]> 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-2422.100.13.tar.gz
[apple/xnu.git] / bsd / netinet / tcp_var.h
1 /*
2 * Copyright (c) 2000-2013 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 #define TCP_RETRANSHZ 1000 /* granularity of TCP timestamps, 1ms */
83 #define TCP_TIMERHZ 100 /* frequency of TCP fast timer, 100 ms */
84
85 /* Minimum time quantum within which the timers are coalesced */
86 #define TCP_FASTTIMER_QUANTUM TCP_TIMERHZ /* fast mode, once every 100ms */
87 #define TCP_SLOWTIMER_QUANTUM (TCP_RETRANSHZ/2) /* slow mode, once every 500ms */
88
89 #define TCP_RETRANSHZ_TO_USEC 1000
90
91 #ifdef KERNEL_PRIVATE
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 extern int tcp_reass_maxseg;
163 extern int tcp_reass_qsize;
164 #ifdef MALLOC_DECLARE
165 MALLOC_DECLARE(M_TSEGQ);
166 #endif
167
168 struct sackblk {
169 tcp_seq start; /* start seq no. of sack block */
170 tcp_seq end; /* end seq no. */
171 };
172
173 struct sackhole {
174 tcp_seq start; /* start seq no. of hole */
175 tcp_seq end; /* end seq no. */
176 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */
177 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */
178 };
179
180 struct sackhint {
181 struct sackhole *nexthole;
182 int sack_bytes_rexmit;
183 };
184
185 struct tcptemp {
186 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
187 struct tcphdr tt_t;
188 };
189
190 struct bwmeas {
191 tcp_seq bw_start; /* start of bw measurement */
192 uint32_t bw_ts; /* timestamp when bw measurement started */
193 uint32_t bw_size; /* burst size in bytes for this bw measurement */
194 uint32_t bw_minsizepkts; /* Min burst size as segments */
195 uint32_t bw_maxsizepkts; /* Max burst size as segments */
196 uint32_t bw_minsize; /* Min size in bytes */
197 uint32_t bw_maxsize; /* Max size in bytes */
198 uint32_t bw_sndbw; /* Measured send bw */
199 };
200
201 /* MPTCP Data sequence map entry */
202 struct mpt_dsn_map {
203 uint64_t mpt_dsn; /* data seq num recvd */
204 uint32_t mpt_sseq; /* relative subflow # */
205 uint16_t mpt_len; /* length of mapping */
206 uint16_t mpt_csum; /* checksum value if on */
207 };
208 #define tcp6cb tcpcb /* for KAME src sync over BSD*'s */
209
210 /*
211 * Tcp control block, one per tcp; fields:
212 * Organized for 16 byte cacheline efficiency.
213 */
214 struct tcpcb {
215 struct tsegqe_head t_segq;
216 int t_dupacks; /* consecutive dup acks recd */
217 uint32_t t_timer[TCPT_NTIMERS]; /* tcp timers */
218 struct tcptimerentry tentry; /* entry in timer list */
219
220 struct inpcb *t_inpcb; /* back pointer to internet pcb */
221 int t_state; /* state of this connection */
222 uint32_t t_flags;
223 #define TF_ACKNOW 0x00001 /* ack peer immediately */
224 #define TF_DELACK 0x00002 /* ack, but try to delay it */
225 #define TF_NODELAY 0x00004 /* don't delay packets to coalesce */
226 #define TF_NOOPT 0x00008 /* don't use tcp options */
227 #define TF_SENTFIN 0x00010 /* have sent FIN */
228 #define TF_REQ_SCALE 0x00020 /* have/will request window scaling */
229 #define TF_RCVD_SCALE 0x00040 /* other side has requested scaling */
230 #define TF_REQ_TSTMP 0x00080 /* have/will request timestamps */
231 #define TF_RCVD_TSTMP 0x00100 /* a timestamp was received in SYN */
232 #define TF_SACK_PERMIT 0x00200 /* other side said I could SACK */
233 #define TF_NEEDSYN 0x00400 /* send SYN (implicit state) */
234 #define TF_NEEDFIN 0x00800 /* send FIN (implicit state) */
235 #define TF_NOPUSH 0x01000 /* don't push */
236 #define TF_REQ_CC 0x02000 /* have/will request CC */
237 #define TF_RCVD_CC 0x04000 /* a CC was received in SYN */
238 #define TF_SENDCCNEW 0x08000 /* send CCnew instead of CC in SYN */
239 #define TF_MORETOCOME 0x10000 /* More data to be appended to sock */
240 #define TF_LOCAL 0x20000 /* connection to a host on local link */
241 #define TF_RXWIN0SENT 0x40000 /* sent a receiver win 0 in response */
242 #define TF_SLOWLINK 0x80000 /* route is a on a modem speed link */
243 #define TF_LASTIDLE 0x100000 /* connection was previously idle */
244 #define TF_FASTRECOVERY 0x200000 /* in NewReno Fast Recovery */
245 #define TF_WASFRECOVERY 0x400000 /* was in NewReno Fast Recovery */
246 #define TF_SIGNATURE 0x800000 /* require MD5 digests (RFC2385) */
247 #define TF_MAXSEGSNT 0x1000000 /* last segment sent was a full segment */
248 #define TF_PMTUD 0x4000000 /* Perform Path MTU Discovery for this connection */
249 #define TF_CLOSING 0x8000000 /* pending tcp close */
250 #define TF_TSO 0x10000000 /* TCP Segment Offloading is enable on this connection */
251 #define TF_BLACKHOLE 0x20000000 /* Path MTU Discovery Black Hole detection */
252 #define TF_TIMER_ONLIST 0x40000000 /* pcb is on tcp_timer_list */
253 #define TF_STRETCHACK 0x80000000 /* receiver is going to delay acks */
254
255 int t_force; /* 1 if forcing out a byte */
256
257 tcp_seq snd_una; /* send unacknowledged */
258 tcp_seq snd_max; /* highest sequence number sent;
259 * used to recognize retransmits
260 */
261 tcp_seq snd_nxt; /* send next */
262 tcp_seq snd_up; /* send urgent pointer */
263
264 tcp_seq snd_wl1; /* window update seg seq number */
265 tcp_seq snd_wl2; /* window update seg ack number */
266 tcp_seq iss; /* initial send sequence number */
267 tcp_seq irs; /* initial receive sequence number */
268
269 tcp_seq rcv_nxt; /* receive next */
270 tcp_seq rcv_adv; /* advertised window */
271 u_int32_t rcv_wnd; /* receive window */
272 tcp_seq rcv_up; /* receive urgent pointer */
273
274 u_int32_t snd_wnd; /* send window */
275 u_int32_t snd_cwnd; /* congestion-controlled window */
276 u_int32_t snd_ssthresh; /* snd_cwnd size threshold for
277 * for slow start exponential to
278 * linear switch
279 */
280 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */
281
282 u_int t_maxopd; /* mss plus options */
283
284 u_int32_t t_rcvtime; /* time at which a packet was received */
285 u_int32_t t_starttime; /* time connection was established */
286 int t_rtttime; /* tcp clock when rtt calculation was started */
287 tcp_seq t_rtseq; /* sequence number being timed */
288
289 u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */
290 u_int32_t rfbuf_cnt; /* recv buffer autoscaling byte count */
291
292 int t_rxtcur; /* current retransmit value (ticks) */
293 u_int t_maxseg; /* maximum segment size */
294 int t_srtt; /* smoothed round-trip time */
295 int t_rttvar; /* variance in round-trip time */
296
297 int t_rxtshift; /* log(2) of rexmt exp. backoff */
298 u_int t_rttmin; /* minimum rtt allowed */
299 u_int t_rttbest; /* best rtt we've seen */
300 u_int t_rttcur; /* most recent value of rtt */
301 u_int32_t t_rttupdated; /* number of times rtt sampled */
302 u_int32_t t_rxt_conndroptime; /* retxmt conn gets dropped after this time, when set */
303 u_int32_t t_rxtstart; /* time at which retransmission started */
304 u_int32_t max_sndwnd; /* largest window peer has offered */
305
306 int t_softerror; /* possible error not yet reported */
307 /* out-of-band data */
308 char t_oobflags; /* have some */
309 char t_iobc; /* input character */
310 #define TCPOOB_HAVEDATA 0x01
311 #define TCPOOB_HADDATA 0x02
312 /* RFC 1323 variables */
313 u_int8_t snd_scale; /* window scaling for send window */
314 u_int8_t rcv_scale; /* window scaling for recv window */
315 u_int8_t request_r_scale; /* pending window scaling */
316 u_int8_t requested_s_scale;
317 u_int8_t tcp_cc_index; /* index of congestion control algorithm */
318 u_int8_t t_adaptive_rtimo; /* Read timeout used as a multiple of RTT */
319 u_int8_t t_adaptive_wtimo; /* Write timeout used as a multiple of RTT */
320 u_int32_t ts_recent; /* timestamp echo data */
321
322 u_int32_t ts_recent_age; /* when last updated */
323 tcp_seq last_ack_sent;
324 /* RFC 1644 variables */
325 tcp_cc cc_send; /* send connection count */
326 tcp_cc cc_recv; /* receive connection count */
327 /* RFC 3465 variables */
328 u_int32_t t_bytes_acked; /* ABC "bytes_acked" parameter */
329
330 int t_lastchain; /* amount of packets chained last time around */
331 u_int16_t t_unacksegs; /* received but unacked segments for delaying acks */
332 u_int8_t t_rexmtthresh; /* duplicate ack threshold for entering fast recovery */
333 u_int8_t t_rtimo_probes; /* number of adaptive rtimo probes sent */
334 u_int32_t t_persist_timeout; /* ZWP persistence limit as set by PERSIST_TIMEOUT */
335 u_int32_t t_persist_stop; /* persistence limit deadline if triggered by ZWP */
336 u_int32_t t_notsent_lowat; /* Low water for not sent data */
337
338 /* Receiver state for stretch-ack algorithm */
339 u_int32_t rcv_unackwin; /* to measure win for stretching acks */
340 u_int32_t rcv_by_unackwin; /* bytes seen during the last ack-stretching win */
341 u_int16_t rcv_waitforss; /* wait for packets during slow-start */
342 u_int16_t ecn_flags;
343 #define TE_SETUPSENT 0x01 /* Indicate we have sent ECN-SETUP SYN or SYN-ACK */
344 #define TE_SETUPRECEIVED 0x02 /* Indicate we have received ECN-SETUP SYN or SYN-ACK */
345 #define TE_SENDIPECT 0x04 /* Indicate we haven't sent or received non-ECN-setup SYN or SYN-ACK */
346 #define TE_SENDCWR 0x08 /* Indicate that the next non-retransmit should have the TCP CWR flag set */
347 #define TE_SENDECE 0x10 /* Indicate that the next packet should have the TCP ECE flag set */
348 #define TE_ECN_ON (TE_SETUPSENT | TE_SETUPRECEIVED) /* Indicate ECN was successfully negotiated on a connection) */
349
350 /* state for bad retransmit recovery */
351 u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */
352 u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
353 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */
354 int t_srtt_prev; /* srtt prior to retransmit */
355 int t_rttvar_prev; /* rttvar prior to retransmit */
356 u_int32_t t_badrexmt_time; /* bad rexmt detection time */
357
358 /* state to limit the number of early retransmits */
359 u_int32_t t_early_rexmt_win; /* window for limiting early retransmits */
360 u_int16_t t_early_rexmt_count; /* number of early rexmts seen in past window */
361
362 /* SACK related state */
363 int16_t snd_numholes; /* number of holes seen by sender */
364 TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
365 /* SACK scoreboard (sorted) */
366 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/
367 int rcv_numsacks; /* # distinct sack blks present */
368 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
369 tcp_seq sack_newdata; /* New data xmitted in this recovery
370 episode starts at this seq number */
371 struct sackhint sackhint; /* SACK scoreboard hint */
372
373 u_int32_t t_pktlist_sentlen; /* total bytes in transmit chain */
374 struct mbuf *t_pktlist_head; /* First packet in transmit chain */
375 struct mbuf *t_pktlist_tail; /* Last packet in transmit chain */
376
377 u_int32_t t_keepidle; /* keepalive idle timer (override global if > 0) */
378 u_int32_t t_keepinit; /* connection timeout, i.e. idle time
379 in SYN_SENT or SYN_RECV state */
380 u_int32_t t_keepintvl; /* interval between keepalives */
381 u_int32_t t_keepcnt; /* number of keepalives before close */
382
383 u_int32_t tso_max_segment_size; /* TCP Segment Offloading maximum segment unit for NIC */
384 u_int t_pmtud_saved_maxopd; /* MSS saved before performing PMTU-D BlackHole detection */
385
386 struct
387 {
388 u_int32_t rxduplicatebytes;
389 u_int32_t rxoutoforderbytes;
390 u_int32_t txretransmitbytes;
391 u_int8_t synrxtshift;
392 u_int8_t unused;
393 u_int16_t unused_pad_to_8;
394 } t_stat;
395
396 /* Background congestion related state */
397 uint32_t rtt_hist[N_RTT_BASE]; /* history of minimum RTT */
398 uint32_t rtt_count; /* Number of RTT samples in recent base history */
399 uint32_t bg_ssthresh; /* Slow start threshold until delay increases */
400 uint32_t t_flagsext; /* Another field to accommodate more flags */
401 #define TF_RXTFINDROP 0x1 /* Drop conn after retransmitting FIN 3 times */
402 #define TF_RCVUNACK_WAITSS 0x2 /* set when the receiver should not stretch acks */
403 #define TF_BWMEAS_INPROGRESS 0x4 /* Indicate BW meas is happening */
404 #define TF_MEASURESNDBW 0x8 /* Measure send bw on this connection */
405 #define TF_LRO_OFFLOADED 0x10 /* Connection LRO offloaded */
406 #define TF_SACK_ENABLE 0x20 /* SACK is enabled */
407 #define TF_RECOMPUTE_RTT 0x40 /* recompute RTT after spurious retransmit */
408 #define TF_DETECT_READSTALL 0x80 /* Used to detect a stall during read operation */
409 #define TF_RECV_THROTTLE 0x100 /* Input throttling active */
410 #define TF_NOSTRETCHACK 0x200 /* ack every other packet */
411 #define TF_STREAMEOW 0x400 /* Last packet was small indicating end of write */
412 #if TRAFFIC_MGT
413 /* Inter-arrival jitter related state */
414 uint32_t iaj_rcv_ts; /* tcp clock when the first packet was received */
415 uint16_t iaj_size; /* Size of packet for iaj measurement */
416 uint16_t iaj_small_pkt; /* Count of packets smaller than iaj_size */
417 uint16_t iaj_pktcnt; /* packet count, to avoid throttling initially */
418 uint16_t acc_iaj; /* Accumulated iaj */
419 tcp_seq iaj_rwintop; /* recent max advertised window */
420 uint32_t avg_iaj; /* Mean */
421 uint32_t std_dev_iaj; /* Standard deviation */
422 #endif /* TRAFFIC_MGT */
423 struct bwmeas *t_bwmeas; /* State for bandwidth measurement */
424 uint32_t t_lropktlen; /* Bytes in a LRO frame */
425 tcp_seq t_idleat; /* rcv_nxt at idle time */
426 TAILQ_ENTRY(tcpcb) t_twentry; /* link for time wait queue */
427 #if MPTCP
428 u_int32_t t_mpflags; /* flags for multipath TCP */
429
430 #define TMPF_PREESTABLISHED 0x00000001 /* conn in pre-established state */
431 #define TMPF_SENT_KEYS 0x00000002 /* indicates that keys were sent */
432 #define TMPF_MPTCP_TRUE 0x00000004 /* negotiated MPTCP successfully */
433 #define TMPF_MPTCP_RCVD_KEY 0x00000008 /* state for 3-way handshake */
434 #define TMPF_SND_MPPRIO 0x00000010 /* send priority of subflow */
435 #define TMPF_SND_REM_ADDR 0x00000020 /* initiate address removal */
436 #define TMPF_UNUSED 0x00000040 /* address addition acked by peer */
437 #define TMPF_JOINED_FLOW 0x00000080 /* Indicates additional flow */
438 #define TMPF_BACKUP_PATH 0x00000100 /* Indicates backup path */
439 #define TMPF_MPTCP_ACKNOW 0x00000200 /* Send Data ACK */
440 #define TMPF_SEND_DSN 0x00000400 /* Send DSN mapping */
441 #define TMPF_SEND_DFIN 0x00000800 /* Send Data FIN */
442 #define TMPF_RECV_DFIN 0x00001000 /* Recv Data FIN */
443 #define TMPF_SENT_JOIN 0x00002000 /* Sent Join */
444 #define TMPF_RECVD_JOIN 0x00004000 /* Received Join */
445 #define TMPF_RESET 0x00008000 /* Send RST */
446 #define TMPF_TCP_FALLBACK 0x00010000 /* Fallback to TCP */
447 #define TMPF_FASTCLOSE 0x00020000 /* Send Fastclose option */
448 #define TMPF_EMBED_DSN 0x00040000 /* tp has DSN mapping */
449 #define TMPF_MPTCP_READY 0x00080000 /* Can send DSS options on data */
450 #define TMPF_INFIN_SENT 0x00100000 /* Sent infinite mapping */
451 #define TMPF_SND_MPFAIL 0x00200000 /* Received mapping csum failure */
452 void *t_mptcb; /* pointer to MPTCP TCB */
453 tcp_seq t_mpuna; /* unacknowledged sequence */
454 struct mpt_dsn_map t_rcv_map; /* Receive mapping list */
455 u_int8_t t_local_aid; /* Addr Id for authentication */
456 u_int8_t t_rem_aid; /* Addr ID of another subflow */
457 u_int8_t t_mprxtshift; /* join retransmission */
458 #endif /* MPTCP */
459 };
460
461 #define IN_FASTRECOVERY(tp) (tp->t_flags & TF_FASTRECOVERY)
462 #define SACK_ENABLED(tp) (tp->t_flagsext & TF_SACK_ENABLE)
463
464 /*
465 * If the connection is in a throttled state due to advisory feedback from
466 * the interface output queue, reset that state. We do this in favor
467 * of entering recovery because the data transfer during recovery
468 * should be just a trickle and it will help to improve performance.
469 * We also do not want to back off twice in the same RTT.
470 */
471 #define ENTER_FASTRECOVERY(_tp_) do { \
472 (_tp_)->t_flags |= TF_FASTRECOVERY; \
473 if (INP_IS_FLOW_CONTROLLED((_tp_)->t_inpcb)) \
474 inp_reset_fc_state((_tp_)->t_inpcb); \
475 } while(0)
476
477 #define EXIT_FASTRECOVERY(_tp_) do { \
478 (_tp_)->t_flags &= ~TF_FASTRECOVERY; \
479 (_tp_)->t_dupacks = 0; \
480 (_tp_)->t_rexmtthresh = tcprexmtthresh; \
481 (_tp_)->t_bytes_acked = 0; \
482 } while(0)
483
484 /*
485 * When the number of duplicate acks received is less than
486 * the retransmit threshold, use Limited Transmit algorithm
487 */
488 extern int tcprexmtthresh;
489 #define ALLOW_LIMITED_TRANSMIT(_tp_) \
490 ((_tp_)->t_dupacks > 0 && \
491 (_tp_)->t_dupacks < (_tp_)->t_rexmtthresh)
492
493 /*
494 * This condition is true is timestamp option is supported
495 * on a connection.
496 */
497 #define TSTMP_SUPPORTED(_tp_) \
498 (((_tp_)->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP)) == \
499 (TF_REQ_TSTMP|TF_RCVD_TSTMP))
500
501 /*
502 * Gives number of bytes acked by this ack
503 */
504 #define BYTES_ACKED(_th_, _tp_) \
505 ((_th_)->th_ack - (_tp_)->snd_una)
506
507 #if CONFIG_DTRACE
508 enum tcp_cc_event {
509 TCP_CC_CWND_INIT,
510 TCP_CC_INSEQ_ACK_RCVD,
511 TCP_CC_ACK_RCVD,
512 TCP_CC_ENTER_FASTRECOVERY,
513 TCP_CC_IN_FASTRECOVERY,
514 TCP_CC_EXIT_FASTRECOVERY,
515 TCP_CC_PARTIAL_ACK,
516 TCP_CC_IDLE_TIMEOUT,
517 TCP_CC_REXMT_TIMEOUT,
518 TCP_CC_ECN_RCVD,
519 TCP_CC_BAD_REXMT_RECOVERY,
520 TCP_CC_OUTPUT_ERROR,
521 TCP_CC_CHANGE_ALGO,
522 TCP_CC_FLOW_CONTROL,
523 TCP_CC_SUSPEND,
524 TCP_CC_LIMITED_TRANSMIT,
525 TCP_CC_EARLY_RETRANSMIT
526 };
527 #endif /* CONFIG_DTRACE */
528
529 /*
530 * Structure to hold TCP options that are only used during segment
531 * processing (in tcp_input), but not held in the tcpcb.
532 * It's basically used to reduce the number of parameters
533 * to tcp_dooptions.
534 */
535 struct tcpopt {
536 u_int32_t to_flags; /* which options are present */
537 #define TOF_TS 0x0001 /* timestamp */
538 #define TOF_MSS 0x0010
539 #define TOF_SCALE 0x0020
540 #define TOF_SIGNATURE 0x0040 /* signature option present */
541 #define TOF_SIGLEN 0x0080 /* signature length valid (RFC2385) */
542 #define TOF_SACK 0x0100 /* Peer sent SACK option */
543 #define TOF_MPTCP 0x0200 /* MPTCP options to be dropped */
544 u_int32_t to_tsval;
545 u_int32_t to_tsecr;
546 u_int16_t to_mss;
547 u_int8_t to_requested_s_scale;
548 u_int8_t to_nsacks; /* number of SACK blocks */
549 u_char *to_sacks; /* pointer to the first SACK blocks */
550 };
551
552 /*
553 * The TAO cache entry which is stored in the protocol family specific
554 * portion of the route metrics.
555 */
556 struct rmxp_tao {
557 tcp_cc tao_cc; /* latest CC in valid SYN */
558 tcp_cc tao_ccsent; /* latest CC sent to peer */
559 u_short tao_mssopt; /* peer's cached MSS */
560 #ifdef notyet
561 u_short tao_flags; /* cache status flags */
562 #define TAOF_DONT 0x0001 /* peer doesn't understand rfc1644 */
563 #define TAOF_OK 0x0002 /* peer does understand rfc1644 */
564 #define TAOF_UNDEF 0 /* we don't know yet */
565 #endif /* notyet */
566 };
567 #define rmx_taop(r) ((struct rmxp_tao *)(r).rmx_filler)
568
569 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb)
570 #define sototcpcb(so) (intotcpcb(sotoinpcb(so)))
571
572 /*
573 * The rtt measured is in milliseconds as the timestamp granularity is
574 * a millisecond. The smoothed round-trip time and estimated variance
575 * are stored as fixed point numbers scaled by the values below.
576 * For convenience, these scales are also used in smoothing the average
577 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
578 * With these scales, srtt has 5 bits to the right of the binary point,
579 * and thus an "ALPHA" of 0.875. rttvar has 4 bits to the right of the
580 * binary point, and is smoothed with an ALPHA of 0.75.
581 */
582 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */
583 #define TCP_RTT_SHIFT 5 /* shift for srtt; 5 bits frac. */
584 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 4 bits */
585 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 4 bits */
586 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
587
588 /*
589 * The initial retransmission should happen at rtt + 4 * rttvar.
590 * Because of the way we do the smoothing, srtt and rttvar
591 * will each average +1/2 tick of bias. When we compute
592 * the retransmit timer, we want 1/2 tick of rounding and
593 * 1 extra tick because of +-1/2 tick uncertainty in the
594 * firing of the timer. The bias will give us exactly the
595 * 1.5 tick we need. But, because the bias is
596 * statistical, we have to test that we don't drop below
597 * the minimum feasible timer (which is 2 ticks).
598 * This version of the macro adapted from a paper by Lawrence
599 * Brakmo and Larry Peterson which outlines a problem caused
600 * by insufficient precision in the original implementation,
601 * which results in inappropriately large RTO values for very
602 * fast networks.
603 */
604 #define TCP_REXMTVAL(tp) \
605 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
606 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
607
608 /*
609 * Jaguar compatible TCP control block, for xtcpcb
610 * Does not have the old fields
611 */
612 struct otcpcb {
613 #else
614 struct tseg_qent;
615 _TCPCB_LIST_HEAD(tsegqe_head, tseg_qent);
616
617 struct tcpcb {
618 #endif /* KERNEL_PRIVATE */
619 #if defined(KERNEL_PRIVATE)
620 u_int32_t t_segq;
621 #else
622 struct tsegqe_head t_segq;
623 #endif /* KERNEL_PRIVATE */
624 int t_dupacks; /* consecutive dup acks recd */
625 u_int32_t unused; /* unused now: was t_template */
626
627 int t_timer[TCPT_NTIMERS_EXT]; /* tcp timers */
628
629 _TCPCB_PTR(struct inpcb *) t_inpcb; /* back pointer to internet pcb */
630 int t_state; /* state of this connection */
631 u_int t_flags;
632 #define TF_ACKNOW 0x00001 /* ack peer immediately */
633 #define TF_DELACK 0x00002 /* ack, but try to delay it */
634 #define TF_NODELAY 0x00004 /* don't delay packets to coalesce */
635 #define TF_NOOPT 0x00008 /* don't use tcp options */
636 #define TF_SENTFIN 0x00010 /* have sent FIN */
637 #define TF_REQ_SCALE 0x00020 /* have/will request window scaling */
638 #define TF_RCVD_SCALE 0x00040 /* other side has requested scaling */
639 #define TF_REQ_TSTMP 0x00080 /* have/will request timestamps */
640 #define TF_RCVD_TSTMP 0x00100 /* a timestamp was received in SYN */
641 #define TF_SACK_PERMIT 0x00200 /* other side said I could SACK */
642 #define TF_NEEDSYN 0x00400 /* send SYN (implicit state) */
643 #define TF_NEEDFIN 0x00800 /* send FIN (implicit state) */
644 #define TF_NOPUSH 0x01000 /* don't push */
645 #define TF_REQ_CC 0x02000 /* have/will request CC */
646 #define TF_RCVD_CC 0x04000 /* a CC was received in SYN */
647 #define TF_SENDCCNEW 0x08000 /* send CCnew instead of CC in SYN */
648 #define TF_MORETOCOME 0x10000 /* More data to be appended to sock */
649 #define TF_LQ_OVERFLOW 0x20000 /* listen queue overflow */
650 #define TF_RXWIN0SENT 0x40000 /* sent a receiver win 0 in response */
651 #define TF_SLOWLINK 0x80000 /* route is a on a modem speed link */
652
653 int t_force; /* 1 if forcing out a byte */
654
655 tcp_seq snd_una; /* send unacknowledged */
656 tcp_seq snd_max; /* highest sequence number sent;
657 * used to recognize retransmits
658 */
659 tcp_seq snd_nxt; /* send next */
660 tcp_seq snd_up; /* send urgent pointer */
661
662 tcp_seq snd_wl1; /* window update seg seq number */
663 tcp_seq snd_wl2; /* window update seg ack number */
664 tcp_seq iss; /* initial send sequence number */
665 tcp_seq irs; /* initial receive sequence number */
666
667 tcp_seq rcv_nxt; /* receive next */
668 tcp_seq rcv_adv; /* advertised window */
669 u_int32_t rcv_wnd; /* receive window */
670 tcp_seq rcv_up; /* receive urgent pointer */
671
672 u_int32_t snd_wnd; /* send window */
673 u_int32_t snd_cwnd; /* congestion-controlled window */
674 u_int32_t snd_ssthresh; /* snd_cwnd size threshold for
675 * for slow start exponential to
676 * linear switch
677 */
678 u_int t_maxopd; /* mss plus options */
679
680 u_int32_t t_rcvtime; /* time at which a packet was received */
681 u_int32_t t_starttime; /* time connection was established */
682 int t_rtttime; /* round trip time */
683 tcp_seq t_rtseq; /* sequence number being timed */
684
685 int t_rxtcur; /* current retransmit value (ticks) */
686 u_int t_maxseg; /* maximum segment size */
687 int t_srtt; /* smoothed round-trip time */
688 int t_rttvar; /* variance in round-trip time */
689
690 int t_rxtshift; /* log(2) of rexmt exp. backoff */
691 u_int t_rttmin; /* minimum rtt allowed */
692 u_int32_t t_rttupdated; /* number of times rtt sampled */
693 u_int32_t max_sndwnd; /* largest window peer has offered */
694
695 int t_softerror; /* possible error not yet reported */
696 /* out-of-band data */
697 char t_oobflags; /* have some */
698 char t_iobc; /* input character */
699 #define TCPOOB_HAVEDATA 0x01
700 #define TCPOOB_HADDATA 0x02
701 /* RFC 1323 variables */
702 u_char snd_scale; /* window scaling for send window */
703 u_char rcv_scale; /* window scaling for recv window */
704 u_char request_r_scale; /* pending window scaling */
705 u_char requested_s_scale;
706 u_int32_t ts_recent; /* timestamp echo data */
707
708 u_int32_t ts_recent_age; /* when last updated */
709 tcp_seq last_ack_sent;
710 /* RFC 1644 variables */
711 tcp_cc cc_send; /* send connection count */
712 tcp_cc cc_recv; /* receive connection count */
713 tcp_seq snd_recover; /* for use in fast recovery */
714 /* experimental */
715 u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */
716 u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
717 u_int32_t t_badrxtwin; /* window for retransmit recovery */
718 };
719
720
721 /*
722 * TCP statistics.
723 * Many of these should be kept per connection,
724 * but that's inconvenient at the moment.
725 */
726 struct tcpstat {
727 u_int32_t tcps_connattempt; /* connections initiated */
728 u_int32_t tcps_accepts; /* connections accepted */
729 u_int32_t tcps_connects; /* connections established */
730 u_int32_t tcps_drops; /* connections dropped */
731 u_int32_t tcps_conndrops; /* embryonic connections dropped */
732 u_int32_t tcps_closed; /* conn. closed (includes drops) */
733 u_int32_t tcps_segstimed; /* segs where we tried to get rtt */
734 u_int32_t tcps_rttupdated; /* times we succeeded */
735 u_int32_t tcps_delack; /* delayed acks sent */
736 u_int32_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
737 u_int32_t tcps_rexmttimeo; /* retransmit timeouts */
738 u_int32_t tcps_persisttimeo; /* persist timeouts */
739 u_int32_t tcps_keeptimeo; /* keepalive timeouts */
740 u_int32_t tcps_keepprobe; /* keepalive probes sent */
741 u_int32_t tcps_keepdrops; /* connections dropped in keepalive */
742
743 u_int32_t tcps_sndtotal; /* total packets sent */
744 u_int32_t tcps_sndpack; /* data packets sent */
745 u_int32_t tcps_sndbyte; /* data bytes sent */
746 u_int32_t tcps_sndrexmitpack; /* data packets retransmitted */
747 u_int32_t tcps_sndrexmitbyte; /* data bytes retransmitted */
748 u_int32_t tcps_sndacks; /* ack-only packets sent */
749 u_int32_t tcps_sndprobe; /* window probes sent */
750 u_int32_t tcps_sndurg; /* packets sent with URG only */
751 u_int32_t tcps_sndwinup; /* window update-only packets sent */
752 u_int32_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
753
754 u_int32_t tcps_rcvtotal; /* total packets received */
755 u_int32_t tcps_rcvpack; /* packets received in sequence */
756 u_int32_t tcps_rcvbyte; /* bytes received in sequence */
757 u_int32_t tcps_rcvbadsum; /* packets received with ccksum errs */
758 u_int32_t tcps_rcvbadoff; /* packets received with bad offset */
759 u_int32_t tcps_rcvmemdrop; /* packets dropped for lack of memory */
760 u_int32_t tcps_rcvshort; /* packets received too short */
761 u_int32_t tcps_rcvduppack; /* duplicate-only packets received */
762 u_int32_t tcps_rcvdupbyte; /* duplicate-only bytes received */
763 u_int32_t tcps_rcvpartduppack; /* packets with some duplicate data */
764 u_int32_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
765 u_int32_t tcps_rcvoopack; /* out-of-order packets received */
766 u_int32_t tcps_rcvoobyte; /* out-of-order bytes received */
767 u_int32_t tcps_rcvpackafterwin; /* packets with data after window */
768 u_int32_t tcps_rcvbyteafterwin; /* bytes rcvd after window */
769 u_int32_t tcps_rcvafterclose; /* packets rcvd after "close" */
770 u_int32_t tcps_rcvwinprobe; /* rcvd window probe packets */
771 u_int32_t tcps_rcvdupack; /* rcvd duplicate acks */
772 u_int32_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */
773 u_int32_t tcps_rcvackpack; /* rcvd ack packets */
774 u_int32_t tcps_rcvackbyte; /* bytes acked by rcvd acks */
775 u_int32_t tcps_rcvwinupd; /* rcvd window update packets */
776 u_int32_t tcps_pawsdrop; /* segments dropped due to PAWS */
777 u_int32_t tcps_predack; /* times hdr predict ok for acks */
778 u_int32_t tcps_preddat; /* times hdr predict ok for data pkts */
779 u_int32_t tcps_pcbcachemiss;
780 u_int32_t tcps_cachedrtt; /* times cached RTT in route updated */
781 u_int32_t tcps_cachedrttvar; /* times cached rttvar updated */
782 u_int32_t tcps_cachedssthresh; /* times cached ssthresh updated */
783 u_int32_t tcps_usedrtt; /* times RTT initialized from route */
784 u_int32_t tcps_usedrttvar; /* times RTTVAR initialized from rt */
785 u_int32_t tcps_usedssthresh; /* times ssthresh initialized from rt*/
786 u_int32_t tcps_persistdrop; /* timeout in persist state */
787 u_int32_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */
788 u_int32_t tcps_mturesent; /* resends due to MTU discovery */
789 u_int32_t tcps_listendrop; /* listen queue overflows */
790
791 /* new stats from FreeBSD 5.4 sync up */
792 u_int32_t tcps_minmssdrops; /* average minmss too low drops */
793 u_int32_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */
794 u_int32_t tcps_badrst; /* ignored RSTs in the window */
795
796 u_int32_t tcps_sc_added; /* entry added to syncache */
797 u_int32_t tcps_sc_retransmitted; /* syncache entry was retransmitted */
798 u_int32_t tcps_sc_dupsyn; /* duplicate SYN packet */
799 u_int32_t tcps_sc_dropped; /* could not reply to packet */
800 u_int32_t tcps_sc_completed; /* successful extraction of entry */
801 u_int32_t tcps_sc_bucketoverflow; /* syncache per-bucket limit hit */
802 u_int32_t tcps_sc_cacheoverflow; /* syncache cache limit hit */
803 u_int32_t tcps_sc_reset; /* RST removed entry from syncache */
804 u_int32_t tcps_sc_stale; /* timed out or listen socket gone */
805 u_int32_t tcps_sc_aborted; /* syncache entry aborted */
806 u_int32_t tcps_sc_badack; /* removed due to bad ACK */
807 u_int32_t tcps_sc_unreach; /* ICMP unreachable received */
808 u_int32_t tcps_sc_zonefail; /* zalloc() failed */
809 u_int32_t tcps_sc_sendcookie; /* SYN cookie sent */
810 u_int32_t tcps_sc_recvcookie; /* SYN cookie received */
811
812 u_int32_t tcps_hc_added; /* entry added to hostcache */
813 u_int32_t tcps_hc_bucketoverflow; /* hostcache per bucket limit hit */
814
815 /* SACK related stats */
816 u_int32_t tcps_sack_recovery_episode; /* SACK recovery episodes */
817 u_int32_t tcps_sack_rexmits; /* SACK rexmit segments */
818 u_int32_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */
819 u_int32_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */
820 u_int32_t tcps_sack_send_blocks; /* SACK blocks (options) sent */
821 u_int32_t tcps_sack_sboverflow; /* SACK sendblock overflow */
822
823 u_int32_t tcps_bg_rcvtotal; /* total background packets received */
824 u_int32_t tcps_rxtfindrop; /* drop conn after retransmitting FIN */
825 u_int32_t tcps_fcholdpacket; /* packets withheld because of flow control */
826
827 /* LRO related stats */
828 u_int32_t tcps_coalesced_pack; /* number of coalesced packets */
829 u_int32_t tcps_flowtbl_full; /* times flow table was full */
830 u_int32_t tcps_flowtbl_collision; /* collisions in flow tbl */
831 u_int32_t tcps_lro_twopack; /* 2 packets coalesced */
832 u_int32_t tcps_lro_multpack; /* 3 or 4 pkts coalesced */
833 u_int32_t tcps_lro_largepack; /* 5 or more pkts coalesced */
834
835 u_int32_t tcps_limited_txt; /* Limited transmit used */
836 u_int32_t tcps_early_rexmt; /* Early retransmit used */
837 u_int32_t tcps_sack_ackadv; /* Cumulative ack advanced along with sack */
838
839 /* Checksum related stats */
840 u_int32_t tcps_rcv_swcsum; /* tcp swcksum (inbound), packets */
841 u_int32_t tcps_rcv_swcsum_bytes; /* tcp swcksum (inbound), bytes */
842 u_int32_t tcps_rcv6_swcsum; /* tcp6 swcksum (inbound), packets */
843 u_int32_t tcps_rcv6_swcsum_bytes; /* tcp6 swcksum (inbound), bytes */
844 u_int32_t tcps_snd_swcsum; /* tcp swcksum (outbound), packets */
845 u_int32_t tcps_snd_swcsum_bytes; /* tcp swcksum (outbound), bytes */
846 u_int32_t tcps_snd6_swcsum; /* tcp6 swcksum (outbound), packets */
847 u_int32_t tcps_snd6_swcsum_bytes; /* tcp6 swcksum (outbound), bytes */
848 u_int32_t tcps_msg_unopkts; /* unordered packet on TCP msg stream */
849 u_int32_t tcps_msg_unoappendfail; /* failed to append unordered pkt */
850 u_int32_t tcps_msg_sndwaithipri; /* send is waiting for high priority data */
851
852 /* MPTCP Related stats */
853 u_int32_t tcps_invalid_mpcap; /* Invalid MPTCP capable opts */
854 u_int32_t tcps_invalid_joins; /* Invalid MPTCP joins */
855 u_int32_t tcps_mpcap_fallback; /* TCP fallback in primary */
856 u_int32_t tcps_join_fallback; /* No MPTCP in secondary */
857 u_int32_t tcps_estab_fallback; /* DSS option dropped */
858 u_int32_t tcps_invalid_opt; /* Catchall error stat */
859 u_int32_t tcps_mp_outofwin; /* Packet lies outside the
860 shared recv window */
861 u_int32_t tcps_mp_reducedwin; /* Reduced subflow window */
862 u_int32_t tcps_mp_badcsum; /* Bad DSS csum */
863 u_int32_t tcps_mp_oodata; /* Out of order data */
864 u_int32_t tcps_mp_switches; /* number of subflow switch */
865 u_int32_t tcps_mp_rcvtotal; /* number of rcvd packets */
866 u_int32_t tcps_mp_rcvbytes; /* number of bytes received */
867 u_int32_t tcps_mp_sndpacks; /* number of data packs sent */
868 u_int32_t tcps_mp_sndbytes; /* number of bytes sent */
869 u_int32_t tcps_join_rxmts; /* join ack retransmits */
870 };
871
872 struct tcpstat_local {
873 u_int64_t badformat;
874 u_int64_t unspecv6;
875 u_int64_t synfin;
876 u_int64_t badformatipsec;
877 u_int64_t noconnnolist;
878 u_int64_t noconnlist;
879 u_int64_t listbadsyn;
880 u_int64_t icmp6unreach;
881 u_int64_t deprecate6;
882 u_int64_t ooopacket;
883 u_int64_t rstinsynrcv;
884 u_int64_t dospacket;
885 u_int64_t cleanup;
886 u_int64_t synwindow;
887 };
888
889 #pragma pack(4)
890
891 /*
892 * TCB structure exported to user-land via sysctl(3).
893 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
894 * included. Not all of our clients do.
895 */
896
897 struct xtcpcb {
898 u_int32_t xt_len;
899 #ifdef KERNEL_PRIVATE
900 struct inpcb_compat xt_inp;
901 #else
902 struct inpcb xt_inp;
903 #endif
904 #ifdef KERNEL_PRIVATE
905 struct otcpcb xt_tp;
906 #else
907 struct tcpcb xt_tp;
908 #endif
909 struct xsocket xt_socket;
910 u_quad_t xt_alignment_hack;
911 };
912
913
914 struct xtcpcb64 {
915 u_int32_t xt_len;
916 struct xinpcb64 xt_inpcb;
917
918 u_int64_t t_segq;
919 int t_dupacks; /* consecutive dup acks recd */
920
921 int t_timer[TCPT_NTIMERS_EXT]; /* tcp timers */
922
923 int t_state; /* state of this connection */
924 u_int t_flags;
925
926 int t_force; /* 1 if forcing out a byte */
927
928 tcp_seq snd_una; /* send unacknowledged */
929 tcp_seq snd_max; /* highest sequence number sent;
930 * used to recognize retransmits
931 */
932 tcp_seq snd_nxt; /* send next */
933 tcp_seq snd_up; /* send urgent pointer */
934
935 tcp_seq snd_wl1; /* window update seg seq number */
936 tcp_seq snd_wl2; /* window update seg ack number */
937 tcp_seq iss; /* initial send sequence number */
938 tcp_seq irs; /* initial receive sequence number */
939
940 tcp_seq rcv_nxt; /* receive next */
941 tcp_seq rcv_adv; /* advertised window */
942 u_int32_t rcv_wnd; /* receive window */
943 tcp_seq rcv_up; /* receive urgent pointer */
944
945 u_int32_t snd_wnd; /* send window */
946 u_int32_t snd_cwnd; /* congestion-controlled window */
947 u_int32_t snd_ssthresh; /* snd_cwnd size threshold for
948 * for slow start exponential to
949 * linear switch
950 */
951 u_int t_maxopd; /* mss plus options */
952
953 u_int32_t t_rcvtime; /* time at which a packet was received */
954 u_int32_t t_starttime; /* time connection was established */
955 int t_rtttime; /* round trip time */
956 tcp_seq t_rtseq; /* sequence number being timed */
957
958 int t_rxtcur; /* current retransmit value (ticks) */
959 u_int t_maxseg; /* maximum segment size */
960 int t_srtt; /* smoothed round-trip time */
961 int t_rttvar; /* variance in round-trip time */
962
963 int t_rxtshift; /* log(2) of rexmt exp. backoff */
964 u_int t_rttmin; /* minimum rtt allowed */
965 u_int32_t t_rttupdated; /* number of times rtt sampled */
966 u_int32_t max_sndwnd; /* largest window peer has offered */
967
968 int t_softerror; /* possible error not yet reported */
969 /* out-of-band data */
970 char t_oobflags; /* have some */
971 char t_iobc; /* input character */
972 /* RFC 1323 variables */
973 u_char snd_scale; /* window scaling for send window */
974 u_char rcv_scale; /* window scaling for recv window */
975 u_char request_r_scale; /* pending window scaling */
976 u_char requested_s_scale;
977 u_int32_t ts_recent; /* timestamp echo data */
978
979 u_int32_t ts_recent_age; /* when last updated */
980 tcp_seq last_ack_sent;
981 /* RFC 1644 variables */
982 tcp_cc cc_send; /* send connection count */
983 tcp_cc cc_recv; /* receive connection count */
984 tcp_seq snd_recover; /* for use in fast recovery */
985 /* experimental */
986 u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */
987 u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
988 u_int32_t t_badrxtwin; /* window for retransmit recovery */
989
990 u_quad_t xt_alignment_hack;
991 };
992
993
994 #ifdef PRIVATE
995
996 struct xtcpcb_n {
997 u_int32_t xt_len;
998 u_int32_t xt_kind; /* XSO_TCPCB */
999
1000 u_int64_t t_segq;
1001 int t_dupacks; /* consecutive dup acks recd */
1002
1003 int t_timer[TCPT_NTIMERS_EXT]; /* tcp timers */
1004
1005 int t_state; /* state of this connection */
1006 u_int t_flags;
1007
1008 int t_force; /* 1 if forcing out a byte */
1009
1010 tcp_seq snd_una; /* send unacknowledged */
1011 tcp_seq snd_max; /* highest sequence number sent;
1012 * used to recognize retransmits
1013 */
1014 tcp_seq snd_nxt; /* send next */
1015 tcp_seq snd_up; /* send urgent pointer */
1016
1017 tcp_seq snd_wl1; /* window update seg seq number */
1018 tcp_seq snd_wl2; /* window update seg ack number */
1019 tcp_seq iss; /* initial send sequence number */
1020 tcp_seq irs; /* initial receive sequence number */
1021
1022 tcp_seq rcv_nxt; /* receive next */
1023 tcp_seq rcv_adv; /* advertised window */
1024 u_int32_t rcv_wnd; /* receive window */
1025 tcp_seq rcv_up; /* receive urgent pointer */
1026
1027 u_int32_t snd_wnd; /* send window */
1028 u_int32_t snd_cwnd; /* congestion-controlled window */
1029 u_int32_t snd_ssthresh; /* snd_cwnd size threshold for
1030 * for slow start exponential to
1031 * linear switch
1032 */
1033 u_int t_maxopd; /* mss plus options */
1034
1035 u_int32_t t_rcvtime; /* time at which a packet was received */
1036 u_int32_t t_starttime; /* time connection was established */
1037 int t_rtttime; /* round trip time */
1038 tcp_seq t_rtseq; /* sequence number being timed */
1039
1040 int t_rxtcur; /* current retransmit value (ticks) */
1041 u_int t_maxseg; /* maximum segment size */
1042 int t_srtt; /* smoothed round-trip time */
1043 int t_rttvar; /* variance in round-trip time */
1044
1045 int t_rxtshift; /* log(2) of rexmt exp. backoff */
1046 u_int t_rttmin; /* minimum rtt allowed */
1047 u_int32_t t_rttupdated; /* number of times rtt sampled */
1048 u_int32_t max_sndwnd; /* largest window peer has offered */
1049
1050 int t_softerror; /* possible error not yet reported */
1051 /* out-of-band data */
1052 char t_oobflags; /* have some */
1053 char t_iobc; /* input character */
1054 /* RFC 1323 variables */
1055 u_char snd_scale; /* window scaling for send window */
1056 u_char rcv_scale; /* window scaling for recv window */
1057 u_char request_r_scale; /* pending window scaling */
1058 u_char requested_s_scale;
1059 u_int32_t ts_recent; /* timestamp echo data */
1060
1061 u_int32_t ts_recent_age; /* when last updated */
1062 tcp_seq last_ack_sent;
1063 /* RFC 1644 variables */
1064 tcp_cc cc_send; /* send connection count */
1065 tcp_cc cc_recv; /* receive connection count */
1066 tcp_seq snd_recover; /* for use in fast recovery */
1067 /* experimental */
1068 u_int32_t snd_cwnd_prev; /* cwnd prior to retransmit */
1069 u_int32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
1070 };
1071
1072 #endif /* PRIVATE */
1073
1074 #pragma pack()
1075
1076 /*
1077 * Names for TCP sysctl objects
1078 */
1079 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
1080 #define TCPCTL_DO_RFC1644 2 /* use RFC-1644 extensions */
1081 #define TCPCTL_MSSDFLT 3 /* MSS default */
1082 #define TCPCTL_STATS 4 /* statistics (read-only) */
1083 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */
1084 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
1085 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
1086 #define TCPCTL_SENDSPACE 8 /* send buffer space */
1087 #define TCPCTL_RECVSPACE 9 /* receive buffer space */
1088 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */
1089 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */
1090 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */
1091 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */
1092 #define TCPCTL_MAXID 14
1093
1094 #ifdef BSD_KERNEL_PRIVATE
1095 #include <sys/bitstring.h>
1096
1097 #define TCP_PKTLIST_CLEAR(tp) { \
1098 (tp)->t_pktlist_head = (tp)->t_pktlist_tail = NULL; \
1099 (tp)->t_lastchain = (tp)->t_pktlist_sentlen = 0; \
1100 }
1101
1102 #define TCPCTL_NAMES { \
1103 { 0, 0 }, \
1104 { "rfc1323", CTLTYPE_INT }, \
1105 { "rfc1644", CTLTYPE_INT }, \
1106 { "mssdflt", CTLTYPE_INT }, \
1107 { "stats", CTLTYPE_STRUCT }, \
1108 { "rttdflt", CTLTYPE_INT }, \
1109 { "keepidle", CTLTYPE_INT }, \
1110 { "keepintvl", CTLTYPE_INT }, \
1111 { "sendspace", CTLTYPE_INT }, \
1112 { "recvspace", CTLTYPE_INT }, \
1113 { "keepinit", CTLTYPE_INT }, \
1114 { "pcblist", CTLTYPE_STRUCT }, \
1115 { "delacktime", CTLTYPE_INT }, \
1116 { "v6mssdflt", CTLTYPE_INT }, \
1117 }
1118
1119 #ifdef SYSCTL_DECL
1120 SYSCTL_DECL(_net_inet_tcp);
1121 #endif /* SYSCTL_DECL */
1122
1123 /*
1124 * Flags for TCP's connectx(2) user-protocol request routine.
1125 */
1126 #if MPTCP
1127 #define TCP_CONNREQF_MPTCP 0x1 /* called internally by MPTCP */
1128 #endif /* MPTCP */
1129
1130 extern struct inpcbhead tcb; /* head of queue of active tcpcb's */
1131 extern struct inpcbinfo tcbinfo;
1132 extern struct tcpstat tcpstat; /* tcp statistics */
1133 extern int tcp_mssdflt; /* XXX */
1134 extern int tcp_minmss;
1135 extern int ss_fltsz;
1136 extern int ss_fltsz_local;
1137 extern int tcp_do_rfc3390; /* Calculate ss_fltsz according to RFC 3390 */
1138 extern int target_qdelay;
1139 #ifdef __APPLE__
1140 extern u_int32_t tcp_now; /* for RFC 1323 timestamps */
1141 extern struct timeval tcp_uptime;
1142 extern lck_spin_t *tcp_uptime_lock;
1143
1144 extern int tcp_delack_enabled;
1145 #endif /* __APPLE__ */
1146
1147 extern int tcp_do_sack; /* SACK enabled/disabled */
1148
1149 #if CONFIG_IFEF_NOWINDOWSCALE
1150 extern int tcp_obey_ifef_nowindowscale;
1151 #endif
1152
1153 struct protosw;
1154 struct domain;
1155
1156 void tcp_canceltimers(struct tcpcb *);
1157 struct tcpcb *
1158 tcp_close(struct tcpcb *);
1159 void tcp_ctlinput(int, struct sockaddr *, void *);
1160 int tcp_ctloutput(struct socket *, struct sockopt *);
1161 struct tcpcb *
1162 tcp_drop(struct tcpcb *, int);
1163 void tcp_drain(void);
1164 void tcp_getrt_rtt(struct tcpcb *tp, struct rtentry *rt);
1165 struct rmxp_tao *
1166 tcp_gettaocache(struct inpcb *);
1167 void tcp_init(struct protosw *, struct domain *);
1168 void tcp_input(struct mbuf *, int);
1169 void tcp_mss(struct tcpcb *, int, unsigned int);
1170 int tcp_mssopt(struct tcpcb *);
1171 void tcp_drop_syn_sent(struct inpcb *, int);
1172 void tcp_mtudisc(struct inpcb *, int);
1173 struct tcpcb *
1174 tcp_newtcpcb(struct inpcb *);
1175 int tcp_output(struct tcpcb *);
1176 void tcp_respond(struct tcpcb *, void *,
1177 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int,
1178 unsigned int, unsigned int);
1179 struct rtentry *tcp_rtlookup(struct inpcb *, unsigned int);
1180 void tcp_setpersist(struct tcpcb *);
1181 void tcp_gc(struct inpcbinfo *);
1182 void tcp_check_timer_state(struct tcpcb *tp);
1183 void tcp_run_timerlist(void *arg1, void *arg2);
1184
1185 struct tcptemp *
1186 tcp_maketemplate(struct tcpcb *);
1187 void tcp_fillheaders(struct tcpcb *, void *, void *);
1188 struct tcpcb *
1189 tcp_timers(struct tcpcb *, int);
1190 void tcp_trace(int, int, struct tcpcb *, void *, struct tcphdr *, int);
1191
1192 void tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq, u_int32_t *);
1193 void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1194 void tcp_clean_sackreport(struct tcpcb *tp);
1195 void tcp_sack_adjust(struct tcpcb *tp);
1196 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1197 void tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1198 void tcp_free_sackholes(struct tcpcb *tp);
1199 int32_t tcp_sbspace(struct tcpcb *tp);
1200 void tcp_set_tso(struct tcpcb *tp, struct ifnet *ifp);
1201 void tcp_reset_stretch_ack(struct tcpcb *tp);
1202 extern void tcp_get_ports_used(u_int32_t, int, u_int32_t, bitstr_t *);
1203 uint32_t tcp_count_opportunistic(unsigned int ifindex, u_int32_t flags);
1204 uint32_t tcp_find_anypcb_byaddr(struct ifaddr *ifa);
1205 void tcp_set_max_rwinscale(struct tcpcb *tp, struct socket *so);
1206 struct bwmeas* tcp_bwmeas_alloc(struct tcpcb *tp);
1207 void tcp_bwmeas_free(struct tcpcb *tp);
1208
1209 extern void tcp_set_background_cc(struct socket *);
1210 extern void tcp_set_foreground_cc(struct socket *);
1211 extern void tcp_set_recv_bg(struct socket *);
1212 extern void tcp_clear_recv_bg(struct socket *);
1213 #define IS_TCP_RECV_BG(_so) \
1214 ((_so)->so_traffic_mgt_flags & TRAFFIC_MGT_TCP_RECVBG)
1215
1216 #if TRAFFIC_MGT
1217 #define CLEAR_IAJ_STATE(_tp_) (_tp_)->iaj_rcv_ts = 0
1218 void reset_acc_iaj(struct tcpcb *tp);
1219 #endif /* TRAFFIC_MGT */
1220
1221 int tcp_lock (struct socket *, int, void *);
1222 int tcp_unlock (struct socket *, int, void *);
1223 void calculate_tcp_clock(void);
1224
1225 extern void mptcp_insert_rmap(struct tcpcb *, struct mbuf *);
1226 extern void tcp_keepalive_reset(struct tcpcb *);
1227
1228 #ifdef _KERN_LOCKS_H_
1229 lck_mtx_t * tcp_getlock (struct socket *, int);
1230 #else
1231 void * tcp_getlock (struct socket *, int);
1232 #endif
1233
1234
1235 extern struct pr_usrreqs tcp_usrreqs;
1236 extern u_int32_t tcp_sendspace;
1237 extern u_int32_t tcp_recvspace;
1238 tcp_seq tcp_new_isn(struct tcpcb *);
1239
1240 extern int tcp_input_checksum(int, struct mbuf *, struct tcphdr *, int, int);
1241 extern void tcp_getconninfo(struct socket *, struct conninfo_tcp *);
1242 #if MPTCP
1243 extern uint16_t mptcp_input_csum(struct tcpcb *, struct mbuf *, int);
1244 extern void mptcp_output_csum(struct tcpcb *, struct mbuf *, int32_t, unsigned,
1245 u_int64_t, u_int32_t *);
1246 extern int mptcp_adj_mss(struct tcpcb *, boolean_t);
1247 #endif
1248 #endif /* BSD_KERNEL_RPIVATE */
1249
1250 #endif /* _NETINET_TCP_VAR_H_ */
mirror of XNU