xref: /freebsd/sys/netinet/tcp_var.h (revision 839fdcfc0c1dba34f728813d9756515ad82ff58a)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1993, 1994, 1995
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
32  * $FreeBSD$
33  */
34 
35 #ifndef _NETINET_TCP_VAR_H_
36 #define _NETINET_TCP_VAR_H_
37 
38 #include <netinet/tcp.h>
39 #include <netinet/tcp_fsm.h>
40 
41 #ifdef _KERNEL
42 #include <net/vnet.h>
43 #include <sys/mbuf.h>
44 #endif
45 
46 #define TCP_END_BYTE_INFO 8	/* Bytes that makeup the "end information array" */
47 /* Types of ending byte info */
48 #define TCP_EI_EMPTY_SLOT	0
49 #define TCP_EI_STATUS_CLIENT_FIN	0x1
50 #define TCP_EI_STATUS_CLIENT_RST	0x2
51 #define TCP_EI_STATUS_SERVER_FIN	0x3
52 #define TCP_EI_STATUS_SERVER_RST	0x4
53 #define TCP_EI_STATUS_RETRAN		0x5
54 #define TCP_EI_STATUS_PROGRESS		0x6
55 #define TCP_EI_STATUS_PERSIST_MAX	0x7
56 #define TCP_EI_STATUS_KEEP_MAX		0x8
57 #define TCP_EI_STATUS_DATA_A_CLOSE	0x9
58 #define TCP_EI_STATUS_RST_IN_FRONT	0xa
59 #define TCP_EI_STATUS_2MSL		0xb
60 #define TCP_EI_STATUS_MAX_VALUE		0xb
61 
62 /************************************************/
63 /* Status bits we track to assure no duplicates,
64  * the bits here are not used by the code but
65  * for human representation. To check a bit we
66  * take and shift over by 1 minus the value (1-8).
67  */
68 /************************************************/
69 #define TCP_EI_BITS_CLIENT_FIN	0x001
70 #define TCP_EI_BITS_CLIENT_RST	0x002
71 #define TCP_EI_BITS_SERVER_FIN	0x004
72 #define TCP_EI_BITS_SERVER_RST	0x008
73 #define TCP_EI_BITS_RETRAN	0x010
74 #define TCP_EI_BITS_PROGRESS	0x020
75 #define TCP_EI_BITS_PRESIST_MAX	0x040
76 #define TCP_EI_BITS_KEEP_MAX	0x080
77 #define TCP_EI_BITS_DATA_A_CLO  0x100
78 #define TCP_EI_BITS_RST_IN_FR	0x200	/* a front state reset */
79 #define TCP_EI_BITS_2MS_TIMER	0x400	/* 2 MSL timer expired */
80 
81 #if defined(_KERNEL) || defined(_WANT_TCPCB)
82 /* TCP segment queue entry */
83 struct tseg_qent {
84 	TAILQ_ENTRY(tseg_qent) tqe_q;
85 	struct	mbuf   *tqe_m;		/* mbuf contains packet */
86 	struct  mbuf   *tqe_last;	/* last mbuf in chain */
87 	tcp_seq tqe_start;		/* TCP Sequence number start */
88 	int	tqe_len;		/* TCP segment data length */
89 	uint32_t tqe_flags;		/* The flags from the th->th_flags */
90 	uint32_t tqe_mbuf_cnt;		/* Count of mbuf overhead */
91 };
92 TAILQ_HEAD(tsegqe_head, tseg_qent);
93 
94 struct sackblk {
95 	tcp_seq start;		/* start seq no. of sack block */
96 	tcp_seq end;		/* end seq no. */
97 };
98 
99 struct sackhole {
100 	tcp_seq start;		/* start seq no. of hole */
101 	tcp_seq end;		/* end seq no. */
102 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
103 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
104 };
105 
106 struct sackhint {
107 	struct sackhole	*nexthole;
108 	int32_t		sack_bytes_rexmit;
109 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
110 
111 	int32_t		delivered_data; /* Newly acked data from last SACK */
112 
113 	int32_t		sacked_bytes;	/* Total sacked bytes reported by the
114 					 * receiver via sack option
115 					 */
116 	uint32_t	recover_fs;	/* Flight Size at the start of Loss recovery */
117 	uint32_t	prr_delivered;	/* Total bytes delivered using PRR */
118 	uint32_t	prr_out;	/* Bytes sent during IN_RECOVERY */
119 };
120 
121 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
122 
123 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
124 
125 /*
126  * Tcp control block, one per tcp; fields:
127  * Organized for 64 byte cacheline efficiency based
128  * on common tcp_input/tcp_output processing.
129  */
130 struct tcpcb {
131 	/* Cache line 1 */
132 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
133 	struct tcp_function_block *t_fb;/* TCP function call block */
134 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
135 	uint32_t t_maxseg:24,		/* maximum segment size */
136 		t_logstate:8;		/* State of "black box" logging */
137 	uint32_t t_port:16,		/* Tunneling (over udp) port */
138 		t_state:4,		/* state of this connection */
139 		t_idle_reduce : 1,
140 		t_delayed_ack: 7,	/* Delayed ack variable */
141 		t_fin_is_rst: 1,	/* Are fin's treated as resets */
142 		t_log_state_set: 1,
143 		bits_spare : 2;
144 	u_int	t_flags;
145 	tcp_seq	snd_una;		/* sent but unacknowledged */
146 	tcp_seq	snd_max;		/* highest sequence number sent;
147 					 * used to recognize retransmits
148 					 */
149 	tcp_seq	snd_nxt;		/* send next */
150 	tcp_seq	snd_up;			/* send urgent pointer */
151 	uint32_t  snd_wnd;		/* send window */
152 	uint32_t  snd_cwnd;		/* congestion-controlled window */
153 	uint32_t t_peakrate_thr; 	/* pre-calculated peak rate threshold */
154 	/* Cache line 2 */
155 	u_int32_t  ts_offset;		/* our timestamp offset */
156 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
157 	int	rcv_numsacks;		/* # distinct sack blks present */
158 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
159 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
160 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
161 	tcp_seq	rcv_nxt;		/* receive next */
162 	tcp_seq	rcv_adv;		/* advertised window */
163 	uint32_t  rcv_wnd;		/* receive window */
164 	u_int	t_flags2;		/* More tcpcb flags storage */
165 	int	t_srtt;			/* smoothed round-trip time */
166 	int	t_rttvar;		/* variance in round-trip time */
167 	u_int32_t  ts_recent;		/* timestamp echo data */
168 	u_char	snd_scale;		/* window scaling for send window */
169 	u_char	rcv_scale;		/* window scaling for recv window */
170 	u_char	snd_limited;		/* segments limited transmitted */
171 	u_char	request_r_scale;	/* pending window scaling */
172 	tcp_seq	last_ack_sent;
173 	u_int	t_rcvtime;		/* inactivity time */
174 	/* Cache line 3 */
175 	tcp_seq	rcv_up;			/* receive urgent pointer */
176 	int	t_segqlen;		/* segment reassembly queue length */
177 	uint32_t t_segqmbuflen;		/* Count of bytes mbufs on all entries */
178 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
179 	struct mbuf      *t_in_pkt;
180 	struct mbuf	 *t_tail_pkt;
181 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
182 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
183 	uint32_t  snd_ssthresh;		/* snd_cwnd size threshold for
184 					 * for slow start exponential to
185 					 * linear switch
186 					 */
187 	tcp_seq	snd_wl1;		/* window update seg seq number */
188 	/* Cache line 4 */
189 	tcp_seq	snd_wl2;		/* window update seg ack number */
190 
191 	tcp_seq	irs;			/* initial receive sequence number */
192 	tcp_seq	iss;			/* initial send sequence number */
193 	u_int	t_acktime;		/* RACK and BBR incoming new data was acked */
194 	u_int	t_sndtime;		/* time last data was sent */
195 	u_int	ts_recent_age;		/* when last updated */
196 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
197 	uint16_t cl4_spare;		/* Spare to adjust CL 4 */
198 	char	t_oobflags;		/* have some */
199 	char	t_iobc;			/* input character */
200 	int	t_rxtcur;		/* current retransmit value (ticks) */
201 
202 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
203 	u_int	t_rtttime;		/* RTT measurement start time */
204 
205 	tcp_seq	t_rtseq;		/* sequence number being timed */
206 	u_int	t_starttime;		/* time connection was established */
207 	u_int	t_fbyte_in;		/* ticks time when first byte queued in */
208 	u_int	t_fbyte_out;		/* ticks time when first byte queued out */
209 
210 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
211 	int	t_blackhole_enter;	/* when to enter blackhole detection */
212 	int	t_blackhole_exit;	/* when to exit blackhole detection */
213 	u_int	t_rttmin;		/* minimum rtt allowed */
214 
215 	u_int	t_rttbest;		/* best rtt we've seen */
216 
217 	int	t_softerror;		/* possible error not yet reported */
218 	uint32_t  max_sndwnd;		/* largest window peer has offered */
219 	/* Cache line 5 */
220 	uint32_t  snd_cwnd_prev;	/* cwnd prior to retransmit */
221 	uint32_t  snd_ssthresh_prev;	/* ssthresh prior to retransmit */
222 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
223 	int	t_sndzerowin;		/* zero-window updates sent */
224 	u_long	t_rttupdated;		/* number of times rtt sampled */
225 	int	snd_numholes;		/* number of holes seen by sender */
226 	u_int	t_badrxtwin;		/* window for retransmit recovery */
227 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
228 					/* SACK scoreboard (sorted) */
229 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
230 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
231 	struct sackhint	sackhint;	/* SACK scoreboard hint */
232 	int	t_rttlow;		/* smallest observerved RTT */
233 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
234 	struct toedev	*tod;		/* toedev handling this connection */
235 	int	t_sndrexmitpack;	/* retransmit packets sent */
236 	int	t_rcvoopack;		/* out-of-order packets received */
237 	void	*t_toe;			/* TOE pcb pointer */
238 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
239 	struct cc_var	*ccv;		/* congestion control specific vars */
240 	struct osd	*osd;		/* storage for Khelp module data */
241 	int	t_bytes_acked;		/* # bytes acked during current RTT */
242 	u_int   t_maxunacktime;
243 	u_int	t_keepinit;		/* time to establish connection */
244 	u_int	t_keepidle;		/* time before keepalive probes begin */
245 	u_int	t_keepintvl;		/* interval between keepalives */
246 	u_int	t_keepcnt;		/* number of keepalives before close */
247 	int	t_dupacks;		/* consecutive dup acks recd */
248 	int	t_lognum;		/* Number of log entries */
249 	int	t_loglimit;		/* Maximum number of log entries */
250 	int64_t	t_pacing_rate;		/* bytes / sec, -1 => unlimited */
251 	struct tcp_log_stailq t_logs;	/* Log buffer */
252 	struct tcp_log_id_node *t_lin;
253 	struct tcp_log_id_bucket *t_lib;
254 	const char *t_output_caller;	/* Function that called tcp_output */
255 	struct statsblob *t_stats;	/* Per-connection stats */
256 	uint32_t t_logsn;		/* Log "serial number" */
257 	uint32_t gput_ts;		/* Time goodput measurement started */
258 	tcp_seq gput_seq;		/* Outbound measurement seq */
259 	tcp_seq gput_ack;		/* Inbound measurement ack */
260 	int32_t t_stats_gput_prev;	/* XXXLAS: Prev gput measurement */
261 	uint8_t t_tfo_client_cookie_len; /* TCP Fast Open client cookie length */
262 	uint32_t t_end_info_status;	/* Status flag of end info */
263 	unsigned int *t_tfo_pending;	/* TCP Fast Open server pending counter */
264 	union {
265 		uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
266 		uint64_t server;
267 	} t_tfo_cookie;			/* TCP Fast Open cookie to send */
268 	union {
269 		uint8_t t_end_info_bytes[TCP_END_BYTE_INFO];
270 		uint64_t t_end_info;
271 	};
272 #ifdef TCPPCAP
273 	struct mbufq t_inpkts;		/* List of saved input packets. */
274 	struct mbufq t_outpkts;		/* List of saved output packets. */
275 #endif
276 };
277 #endif	/* _KERNEL || _WANT_TCPCB */
278 
279 #ifdef _KERNEL
280 struct tcptemp {
281 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
282 	struct	tcphdr tt_t;
283 };
284 
285 /* Minimum map entries limit value, if set */
286 #define TCP_MIN_MAP_ENTRIES_LIMIT	128
287 
288 /*
289  * TODO: We yet need to brave plowing in
290  * to tcp_input() and the pru_usrreq() block.
291  * Right now these go to the old standards which
292  * are somewhat ok, but in the long term may
293  * need to be changed. If we do tackle tcp_input()
294  * then we need to get rid of the tcp_do_segment()
295  * function below.
296  */
297 /* Flags for tcp functions */
298 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
299 
300 /*
301  * If defining the optional tcp_timers, in the
302  * tfb_tcp_timer_stop call you must use the
303  * callout_async_drain() function with the
304  * tcp_timer_discard callback. You should check
305  * the return of callout_async_drain() and if 0
306  * increment tt_draincnt. Since the timer sub-system
307  * does not know your callbacks you must provide a
308  * stop_all function that loops through and calls
309  * tcp_timer_stop() with each of your defined timers.
310  * Adding a tfb_tcp_handoff_ok function allows the socket
311  * option to change stacks to query you even if the
312  * connection is in a later stage. You return 0 to
313  * say you can take over and run your stack, you return
314  * non-zero (an error number) to say no you can't.
315  * If the function is undefined you can only change
316  * in the early states (before connect or listen).
317  * tfb_tcp_fb_fini is changed to add a flag to tell
318  * the old stack if the tcb is being destroyed or
319  * not. A one in the flag means the TCB is being
320  * destroyed, a zero indicates its transitioning to
321  * another stack (via socket option).
322  */
323 struct tcp_function_block {
324 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
325 	int	(*tfb_tcp_output)(struct tcpcb *);
326 	int	(*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *);
327 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
328 			    struct socket *, struct tcpcb *,
329 		        int, int, uint8_t);
330 	int     (*tfb_do_queued_segments)(struct socket *, struct tcpcb *, int);
331 	int      (*tfb_do_segment_nounlock)(struct mbuf *, struct tcphdr *,
332 			    struct socket *, struct tcpcb *,
333 			    int, int, uint8_t,
334 			    int, struct timeval *);
335 	void	(*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *,
336 			    struct socket *, struct tcpcb *,
337 			    int, int, uint8_t,
338 			    int, struct timeval *);
339 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
340 			    struct inpcb *inp, struct tcpcb *tp);
341 	/* Optional memory allocation/free routine */
342 	int	(*tfb_tcp_fb_init)(struct tcpcb *);
343 	void	(*tfb_tcp_fb_fini)(struct tcpcb *, int);
344 	/* Optional timers, must define all if you define one */
345 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
346 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
347 			    uint32_t, u_int);
348 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
349 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
350 	void	(*tfb_tcp_rexmit_tmr)(struct tcpcb *);
351 	int	(*tfb_tcp_handoff_ok)(struct tcpcb *);
352 	void	(*tfb_tcp_mtu_chg)(struct tcpcb *);
353 	int	(*tfb_pru_options)(struct tcpcb *, int);
354 	volatile uint32_t tfb_refcnt;
355 	uint32_t  tfb_flags;
356 	uint8_t	tfb_id;
357 };
358 
359 struct tcp_function {
360 	TAILQ_ENTRY(tcp_function)	tf_next;
361 	char				tf_name[TCP_FUNCTION_NAME_LEN_MAX];
362 	struct tcp_function_block	*tf_fb;
363 };
364 
365 TAILQ_HEAD(tcp_funchead, tcp_function);
366 #endif	/* _KERNEL */
367 
368 /*
369  * Flags and utility macros for the t_flags field.
370  */
371 #define	TF_ACKNOW	0x00000001	/* ack peer immediately */
372 #define	TF_DELACK	0x00000002	/* ack, but try to delay it */
373 #define	TF_NODELAY	0x00000004	/* don't delay packets to coalesce */
374 #define	TF_NOOPT	0x00000008	/* don't use tcp options */
375 #define	TF_SENTFIN	0x00000010	/* have sent FIN */
376 #define	TF_REQ_SCALE	0x00000020	/* have/will request window scaling */
377 #define	TF_RCVD_SCALE	0x00000040	/* other side has requested scaling */
378 #define	TF_REQ_TSTMP	0x00000080	/* have/will request timestamps */
379 #define	TF_RCVD_TSTMP	0x00000100	/* a timestamp was received in SYN */
380 #define	TF_SACK_PERMIT	0x00000200	/* other side said I could SACK */
381 #define	TF_NEEDSYN	0x00000400	/* send SYN (implicit state) */
382 #define	TF_NEEDFIN	0x00000800	/* send FIN (implicit state) */
383 #define	TF_NOPUSH	0x00001000	/* don't push */
384 #define	TF_PREVVALID	0x00002000	/* saved values for bad rxmit valid */
385 #define	TF_WAKESOR	0x00004000	/* wake up receive socket */
386 #define	TF_GPUTINPROG	0x00008000	/* Goodput measurement in progress */
387 #define	TF_MORETOCOME	0x00010000	/* More data to be appended to sock */
388 #define	TF_LQ_OVERFLOW	0x00020000	/* listen queue overflow */
389 #define	TF_LASTIDLE	0x00040000	/* connection was previously idle */
390 #define	TF_RXWIN0SENT	0x00080000	/* sent a receiver win 0 in response */
391 #define	TF_FASTRECOVERY	0x00100000	/* in NewReno Fast Recovery */
392 #define	TF_WASFRECOVERY	0x00200000	/* was in NewReno Fast Recovery */
393 #define	TF_SIGNATURE	0x00400000	/* require MD5 digests (RFC2385) */
394 #define	TF_FORCEDATA	0x00800000	/* force out a byte */
395 #define	TF_TSO		0x01000000	/* TSO enabled on this connection */
396 #define	TF_TOE		0x02000000	/* this connection is offloaded */
397 #define	TF_WAKESOW	0x04000000	/* wake up send socket */
398 #define	TF_UNUSED1	0x08000000	/* unused */
399 #define	TF_UNUSED2	0x10000000	/* unused */
400 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
401 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
402 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
403 
404 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
405 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
406 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
407 
408 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
409 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
410 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
411 
412 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
413 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
414 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
415 
416 #if defined(_KERNEL) && !defined(TCP_RFC7413)
417 #define	IS_FASTOPEN(t_flags)		(false)
418 #else
419 #define	IS_FASTOPEN(t_flags)		(t_flags & TF_FASTOPEN)
420 #endif
421 
422 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
423 
424 /*
425  * Flags for the t_oobflags field.
426  */
427 #define	TCPOOB_HAVEDATA	0x01
428 #define	TCPOOB_HADDATA	0x02
429 
430 /*
431  * Flags for the extended TCP flags field, t_flags2
432  */
433 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
434 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
435 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
436 #define	TF2_LOG_AUTO		0x00000008 /* Session is auto-logging. */
437 #define TF2_DROP_AF_DATA 	0x00000010 /* Drop after all data ack'd */
438 #define	TF2_ECN_PERMIT		0x00000020 /* connection ECN-ready */
439 #define	TF2_ECN_SND_CWR		0x00000040 /* ECN CWR in queue */
440 #define	TF2_ECN_SND_ECE		0x00000080 /* ECN ECE in queue */
441 #define	TF2_ACE_PERMIT		0x00000100 /* Accurate ECN mode */
442 #define TF2_FBYTES_COMPLETE	0x00000400 /* We have first bytes in and out */
443 /*
444  * Structure to hold TCP options that are only used during segment
445  * processing (in tcp_input), but not held in the tcpcb.
446  * It's basically used to reduce the number of parameters
447  * to tcp_dooptions and tcp_addoptions.
448  * The binary order of the to_flags is relevant for packing of the
449  * options in tcp_addoptions.
450  */
451 struct tcpopt {
452 	u_int32_t	to_flags;	/* which options are present */
453 #define	TOF_MSS		0x0001		/* maximum segment size */
454 #define	TOF_SCALE	0x0002		/* window scaling */
455 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
456 #define	TOF_TS		0x0010		/* timestamp */
457 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
458 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
459 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
460 #define	TOF_MAXOPT	0x0200
461 	u_int32_t	to_tsval;	/* new timestamp */
462 	u_int32_t	to_tsecr;	/* reflected timestamp */
463 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
464 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
465 	u_int8_t	*to_tfo_cookie; /* pointer to the TFO cookie */
466 	u_int16_t	to_mss;		/* maximum segment size */
467 	u_int8_t	to_wscale;	/* window scaling */
468 	u_int8_t	to_nsacks;	/* number of SACK blocks */
469 	u_int8_t	to_tfo_len;	/* TFO cookie length */
470 	u_int32_t	to_spare;	/* UTO */
471 };
472 
473 /*
474  * Flags for tcp_dooptions.
475  */
476 #define	TO_SYN		0x01		/* parse SYN-only options */
477 
478 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
479 	uint32_t	rmx_mtu;	/* MTU for this path */
480 	uint32_t	rmx_ssthresh;	/* outbound gateway buffer limit */
481 	uint32_t	rmx_rtt;	/* estimated round trip time */
482 	uint32_t	rmx_rttvar;	/* estimated rtt variance */
483 	uint32_t	rmx_cwnd;	/* congestion window */
484 	uint32_t	rmx_sendpipe;   /* outbound delay-bandwidth product */
485 	uint32_t	rmx_recvpipe;   /* inbound delay-bandwidth product */
486 };
487 
488 /*
489  * Used by tcp_maxmtu() to communicate interface specific features
490  * and limits at the time of connection setup.
491  */
492 struct tcp_ifcap {
493 	int	ifcap;
494 	u_int	tsomax;
495 	u_int	tsomaxsegcount;
496 	u_int	tsomaxsegsize;
497 };
498 
499 #ifndef _NETINET_IN_PCB_H_
500 struct in_conninfo;
501 #endif /* _NETINET_IN_PCB_H_ */
502 
503 struct tcptw {
504 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
505 	tcp_seq		snd_nxt;
506 	tcp_seq		rcv_nxt;
507 	tcp_seq		iss;
508 	tcp_seq		irs;
509 	u_short		last_win;	/* cached window value */
510 	short		tw_so_options;	/* copy of so_options */
511 	struct ucred	*tw_cred;	/* user credentials */
512 	u_int32_t	t_recent;
513 	u_int32_t	ts_offset;	/* our timestamp offset */
514 	u_int		t_starttime;
515 	int		tw_time;
516 	TAILQ_ENTRY(tcptw) tw_2msl;
517 	void		*tw_pspare;	/* TCP_SIGNATURE */
518 	u_int		*tw_spare;	/* TCP_SIGNATURE */
519 };
520 
521 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
522 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
523 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
524 
525 /*
526  * The smoothed round-trip time and estimated variance
527  * are stored as fixed point numbers scaled by the values below.
528  * For convenience, these scales are also used in smoothing the average
529  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
530  * With these scales, srtt has 3 bits to the right of the binary point,
531  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
532  * binary point, and is smoothed with an ALPHA of 0.75.
533  */
534 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
535 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
536 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
537 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
538 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
539 
540 /*
541  * The initial retransmission should happen at rtt + 4 * rttvar.
542  * Because of the way we do the smoothing, srtt and rttvar
543  * will each average +1/2 tick of bias.  When we compute
544  * the retransmit timer, we want 1/2 tick of rounding and
545  * 1 extra tick because of +-1/2 tick uncertainty in the
546  * firing of the timer.  The bias will give us exactly the
547  * 1.5 tick we need.  But, because the bias is
548  * statistical, we have to test that we don't drop below
549  * the minimum feasible timer (which is 2 ticks).
550  * This version of the macro adapted from a paper by Lawrence
551  * Brakmo and Larry Peterson which outlines a problem caused
552  * by insufficient precision in the original implementation,
553  * which results in inappropriately large RTO values for very
554  * fast networks.
555  */
556 #define	TCP_REXMTVAL(tp) \
557 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
558 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
559 
560 /*
561  * TCP statistics.
562  * Many of these should be kept per connection,
563  * but that's inconvenient at the moment.
564  */
565 struct	tcpstat {
566 	uint64_t tcps_connattempt;	/* connections initiated */
567 	uint64_t tcps_accepts;		/* connections accepted */
568 	uint64_t tcps_connects;		/* connections established */
569 	uint64_t tcps_drops;		/* connections dropped */
570 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
571 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
572 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
573 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
574 	uint64_t tcps_rttupdated;	/* times we succeeded */
575 	uint64_t tcps_delack;		/* delayed acks sent */
576 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
577 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
578 	uint64_t tcps_persisttimeo;	/* persist timeouts */
579 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
580 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
581 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
582 
583 	uint64_t tcps_sndtotal;		/* total packets sent */
584 	uint64_t tcps_sndpack;		/* data packets sent */
585 	uint64_t tcps_sndbyte;		/* data bytes sent */
586 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
587 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
588 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
589 	uint64_t tcps_sndacks;		/* ack-only packets sent */
590 	uint64_t tcps_sndprobe;		/* window probes sent */
591 	uint64_t tcps_sndurg;		/* packets sent with URG only */
592 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
593 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
594 
595 	uint64_t tcps_rcvtotal;		/* total packets received */
596 	uint64_t tcps_rcvpack;		/* packets received in sequence */
597 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
598 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
599 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
600 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
601 	uint64_t tcps_rcvshort;		/* packets received too short */
602 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
603 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
604 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
605 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
606 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
607 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
608 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
609 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
610 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
611 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
612 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
613 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
614 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
615 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
616 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
617 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
618 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
619 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
620 	uint64_t tcps_pcbcachemiss;
621 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
622 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
623 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
624 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
625 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
626 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
627 	uint64_t tcps_persistdrop;	/* timeout in persist state */
628 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
629 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
630 	uint64_t tcps_listendrop;	/* listen queue overflows */
631 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
632 
633 	uint64_t tcps_sc_added;		/* entry added to syncache */
634 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
635 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
636 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
637 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
638 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
639 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
640 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
641 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
642 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
643 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
644 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
645 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
646 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
647 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
648 
649 	uint64_t tcps_hc_added;		/* entry added to hostcache */
650 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
651 
652 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
653 
654 	/* SACK related stats */
655 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
656 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
657 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
658 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
659 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
660 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
661 
662 	/* ECN related stats */
663 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
664 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
665 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
666 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
667 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
668 
669 	/* TCP_SIGNATURE related stats */
670 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
671 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
672 	uint64_t tcps_sig_err_buildsig;	/* Failed to make signature */
673 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
674 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
675 
676 	/* Path MTU Discovery Black Hole Detection related stats */
677 	uint64_t tcps_pmtud_blackhole_activated;	 /* Black Hole Count */
678 	uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
679 	uint64_t tcps_pmtud_blackhole_failed;		 /* Black Hole Failure Count */
680 
681 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
682 };
683 
684 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
685 
686 #ifdef _KERNEL
687 #define	TI_UNLOCKED	1
688 #define	TI_RLOCKED	2
689 #include <sys/counter.h>
690 
691 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
692 /*
693  * In-kernel consumers can use these accessor macros directly to update
694  * stats.
695  */
696 #define	TCPSTAT_ADD(name, val)	\
697     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
698 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
699 
700 /*
701  * Kernel module consumers must use this accessor macro.
702  */
703 void	kmod_tcpstat_add(int statnum, int val);
704 #define	KMOD_TCPSTAT_ADD(name, val)					\
705     kmod_tcpstat_add(offsetof(struct tcpstat, name) / sizeof(uint64_t), val)
706 #define	KMOD_TCPSTAT_INC(name)	KMOD_TCPSTAT_ADD(name, 1)
707 
708 /*
709  * Running TCP connection count by state.
710  */
711 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
712 #define	V_tcps_states	VNET(tcps_states)
713 #define	TCPSTATES_INC(state)	counter_u64_add(V_tcps_states[state], 1)
714 #define	TCPSTATES_DEC(state)	counter_u64_add(V_tcps_states[state], -1)
715 
716 /*
717  * TCP specific helper hook point identifiers.
718  */
719 #define	HHOOK_TCP_EST_IN		0
720 #define	HHOOK_TCP_EST_OUT		1
721 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
722 
723 struct tcp_hhook_data {
724 	struct tcpcb	*tp;
725 	struct tcphdr	*th;
726 	struct tcpopt	*to;
727 	uint32_t	len;
728 	int		tso;
729 	tcp_seq		curack;
730 };
731 #ifdef TCP_HHOOK
732 void hhook_run_tcp_est_out(struct tcpcb *tp,
733 	struct tcphdr *th, struct tcpopt *to,
734 	uint32_t len, int tso);
735 #endif
736 #endif
737 
738 /*
739  * TCB structure exported to user-land via sysctl(3).
740  *
741  * Fields prefixed with "xt_" are unique to the export structure, and fields
742  * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
743  *
744  * Legend:
745  * (s) - used by userland utilities in src
746  * (p) - used by utilities in ports
747  * (3) - is known to be used by third party software not in ports
748  * (n) - no known usage
749  *
750  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
751  * included.  Not all of our clients do.
752  */
753 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
754 struct xtcpcb {
755 	ksize_t	xt_len;		/* length of this structure */
756 	struct xinpcb	xt_inp;
757 	char		xt_stack[TCP_FUNCTION_NAME_LEN_MAX];	/* (s) */
758 	char		xt_logid[TCP_LOG_ID_LEN];	/* (s) */
759 	char		xt_cc[TCP_CA_NAME_MAX];	/* (s) */
760 	int64_t		spare64[6];
761 	int32_t		t_state;		/* (s,p) */
762 	uint32_t	t_flags;		/* (s,p) */
763 	int32_t		t_sndzerowin;		/* (s) */
764 	int32_t		t_sndrexmitpack;	/* (s) */
765 	int32_t		t_rcvoopack;		/* (s) */
766 	int32_t		t_rcvtime;		/* (s) */
767 	int32_t		tt_rexmt;		/* (s) */
768 	int32_t		tt_persist;		/* (s) */
769 	int32_t		tt_keep;		/* (s) */
770 	int32_t		tt_2msl;		/* (s) */
771 	int32_t		tt_delack;		/* (s) */
772 	int32_t		t_logstate;		/* (3) */
773 	uint32_t	t_snd_cwnd;		/* (s) */
774 	uint32_t	t_snd_ssthresh;		/* (s) */
775 	uint32_t	t_maxseg;		/* (s) */
776 	uint32_t	t_rcv_wnd;		/* (s) */
777 	uint32_t	t_snd_wnd;		/* (s) */
778 	uint32_t	xt_ecn;			/* (s) */
779 	int32_t		spare32[26];
780 } __aligned(8);
781 
782 #ifdef _KERNEL
783 void	tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
784 #endif
785 #endif
786 
787 /*
788  * TCP function information (name-to-id mapping, aliases, and refcnt)
789  * exported to user-land via sysctl(3).
790  */
791 struct tcp_function_info {
792 	uint32_t	tfi_refcnt;
793 	uint8_t		tfi_id;
794 	char		tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
795 	char		tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
796 };
797 
798 /*
799  * Identifiers for TCP sysctl nodes
800  */
801 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
802 #define	TCPCTL_MSSDFLT		3	/* MSS default */
803 #define TCPCTL_STATS		4	/* statistics */
804 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
805 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
806 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
807 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
808 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
809 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
810 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
811 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
812 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
813 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
814 #define	TCPCTL_DROP		15	/* drop tcp connection */
815 #define	TCPCTL_STATES		16	/* connection counts by TCP state */
816 
817 #ifdef _KERNEL
818 #ifdef SYSCTL_DECL
819 SYSCTL_DECL(_net_inet_tcp);
820 SYSCTL_DECL(_net_inet_tcp_sack);
821 MALLOC_DECLARE(M_TCPLOG);
822 #endif
823 
824 VNET_DECLARE(int, tcp_log_in_vain);
825 #define	V_tcp_log_in_vain		VNET(tcp_log_in_vain)
826 
827 /*
828  * Global TCP tunables shared between different stacks.
829  * Please keep the list sorted.
830  */
831 VNET_DECLARE(int, drop_synfin);
832 VNET_DECLARE(int, path_mtu_discovery);
833 VNET_DECLARE(int, tcp_abc_l_var);
834 VNET_DECLARE(int, tcp_autorcvbuf_max);
835 VNET_DECLARE(int, tcp_autosndbuf_inc);
836 VNET_DECLARE(int, tcp_autosndbuf_max);
837 VNET_DECLARE(int, tcp_delack_enabled);
838 VNET_DECLARE(int, tcp_do_autorcvbuf);
839 VNET_DECLARE(int, tcp_do_autosndbuf);
840 VNET_DECLARE(int, tcp_do_ecn);
841 VNET_DECLARE(int, tcp_do_newcwv);
842 VNET_DECLARE(int, tcp_do_rfc1323);
843 VNET_DECLARE(int, tcp_tolerate_missing_ts);
844 VNET_DECLARE(int, tcp_do_rfc3042);
845 VNET_DECLARE(int, tcp_do_rfc3390);
846 VNET_DECLARE(int, tcp_do_rfc3465);
847 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
848 VNET_DECLARE(int, tcp_do_sack);
849 VNET_DECLARE(int, tcp_do_tso);
850 VNET_DECLARE(int, tcp_ecn_maxretries);
851 VNET_DECLARE(int, tcp_initcwnd_segments);
852 VNET_DECLARE(int, tcp_insecure_rst);
853 VNET_DECLARE(int, tcp_insecure_syn);
854 VNET_DECLARE(uint32_t, tcp_map_entries_limit);
855 VNET_DECLARE(uint32_t, tcp_map_split_limit);
856 VNET_DECLARE(int, tcp_minmss);
857 VNET_DECLARE(int, tcp_mssdflt);
858 #ifdef STATS
859 VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl);
860 VNET_DECLARE(int, tcp_perconn_stats_enable);
861 #endif /* STATS */
862 VNET_DECLARE(int, tcp_recvspace);
863 VNET_DECLARE(int, tcp_sack_globalholes);
864 VNET_DECLARE(int, tcp_sack_globalmaxholes);
865 VNET_DECLARE(int, tcp_sack_maxholes);
866 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
867 VNET_DECLARE(int, tcp_sendspace);
868 VNET_DECLARE(struct inpcbhead, tcb);
869 VNET_DECLARE(struct inpcbinfo, tcbinfo);
870 
871 #define	V_tcp_do_prr			VNET(tcp_do_prr)
872 #define	V_tcp_do_prr_conservative	VNET(tcp_do_prr_conservative)
873 #define	V_tcp_do_newcwv			VNET(tcp_do_newcwv)
874 #define	V_drop_synfin			VNET(drop_synfin)
875 #define	V_path_mtu_discovery		VNET(path_mtu_discovery)
876 #define	V_tcb				VNET(tcb)
877 #define	V_tcbinfo			VNET(tcbinfo)
878 #define	V_tcp_abc_l_var			VNET(tcp_abc_l_var)
879 #define	V_tcp_autorcvbuf_max		VNET(tcp_autorcvbuf_max)
880 #define	V_tcp_autosndbuf_inc		VNET(tcp_autosndbuf_inc)
881 #define	V_tcp_autosndbuf_max		VNET(tcp_autosndbuf_max)
882 #define	V_tcp_delack_enabled		VNET(tcp_delack_enabled)
883 #define	V_tcp_do_autorcvbuf		VNET(tcp_do_autorcvbuf)
884 #define	V_tcp_do_autosndbuf		VNET(tcp_do_autosndbuf)
885 #define	V_tcp_do_ecn			VNET(tcp_do_ecn)
886 #define	V_tcp_do_rfc1323		VNET(tcp_do_rfc1323)
887 #define	V_tcp_tolerate_missing_ts	VNET(tcp_tolerate_missing_ts)
888 #define V_tcp_ts_offset_per_conn	VNET(tcp_ts_offset_per_conn)
889 #define	V_tcp_do_rfc3042		VNET(tcp_do_rfc3042)
890 #define	V_tcp_do_rfc3390		VNET(tcp_do_rfc3390)
891 #define	V_tcp_do_rfc3465		VNET(tcp_do_rfc3465)
892 #define	V_tcp_do_rfc6675_pipe		VNET(tcp_do_rfc6675_pipe)
893 #define	V_tcp_do_sack			VNET(tcp_do_sack)
894 #define	V_tcp_do_tso			VNET(tcp_do_tso)
895 #define	V_tcp_ecn_maxretries		VNET(tcp_ecn_maxretries)
896 #define	V_tcp_initcwnd_segments		VNET(tcp_initcwnd_segments)
897 #define	V_tcp_insecure_rst		VNET(tcp_insecure_rst)
898 #define	V_tcp_insecure_syn		VNET(tcp_insecure_syn)
899 #define	V_tcp_map_entries_limit		VNET(tcp_map_entries_limit)
900 #define	V_tcp_map_split_limit		VNET(tcp_map_split_limit)
901 #define	V_tcp_minmss			VNET(tcp_minmss)
902 #define	V_tcp_mssdflt			VNET(tcp_mssdflt)
903 #ifdef STATS
904 #define	V_tcp_perconn_stats_dflt_tpl	VNET(tcp_perconn_stats_dflt_tpl)
905 #define	V_tcp_perconn_stats_enable	VNET(tcp_perconn_stats_enable)
906 #endif /* STATS */
907 #define	V_tcp_recvspace			VNET(tcp_recvspace)
908 #define	V_tcp_sack_globalholes		VNET(tcp_sack_globalholes)
909 #define	V_tcp_sack_globalmaxholes	VNET(tcp_sack_globalmaxholes)
910 #define	V_tcp_sack_maxholes		VNET(tcp_sack_maxholes)
911 #define	V_tcp_sc_rst_sock_fail		VNET(tcp_sc_rst_sock_fail)
912 #define	V_tcp_sendspace			VNET(tcp_sendspace)
913 #define	V_tcp_udp_tunneling_overhead	VNET(tcp_udp_tunneling_overhead)
914 #define	V_tcp_udp_tunneling_port	VNET(tcp_udp_tunneling_port)
915 
916 #ifdef TCP_HHOOK
917 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
918 #define	V_tcp_hhh		VNET(tcp_hhh)
919 #endif
920 
921 int	 tcp_addoptions(struct tcpopt *, u_char *);
922 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
923 struct tcpcb *
924 	 tcp_close(struct tcpcb *);
925 void	 tcp_discardcb(struct tcpcb *);
926 void	 tcp_twstart(struct tcpcb *);
927 void	 tcp_twclose(struct tcptw *, int);
928 void	 tcp_ctlinput(int, struct sockaddr *, void *);
929 int	 tcp_ctloutput(struct socket *, struct sockopt *);
930 struct tcpcb *
931 	 tcp_drop(struct tcpcb *, int);
932 void	 tcp_drain(void);
933 void	 tcp_init(void);
934 void	 tcp_fini(void *);
935 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
936 	    const void *);
937 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
938 	    const void *);
939 int	 tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *,
940 	    struct mbuf *);
941 void	 tcp_reass_global_init(void);
942 void	 tcp_reass_flush(struct tcpcb *);
943 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
944 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
945 		     struct tcpcb *, int, int);
946 void	tcp_pulloutofband(struct socket *,
947 		     struct tcphdr *, struct mbuf *, int);
948 void	tcp_xmit_timer(struct tcpcb *, int);
949 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
950 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
951 			    uint16_t nsegs, uint16_t type);
952 void 	cc_conn_init(struct tcpcb *tp);
953 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
954 void    cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos);
955 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
956 #ifdef TCP_HHOOK
957 void	hhook_run_tcp_est_in(struct tcpcb *tp,
958 			    struct tcphdr *th, struct tcpopt *to);
959 #endif
960 
961 int	 tcp_input(struct mbuf **, int *, int);
962 int	 tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
963 	    struct tcpcb *, int);
964 void	 tcp_handle_wakeup(struct tcpcb *, struct socket *);
965 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
966 			struct socket *, struct tcpcb *, int, int, uint8_t);
967 
968 int register_tcp_functions(struct tcp_function_block *blk, int wait);
969 int register_tcp_functions_as_names(struct tcp_function_block *blk,
970     int wait, const char *names[], int *num_names);
971 int register_tcp_functions_as_name(struct tcp_function_block *blk,
972     const char *name, int wait);
973 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
974     bool force);
975 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
976 void tcp_switch_back_to_default(struct tcpcb *tp);
977 struct tcp_function_block *
978 find_and_ref_tcp_fb(struct tcp_function_block *fs);
979 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
980 
981 extern counter_u64_t tcp_inp_lro_direct_queue;
982 extern counter_u64_t tcp_inp_lro_wokeup_queue;
983 extern counter_u64_t tcp_inp_lro_compressed;
984 extern counter_u64_t tcp_inp_lro_single_push;
985 extern counter_u64_t tcp_inp_lro_locks_taken;
986 extern counter_u64_t tcp_inp_lro_sack_wake;
987 extern counter_u64_t tcp_extra_mbuf;
988 extern counter_u64_t tcp_would_have_but;
989 extern counter_u64_t tcp_comp_total;
990 extern counter_u64_t tcp_uncomp_total;
991 extern counter_u64_t tcp_csum_hardware;
992 extern counter_u64_t tcp_csum_hardware_w_ph;
993 extern counter_u64_t tcp_csum_software;
994 
995 #ifdef NETFLIX_EXP_DETECTION
996 /* Various SACK attack thresholds */
997 extern int32_t tcp_force_detection;
998 extern int32_t tcp_sack_to_ack_thresh;
999 extern int32_t tcp_sack_to_move_thresh;
1000 extern int32_t tcp_restoral_thresh;
1001 extern int32_t tcp_sad_decay_val;
1002 extern int32_t tcp_sad_pacing_interval;
1003 extern int32_t tcp_sad_low_pps;
1004 extern int32_t tcp_map_minimum;
1005 extern int32_t tcp_attack_on_turns_on_logging;
1006 #endif
1007 
1008 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
1009 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
1010 u_int	 tcp_maxseg(const struct tcpcb *);
1011 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
1012 	    struct tcp_ifcap *);
1013 void	 tcp_mss(struct tcpcb *, int);
1014 int	 tcp_mssopt(struct in_conninfo *);
1015 struct inpcb *
1016 	 tcp_drop_syn_sent(struct inpcb *, int);
1017 struct tcpcb *
1018 	 tcp_newtcpcb(struct inpcb *);
1019 int	 tcp_output(struct tcpcb *);
1020 void	 tcp_state_change(struct tcpcb *, int);
1021 void	 tcp_respond(struct tcpcb *, void *,
1022 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
1023 void	 tcp_tw_init(void);
1024 #ifdef VIMAGE
1025 void	 tcp_tw_destroy(void);
1026 #endif
1027 void	 tcp_tw_zone_change(void);
1028 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
1029 	    struct mbuf *, int);
1030 void	 tcp_setpersist(struct tcpcb *);
1031 void	 tcp_slowtimo(void);
1032 struct tcptemp *
1033 	 tcpip_maketemplate(struct inpcb *);
1034 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
1035 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
1036 int	 tcp_timer_suspend(struct tcpcb *, uint32_t);
1037 void	 tcp_timers_unsuspend(struct tcpcb *, uint32_t);
1038 int	 tcp_timer_active(struct tcpcb *, uint32_t);
1039 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
1040 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
1041 int	 inp_to_cpuid(struct inpcb *inp);
1042 /*
1043  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
1044  */
1045 void	 tcp_hc_init(void);
1046 #ifdef VIMAGE
1047 void	 tcp_hc_destroy(void);
1048 #endif
1049 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
1050 uint32_t tcp_hc_getmtu(struct in_conninfo *);
1051 void	 tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
1052 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
1053 
1054 extern	struct pr_usrreqs tcp_usrreqs;
1055 
1056 uint32_t tcp_new_ts_offset(struct in_conninfo *);
1057 tcp_seq	 tcp_new_isn(struct in_conninfo *);
1058 
1059 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
1060 void	 tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq);
1061 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1062 void	 tcp_clean_dsack_blocks(struct tcpcb *tp);
1063 void	 tcp_clean_sackreport(struct tcpcb *tp);
1064 void	 tcp_sack_adjust(struct tcpcb *tp);
1065 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1066 void	 tcp_prr_partialack(struct tcpcb *, struct tcphdr *);
1067 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1068 void	 tcp_free_sackholes(struct tcpcb *tp);
1069 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
1070 int	 tcp_compute_pipe(struct tcpcb *);
1071 uint32_t tcp_compute_initwnd(uint32_t);
1072 void	 tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
1073 int	 tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes,
1074     size_t seed_len);
1075 struct mbuf *
1076 	 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen,
1077 	   int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls);
1078 
1079 int	tcp_stats_init(void);
1080 void tcp_log_end_status(struct tcpcb *tp, uint8_t status);
1081 
1082 static inline void
1083 tcp_fields_to_host(struct tcphdr *th)
1084 {
1085 
1086 	th->th_seq = ntohl(th->th_seq);
1087 	th->th_ack = ntohl(th->th_ack);
1088 	th->th_win = ntohs(th->th_win);
1089 	th->th_urp = ntohs(th->th_urp);
1090 }
1091 
1092 static inline void
1093 tcp_fields_to_net(struct tcphdr *th)
1094 {
1095 
1096 	th->th_seq = htonl(th->th_seq);
1097 	th->th_ack = htonl(th->th_ack);
1098 	th->th_win = htons(th->th_win);
1099 	th->th_urp = htons(th->th_urp);
1100 }
1101 #endif /* _KERNEL */
1102 
1103 #endif /* _NETINET_TCP_VAR_H_ */
1104