xref: /freebsd/sys/netinet/tcp_var.h (revision b6a05070fa77edc7ce6e60b61623fd806e807be6)
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 
38 #ifdef _KERNEL
39 #include <net/vnet.h>
40 #include <sys/mbuf.h>
41 
42 /*
43  * Kernel variables for tcp.
44  */
45 VNET_DECLARE(int, tcp_do_rfc1323);
46 #define	V_tcp_do_rfc1323	VNET(tcp_do_rfc1323)
47 
48 #endif /* _KERNEL */
49 
50 /* TCP segment queue entry */
51 struct tseg_qent {
52 	LIST_ENTRY(tseg_qent) tqe_q;
53 	int	tqe_len;		/* TCP segment data length */
54 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
55 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
56 };
57 LIST_HEAD(tsegqe_head, tseg_qent);
58 
59 struct sackblk {
60 	tcp_seq start;		/* start seq no. of sack block */
61 	tcp_seq end;		/* end seq no. */
62 };
63 
64 struct sackhole {
65 	tcp_seq start;		/* start seq no. of hole */
66 	tcp_seq end;		/* end seq no. */
67 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
68 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
69 };
70 
71 struct sackhint {
72 	struct sackhole	*nexthole;
73 	int		sack_bytes_rexmit;
74 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
75 
76 	int		ispare;		/* explicit pad for 64bit alignment */
77 	int             sacked_bytes;	/*
78 					 * Total sacked bytes reported by the
79 					 * receiver via sack option
80 					 */
81 	uint32_t	_pad1[1];	/* TBD */
82 	uint64_t	_pad[1];	/* TBD */
83 };
84 
85 struct tcptemp {
86 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
87 	struct	tcphdr tt_t;
88 };
89 
90 #define tcp6cb		tcpcb  /* for KAME src sync over BSD*'s */
91 
92 /*
93  * TODO: We yet need to brave plowing in
94  * to tcp_input() and the pru_usrreq() block.
95  * Right now these go to the old standards which
96  * are somewhat ok, but in the long term may
97  * need to be changed. If we do tackle tcp_input()
98  * then we need to get rid of the tcp_do_segment()
99  * function below.
100  */
101 /* Flags for tcp functions */
102 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
103 struct tcpcb;
104 struct inpcb;
105 struct sockopt;
106 struct socket;
107 
108 struct tcp_function_block {
109 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
110 	int	(*tfb_tcp_output)(struct tcpcb *);
111 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
112 			    struct socket *, struct tcpcb *,
113 			    int, int, uint8_t,
114 			    int);
115 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
116 			    struct inpcb *inp, struct tcpcb *tp);
117 	/* Optional memory allocation/free routine */
118 	void	(*tfb_tcp_fb_init)(struct tcpcb *);
119 	void	(*tfb_tcp_fb_fini)(struct tcpcb *);
120 	/* Optional timers, must define all if you define one */
121 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
122 	int	(*tfb_tcp_timers_left)(struct tcpcb *);
123 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
124 			    uint32_t, u_int);
125 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
126 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
127 	volatile uint32_t tfb_refcnt;
128 	uint32_t  tfb_flags;
129 };
130 
131 struct tcp_function {
132 	TAILQ_ENTRY(tcp_function) tf_next;
133 	struct tcp_function_block *tf_fb;
134 };
135 
136 TAILQ_HEAD(tcp_funchead, tcp_function);
137 
138 /*
139  * Tcp control block, one per tcp; fields:
140  * Organized for 16 byte cacheline efficiency.
141  */
142 struct tcpcb {
143 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
144 	void	*t_pspare[2];		/* new reassembly queue */
145 	int	t_segqlen;		/* segment reassembly queue length */
146 	int	t_dupacks;		/* consecutive dup acks recd */
147 
148 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
149 
150 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
151 	int	t_state;		/* state of this connection */
152 	u_int	t_flags;
153 
154 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
155 
156 	tcp_seq	snd_una;		/* sent but unacknowledged */
157 	tcp_seq	snd_max;		/* highest sequence number sent;
158 					 * used to recognize retransmits
159 					 */
160 	tcp_seq	snd_nxt;		/* send next */
161 	tcp_seq	snd_up;			/* send urgent pointer */
162 
163 	tcp_seq	snd_wl1;		/* window update seg seq number */
164 	tcp_seq	snd_wl2;		/* window update seg ack number */
165 	tcp_seq	iss;			/* initial send sequence number */
166 	tcp_seq	irs;			/* initial receive sequence number */
167 
168 	tcp_seq	rcv_nxt;		/* receive next */
169 	tcp_seq	rcv_adv;		/* advertised window */
170 	u_long	rcv_wnd;		/* receive window */
171 	tcp_seq	rcv_up;			/* receive urgent pointer */
172 
173 	u_long	snd_wnd;		/* send window */
174 	u_long	snd_cwnd;		/* congestion-controlled window */
175 	u_long	snd_spare1;		/* unused */
176 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
177 					 * for slow start exponential to
178 					 * linear switch
179 					 */
180 	u_long	snd_spare2;		/* unused */
181 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
182 
183 	u_int	t_maxopd;		/* mss plus options */
184 
185 	u_int	t_rcvtime;		/* inactivity time */
186 	u_int	t_starttime;		/* time connection was established */
187 	u_int	t_rtttime;		/* RTT measurement start time */
188 	tcp_seq	t_rtseq;		/* sequence number being timed */
189 
190 	u_int	t_bw_spare1;		/* unused */
191 	tcp_seq	t_bw_spare2;		/* unused */
192 
193 	int	t_rxtcur;		/* current retransmit value (ticks) */
194 	u_int	t_maxseg;		/* maximum segment size */
195 	int	t_srtt;			/* smoothed round-trip time */
196 	int	t_rttvar;		/* variance in round-trip time */
197 
198 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
199 	u_int	t_rttmin;		/* minimum rtt allowed */
200 	u_int	t_rttbest;		/* best rtt we've seen */
201 	u_long	t_rttupdated;		/* number of times rtt sampled */
202 	u_long	max_sndwnd;		/* largest window peer has offered */
203 
204 	int	t_softerror;		/* possible error not yet reported */
205 /* out-of-band data */
206 	char	t_oobflags;		/* have some */
207 	char	t_iobc;			/* input character */
208 /* RFC 1323 variables */
209 	u_char	snd_scale;		/* window scaling for send window */
210 	u_char	rcv_scale;		/* window scaling for recv window */
211 	u_char	request_r_scale;	/* pending window scaling */
212 	u_int32_t  ts_recent;		/* timestamp echo data */
213 	u_int	ts_recent_age;		/* when last updated */
214 	u_int32_t  ts_offset;		/* our timestamp offset */
215 
216 	tcp_seq	last_ack_sent;
217 /* experimental */
218 	u_long	snd_cwnd_prev;		/* cwnd prior to retransmit */
219 	u_long	snd_ssthresh_prev;	/* ssthresh prior to retransmit */
220 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
221 	int	t_sndzerowin;		/* zero-window updates sent */
222 	u_int	t_badrxtwin;		/* window for retransmit recovery */
223 	u_char	snd_limited;		/* segments limited transmitted */
224 /* SACK related state */
225 	int	snd_numholes;		/* number of holes seen by sender */
226 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
227 					/* SACK scoreboard (sorted) */
228 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
229 	int	rcv_numsacks;		/* # distinct sack blks present */
230 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
231 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
232 					   episode starts at this seq number */
233 	struct sackhint	sackhint;	/* SACK scoreboard hint */
234 	int	t_rttlow;		/* smallest observerved RTT */
235 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
236 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
237 	struct toedev	*tod;		/* toedev handling this connection */
238 	int	t_sndrexmitpack;	/* retransmit packets sent */
239 	int	t_rcvoopack;		/* out-of-order packets received */
240 	void	*t_toe;			/* TOE pcb pointer */
241 	int	t_bytes_acked;		/* # bytes acked during current RTT */
242 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
243 	struct cc_var	*ccv;		/* congestion control specific vars */
244 	struct osd	*osd;		/* storage for Khelp module data */
245 
246 	u_int	t_keepinit;		/* time to establish connection */
247 	u_int	t_keepidle;		/* time before keepalive probes begin */
248 	u_int	t_keepintvl;		/* interval between keepalives */
249 	u_int	t_keepcnt;		/* number of keepalives before close */
250 
251 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
252 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
253 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
254 	u_int	t_pmtud_saved_maxopd;	/* pre-blackhole MSS */
255 	u_int	t_flags2;		/* More tcpcb flags storage */
256 	uint32_t t_ispare[8];		/* 5 UTO, 3 TBD */
257 	struct tcp_function_block *t_fb;/* TCP function call block */
258 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
259 	void	*t_pspare2[2];		/* 1 TCP_SIGNATURE, 1 TBD */
260 #if defined(_KERNEL) && defined(TCPPCAP)
261 	struct mbufq t_inpkts;		/* List of saved input packets. */
262 	struct mbufq t_outpkts;		/* List of saved output packets. */
263 #ifdef _LP64
264 	uint64_t _pad[0];		/* all used! */
265 #else
266 	uint64_t _pad[2];		/* 2 are available */
267 #endif /* _LP64 */
268 #else
269 	uint64_t _pad[6];
270 #endif /* defined(_KERNEL) && defined(TCPPCAP) */
271 };
272 
273 /*
274  * Flags and utility macros for the t_flags field.
275  */
276 #define	TF_ACKNOW	0x000001	/* ack peer immediately */
277 #define	TF_DELACK	0x000002	/* ack, but try to delay it */
278 #define	TF_NODELAY	0x000004	/* don't delay packets to coalesce */
279 #define	TF_NOOPT	0x000008	/* don't use tcp options */
280 #define	TF_SENTFIN	0x000010	/* have sent FIN */
281 #define	TF_REQ_SCALE	0x000020	/* have/will request window scaling */
282 #define	TF_RCVD_SCALE	0x000040	/* other side has requested scaling */
283 #define	TF_REQ_TSTMP	0x000080	/* have/will request timestamps */
284 #define	TF_RCVD_TSTMP	0x000100	/* a timestamp was received in SYN */
285 #define	TF_SACK_PERMIT	0x000200	/* other side said I could SACK */
286 #define	TF_NEEDSYN	0x000400	/* send SYN (implicit state) */
287 #define	TF_NEEDFIN	0x000800	/* send FIN (implicit state) */
288 #define	TF_NOPUSH	0x001000	/* don't push */
289 #define	TF_PREVVALID	0x002000	/* saved values for bad rxmit valid */
290 #define	TF_MORETOCOME	0x010000	/* More data to be appended to sock */
291 #define	TF_LQ_OVERFLOW	0x020000	/* listen queue overflow */
292 #define	TF_LASTIDLE	0x040000	/* connection was previously idle */
293 #define	TF_RXWIN0SENT	0x080000	/* sent a receiver win 0 in response */
294 #define	TF_FASTRECOVERY	0x100000	/* in NewReno Fast Recovery */
295 #define	TF_WASFRECOVERY	0x200000	/* was in NewReno Fast Recovery */
296 #define	TF_SIGNATURE	0x400000	/* require MD5 digests (RFC2385) */
297 #define	TF_FORCEDATA	0x800000	/* force out a byte */
298 #define	TF_TSO		0x1000000	/* TSO enabled on this connection */
299 #define	TF_TOE		0x2000000	/* this connection is offloaded */
300 #define	TF_ECN_PERMIT	0x4000000	/* connection ECN-ready */
301 #define	TF_ECN_SND_CWR	0x8000000	/* ECN CWR in queue */
302 #define	TF_ECN_SND_ECE	0x10000000	/* ECN ECE in queue */
303 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
304 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
305 
306 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
307 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
308 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
309 
310 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
311 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
312 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
313 
314 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
315 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
316 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
317 
318 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
319 
320 /*
321  * Flags for the t_oobflags field.
322  */
323 #define	TCPOOB_HAVEDATA	0x01
324 #define	TCPOOB_HADDATA	0x02
325 
326 #ifdef TCP_SIGNATURE
327 /*
328  * Defines which are needed by the xform_tcp module and tcp_[in|out]put
329  * for SADB verification and lookup.
330  */
331 #define	TCP_SIGLEN	16	/* length of computed digest in bytes */
332 #define	TCP_KEYLEN_MIN	1	/* minimum length of TCP-MD5 key */
333 #define	TCP_KEYLEN_MAX	80	/* maximum length of TCP-MD5 key */
334 /*
335  * Only a single SA per host may be specified at this time. An SPI is
336  * needed in order for the KEY_ALLOCSA() lookup to work.
337  */
338 #define	TCP_SIG_SPI	0x1000
339 #endif /* TCP_SIGNATURE */
340 
341 /*
342  * Flags for PLPMTU handling, t_flags2
343  */
344 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
345 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
346 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
347 
348 /*
349  * Structure to hold TCP options that are only used during segment
350  * processing (in tcp_input), but not held in the tcpcb.
351  * It's basically used to reduce the number of parameters
352  * to tcp_dooptions and tcp_addoptions.
353  * The binary order of the to_flags is relevant for packing of the
354  * options in tcp_addoptions.
355  */
356 struct tcpopt {
357 	u_int64_t	to_flags;	/* which options are present */
358 #define	TOF_MSS		0x0001		/* maximum segment size */
359 #define	TOF_SCALE	0x0002		/* window scaling */
360 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
361 #define	TOF_TS		0x0010		/* timestamp */
362 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
363 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
364 #define	TOF_MAXOPT	0x0100
365 	u_int32_t	to_tsval;	/* new timestamp */
366 	u_int32_t	to_tsecr;	/* reflected timestamp */
367 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
368 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
369 	u_int16_t	to_mss;		/* maximum segment size */
370 	u_int8_t	to_wscale;	/* window scaling */
371 	u_int8_t	to_nsacks;	/* number of SACK blocks */
372 	u_int32_t	to_spare;	/* UTO */
373 };
374 
375 /*
376  * Flags for tcp_dooptions.
377  */
378 #define	TO_SYN		0x01		/* parse SYN-only options */
379 
380 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
381 	u_long	rmx_mtu;	/* MTU for this path */
382 	u_long	rmx_ssthresh;	/* outbound gateway buffer limit */
383 	u_long	rmx_rtt;	/* estimated round trip time */
384 	u_long	rmx_rttvar;	/* estimated rtt variance */
385 	u_long	rmx_cwnd;	/* congestion window */
386 	u_long	rmx_sendpipe;   /* outbound delay-bandwidth product */
387 	u_long	rmx_recvpipe;   /* inbound delay-bandwidth product */
388 };
389 
390 /*
391  * Used by tcp_maxmtu() to communicate interface specific features
392  * and limits at the time of connection setup.
393  */
394 struct tcp_ifcap {
395 	int	ifcap;
396 	u_int	tsomax;
397 	u_int	tsomaxsegcount;
398 	u_int	tsomaxsegsize;
399 };
400 
401 #ifndef _NETINET_IN_PCB_H_
402 struct in_conninfo;
403 #endif /* _NETINET_IN_PCB_H_ */
404 
405 struct tcptw {
406 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
407 	tcp_seq		snd_nxt;
408 	tcp_seq		rcv_nxt;
409 	tcp_seq		iss;
410 	tcp_seq		irs;
411 	u_short		last_win;	/* cached window value */
412 	u_short		tw_so_options;	/* copy of so_options */
413 	struct ucred	*tw_cred;	/* user credentials */
414 	u_int32_t	t_recent;
415 	u_int32_t	ts_offset;	/* our timestamp offset */
416 	u_int		t_starttime;
417 	int		tw_time;
418 	TAILQ_ENTRY(tcptw) tw_2msl;
419 	void		*tw_pspare;	/* TCP_SIGNATURE */
420 	u_int		*tw_spare;	/* TCP_SIGNATURE */
421 };
422 
423 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
424 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
425 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
426 
427 /*
428  * The smoothed round-trip time and estimated variance
429  * are stored as fixed point numbers scaled by the values below.
430  * For convenience, these scales are also used in smoothing the average
431  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
432  * With these scales, srtt has 3 bits to the right of the binary point,
433  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
434  * binary point, and is smoothed with an ALPHA of 0.75.
435  */
436 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
437 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
438 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
439 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
440 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
441 
442 /*
443  * The initial retransmission should happen at rtt + 4 * rttvar.
444  * Because of the way we do the smoothing, srtt and rttvar
445  * will each average +1/2 tick of bias.  When we compute
446  * the retransmit timer, we want 1/2 tick of rounding and
447  * 1 extra tick because of +-1/2 tick uncertainty in the
448  * firing of the timer.  The bias will give us exactly the
449  * 1.5 tick we need.  But, because the bias is
450  * statistical, we have to test that we don't drop below
451  * the minimum feasible timer (which is 2 ticks).
452  * This version of the macro adapted from a paper by Lawrence
453  * Brakmo and Larry Peterson which outlines a problem caused
454  * by insufficient precision in the original implementation,
455  * which results in inappropriately large RTO values for very
456  * fast networks.
457  */
458 #define	TCP_REXMTVAL(tp) \
459 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
460 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
461 
462 /*
463  * TCP statistics.
464  * Many of these should be kept per connection,
465  * but that's inconvenient at the moment.
466  */
467 struct	tcpstat {
468 	uint64_t tcps_connattempt;	/* connections initiated */
469 	uint64_t tcps_accepts;		/* connections accepted */
470 	uint64_t tcps_connects;		/* connections established */
471 	uint64_t tcps_drops;		/* connections dropped */
472 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
473 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
474 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
475 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
476 	uint64_t tcps_rttupdated;	/* times we succeeded */
477 	uint64_t tcps_delack;		/* delayed acks sent */
478 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
479 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
480 	uint64_t tcps_persisttimeo;	/* persist timeouts */
481 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
482 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
483 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
484 
485 	uint64_t tcps_sndtotal;		/* total packets sent */
486 	uint64_t tcps_sndpack;		/* data packets sent */
487 	uint64_t tcps_sndbyte;		/* data bytes sent */
488 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
489 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
490 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
491 	uint64_t tcps_sndacks;		/* ack-only packets sent */
492 	uint64_t tcps_sndprobe;		/* window probes sent */
493 	uint64_t tcps_sndurg;		/* packets sent with URG only */
494 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
495 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
496 
497 	uint64_t tcps_rcvtotal;		/* total packets received */
498 	uint64_t tcps_rcvpack;		/* packets received in sequence */
499 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
500 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
501 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
502 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
503 	uint64_t tcps_rcvshort;		/* packets received too short */
504 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
505 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
506 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
507 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
508 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
509 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
510 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
511 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
512 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
513 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
514 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
515 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
516 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
517 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
518 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
519 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
520 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
521 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
522 	uint64_t tcps_pcbcachemiss;
523 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
524 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
525 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
526 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
527 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
528 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
529 	uint64_t tcps_persistdrop;	/* timeout in persist state */
530 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
531 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
532 	uint64_t tcps_listendrop;	/* listen queue overflows */
533 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
534 
535 	uint64_t tcps_sc_added;		/* entry added to syncache */
536 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
537 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
538 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
539 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
540 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
541 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
542 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
543 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
544 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
545 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
546 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
547 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
548 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
549 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
550 
551 	uint64_t tcps_hc_added;		/* entry added to hostcache */
552 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
553 
554 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
555 
556 	/* SACK related stats */
557 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
558 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
559 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
560 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
561 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
562 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
563 
564 	/* ECN related stats */
565 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
566 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
567 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
568 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
569 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
570 
571 	/* TCP_SIGNATURE related stats */
572 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
573 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
574 	uint64_t tcps_sig_err_buildsig;	/* Mismatching signature received */
575 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
576 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
577 
578 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
579 };
580 
581 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
582 
583 #ifdef _KERNEL
584 #define	TI_UNLOCKED	1
585 #define	TI_RLOCKED	2
586 #include <sys/counter.h>
587 
588 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
589 /*
590  * In-kernel consumers can use these accessor macros directly to update
591  * stats.
592  */
593 #define	TCPSTAT_ADD(name, val)	\
594     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
595 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
596 
597 /*
598  * Kernel module consumers must use this accessor macro.
599  */
600 void	kmod_tcpstat_inc(int statnum);
601 #define	KMOD_TCPSTAT_INC(name)						\
602     kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
603 
604 /*
605  * TCP specific helper hook point identifiers.
606  */
607 #define	HHOOK_TCP_EST_IN		0
608 #define	HHOOK_TCP_EST_OUT		1
609 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
610 
611 struct tcp_hhook_data {
612 	struct tcpcb	*tp;
613 	struct tcphdr	*th;
614 	struct tcpopt	*to;
615 	long		len;
616 	int		tso;
617 	tcp_seq		curack;
618 };
619 #endif
620 
621 /*
622  * TCB structure exported to user-land via sysctl(3).
623  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
624  * included.  Not all of our clients do.
625  */
626 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
627 struct xtcp_timer {
628 	int tt_rexmt;	/* retransmit timer */
629 	int tt_persist;	/* retransmit persistence */
630 	int tt_keep;	/* keepalive */
631 	int tt_2msl;	/* 2*msl TIME_WAIT timer */
632 	int tt_delack;	/* delayed ACK timer */
633 	int t_rcvtime;	/* Time since last packet received */
634 };
635 struct	xtcpcb {
636 	size_t	xt_len;
637 	struct	inpcb	xt_inp;
638 	struct	tcpcb	xt_tp;
639 	struct	xsocket	xt_socket;
640 	struct	xtcp_timer xt_timer;
641 	u_quad_t	xt_alignment_hack;
642 };
643 #endif
644 
645 /*
646  * Identifiers for TCP sysctl nodes
647  */
648 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
649 #define	TCPCTL_MSSDFLT		3	/* MSS default */
650 #define TCPCTL_STATS		4	/* statistics (read-only) */
651 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
652 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
653 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
654 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
655 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
656 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
657 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
658 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
659 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
660 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
661 #define	TCPCTL_DROP		15	/* drop tcp connection */
662 
663 #ifdef _KERNEL
664 #ifdef SYSCTL_DECL
665 SYSCTL_DECL(_net_inet_tcp);
666 SYSCTL_DECL(_net_inet_tcp_sack);
667 MALLOC_DECLARE(M_TCPLOG);
668 #endif
669 
670 VNET_DECLARE(struct inpcbhead, tcb);		/* queue of active tcpcb's */
671 VNET_DECLARE(struct inpcbinfo, tcbinfo);
672 extern	int tcp_log_in_vain;
673 VNET_DECLARE(int, tcp_mssdflt);	/* XXX */
674 VNET_DECLARE(int, tcp_minmss);
675 VNET_DECLARE(int, tcp_delack_enabled);
676 VNET_DECLARE(int, tcp_do_rfc3390);
677 VNET_DECLARE(int, tcp_initcwnd_segments);
678 VNET_DECLARE(int, tcp_sendspace);
679 VNET_DECLARE(int, tcp_recvspace);
680 VNET_DECLARE(int, path_mtu_discovery);
681 VNET_DECLARE(int, tcp_do_rfc3465);
682 VNET_DECLARE(int, tcp_abc_l_var);
683 #define	V_tcb			VNET(tcb)
684 #define	V_tcbinfo		VNET(tcbinfo)
685 #define	V_tcp_mssdflt		VNET(tcp_mssdflt)
686 #define	V_tcp_minmss		VNET(tcp_minmss)
687 #define	V_tcp_delack_enabled	VNET(tcp_delack_enabled)
688 #define	V_tcp_do_rfc3390	VNET(tcp_do_rfc3390)
689 #define	V_tcp_initcwnd_segments	VNET(tcp_initcwnd_segments)
690 #define	V_tcp_sendspace		VNET(tcp_sendspace)
691 #define	V_tcp_recvspace		VNET(tcp_recvspace)
692 #define	V_path_mtu_discovery	VNET(path_mtu_discovery)
693 #define	V_tcp_do_rfc3465	VNET(tcp_do_rfc3465)
694 #define	V_tcp_abc_l_var		VNET(tcp_abc_l_var)
695 
696 VNET_DECLARE(int, tcp_do_sack);			/* SACK enabled/disabled */
697 VNET_DECLARE(int, tcp_sc_rst_sock_fail);	/* RST on sock alloc failure */
698 #define	V_tcp_do_sack		VNET(tcp_do_sack)
699 #define	V_tcp_sc_rst_sock_fail	VNET(tcp_sc_rst_sock_fail)
700 
701 VNET_DECLARE(int, tcp_do_ecn);			/* TCP ECN enabled/disabled */
702 VNET_DECLARE(int, tcp_ecn_maxretries);
703 #define	V_tcp_do_ecn		VNET(tcp_do_ecn)
704 #define	V_tcp_ecn_maxretries	VNET(tcp_ecn_maxretries)
705 
706 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
707 #define	V_tcp_hhh		VNET(tcp_hhh)
708 
709 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
710 #define V_tcp_do_rfc6675_pipe	VNET(tcp_do_rfc6675_pipe)
711 
712 int	 tcp_addoptions(struct tcpopt *, u_char *);
713 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
714 struct tcpcb *
715 	 tcp_close(struct tcpcb *);
716 void	 tcp_discardcb(struct tcpcb *);
717 void	 tcp_twstart(struct tcpcb *);
718 void	 tcp_twclose(struct tcptw *, int);
719 void	 tcp_ctlinput(int, struct sockaddr *, void *);
720 int	 tcp_ctloutput(struct socket *, struct sockopt *);
721 struct tcpcb *
722 	 tcp_drop(struct tcpcb *, int);
723 void	 tcp_drain(void);
724 void	 tcp_init(void);
725 #ifdef VIMAGE
726 void	 tcp_destroy(void);
727 #endif
728 void	 tcp_fini(void *);
729 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
730 	    const void *);
731 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
732 	    const void *);
733 int	 tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *);
734 void	 tcp_reass_global_init(void);
735 void	 tcp_reass_flush(struct tcpcb *);
736 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
737 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
738 		     struct tcpcb *, int, int);
739 void	tcp_pulloutofband(struct socket *,
740 		     struct tcphdr *, struct mbuf *, int);
741 void	tcp_xmit_timer(struct tcpcb *, int);
742 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
743 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
744 			    uint16_t type);
745 void 	cc_conn_init(struct tcpcb *tp);
746 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
747 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
748 void	hhook_run_tcp_est_in(struct tcpcb *tp,
749 			    struct tcphdr *th, struct tcpopt *to);
750 
751 int	 tcp_input(struct mbuf **, int *, int);
752 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
753 			struct socket *, struct tcpcb *, int, int, uint8_t,
754 			int);
755 
756 int register_tcp_functions(struct tcp_function_block *blk, int wait);
757 int deregister_tcp_functions(struct tcp_function_block *blk);
758 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
759 struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk);
760 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
761 
762 u_long	 tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
763 u_long	 tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
764 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
765 	    struct tcp_ifcap *);
766 void	 tcp_mss(struct tcpcb *, int);
767 int	 tcp_mssopt(struct in_conninfo *);
768 struct inpcb *
769 	 tcp_drop_syn_sent(struct inpcb *, int);
770 struct tcpcb *
771 	 tcp_newtcpcb(struct inpcb *);
772 int	 tcp_output(struct tcpcb *);
773 void	 tcp_state_change(struct tcpcb *, int);
774 void	 tcp_respond(struct tcpcb *, void *,
775 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
776 void	 tcp_tw_init(void);
777 #ifdef VIMAGE
778 void	 tcp_tw_destroy(void);
779 #endif
780 void	 tcp_tw_zone_change(void);
781 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
782 	    struct mbuf *, int);
783 void	 tcp_setpersist(struct tcpcb *);
784 #ifdef TCP_SIGNATURE
785 struct secasvar;
786 struct secasvar *tcp_get_sav(struct mbuf *, u_int);
787 int	 tcp_signature_do_compute(struct mbuf *, int, int, u_char *,
788 	    struct secasvar *);
789 int	 tcp_signature_compute(struct mbuf *, int, int, int, u_char *, u_int);
790 int	 tcp_signature_verify(struct mbuf *, int, int, int, struct tcpopt *,
791 	    struct tcphdr *, u_int);
792 int	tcp_signature_check(struct mbuf *m, int off0, int tlen, int optlen,
793 	    struct tcpopt *to, struct tcphdr *th, u_int tcpbflag);
794 #endif
795 void	 tcp_slowtimo(void);
796 struct tcptemp *
797 	 tcpip_maketemplate(struct inpcb *);
798 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
799 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
800 int	 tcp_timer_active(struct tcpcb *, uint32_t);
801 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
802 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
803 /*
804  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
805  */
806 void	 tcp_hc_init(void);
807 #ifdef VIMAGE
808 void	 tcp_hc_destroy(void);
809 #endif
810 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
811 u_long	 tcp_hc_getmtu(struct in_conninfo *);
812 void	 tcp_hc_updatemtu(struct in_conninfo *, u_long);
813 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
814 
815 extern	struct pr_usrreqs tcp_usrreqs;
816 tcp_seq tcp_new_isn(struct tcpcb *);
817 
818 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
819 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
820 void	 tcp_clean_sackreport(struct tcpcb *tp);
821 void	 tcp_sack_adjust(struct tcpcb *tp);
822 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
823 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
824 void	 tcp_free_sackholes(struct tcpcb *tp);
825 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
826 u_long	 tcp_seq_subtract(u_long, u_long );
827 int	 tcp_compute_pipe(struct tcpcb *);
828 
829 static inline void
830 tcp_fields_to_host(struct tcphdr *th)
831 {
832 
833 	th->th_seq = ntohl(th->th_seq);
834 	th->th_ack = ntohl(th->th_ack);
835 	th->th_win = ntohs(th->th_win);
836 	th->th_urp = ntohs(th->th_urp);
837 }
838 
839 #ifdef TCP_SIGNATURE
840 static inline void
841 tcp_fields_to_net(struct tcphdr *th)
842 {
843 
844 	th->th_seq = htonl(th->th_seq);
845 	th->th_ack = htonl(th->th_ack);
846 	th->th_win = htons(th->th_win);
847 	th->th_urp = htons(th->th_urp);
848 }
849 #endif
850 #endif /* _KERNEL */
851 
852 #endif /* _NETINET_TCP_VAR_H_ */
853