1 /*-
2 * Copyright (c) 1982, 1986, 1993, 1994, 1995
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)tcp_var.h 8.4 (Berkeley) 5/24/95
30 * $FreeBSD$
31 */
32
33 #ifndef _NETINET_TCP_VAR_H_
34 #define _NETINET_TCP_VAR_H_
35
36 #include <netinet/tcp.h>
37 #include <netinet/tcp_fsm.h>
38
39 #ifdef _KERNEL
40 #include <net/vnet.h>
41 #include <sys/mbuf.h>
42
43 /*
44 * Kernel variables for tcp.
45 */
46 VNET_DECLARE(int, tcp_do_rfc1323);
47 #define V_tcp_do_rfc1323 VNET(tcp_do_rfc1323)
48
49 #endif /* _KERNEL */
50
51 /* TCP segment queue entry */
52 struct tseg_qent {
53 TAILQ_ENTRY(tseg_qent) tqe_q;
54 struct mbuf *tqe_m; /* mbuf contains packet */
55 struct mbuf *tqe_last; /* last mbuf in chain */
56 tcp_seq tqe_start; /* TCP Sequence number start */
57 int tqe_len; /* TCP segment data length */
58 uint32_t tqe_flags; /* The flags from the th->th_flags */
59 uint32_t tqe_mbuf_cnt; /* Count of mbuf overhead */
60 };
61 TAILQ_HEAD(tsegqe_head, tseg_qent);
62
63 struct sackblk {
64 tcp_seq start; /* start seq no. of sack block */
65 tcp_seq end; /* end seq no. */
66 };
67
68 struct sackhole {
69 tcp_seq start; /* start seq no. of hole */
70 tcp_seq end; /* end seq no. */
71 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */
72 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */
73 };
74
75 struct sackhint {
76 struct sackhole *nexthole;
77 int sack_bytes_rexmit;
78 tcp_seq last_sack_ack; /* Most recent/largest sacked ack */
79
80 int ispare; /* explicit pad for 64bit alignment */
81 int sacked_bytes; /*
82 * Total sacked bytes reported by the
83 * receiver via sack option
84 */
85 uint32_t _pad1[1]; /* TBD */
86 uint64_t _pad[1]; /* TBD */
87 };
88
89 struct tcptemp {
90 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
91 struct tcphdr tt_t;
92 };
93
94 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
95
96 #define tcp6cb tcpcb /* for KAME src sync over BSD*'s */
97
98 /*
99 * TODO: We yet need to brave plowing in
100 * to tcp_input() and the pru_usrreq() block.
101 * Right now these go to the old standards which
102 * are somewhat ok, but in the long term may
103 * need to be changed. If we do tackle tcp_input()
104 * then we need to get rid of the tcp_do_segment()
105 * function below.
106 */
107 /* Flags for tcp functions */
108 #define TCP_FUNC_BEING_REMOVED 0x01 /* Can no longer be referenced */
109 struct tcpcb;
110 struct inpcb;
111 struct sockopt;
112 struct socket;
113
114 /*
115 * If defining the optional tcp_timers, in the
116 * tfb_tcp_timer_stop call you must use the
117 * callout_async_drain() function with the
118 * tcp_timer_discard callback. You should check
119 * the return of callout_async_drain() and if 0
120 * increment tt_draincnt. Since the timer sub-system
121 * does not know your callbacks you must provide a
122 * stop_all function that loops through and calls
123 * tcp_timer_stop() with each of your defined timers.
124 */
125 struct tcp_function_block {
126 char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
127 int (*tfb_tcp_output)(struct tcpcb *);
128 void (*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
129 struct socket *, struct tcpcb *,
130 int, int, uint8_t,
131 int);
132 int (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
133 struct inpcb *inp, struct tcpcb *tp);
134 /* Optional memory allocation/free routine */
135 void (*tfb_tcp_fb_init)(struct tcpcb *);
136 void (*tfb_tcp_fb_fini)(struct tcpcb *);
137 /* Optional timers, must define all if you define one */
138 int (*tfb_tcp_timer_stop_all)(struct tcpcb *);
139 void (*tfb_tcp_timer_activate)(struct tcpcb *,
140 uint32_t, u_int);
141 int (*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
142 void (*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
143 void (*tfb_tcp_rexmit_tmr)(struct tcpcb *);
144 volatile uint32_t tfb_refcnt;
145 uint32_t tfb_flags;
146 };
147
148 struct tcp_function {
149 TAILQ_ENTRY(tcp_function) tf_next;
150 struct tcp_function_block *tf_fb;
151 };
152
153 TAILQ_HEAD(tcp_funchead, tcp_function);
154
155 /*
156 * Tcp control block, one per tcp; fields:
157 * Organized for 16 byte cacheline efficiency.
158 */
159 struct tcpcb {
160 struct tsegqe_head t_segq; /* segment reassembly queue */
161 void *t_pspare[2]; /* new reassembly queue */
162 int t_segqlen; /* segment reassembly queue length */
163 int t_dupacks; /* consecutive dup acks recd */
164
165 struct tcp_timer *t_timers; /* All the TCP timers in one struct */
166
167 struct inpcb *t_inpcb; /* back pointer to internet pcb */
168 int t_state; /* state of this connection */
169 u_int t_flags;
170
171 struct vnet *t_vnet; /* back pointer to parent vnet */
172
173 tcp_seq snd_una; /* sent but unacknowledged */
174 tcp_seq snd_max; /* highest sequence number sent;
175 * used to recognize retransmits
176 */
177 tcp_seq snd_nxt; /* send next */
178 tcp_seq snd_up; /* send urgent pointer */
179
180 tcp_seq snd_wl1; /* window update seg seq number */
181 tcp_seq snd_wl2; /* window update seg ack number */
182 tcp_seq iss; /* initial send sequence number */
183 tcp_seq irs; /* initial receive sequence number */
184
185 tcp_seq rcv_nxt; /* receive next */
186 tcp_seq rcv_adv; /* advertised window */
187 u_long rcv_wnd; /* receive window */
188 tcp_seq rcv_up; /* receive urgent pointer */
189
190 u_long snd_wnd; /* send window */
191 u_long snd_cwnd; /* congestion-controlled window */
192 u_long snd_spare1; /* unused */
193 u_long snd_ssthresh; /* snd_cwnd size threshold for
194 * for slow start exponential to
195 * linear switch
196 */
197 u_long snd_spare2; /* unused */
198 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */
199
200 u_int t_rcvtime; /* inactivity time */
201 u_int t_starttime; /* time connection was established */
202 u_int t_rtttime; /* RTT measurement start time */
203 tcp_seq t_rtseq; /* sequence number being timed */
204
205 u_int t_bw_spare1; /* unused */
206 tcp_seq t_bw_spare2; /* unused */
207
208 int t_rxtcur; /* current retransmit value (ticks) */
209 u_int t_maxseg; /* maximum segment size */
210 u_int t_pmtud_saved_maxseg; /* pre-blackhole MSS */
211 int t_srtt; /* smoothed round-trip time */
212 int t_rttvar; /* variance in round-trip time */
213
214 int t_rxtshift; /* log(2) of rexmt exp. backoff */
215 u_int t_rttmin; /* minimum rtt allowed */
216 u_int t_rttbest; /* best rtt we've seen */
217 u_long t_rttupdated; /* number of times rtt sampled */
218 u_long max_sndwnd; /* largest window peer has offered */
219
220 int t_softerror; /* possible error not yet reported */
221 /* out-of-band data */
222 char t_oobflags; /* have some */
223 char t_iobc; /* input character */
224 /* RFC 1323 variables */
225 u_char snd_scale; /* window scaling for send window */
226 u_char rcv_scale; /* window scaling for recv window */
227 u_char request_r_scale; /* pending window scaling */
228 u_int32_t ts_recent; /* timestamp echo data */
229 u_int ts_recent_age; /* when last updated */
230 u_int32_t ts_offset; /* our timestamp offset */
231
232 tcp_seq last_ack_sent;
233 /* experimental */
234 u_long snd_cwnd_prev; /* cwnd prior to retransmit */
235 u_long snd_ssthresh_prev; /* ssthresh prior to retransmit */
236 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */
237 int t_sndzerowin; /* zero-window updates sent */
238 u_int t_badrxtwin; /* window for retransmit recovery */
239 u_char snd_limited; /* segments limited transmitted */
240 /* SACK related state */
241 int snd_numholes; /* number of holes seen by sender */
242 TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
243 /* SACK scoreboard (sorted) */
244 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/
245 int rcv_numsacks; /* # distinct sack blks present */
246 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
247 tcp_seq sack_newdata; /* New data xmitted in this recovery
248 episode starts at this seq number */
249 struct sackhint sackhint; /* SACK scoreboard hint */
250 int t_rttlow; /* smallest observerved RTT */
251 u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */
252 int rfbuf_cnt; /* recv buffer autoscaling byte count */
253 struct toedev *tod; /* toedev handling this connection */
254 int t_sndrexmitpack; /* retransmit packets sent */
255 int t_rcvoopack; /* out-of-order packets received */
256 void *t_toe; /* TOE pcb pointer */
257 int t_bytes_acked; /* # bytes acked during current RTT */
258 struct cc_algo *cc_algo; /* congestion control algorithm */
259 struct cc_var *ccv; /* congestion control specific vars */
260 struct osd *osd; /* storage for Khelp module data */
261
262 u_int t_keepinit; /* time to establish connection */
263 u_int t_keepidle; /* time before keepalive probes begin */
264 u_int t_keepintvl; /* interval between keepalives */
265 u_int t_keepcnt; /* number of keepalives before close */
266
267 u_int t_tsomax; /* TSO total burst length limit in bytes */
268 u_int t_tsomaxsegcount; /* TSO maximum segment count */
269 u_int t_tsomaxsegsize; /* TSO maximum segment size in bytes */
270 u_int t_flags2; /* More tcpcb flags storage */
271 uint32_t t_segqmbuflen; /* Count of bytes mbufs on all entries */
272 #if defined(_KERNEL) && defined(TCP_RFC7413)
273 uint32_t t_ispare[5]; /* 5 UTO, 0 TBD */
274 uint64_t t_tfo_cookie; /* TCP Fast Open cookie */
275 #else
276 uint32_t t_ispare[7]; /* 5 UTO, 2 TBD */
277 #endif
278 struct tcp_function_block *t_fb;/* TCP function call block */
279 void *t_fb_ptr; /* Pointer to t_fb specific data */
280 #if defined(_KERNEL) && defined(TCP_RFC7413)
281 unsigned int *t_tfo_pending; /* TCP Fast Open pending counter */
282 void *t_pspare2[1]; /* 1 TCP_SIGNATURE */
283 #else
284 void *t_pspare2[2]; /* 1 TCP_SIGNATURE, 1 TBD */
285 #endif
286 #if defined(_KERNEL) && defined(TCPPCAP)
287 struct mbufq t_inpkts; /* List of saved input packets. */
288 struct mbufq t_outpkts; /* List of saved output packets. */
289 #ifdef _LP64
290 uint64_t _pad[0]; /* all used! */
291 #else
292 uint64_t _pad[2]; /* 2 are available */
293 #endif /* _LP64 */
294 #else
295 uint64_t _pad[6];
296 #endif /* defined(_KERNEL) && defined(TCPPCAP) */
297 };
298
299 /*
300 * Flags and utility macros for the t_flags field.
301 */
302 #define TF_ACKNOW 0x000001 /* ack peer immediately */
303 #define TF_DELACK 0x000002 /* ack, but try to delay it */
304 #define TF_NODELAY 0x000004 /* don't delay packets to coalesce */
305 #define TF_NOOPT 0x000008 /* don't use tcp options */
306 #define TF_SENTFIN 0x000010 /* have sent FIN */
307 #define TF_REQ_SCALE 0x000020 /* have/will request window scaling */
308 #define TF_RCVD_SCALE 0x000040 /* other side has requested scaling */
309 #define TF_REQ_TSTMP 0x000080 /* have/will request timestamps */
310 #define TF_RCVD_TSTMP 0x000100 /* a timestamp was received in SYN */
311 #define TF_SACK_PERMIT 0x000200 /* other side said I could SACK */
312 #define TF_NEEDSYN 0x000400 /* send SYN (implicit state) */
313 #define TF_NEEDFIN 0x000800 /* send FIN (implicit state) */
314 #define TF_NOPUSH 0x001000 /* don't push */
315 #define TF_PREVVALID 0x002000 /* saved values for bad rxmit valid */
316 #define TF_MORETOCOME 0x010000 /* More data to be appended to sock */
317 #define TF_LQ_OVERFLOW 0x020000 /* listen queue overflow */
318 #define TF_LASTIDLE 0x040000 /* connection was previously idle */
319 #define TF_RXWIN0SENT 0x080000 /* sent a receiver win 0 in response */
320 #define TF_FASTRECOVERY 0x100000 /* in NewReno Fast Recovery */
321 #define TF_WASFRECOVERY 0x200000 /* was in NewReno Fast Recovery */
322 #define TF_SIGNATURE 0x400000 /* require MD5 digests (RFC2385) */
323 #define TF_FORCEDATA 0x800000 /* force out a byte */
324 #define TF_TSO 0x1000000 /* TSO enabled on this connection */
325 #define TF_TOE 0x2000000 /* this connection is offloaded */
326 #define TF_ECN_PERMIT 0x4000000 /* connection ECN-ready */
327 #define TF_ECN_SND_CWR 0x8000000 /* ECN CWR in queue */
328 #define TF_ECN_SND_ECE 0x10000000 /* ECN ECE in queue */
329 #define TF_CONGRECOVERY 0x20000000 /* congestion recovery mode */
330 #define TF_WASCRECOVERY 0x40000000 /* was in congestion recovery */
331 #define TF_FASTOPEN 0x80000000 /* TCP Fast Open indication */
332
333 #define IN_FASTRECOVERY(t_flags) (t_flags & TF_FASTRECOVERY)
334 #define ENTER_FASTRECOVERY(t_flags) t_flags |= TF_FASTRECOVERY
335 #define EXIT_FASTRECOVERY(t_flags) t_flags &= ~TF_FASTRECOVERY
336
337 #define IN_CONGRECOVERY(t_flags) (t_flags & TF_CONGRECOVERY)
338 #define ENTER_CONGRECOVERY(t_flags) t_flags |= TF_CONGRECOVERY
339 #define EXIT_CONGRECOVERY(t_flags) t_flags &= ~TF_CONGRECOVERY
340
341 #define IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
342 #define ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
343 #define EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
344
345 #define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una)
346
347 /*
348 * Flags for the t_oobflags field.
349 */
350 #define TCPOOB_HAVEDATA 0x01
351 #define TCPOOB_HADDATA 0x02
352
353 /*
354 * Flags for PLPMTU handling, t_flags2
355 */
356 #define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */
357 #define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */
358 #define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */
359
360 /*
361 * Structure to hold TCP options that are only used during segment
362 * processing (in tcp_input), but not held in the tcpcb.
363 * It's basically used to reduce the number of parameters
364 * to tcp_dooptions and tcp_addoptions.
365 * The binary order of the to_flags is relevant for packing of the
366 * options in tcp_addoptions.
367 */
368 struct tcpopt {
369 u_int32_t to_flags; /* which options are present */
370 #define TOF_MSS 0x0001 /* maximum segment size */
371 #define TOF_SCALE 0x0002 /* window scaling */
372 #define TOF_SACKPERM 0x0004 /* SACK permitted */
373 #define TOF_TS 0x0010 /* timestamp */
374 #define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */
375 #define TOF_SACK 0x0080 /* Peer sent SACK option */
376 #define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */
377 #define TOF_MAXOPT 0x0200
378 u_int32_t to_tsval; /* new timestamp */
379 u_int32_t to_tsecr; /* reflected timestamp */
380 u_char *to_sacks; /* pointer to the first SACK blocks */
381 u_char *to_signature; /* pointer to the TCP-MD5 signature */
382 u_char *to_tfo_cookie; /* pointer to the TFO cookie */
383 u_int16_t to_mss; /* maximum segment size */
384 u_int8_t to_wscale; /* window scaling */
385 u_int8_t to_nsacks; /* number of SACK blocks */
386 u_int8_t to_tfo_len; /* TFO cookie length */
387 u_int32_t to_spare; /* UTO */
388 };
389
390 /*
391 * Flags for tcp_dooptions.
392 */
393 #define TO_SYN 0x01 /* parse SYN-only options */
394
395 struct hc_metrics_lite { /* must stay in sync with hc_metrics */
396 u_long rmx_mtu; /* MTU for this path */
397 u_long rmx_ssthresh; /* outbound gateway buffer limit */
398 u_long rmx_rtt; /* estimated round trip time */
399 u_long rmx_rttvar; /* estimated rtt variance */
400 u_long rmx_cwnd; /* congestion window */
401 u_long rmx_sendpipe; /* outbound delay-bandwidth product */
402 u_long rmx_recvpipe; /* inbound delay-bandwidth product */
403 };
404
405 /*
406 * Used by tcp_maxmtu() to communicate interface specific features
407 * and limits at the time of connection setup.
408 */
409 struct tcp_ifcap {
410 int ifcap;
411 u_int tsomax;
412 u_int tsomaxsegcount;
413 u_int tsomaxsegsize;
414 };
415
416 #ifndef _NETINET_IN_PCB_H_
417 struct in_conninfo;
418 #endif /* _NETINET_IN_PCB_H_ */
419
420 struct tcptw {
421 struct inpcb *tw_inpcb; /* XXX back pointer to internet pcb */
422 tcp_seq snd_nxt;
423 tcp_seq rcv_nxt;
424 tcp_seq iss;
425 tcp_seq irs;
426 u_short last_win; /* cached window value */
427 u_short tw_so_options; /* copy of so_options */
428 struct ucred *tw_cred; /* user credentials */
429 u_int32_t t_recent;
430 u_int32_t ts_offset; /* our timestamp offset */
431 u_int t_starttime;
432 int tw_time;
433 TAILQ_ENTRY(tcptw) tw_2msl;
434 void *tw_pspare; /* TCP_SIGNATURE */
435 u_int *tw_spare; /* TCP_SIGNATURE */
436 };
437
438 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb)
439 #define intotw(ip) ((struct tcptw *)(ip)->inp_ppcb)
440 #define sototcpcb(so) (intotcpcb(sotoinpcb(so)))
441
442 /*
443 * The smoothed round-trip time and estimated variance
444 * are stored as fixed point numbers scaled by the values below.
445 * For convenience, these scales are also used in smoothing the average
446 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
447 * With these scales, srtt has 3 bits to the right of the binary point,
448 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the
449 * binary point, and is smoothed with an ALPHA of 0.75.
450 */
451 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */
452 #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */
453 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */
454 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */
455 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
456
457 /*
458 * The initial retransmission should happen at rtt + 4 * rttvar.
459 * Because of the way we do the smoothing, srtt and rttvar
460 * will each average +1/2 tick of bias. When we compute
461 * the retransmit timer, we want 1/2 tick of rounding and
462 * 1 extra tick because of +-1/2 tick uncertainty in the
463 * firing of the timer. The bias will give us exactly the
464 * 1.5 tick we need. But, because the bias is
465 * statistical, we have to test that we don't drop below
466 * the minimum feasible timer (which is 2 ticks).
467 * This version of the macro adapted from a paper by Lawrence
468 * Brakmo and Larry Peterson which outlines a problem caused
469 * by insufficient precision in the original implementation,
470 * which results in inappropriately large RTO values for very
471 * fast networks.
472 */
473 #define TCP_REXMTVAL(tp) \
474 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
475 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
476
477 /*
478 * TCP statistics.
479 * Many of these should be kept per connection,
480 * but that's inconvenient at the moment.
481 */
482 struct tcpstat {
483 uint64_t tcps_connattempt; /* connections initiated */
484 uint64_t tcps_accepts; /* connections accepted */
485 uint64_t tcps_connects; /* connections established */
486 uint64_t tcps_drops; /* connections dropped */
487 uint64_t tcps_conndrops; /* embryonic connections dropped */
488 uint64_t tcps_minmssdrops; /* average minmss too low drops */
489 uint64_t tcps_closed; /* conn. closed (includes drops) */
490 uint64_t tcps_segstimed; /* segs where we tried to get rtt */
491 uint64_t tcps_rttupdated; /* times we succeeded */
492 uint64_t tcps_delack; /* delayed acks sent */
493 uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
494 uint64_t tcps_rexmttimeo; /* retransmit timeouts */
495 uint64_t tcps_persisttimeo; /* persist timeouts */
496 uint64_t tcps_keeptimeo; /* keepalive timeouts */
497 uint64_t tcps_keepprobe; /* keepalive probes sent */
498 uint64_t tcps_keepdrops; /* connections dropped in keepalive */
499
500 uint64_t tcps_sndtotal; /* total packets sent */
501 uint64_t tcps_sndpack; /* data packets sent */
502 uint64_t tcps_sndbyte; /* data bytes sent */
503 uint64_t tcps_sndrexmitpack; /* data packets retransmitted */
504 uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */
505 uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */
506 uint64_t tcps_sndacks; /* ack-only packets sent */
507 uint64_t tcps_sndprobe; /* window probes sent */
508 uint64_t tcps_sndurg; /* packets sent with URG only */
509 uint64_t tcps_sndwinup; /* window update-only packets sent */
510 uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
511
512 uint64_t tcps_rcvtotal; /* total packets received */
513 uint64_t tcps_rcvpack; /* packets received in sequence */
514 uint64_t tcps_rcvbyte; /* bytes received in sequence */
515 uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */
516 uint64_t tcps_rcvbadoff; /* packets received with bad offset */
517 uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */
518 uint64_t tcps_rcvshort; /* packets received too short */
519 uint64_t tcps_rcvduppack; /* duplicate-only packets received */
520 uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */
521 uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */
522 uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
523 uint64_t tcps_rcvoopack; /* out-of-order packets received */
524 uint64_t tcps_rcvoobyte; /* out-of-order bytes received */
525 uint64_t tcps_rcvpackafterwin; /* packets with data after window */
526 uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */
527 uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */
528 uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */
529 uint64_t tcps_rcvdupack; /* rcvd duplicate acks */
530 uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */
531 uint64_t tcps_rcvackpack; /* rcvd ack packets */
532 uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */
533 uint64_t tcps_rcvwinupd; /* rcvd window update packets */
534 uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */
535 uint64_t tcps_predack; /* times hdr predict ok for acks */
536 uint64_t tcps_preddat; /* times hdr predict ok for data pkts */
537 uint64_t tcps_pcbcachemiss;
538 uint64_t tcps_cachedrtt; /* times cached RTT in route updated */
539 uint64_t tcps_cachedrttvar; /* times cached rttvar updated */
540 uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */
541 uint64_t tcps_usedrtt; /* times RTT initialized from route */
542 uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */
543 uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/
544 uint64_t tcps_persistdrop; /* timeout in persist state */
545 uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */
546 uint64_t tcps_mturesent; /* resends due to MTU discovery */
547 uint64_t tcps_listendrop; /* listen queue overflows */
548 uint64_t tcps_badrst; /* ignored RSTs in the window */
549
550 uint64_t tcps_sc_added; /* entry added to syncache */
551 uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */
552 uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */
553 uint64_t tcps_sc_dropped; /* could not reply to packet */
554 uint64_t tcps_sc_completed; /* successful extraction of entry */
555 uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
556 uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */
557 uint64_t tcps_sc_reset; /* RST removed entry from syncache */
558 uint64_t tcps_sc_stale; /* timed out or listen socket gone */
559 uint64_t tcps_sc_aborted; /* syncache entry aborted */
560 uint64_t tcps_sc_badack; /* removed due to bad ACK */
561 uint64_t tcps_sc_unreach; /* ICMP unreachable received */
562 uint64_t tcps_sc_zonefail; /* zalloc() failed */
563 uint64_t tcps_sc_sendcookie; /* SYN cookie sent */
564 uint64_t tcps_sc_recvcookie; /* SYN cookie received */
565
566 uint64_t tcps_hc_added; /* entry added to hostcache */
567 uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
568
569 uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */
570
571 /* SACK related stats */
572 uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
573 uint64_t tcps_sack_rexmits; /* SACK rexmit segments */
574 uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */
575 uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */
576 uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */
577 uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */
578
579 /* ECN related stats */
580 uint64_t tcps_ecn_ce; /* ECN Congestion Experienced */
581 uint64_t tcps_ecn_ect0; /* ECN Capable Transport */
582 uint64_t tcps_ecn_ect1; /* ECN Capable Transport */
583 uint64_t tcps_ecn_shs; /* ECN successful handshakes */
584 uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */
585
586 /* TCP_SIGNATURE related stats */
587 uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */
588 uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */
589 uint64_t tcps_sig_err_buildsig; /* Failed to make signature */
590 uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */
591 uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */
592
593 uint64_t _pad[12]; /* 6 UTO, 6 TBD */
594 };
595
596 #define tcps_rcvmemdrop tcps_rcvreassfull /* compat */
597
598 #ifdef _KERNEL
599 #define TI_UNLOCKED 1
600 #define TI_RLOCKED 2
601 #include <sys/counter.h>
602
603 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */
604 /*
605 * In-kernel consumers can use these accessor macros directly to update
606 * stats.
607 */
608 #define TCPSTAT_ADD(name, val) \
609 VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
610 #define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1)
611
612 /*
613 * Kernel module consumers must use this accessor macro.
614 */
615 void kmod_tcpstat_inc(int statnum);
616 #define KMOD_TCPSTAT_INC(name) \
617 kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
618
619 /*
620 * Running TCP connection count by state.
621 */
622 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
623 #define V_tcps_states VNET(tcps_states)
624 #define TCPSTATES_INC(state) counter_u64_add(V_tcps_states[state], 1)
625 #define TCPSTATES_DEC(state) counter_u64_add(V_tcps_states[state], -1)
626
627 /*
628 * TCP specific helper hook point identifiers.
629 */
630 #define HHOOK_TCP_EST_IN 0
631 #define HHOOK_TCP_EST_OUT 1
632 #define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT
633
634 struct tcp_hhook_data {
635 struct tcpcb *tp;
636 struct tcphdr *th;
637 struct tcpopt *to;
638 long len;
639 int tso;
640 tcp_seq curack;
641 };
642 #endif
643
644 /*
645 * TCB structure exported to user-land via sysctl(3).
646 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
647 * included. Not all of our clients do.
648 */
649 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
650 struct xtcp_timer {
651 int tt_rexmt; /* retransmit timer */
652 int tt_persist; /* retransmit persistence */
653 int tt_keep; /* keepalive */
654 int tt_2msl; /* 2*msl TIME_WAIT timer */
655 int tt_delack; /* delayed ACK timer */
656 int t_rcvtime; /* Time since last packet received */
657 };
658 struct xtcpcb {
659 size_t xt_len;
660 struct inpcb xt_inp;
661 struct tcpcb xt_tp;
662 struct xsocket xt_socket;
663 struct xtcp_timer xt_timer;
664 u_quad_t xt_alignment_hack;
665 };
666 #endif
667
668 /*
669 * Identifiers for TCP sysctl nodes
670 */
671 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
672 #define TCPCTL_MSSDFLT 3 /* MSS default */
673 #define TCPCTL_STATS 4 /* statistics */
674 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */
675 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
676 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
677 #define TCPCTL_SENDSPACE 8 /* send buffer space */
678 #define TCPCTL_RECVSPACE 9 /* receive buffer space */
679 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */
680 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */
681 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */
682 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */
683 #define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */
684 #define TCPCTL_DROP 15 /* drop tcp connection */
685 #define TCPCTL_STATES 16 /* connection counts by TCP state */
686
687 #ifdef _KERNEL
688 #ifdef SYSCTL_DECL
689 SYSCTL_DECL(_net_inet_tcp);
690 SYSCTL_DECL(_net_inet_tcp_sack);
691 MALLOC_DECLARE(M_TCPLOG);
692 #endif
693
694 VNET_DECLARE(struct inpcbhead, tcb); /* queue of active tcpcb's */
695 VNET_DECLARE(struct inpcbinfo, tcbinfo);
696 extern int tcp_log_in_vain;
697 VNET_DECLARE(int, tcp_mssdflt); /* XXX */
698 VNET_DECLARE(int, tcp_minmss);
699 VNET_DECLARE(int, tcp_delack_enabled);
700 VNET_DECLARE(int, tcp_do_rfc3390);
701 VNET_DECLARE(int, tcp_initcwnd_segments);
702 VNET_DECLARE(int, tcp_sendspace);
703 VNET_DECLARE(int, tcp_recvspace);
704 VNET_DECLARE(int, path_mtu_discovery);
705 VNET_DECLARE(int, tcp_do_rfc3465);
706 VNET_DECLARE(int, tcp_abc_l_var);
707 #define V_tcb VNET(tcb)
708 #define V_tcbinfo VNET(tcbinfo)
709 #define V_tcp_mssdflt VNET(tcp_mssdflt)
710 #define V_tcp_minmss VNET(tcp_minmss)
711 #define V_tcp_delack_enabled VNET(tcp_delack_enabled)
712 #define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390)
713 #define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments)
714 #define V_tcp_sendspace VNET(tcp_sendspace)
715 #define V_tcp_recvspace VNET(tcp_recvspace)
716 #define V_path_mtu_discovery VNET(path_mtu_discovery)
717 #define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465)
718 #define V_tcp_abc_l_var VNET(tcp_abc_l_var)
719
720 VNET_DECLARE(int, tcp_do_sack); /* SACK enabled/disabled */
721 VNET_DECLARE(int, tcp_sc_rst_sock_fail); /* RST on sock alloc failure */
722 #define V_tcp_do_sack VNET(tcp_do_sack)
723 #define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail)
724
725 VNET_DECLARE(int, tcp_do_ecn); /* TCP ECN enabled/disabled */
726 VNET_DECLARE(int, tcp_ecn_maxretries);
727 #define V_tcp_do_ecn VNET(tcp_do_ecn)
728 #define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries)
729
730 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
731 #define V_tcp_hhh VNET(tcp_hhh)
732
733 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
734 #define V_tcp_do_rfc6675_pipe VNET(tcp_do_rfc6675_pipe)
735
736 int tcp_addoptions(struct tcpopt *, u_char *);
737 int tcp_ccalgounload(struct cc_algo *unload_algo);
738 struct tcpcb *
739 tcp_close(struct tcpcb *);
740 void tcp_discardcb(struct tcpcb *);
741 void tcp_twstart(struct tcpcb *);
742 void tcp_twclose(struct tcptw *, int);
743 void tcp_ctlinput(int, struct sockaddr *, void *);
744 int tcp_ctloutput(struct socket *, struct sockopt *);
745 struct tcpcb *
746 tcp_drop(struct tcpcb *, int);
747 void tcp_drain(void);
748 void tcp_init(void);
749 void tcp_fini(void *);
750 char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
751 const void *);
752 char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
753 const void *);
754 int tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *, struct mbuf *);
755 void tcp_reass_global_init(void);
756 void tcp_reass_flush(struct tcpcb *);
757 void tcp_dooptions(struct tcpopt *, u_char *, int, int);
758 void tcp_dropwithreset(struct mbuf *, struct tcphdr *,
759 struct tcpcb *, int, int);
760 void tcp_pulloutofband(struct socket *,
761 struct tcphdr *, struct mbuf *, int);
762 void tcp_xmit_timer(struct tcpcb *, int);
763 void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
764 void cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
765 uint16_t type);
766 void cc_conn_init(struct tcpcb *tp);
767 void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
768 void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
769 void hhook_run_tcp_est_in(struct tcpcb *tp,
770 struct tcphdr *th, struct tcpopt *to);
771
772 int tcp_input(struct mbuf **, int *, int);
773 int tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
774 struct tcpcb *, int);
775 void tcp_do_segment(struct mbuf *, struct tcphdr *,
776 struct socket *, struct tcpcb *, int, int, uint8_t,
777 int);
778
779 int register_tcp_functions(struct tcp_function_block *blk, int wait);
780 int deregister_tcp_functions(struct tcp_function_block *blk);
781 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
782 struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk);
783 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
784
785 u_long tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
786 u_long tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
787 u_int tcp_maxseg(const struct tcpcb *);
788 void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
789 struct tcp_ifcap *);
790 void tcp_mss(struct tcpcb *, int);
791 int tcp_mssopt(struct in_conninfo *);
792 struct inpcb *
793 tcp_drop_syn_sent(struct inpcb *, int);
794 struct tcpcb *
795 tcp_newtcpcb(struct inpcb *);
796 int tcp_output(struct tcpcb *);
797 void tcp_state_change(struct tcpcb *, int);
798 void tcp_respond(struct tcpcb *, void *,
799 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
800 void tcp_tw_init(void);
801 #ifdef VIMAGE
802 void tcp_tw_destroy(void);
803 #endif
804 void tcp_tw_zone_change(void);
805 int tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
806 struct mbuf *, int);
807 void tcp_setpersist(struct tcpcb *);
808 void tcp_slowtimo(void);
809 struct tcptemp *
810 tcpip_maketemplate(struct inpcb *);
811 void tcpip_fillheaders(struct inpcb *, void *, void *);
812 void tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
813 int tcp_timer_active(struct tcpcb *, uint32_t);
814 void tcp_timer_stop(struct tcpcb *, uint32_t);
815 void tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
816 /*
817 * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
818 */
819 void tcp_hc_init(void);
820 #ifdef VIMAGE
821 void tcp_hc_destroy(void);
822 #endif
823 void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
824 u_long tcp_hc_getmtu(struct in_conninfo *);
825 void tcp_hc_updatemtu(struct in_conninfo *, u_long);
826 void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
827
828 extern struct pr_usrreqs tcp_usrreqs;
829
830 uint32_t tcp_new_ts_offset(struct in_conninfo *);
831 tcp_seq tcp_new_isn(struct in_conninfo *);
832
833 int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
834 void tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq);
835 void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
836 void tcp_clean_dsack_blocks(struct tcpcb *tp);
837 void tcp_clean_sackreport(struct tcpcb *tp);
838 void tcp_sack_adjust(struct tcpcb *tp);
839 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
840 void tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
841 void tcp_free_sackholes(struct tcpcb *tp);
842 int tcp_newreno(struct tcpcb *, struct tcphdr *);
843 int tcp_compute_pipe(struct tcpcb *);
844
845 static inline void
tcp_fields_to_host(struct tcphdr * th)846 tcp_fields_to_host(struct tcphdr *th)
847 {
848
849 th->th_seq = ntohl(th->th_seq);
850 th->th_ack = ntohl(th->th_ack);
851 th->th_win = ntohs(th->th_win);
852 th->th_urp = ntohs(th->th_urp);
853 }
854
855 static inline void
tcp_fields_to_net(struct tcphdr * th)856 tcp_fields_to_net(struct tcphdr *th)
857 {
858
859 th->th_seq = htonl(th->th_seq);
860 th->th_ack = htonl(th->th_ack);
861 th->th_win = htons(th->th_win);
862 th->th_urp = htons(th->th_urp);
863 }
864 #endif /* _KERNEL */
865
866 #endif /* _NETINET_TCP_VAR_H_ */
867