xref: /freebsd/sys/netinet/tcp_var.h (revision 076ad2f836d5f49dc1375f1677335a48fe0d4b82)
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  * 3. 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 	LIST_ENTRY(tseg_qent) tqe_q;
54 	int	tqe_len;		/* TCP segment data length */
55 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
56 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
57 };
58 LIST_HEAD(tsegqe_head, tseg_qent);
59 
60 struct sackblk {
61 	tcp_seq start;		/* start seq no. of sack block */
62 	tcp_seq end;		/* end seq no. */
63 };
64 
65 struct sackhole {
66 	tcp_seq start;		/* start seq no. of hole */
67 	tcp_seq end;		/* end seq no. */
68 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
69 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
70 };
71 
72 struct sackhint {
73 	struct sackhole	*nexthole;
74 	int		sack_bytes_rexmit;
75 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
76 
77 	int		ispare;		/* explicit pad for 64bit alignment */
78 	int             sacked_bytes;	/*
79 					 * Total sacked bytes reported by the
80 					 * receiver via sack option
81 					 */
82 	uint32_t	_pad1[1];	/* TBD */
83 	uint64_t	_pad[1];	/* TBD */
84 };
85 
86 struct tcptemp {
87 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
88 	struct	tcphdr tt_t;
89 };
90 
91 #define tcp6cb		tcpcb  /* for KAME src sync over BSD*'s */
92 
93 /*
94  * TODO: We yet need to brave plowing in
95  * to tcp_input() and the pru_usrreq() block.
96  * Right now these go to the old standards which
97  * are somewhat ok, but in the long term may
98  * need to be changed. If we do tackle tcp_input()
99  * then we need to get rid of the tcp_do_segment()
100  * function below.
101  */
102 /* Flags for tcp functions */
103 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
104 struct tcpcb;
105 struct inpcb;
106 struct sockopt;
107 struct socket;
108 
109 /*
110  * If defining the optional tcp_timers, in the
111  * tfb_tcp_timer_stop call you must use the
112  * callout_async_drain() function with the
113  * tcp_timer_discard callback. You should check
114  * the return of callout_async_drain() and if 0
115  * increment tt_draincnt. Since the timer sub-system
116  * does not know your callbacks you must provide a
117  * stop_all function that loops through and calls
118  * tcp_timer_stop() with each of your defined timers.
119  * Adding a tfb_tcp_handoff_ok function allows the socket
120  * option to change stacks to query you even if the
121  * connection is in a later stage. You return 0 to
122  * say you can take over and run your stack, you return
123  * non-zero (an error number) to say no you can't.
124  * If the function is undefined you can only change
125  * in the early states (before connect or listen).
126  * tfb_tcp_fb_fini is changed to add a flag to tell
127  * the old stack if the tcb is being destroyed or
128  * not. A one in the flag means the TCB is being
129  * destroyed, a zero indicates its transitioning to
130  * another stack (via socket option).
131  */
132 struct tcp_function_block {
133 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
134 	int	(*tfb_tcp_output)(struct tcpcb *);
135 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
136 			    struct socket *, struct tcpcb *,
137 			    int, int, uint8_t,
138 			    int);
139 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
140 			    struct inpcb *inp, struct tcpcb *tp);
141 	/* Optional memory allocation/free routine */
142 	void	(*tfb_tcp_fb_init)(struct tcpcb *);
143 	void	(*tfb_tcp_fb_fini)(struct tcpcb *, int);
144 	/* Optional timers, must define all if you define one */
145 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
146 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
147 			    uint32_t, u_int);
148 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
149 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
150 	void	(*tfb_tcp_rexmit_tmr)(struct tcpcb *);
151 	int	(*tfb_tcp_handoff_ok)(struct tcpcb *);
152 	volatile uint32_t tfb_refcnt;
153 	uint32_t  tfb_flags;
154 };
155 
156 struct tcp_function {
157 	TAILQ_ENTRY(tcp_function) tf_next;
158 	struct tcp_function_block *tf_fb;
159 };
160 
161 TAILQ_HEAD(tcp_funchead, tcp_function);
162 
163 /*
164  * Tcp control block, one per tcp; fields:
165  * Organized for 16 byte cacheline efficiency.
166  */
167 struct tcpcb {
168 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
169 	void	*t_pspare[2];		/* new reassembly queue */
170 	int	t_segqlen;		/* segment reassembly queue length */
171 	int	t_dupacks;		/* consecutive dup acks recd */
172 
173 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
174 
175 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
176 	int	t_state;		/* state of this connection */
177 	u_int	t_flags;
178 
179 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
180 
181 	tcp_seq	snd_una;		/* sent but unacknowledged */
182 	tcp_seq	snd_max;		/* highest sequence number sent;
183 					 * used to recognize retransmits
184 					 */
185 	tcp_seq	snd_nxt;		/* send next */
186 	tcp_seq	snd_up;			/* send urgent pointer */
187 
188 	tcp_seq	snd_wl1;		/* window update seg seq number */
189 	tcp_seq	snd_wl2;		/* window update seg ack number */
190 	tcp_seq	iss;			/* initial send sequence number */
191 	tcp_seq	irs;			/* initial receive sequence number */
192 
193 	tcp_seq	rcv_nxt;		/* receive next */
194 	tcp_seq	rcv_adv;		/* advertised window */
195 	uint32_t  rcv_wnd;		/* receive window */
196 	tcp_seq	rcv_up;			/* receive urgent pointer */
197 
198 	uint32_t  snd_wnd;		/* send window */
199 	uint32_t  snd_cwnd;		/* congestion-controlled window */
200 	u_long	snd_spare1;		/* unused */
201 	uint32_t  snd_ssthresh;		/* snd_cwnd size threshold for
202 					 * for slow start exponential to
203 					 * linear switch
204 					 */
205 	u_long	snd_spare2;		/* unused */
206 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
207 
208 	u_int	t_rcvtime;		/* inactivity time */
209 	u_int	t_starttime;		/* time connection was established */
210 	u_int	t_rtttime;		/* RTT measurement start time */
211 	tcp_seq	t_rtseq;		/* sequence number being timed */
212 
213 	u_int	t_bw_spare1;		/* unused */
214 	tcp_seq	t_bw_spare2;		/* unused */
215 
216 	int	t_rxtcur;		/* current retransmit value (ticks) */
217 	u_int	t_maxseg;		/* maximum segment size */
218 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
219 	int	t_srtt;			/* smoothed round-trip time */
220 	int	t_rttvar;		/* variance in round-trip time */
221 
222 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
223 	u_int	t_rttmin;		/* minimum rtt allowed */
224 	u_int	t_rttbest;		/* best rtt we've seen */
225 	u_long	t_rttupdated;		/* number of times rtt sampled */
226 	uint32_t  max_sndwnd;		/* largest window peer has offered */
227 
228 	int	t_softerror;		/* possible error not yet reported */
229 /* out-of-band data */
230 	char	t_oobflags;		/* have some */
231 	char	t_iobc;			/* input character */
232 /* RFC 1323 variables */
233 	u_char	snd_scale;		/* window scaling for send window */
234 	u_char	rcv_scale;		/* window scaling for recv window */
235 	u_char	request_r_scale;	/* pending window scaling */
236 	u_int32_t  ts_recent;		/* timestamp echo data */
237 	u_int	ts_recent_age;		/* when last updated */
238 	u_int32_t  ts_offset;		/* our timestamp offset */
239 
240 	tcp_seq	last_ack_sent;
241 /* experimental */
242 	uint32_t  snd_cwnd_prev;	/* cwnd prior to retransmit */
243 	uint32_t  snd_ssthresh_prev;	/* ssthresh prior to retransmit */
244 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
245 	int	t_sndzerowin;		/* zero-window updates sent */
246 	u_int	t_badrxtwin;		/* window for retransmit recovery */
247 	u_char	snd_limited;		/* segments limited transmitted */
248 /* SACK related state */
249 	int	snd_numholes;		/* number of holes seen by sender */
250 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
251 					/* SACK scoreboard (sorted) */
252 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
253 	int	rcv_numsacks;		/* # distinct sack blks present */
254 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
255 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
256 					   episode starts at this seq number */
257 	struct sackhint	sackhint;	/* SACK scoreboard hint */
258 	int	t_rttlow;		/* smallest observerved RTT */
259 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
260 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
261 	struct toedev	*tod;		/* toedev handling this connection */
262 	int	t_sndrexmitpack;	/* retransmit packets sent */
263 	int	t_rcvoopack;		/* out-of-order packets received */
264 	void	*t_toe;			/* TOE pcb pointer */
265 	int	t_bytes_acked;		/* # bytes acked during current RTT */
266 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
267 	struct cc_var	*ccv;		/* congestion control specific vars */
268 	struct osd	*osd;		/* storage for Khelp module data */
269 
270 	u_int	t_keepinit;		/* time to establish connection */
271 	u_int	t_keepidle;		/* time before keepalive probes begin */
272 	u_int	t_keepintvl;		/* interval between keepalives */
273 	u_int	t_keepcnt;		/* number of keepalives before close */
274 
275 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
276 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
277 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
278 	u_int	t_flags2;		/* More tcpcb flags storage */
279 #if defined(_KERNEL) && defined(TCP_RFC7413)
280 	uint32_t t_ispare[6];		/* 5 UTO, 1 TBD */
281 	uint64_t t_tfo_cookie;		/* TCP Fast Open cookie */
282 #else
283 	uint32_t t_ispare[8];		/* 5 UTO, 3 TBD */
284 #endif
285 	struct tcp_function_block *t_fb;/* TCP function call block */
286 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
287 #if defined(_KERNEL) && defined(TCP_RFC7413)
288 	unsigned int *t_tfo_pending;	/* TCP Fast Open pending counter */
289 	void	*t_pspare2[1];		/* 1 TCP_SIGNATURE */
290 #else
291 	void	*t_pspare2[2];		/* 1 TCP_SIGNATURE, 1 TBD */
292 #endif
293 #if defined(_KERNEL) && defined(TCPPCAP)
294 	struct mbufq t_inpkts;		/* List of saved input packets. */
295 	struct mbufq t_outpkts;		/* List of saved output packets. */
296 #ifdef _LP64
297 	uint64_t _pad[0];		/* all used! */
298 #else
299 	uint64_t _pad[2];		/* 2 are available */
300 #endif /* _LP64 */
301 #else
302 	uint64_t _pad[6];
303 #endif /* defined(_KERNEL) && defined(TCPPCAP) */
304 };
305 
306 /*
307  * Flags and utility macros for the t_flags field.
308  */
309 #define	TF_ACKNOW	0x000001	/* ack peer immediately */
310 #define	TF_DELACK	0x000002	/* ack, but try to delay it */
311 #define	TF_NODELAY	0x000004	/* don't delay packets to coalesce */
312 #define	TF_NOOPT	0x000008	/* don't use tcp options */
313 #define	TF_SENTFIN	0x000010	/* have sent FIN */
314 #define	TF_REQ_SCALE	0x000020	/* have/will request window scaling */
315 #define	TF_RCVD_SCALE	0x000040	/* other side has requested scaling */
316 #define	TF_REQ_TSTMP	0x000080	/* have/will request timestamps */
317 #define	TF_RCVD_TSTMP	0x000100	/* a timestamp was received in SYN */
318 #define	TF_SACK_PERMIT	0x000200	/* other side said I could SACK */
319 #define	TF_NEEDSYN	0x000400	/* send SYN (implicit state) */
320 #define	TF_NEEDFIN	0x000800	/* send FIN (implicit state) */
321 #define	TF_NOPUSH	0x001000	/* don't push */
322 #define	TF_PREVVALID	0x002000	/* saved values for bad rxmit valid */
323 #define	TF_MORETOCOME	0x010000	/* More data to be appended to sock */
324 #define	TF_LQ_OVERFLOW	0x020000	/* listen queue overflow */
325 #define	TF_LASTIDLE	0x040000	/* connection was previously idle */
326 #define	TF_RXWIN0SENT	0x080000	/* sent a receiver win 0 in response */
327 #define	TF_FASTRECOVERY	0x100000	/* in NewReno Fast Recovery */
328 #define	TF_WASFRECOVERY	0x200000	/* was in NewReno Fast Recovery */
329 #define	TF_SIGNATURE	0x400000	/* require MD5 digests (RFC2385) */
330 #define	TF_FORCEDATA	0x800000	/* force out a byte */
331 #define	TF_TSO		0x1000000	/* TSO enabled on this connection */
332 #define	TF_TOE		0x2000000	/* this connection is offloaded */
333 #define	TF_ECN_PERMIT	0x4000000	/* connection ECN-ready */
334 #define	TF_ECN_SND_CWR	0x8000000	/* ECN CWR in queue */
335 #define	TF_ECN_SND_ECE	0x10000000	/* ECN ECE in queue */
336 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
337 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
338 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
339 
340 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
341 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
342 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
343 
344 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
345 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
346 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
347 
348 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
349 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
350 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
351 
352 #if defined(_KERNEL) && !defined(TCP_RFC7413)
353 #define	IS_FASTOPEN(t_flags)		(false)
354 #else
355 #define	IS_FASTOPEN(t_flags)		(t_flags & TF_FASTOPEN)
356 #endif
357 
358 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
359 
360 /*
361  * Flags for the t_oobflags field.
362  */
363 #define	TCPOOB_HAVEDATA	0x01
364 #define	TCPOOB_HADDATA	0x02
365 
366 /*
367  * Flags for PLPMTU handling, t_flags2
368  */
369 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
370 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
371 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
372 
373 /*
374  * Structure to hold TCP options that are only used during segment
375  * processing (in tcp_input), but not held in the tcpcb.
376  * It's basically used to reduce the number of parameters
377  * to tcp_dooptions and tcp_addoptions.
378  * The binary order of the to_flags is relevant for packing of the
379  * options in tcp_addoptions.
380  */
381 struct tcpopt {
382 	u_int32_t	to_flags;	/* which options are present */
383 #define	TOF_MSS		0x0001		/* maximum segment size */
384 #define	TOF_SCALE	0x0002		/* window scaling */
385 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
386 #define	TOF_TS		0x0010		/* timestamp */
387 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
388 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
389 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
390 #define	TOF_MAXOPT	0x0200
391 	u_int32_t	to_tsval;	/* new timestamp */
392 	u_int32_t	to_tsecr;	/* reflected timestamp */
393 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
394 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
395 	u_char		*to_tfo_cookie; /* pointer to the TFO cookie */
396 	u_int16_t	to_mss;		/* maximum segment size */
397 	u_int8_t	to_wscale;	/* window scaling */
398 	u_int8_t	to_nsacks;	/* number of SACK blocks */
399 	u_int8_t	to_tfo_len;	/* TFO cookie length */
400 	u_int32_t	to_spare;	/* UTO */
401 };
402 
403 /*
404  * Flags for tcp_dooptions.
405  */
406 #define	TO_SYN		0x01		/* parse SYN-only options */
407 
408 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
409 	uint32_t	rmx_mtu;	/* MTU for this path */
410 	uint32_t	rmx_ssthresh;	/* outbound gateway buffer limit */
411 	uint32_t	rmx_rtt;	/* estimated round trip time */
412 	uint32_t	rmx_rttvar;	/* estimated rtt variance */
413 	uint32_t	rmx_cwnd;	/* congestion window */
414 	uint32_t	rmx_sendpipe;   /* outbound delay-bandwidth product */
415 	uint32_t	rmx_recvpipe;   /* inbound delay-bandwidth product */
416 };
417 
418 /*
419  * Used by tcp_maxmtu() to communicate interface specific features
420  * and limits at the time of connection setup.
421  */
422 struct tcp_ifcap {
423 	int	ifcap;
424 	u_int	tsomax;
425 	u_int	tsomaxsegcount;
426 	u_int	tsomaxsegsize;
427 };
428 
429 #ifndef _NETINET_IN_PCB_H_
430 struct in_conninfo;
431 #endif /* _NETINET_IN_PCB_H_ */
432 
433 struct tcptw {
434 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
435 	tcp_seq		snd_nxt;
436 	tcp_seq		rcv_nxt;
437 	tcp_seq		iss;
438 	tcp_seq		irs;
439 	u_short		last_win;	/* cached window value */
440 	short		tw_so_options;	/* copy of so_options */
441 	struct ucred	*tw_cred;	/* user credentials */
442 	u_int32_t	t_recent;
443 	u_int32_t	ts_offset;	/* our timestamp offset */
444 	u_int		t_starttime;
445 	int		tw_time;
446 	TAILQ_ENTRY(tcptw) tw_2msl;
447 	void		*tw_pspare;	/* TCP_SIGNATURE */
448 	u_int		*tw_spare;	/* TCP_SIGNATURE */
449 };
450 
451 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
452 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
453 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
454 
455 /*
456  * The smoothed round-trip time and estimated variance
457  * are stored as fixed point numbers scaled by the values below.
458  * For convenience, these scales are also used in smoothing the average
459  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
460  * With these scales, srtt has 3 bits to the right of the binary point,
461  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
462  * binary point, and is smoothed with an ALPHA of 0.75.
463  */
464 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
465 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
466 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
467 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
468 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
469 
470 /*
471  * The initial retransmission should happen at rtt + 4 * rttvar.
472  * Because of the way we do the smoothing, srtt and rttvar
473  * will each average +1/2 tick of bias.  When we compute
474  * the retransmit timer, we want 1/2 tick of rounding and
475  * 1 extra tick because of +-1/2 tick uncertainty in the
476  * firing of the timer.  The bias will give us exactly the
477  * 1.5 tick we need.  But, because the bias is
478  * statistical, we have to test that we don't drop below
479  * the minimum feasible timer (which is 2 ticks).
480  * This version of the macro adapted from a paper by Lawrence
481  * Brakmo and Larry Peterson which outlines a problem caused
482  * by insufficient precision in the original implementation,
483  * which results in inappropriately large RTO values for very
484  * fast networks.
485  */
486 #define	TCP_REXMTVAL(tp) \
487 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
488 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
489 
490 /*
491  * TCP statistics.
492  * Many of these should be kept per connection,
493  * but that's inconvenient at the moment.
494  */
495 struct	tcpstat {
496 	uint64_t tcps_connattempt;	/* connections initiated */
497 	uint64_t tcps_accepts;		/* connections accepted */
498 	uint64_t tcps_connects;		/* connections established */
499 	uint64_t tcps_drops;		/* connections dropped */
500 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
501 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
502 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
503 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
504 	uint64_t tcps_rttupdated;	/* times we succeeded */
505 	uint64_t tcps_delack;		/* delayed acks sent */
506 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
507 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
508 	uint64_t tcps_persisttimeo;	/* persist timeouts */
509 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
510 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
511 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
512 
513 	uint64_t tcps_sndtotal;		/* total packets sent */
514 	uint64_t tcps_sndpack;		/* data packets sent */
515 	uint64_t tcps_sndbyte;		/* data bytes sent */
516 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
517 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
518 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
519 	uint64_t tcps_sndacks;		/* ack-only packets sent */
520 	uint64_t tcps_sndprobe;		/* window probes sent */
521 	uint64_t tcps_sndurg;		/* packets sent with URG only */
522 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
523 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
524 
525 	uint64_t tcps_rcvtotal;		/* total packets received */
526 	uint64_t tcps_rcvpack;		/* packets received in sequence */
527 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
528 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
529 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
530 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
531 	uint64_t tcps_rcvshort;		/* packets received too short */
532 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
533 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
534 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
535 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
536 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
537 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
538 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
539 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
540 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
541 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
542 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
543 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
544 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
545 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
546 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
547 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
548 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
549 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
550 	uint64_t tcps_pcbcachemiss;
551 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
552 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
553 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
554 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
555 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
556 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
557 	uint64_t tcps_persistdrop;	/* timeout in persist state */
558 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
559 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
560 	uint64_t tcps_listendrop;	/* listen queue overflows */
561 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
562 
563 	uint64_t tcps_sc_added;		/* entry added to syncache */
564 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
565 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
566 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
567 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
568 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
569 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
570 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
571 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
572 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
573 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
574 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
575 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
576 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
577 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
578 
579 	uint64_t tcps_hc_added;		/* entry added to hostcache */
580 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
581 
582 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
583 
584 	/* SACK related stats */
585 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
586 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
587 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
588 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
589 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
590 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
591 
592 	/* ECN related stats */
593 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
594 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
595 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
596 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
597 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
598 
599 	/* TCP_SIGNATURE related stats */
600 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
601 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
602 	uint64_t tcps_sig_err_buildsig;	/* Failed to make signature */
603 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
604 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
605 
606 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
607 };
608 
609 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
610 
611 #ifdef _KERNEL
612 #define	TI_UNLOCKED	1
613 #define	TI_RLOCKED	2
614 #include <sys/counter.h>
615 
616 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
617 /*
618  * In-kernel consumers can use these accessor macros directly to update
619  * stats.
620  */
621 #define	TCPSTAT_ADD(name, val)	\
622     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
623 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
624 
625 /*
626  * Kernel module consumers must use this accessor macro.
627  */
628 void	kmod_tcpstat_inc(int statnum);
629 #define	KMOD_TCPSTAT_INC(name)						\
630     kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
631 
632 /*
633  * Running TCP connection count by state.
634  */
635 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
636 #define	V_tcps_states	VNET(tcps_states)
637 #define	TCPSTATES_INC(state)	counter_u64_add(V_tcps_states[state], 1)
638 #define	TCPSTATES_DEC(state)	counter_u64_add(V_tcps_states[state], -1)
639 
640 /*
641  * TCP specific helper hook point identifiers.
642  */
643 #define	HHOOK_TCP_EST_IN		0
644 #define	HHOOK_TCP_EST_OUT		1
645 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
646 
647 struct tcp_hhook_data {
648 	struct tcpcb	*tp;
649 	struct tcphdr	*th;
650 	struct tcpopt	*to;
651 	uint32_t	len;
652 	int		tso;
653 	tcp_seq		curack;
654 };
655 #endif
656 
657 /*
658  * TCB structure exported to user-land via sysctl(3).
659  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
660  * included.  Not all of our clients do.
661  */
662 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
663 struct xtcp_timer {
664 	int tt_rexmt;	/* retransmit timer */
665 	int tt_persist;	/* retransmit persistence */
666 	int tt_keep;	/* keepalive */
667 	int tt_2msl;	/* 2*msl TIME_WAIT timer */
668 	int tt_delack;	/* delayed ACK timer */
669 	int t_rcvtime;	/* Time since last packet received */
670 };
671 struct	xtcpcb {
672 	size_t	xt_len;
673 	struct	inpcb	xt_inp;
674 	struct	tcpcb	xt_tp;
675 	struct	xsocket	xt_socket;
676 	struct	xtcp_timer xt_timer;
677 	u_quad_t	xt_alignment_hack;
678 };
679 #endif
680 
681 /*
682  * Identifiers for TCP sysctl nodes
683  */
684 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
685 #define	TCPCTL_MSSDFLT		3	/* MSS default */
686 #define TCPCTL_STATS		4	/* statistics */
687 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
688 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
689 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
690 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
691 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
692 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
693 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
694 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
695 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
696 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
697 #define	TCPCTL_DROP		15	/* drop tcp connection */
698 #define	TCPCTL_STATES		16	/* connection counts by TCP state */
699 
700 #ifdef _KERNEL
701 #ifdef SYSCTL_DECL
702 SYSCTL_DECL(_net_inet_tcp);
703 SYSCTL_DECL(_net_inet_tcp_sack);
704 MALLOC_DECLARE(M_TCPLOG);
705 #endif
706 
707 VNET_DECLARE(struct inpcbhead, tcb);		/* queue of active tcpcb's */
708 VNET_DECLARE(struct inpcbinfo, tcbinfo);
709 extern	int tcp_log_in_vain;
710 VNET_DECLARE(int, tcp_mssdflt);	/* XXX */
711 VNET_DECLARE(int, tcp_minmss);
712 VNET_DECLARE(int, tcp_delack_enabled);
713 VNET_DECLARE(int, tcp_do_rfc3390);
714 VNET_DECLARE(int, tcp_initcwnd_segments);
715 VNET_DECLARE(int, tcp_sendspace);
716 VNET_DECLARE(int, tcp_recvspace);
717 VNET_DECLARE(int, path_mtu_discovery);
718 VNET_DECLARE(int, tcp_do_rfc3465);
719 VNET_DECLARE(int, tcp_abc_l_var);
720 #define	V_tcb			VNET(tcb)
721 #define	V_tcbinfo		VNET(tcbinfo)
722 #define	V_tcp_mssdflt		VNET(tcp_mssdflt)
723 #define	V_tcp_minmss		VNET(tcp_minmss)
724 #define	V_tcp_delack_enabled	VNET(tcp_delack_enabled)
725 #define	V_tcp_do_rfc3390	VNET(tcp_do_rfc3390)
726 #define	V_tcp_initcwnd_segments	VNET(tcp_initcwnd_segments)
727 #define	V_tcp_sendspace		VNET(tcp_sendspace)
728 #define	V_tcp_recvspace		VNET(tcp_recvspace)
729 #define	V_path_mtu_discovery	VNET(path_mtu_discovery)
730 #define	V_tcp_do_rfc3465	VNET(tcp_do_rfc3465)
731 #define	V_tcp_abc_l_var		VNET(tcp_abc_l_var)
732 
733 VNET_DECLARE(int, tcp_do_sack);			/* SACK enabled/disabled */
734 VNET_DECLARE(int, tcp_sc_rst_sock_fail);	/* RST on sock alloc failure */
735 #define	V_tcp_do_sack		VNET(tcp_do_sack)
736 #define	V_tcp_sc_rst_sock_fail	VNET(tcp_sc_rst_sock_fail)
737 
738 VNET_DECLARE(int, tcp_do_ecn);			/* TCP ECN enabled/disabled */
739 VNET_DECLARE(int, tcp_ecn_maxretries);
740 #define	V_tcp_do_ecn		VNET(tcp_do_ecn)
741 #define	V_tcp_ecn_maxretries	VNET(tcp_ecn_maxretries)
742 
743 #ifdef TCP_HHOOK
744 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
745 #define	V_tcp_hhh		VNET(tcp_hhh)
746 #endif
747 
748 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
749 #define V_tcp_do_rfc6675_pipe	VNET(tcp_do_rfc6675_pipe)
750 
751 int	 tcp_addoptions(struct tcpopt *, u_char *);
752 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
753 struct tcpcb *
754 	 tcp_close(struct tcpcb *);
755 void	 tcp_discardcb(struct tcpcb *);
756 void	 tcp_twstart(struct tcpcb *);
757 void	 tcp_twclose(struct tcptw *, int);
758 void	 tcp_ctlinput(int, struct sockaddr *, void *);
759 int	 tcp_ctloutput(struct socket *, struct sockopt *);
760 struct tcpcb *
761 	 tcp_drop(struct tcpcb *, int);
762 void	 tcp_drain(void);
763 void	 tcp_init(void);
764 void	 tcp_fini(void *);
765 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
766 	    const void *);
767 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
768 	    const void *);
769 int	 tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *);
770 void	 tcp_reass_global_init(void);
771 void	 tcp_reass_flush(struct tcpcb *);
772 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
773 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
774 		     struct tcpcb *, int, int);
775 void	tcp_pulloutofband(struct socket *,
776 		     struct tcphdr *, struct mbuf *, int);
777 void	tcp_xmit_timer(struct tcpcb *, int);
778 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
779 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
780 			    uint16_t nsegs, uint16_t type);
781 void 	cc_conn_init(struct tcpcb *tp);
782 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
783 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
784 #ifdef TCP_HHOOK
785 void	hhook_run_tcp_est_in(struct tcpcb *tp,
786 			    struct tcphdr *th, struct tcpopt *to);
787 #endif
788 
789 int	 tcp_input(struct mbuf **, int *, int);
790 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
791 			struct socket *, struct tcpcb *, int, int, uint8_t,
792 			int);
793 
794 int register_tcp_functions(struct tcp_function_block *blk, int wait);
795 int deregister_tcp_functions(struct tcp_function_block *blk);
796 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
797 struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk);
798 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
799 
800 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
801 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
802 u_int	 tcp_maxseg(const struct tcpcb *);
803 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
804 	    struct tcp_ifcap *);
805 void	 tcp_mss(struct tcpcb *, int);
806 int	 tcp_mssopt(struct in_conninfo *);
807 struct inpcb *
808 	 tcp_drop_syn_sent(struct inpcb *, int);
809 struct tcpcb *
810 	 tcp_newtcpcb(struct inpcb *);
811 int	 tcp_output(struct tcpcb *);
812 void	 tcp_state_change(struct tcpcb *, int);
813 void	 tcp_respond(struct tcpcb *, void *,
814 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
815 void	 tcp_tw_init(void);
816 #ifdef VIMAGE
817 void	 tcp_tw_destroy(void);
818 #endif
819 void	 tcp_tw_zone_change(void);
820 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
821 	    struct mbuf *, int);
822 void	 tcp_setpersist(struct tcpcb *);
823 void	 tcp_slowtimo(void);
824 struct tcptemp *
825 	 tcpip_maketemplate(struct inpcb *);
826 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
827 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
828 int	 tcp_timer_active(struct tcpcb *, uint32_t);
829 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
830 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
831 /*
832  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
833  */
834 void	 tcp_hc_init(void);
835 #ifdef VIMAGE
836 void	 tcp_hc_destroy(void);
837 #endif
838 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
839 uint32_t tcp_hc_getmtu(struct in_conninfo *);
840 void	 tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
841 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
842 
843 extern	struct pr_usrreqs tcp_usrreqs;
844 tcp_seq tcp_new_isn(struct tcpcb *);
845 
846 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
847 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
848 void	 tcp_clean_sackreport(struct tcpcb *tp);
849 void	 tcp_sack_adjust(struct tcpcb *tp);
850 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
851 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
852 void	 tcp_free_sackholes(struct tcpcb *tp);
853 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
854 int	 tcp_compute_pipe(struct tcpcb *);
855 
856 static inline void
857 tcp_fields_to_host(struct tcphdr *th)
858 {
859 
860 	th->th_seq = ntohl(th->th_seq);
861 	th->th_ack = ntohl(th->th_ack);
862 	th->th_win = ntohs(th->th_win);
863 	th->th_urp = ntohs(th->th_urp);
864 }
865 
866 static inline void
867 tcp_fields_to_net(struct tcphdr *th)
868 {
869 
870 	th->th_seq = htonl(th->th_seq);
871 	th->th_ack = htonl(th->th_ack);
872 	th->th_win = htons(th->th_win);
873 	th->th_urp = htons(th->th_urp);
874 }
875 #endif /* _KERNEL */
876 
877 #endif /* _NETINET_TCP_VAR_H_ */
878