq snd_wl1; /* window update seg seq number */
1004 tcp_seq snd_wl2; /* window update seg ack number */
1005 tcp_seq iss; /* initial send sequence number */
1006 tcp_seq irs; /* initial receive sequence number */
1007
1008 tcp_seq rcv_nxt; /* receive next */
1009 tcp_seq rcv_adv; /* advertised window */
1010 u_int32_t rcv_wnd; /* receive window */
1011 tcp_seq rcv_up; /* receive urgent pointer */
1012
1013 u_int32_t snd_wnd; /* send window */
1014 u_int32_t snd_cwnd; /* congestion-controlled window */
1015 u_int32_t snd_ssthresh; /* snd_cwnd size threshold for
1016 * for slow start exponential to
1017 * linear switch
1018 */
1019 u_int t_maxopd; /* mss plus options */
1020
1021 u_int32_t t_rcvtime; /* time at which a packet was received */
1022 u_int32_t t_starttime; /* time connection was established */
1023 int t_rtttime; /* round trip time */
1024 tcp_seq t_rtseq; /* sequence number being timed */
1025
1026 int t_rxtcur; /* current retransmit value (ticks) */
1027 u_int t_maxseg; /* maximum segment size */
1028 int t_srtt; /* smoothed round-trip time */
1029 int t_rttvar; /* variance in round-trip time */
1030
1031 int t_rxtshift; /* log(2) of rexmt exp. backoff */
1032 u_int t_rttmin; /* minimum rtt allowed */
1033 u_int32_t t_rttupdated; /* number of times rtt sampled */
1034 u_int32_t max_sndwnd; /* largest window peer has offered */
1035
1036 int t_softerror; /* possible error not yet reported */
1037 /* out-of-band data */
1038 char t_oobflags; /* have some */
1039 char t_iobc; /* input character */
1040 /* RFC 1323 variables */
1041 u_char snd_scale; /* window scaling for send window */
1042 u_char rcv_scale; /* window scaling for recv window */
1043 u_char request_r_scale; /* pending window scaling */
1044 u_char requested_s_scale;
1045 u_int32_t ts_recent; /* timestamp echo data */
1046
1047 u_int32_t ts_recent_age; /* when last updated */
1048 tcp_seq last_ack_sent;
1049 /* RFC 1644 variables */
1050 tcp_cc cc_send; /* send connection count */
1051 tcp_cc cc_recv; /* receive connection count */
1052 tcp_seq snd_recover; /* for use in fast recovery */
1053 /* experimental */
1054 u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */
1055 u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
1056 u_int32_t t_badrxtwin; /* window for retransmit recovery */
1057
1058 u_quad_t xt_alignment_hack;
1059 };
1060
1061
1062 #ifdef PRIVATE
1063
1064 struct xtcpcb_n {
1065 u_int32_t xt_len;
1066 u_int32_t xt_kind; /* XSO_TCPCB */
1067
1068 u_int64_t t_segq;
1069 int t_dupacks; /* consecutive dup acks recd */
1070
1071 int t_timer[TCPT_NTIMERS_EXT]; /* tcp timers */
1072
1073 int t_state; /* state of this connection */
1074 u_int t_flags;
1075
1076 int t_force; /* 1 if forcing out a byte */
1077
1078 tcp_seq snd_una; /* send unacknowledged */
1079 tcp_seq snd_max; /* highest sequence number sent;
1080 * used to recognize retransmits
1081 */
1082 tcp_seq snd_nxt; /* send next */
1083 tcp_seq snd_up; /* send urgent pointer */
1084
1085 tcp_seq snd_wl1; /* window update seg seq number */
1086 tcp_seq snd_wl2; /* window update seg ack number */
1087 tcp_seq iss; /* initial send sequence number */
1088 tcp_seq irs; /* initial receive sequence number */
1089
1090 tcp_seq rcv_nxt; /* receive next */
1091 tcp_seq rcv_adv; /* advertised window */
1092 u_int32_t rcv_wnd; /* receive window */
1093 tcp_seq rcv_up; /* receive urgent pointer */
1094
1095 u_int32_t snd_wnd; /* send window */
1096 u_int32_t snd_cwnd; /* congestion-controlled window */
1097 u_int32_t snd_ssthresh; /* snd_cwnd size threshold for
1098 * for slow start exponential to
1099 * linear switch
1100 */
1101 u_int t_maxopd; /* mss plus options */
1102
1103 u_int32_t t_rcvtime; /* time at which a packet was received */
1104 u_int32_t t_starttime; /* time connection was established */
1105 int t_rtttime; /* round trip time */
1106 tcp_seq t_rtseq; /* sequence number being timed */
1107
1108 int t_rxtcur; /* current retransmit value (ticks) */
1109 u_int t_maxseg; /* maximum segment size */
1110 int t_srtt; /* smoothed round-trip time */
1111 int t_rttvar; /* variance in round-trip time */
1112
1113 int t_rxtshift; /* log(2) of rexmt exp. backoff */
1114 u_int t_rttmin; /* minimum rtt allowed */
1115 u_int32_t t_rttupdated; /* number of times rtt sampled */
1116 u_int32_t max_sndwnd; /* largest window peer has offered */
1117
1118 int t_softerror; /* possible error not yet reported */
1119 /* out-of-band data */
1120 char t_oobflags; /* have some */
1121 char t_iobc; /* input character */
1122 /* RFC 1323 variables */
1123 u_char snd_scale; /* window scaling for send window */
1124 u_char rcv_scale; /* window scaling for recv window */
1125 u_char request_r_scale; /* pending window scaling */
1126 u_char requested_s_scale;
1127 u_int32_t ts_recent; /* timestamp echo data */
1128
1129 u_int32_t ts_recent_age; /* when last updated */
1130 tcp_seq last_ack_sent;
1131 /* RFC 1644 variables */
1132 tcp_cc cc_send; /* send connection count */
1133 tcp_cc cc_recv; /* receive connection count */
1134 tcp_seq snd_recover; /* for use in fast recovery */
1135 /* experimental */
1136 u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */
1137 u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
1138 };
1139
1140 #endif /* PRIVATE */
1141
1142 #pragma pack()
1143
1144 /*
1145 * Names for TCP sysctl objects
1146 */
1147 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
1148 #define TCPCTL_DO_RFC1644 2 /* use RFC-1644 extensions */
1149 #define TCPCTL_MSSDFLT 3 /* MSS default */
1150 #define TCPCTL_STATS 4 /* statistics (read-only) */
1151 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */
1152 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
1153 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
1154 #define TCPCTL_SENDSPACE 8 /* send buffer space */
1155 #define TCPCTL_RECVSPACE 9 /* receive buffer space */
1156 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */
1157 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */
1158 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */
1159 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */
1160 #define TCPCTL_MAXID 14
1161
1162 #ifdef BSD_KERNEL_PRIVATE
1163 #include <sys/bitstring.h>
1164
1165 #define TCP_PKTLIST_CLEAR(tp) { \
1166 (tp)->t_pktlist_head = (tp)->t_pktlist_tail = NULL; \
1167 (tp)->t_lastchain = (tp)->t_pktlist_sentlen = 0; \
1168 }
1169
1170 #define TCPCTL_NAMES { \
1171 { 0, 0 }, \
1172 { "rfc1323", CTLTYPE_INT }, \
1173 { "rfc1644", CTLTYPE_INT }, \
1174 { "mssdflt", CTLTYPE_INT }, \
1175 { "stats", CTLTYPE_STRUCT }, \
1176 { "rttdflt", CTLTYPE_INT }, \
1177 { "keepidle", CTLTYPE_INT }, \
1178 { "keepintvl", CTLTYPE_INT }, \
1179 { "sendspace", CTLTYPE_INT }, \
1180 { "recvspace", CTLTYPE_INT }, \
1181 { "keepinit", CTLTYPE_INT }, \
1182 { "pcblist", CTLTYPE_STRUCT }, \
1183 { "delacktime", CTLTYPE_INT }, \
1184 { "v6mssdflt", CTLTYPE_INT }, \
1185 }
1186
1187 #ifdef SYSCTL_DECL
1188 SYSCTL_DECL(_net_inet_tcp);
1189 #endif /* SYSCTL_DECL */
1190
1191 /*
1192 * Flags for TCP's connectx(2) user-protocol request routine.
1193 */
1194 #if MPTCP
1195 #define TCP_CONNREQF_MPTCP 0x1 /* called internally by MPTCP */
1196 #endif /* MPTCP */
1197
1198 extern struct inpcbhead tcb; /* head of queue of active tcpcb's */
1199 extern struct inpcbinfo tcbinfo;
1200 extern struct tcpstat tcpstat; /* tcp statistics */
1201 extern int tcp_mssdflt; /* XXX */
1202 extern int tcp_minmss;
1203 extern int ss_fltsz;
1204 extern int ss_fltsz_local;
1205 extern int tcp_do_rfc3390; /* Calculate ss_fltsz according to RFC 3390 */
1206 extern int target_qdelay;
1207 extern u_int32_t tcp_now; /* for RFC 1323 timestamps */
1208 extern struct timeval tcp_uptime;
1209 extern lck_spin_t *tcp_uptime_lock;
1210 extern int tcp_delack_enabled;
1211 extern int tcp_do_sack; /* SACK enabled/disabled */
1212 extern int tcp_do_rfc3465;
1213 extern int tcp_do_rfc3465_lim2;
1214 extern int maxseg_unacked;
1215 extern int tcp_use_newreno;
1216
1217
1218 #if CONFIG_IFEF_NOWINDOWSCALE
1219 extern int tcp_obey_ifef_nowindowscale;
1220 #endif
1221
1222 struct protosw;
1223 struct domain;
1224
1225 struct tcp_respond_args {
1226 unsigned int ifscope;
1227 unsigned int nocell:1,
1228 noexpensive:1,
1229 awdl_unrestricted:1;
1230 };
1231
1232 void tcp_canceltimers(struct tcpcb *);
1233 struct tcpcb *
1234 tcp_close(struct tcpcb *);
1235 void tcp_ctlinput(int, struct sockaddr *, void *);
1236 int tcp_ctloutput(struct socket *, struct sockopt *);
1237 struct tcpcb *
1238 tcp_drop(struct tcpcb *, int);
1239 void tcp_drain(void);
1240 void tcp_getrt_rtt(struct tcpcb *tp, struct rtentry *rt);
1241 struct rmxp_tao *
1242 tcp_gettaocache(struct inpcb *);
1243 void tcp_init(struct protosw *, struct domain *);
1244 void tcp_input(struct mbuf *, int);
1245 void tcp_mss(struct tcpcb *, int, unsigned int);
1246 int tcp_mssopt(struct tcpcb *);
1247 void tcp_drop_syn_sent(struct inpcb *, int);
1248 void tcp_mtudisc(struct inpcb *, int);
1249 struct tcpcb *
1250 tcp_newtcpcb(struct inpcb *);
1251 int tcp_output(struct tcpcb *);
1252 void tcp_respond(struct tcpcb *, void *, struct tcphdr *, struct mbuf *,
1253 tcp_seq, tcp_seq, int, struct tcp_respond_args *);
1254 struct rtentry *
1255 tcp_rtlookup(struct inpcb *, unsigned int);
1256 void tcp_setpersist(struct tcpcb *);
1257 void tcp_gc(struct inpcbinfo *);
1258 void tcp_check_timer_state(struct tcpcb *tp);
1259 void tcp_run_timerlist(void *arg1, void *arg2);
1260
1261 struct tcptemp *tcp_maketemplate(struct tcpcb *);
1262 void tcp_fillheaders(struct tcpcb *, void *, void *);
1263 struct tcpcb *tcp_timers(struct tcpcb *, int);
1264 void tcp_trace(int, int, struct tcpcb *, void *, struct tcphdr *, int);
1265
1266 void tcp_sack_doack(struct tcpcb *, struct tcpopt *, struct tcphdr *,
1267 u_int32_t *);
1268 int tcp_detect_bad_rexmt(struct tcpcb *, struct tcphdr *, struct tcpopt *,
1269 u_int32_t rxtime);
1270 void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1271 void tcp_clean_sackreport(struct tcpcb *tp);
1272 void tcp_sack_adjust(struct tcpcb *tp);
1273 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1274 void tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1275 void tcp_free_sackholes(struct tcpcb *tp);
1276 int32_t tcp_sbspace(struct tcpcb *tp);
1277 void tcp_set_tso(struct tcpcb *tp, struct ifnet *ifp);
1278 void tcp_reset_stretch_ack(struct tcpcb *tp);
1279 extern void tcp_get_ports_used(u_int32_t, int, u_int32_t, bitstr_t *);
1280 uint32_t tcp_count_opportunistic(unsigned int ifindex, u_int32_t flags);
1281 uint32_t tcp_find_anypcb_byaddr(struct ifaddr *ifa);
1282 void tcp_set_max_rwinscale(struct tcpcb *tp, struct socket *so);
1283 struct bwmeas* tcp_bwmeas_alloc(struct tcpcb *tp);
1284 void tcp_bwmeas_free(struct tcpcb *tp);
1285 extern int32_t timer_diff(uint32_t t1, uint32_t toff1, uint32_t t2, uint32_t toff2);
1286
1287 extern void tcp_set_background_cc(struct socket *);
1288 extern void tcp_set_foreground_cc(struct socket *);
1289 extern void tcp_set_recv_bg(struct socket *);
1290 extern void tcp_clear_recv_bg(struct socket *);
1291 extern boolean_t tcp_sack_byte_islost(struct tcpcb *tp);
1292 #define IS_TCP_RECV_BG(_so) \
1293 ((_so)->so_traffic_mgt_flags & TRAFFIC_MGT_TCP_RECVBG)
1294
1295 #if TRAFFIC_MGT
1296 #define CLEAR_IAJ_STATE(_tp_) (_tp_)->iaj_rcv_ts = 0
1297 void reset_acc_iaj(struct tcpcb *tp);
1298 #endif /* TRAFFIC_MGT */
1299
1300 int tcp_lock (struct socket *, int, void *);
1301 int tcp_unlock (struct socket *, int, void *);
1302 void calculate_tcp_clock(void);
1303
1304 extern void tcp_keepalive_reset(struct tcpcb *);
1305 extern uint32_t get_base_rtt(struct tcpcb *tp);
1306
1307 #ifdef _KERN_LOCKS_H_
1308 lck_mtx_t * tcp_getlock (struct socket *, int);
1309 #else
1310 void * tcp_getlock (struct socket *, int);
1311 #endif
1312
1313
1314 extern struct pr_usrreqs tcp_usrreqs;
1315 extern u_int32_t tcp_sendspace;
1316 extern u_int32_t tcp_recvspace;
1317 tcp_seq tcp_new_isn(struct tcpcb *);
1318
1319 extern int tcp_input_checksum(int, struct mbuf *, struct tcphdr *, int, int);
1320 extern void tcp_getconninfo(struct socket *, struct conninfo_tcp *);
1321 extern void add_to_time_wait(struct tcpcb *, uint32_t delay);
1322 extern void tcp_pmtud_revert_segment_size(struct tcpcb *tp);
1323 #if MPTCP
1324 extern uint16_t mptcp_input_csum(struct tcpcb *, struct mbuf *, int);
1325 extern void mptcp_output_csum(struct tcpcb *, struct mbuf *, int32_t, unsigned,
1326 u_int64_t, u_int32_t *);
1327 extern int mptcp_adj_mss(struct tcpcb *, boolean_t);
1328 extern void mptcp_insert_rmap(struct tcpcb *, struct mbuf *);
1329 #endif
1330 #endif /* BSD_KERNEL_RPIVATE */
1331
1332 #endif /* _NETINET_TCP_VAR_H_ */
mirror of XNU