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 */
33
34 #ifndef _NETINET_TCP_VAR_H_
35 #define _NETINET_TCP_VAR_H_
36
37 #include <netinet/tcp.h>
38 #include <netinet/tcp_fsm.h>
39
40 #ifdef _KERNEL
41 #include <net/vnet.h>
42 #include <sys/mbuf.h>
43 #endif
44
45 #define TCP_END_BYTE_INFO 8 /* Bytes that makeup the "end information array" */
46 /* Types of ending byte info */
47 #define TCP_EI_EMPTY_SLOT 0
48 #define TCP_EI_STATUS_CLIENT_FIN 0x1
49 #define TCP_EI_STATUS_CLIENT_RST 0x2
50 #define TCP_EI_STATUS_SERVER_FIN 0x3
51 #define TCP_EI_STATUS_SERVER_RST 0x4
52 #define TCP_EI_STATUS_RETRAN 0x5
53 #define TCP_EI_STATUS_PROGRESS 0x6
54 #define TCP_EI_STATUS_PERSIST_MAX 0x7
55 #define TCP_EI_STATUS_KEEP_MAX 0x8
56 #define TCP_EI_STATUS_DATA_A_CLOSE 0x9
57 #define TCP_EI_STATUS_RST_IN_FRONT 0xa
58 #define TCP_EI_STATUS_2MSL 0xb
59 #define TCP_EI_STATUS_MAX_VALUE 0xb
60
61 /************************************************/
62 /* Status bits we track to assure no duplicates,
63 * the bits here are not used by the code but
64 * for human representation. To check a bit we
65 * take and shift over by 1 minus the value (1-8).
66 */
67 /************************************************/
68 #define TCP_EI_BITS_CLIENT_FIN 0x001
69 #define TCP_EI_BITS_CLIENT_RST 0x002
70 #define TCP_EI_BITS_SERVER_FIN 0x004
71 #define TCP_EI_BITS_SERVER_RST 0x008
72 #define TCP_EI_BITS_RETRAN 0x010
73 #define TCP_EI_BITS_PROGRESS 0x020
74 #define TCP_EI_BITS_PRESIST_MAX 0x040
75 #define TCP_EI_BITS_KEEP_MAX 0x080
76 #define TCP_EI_BITS_DATA_A_CLO 0x100
77 #define TCP_EI_BITS_RST_IN_FR 0x200 /* a front state reset */
78 #define TCP_EI_BITS_2MS_TIMER 0x400 /* 2 MSL timer expired */
79
80 #if defined(_KERNEL) || defined(_WANT_TCPCB)
81 /* TCP segment queue entry */
82 struct tseg_qent {
83 TAILQ_ENTRY(tseg_qent) tqe_q;
84 struct mbuf *tqe_m; /* mbuf contains packet */
85 struct mbuf *tqe_last; /* last mbuf in chain */
86 tcp_seq tqe_start; /* TCP Sequence number start */
87 int tqe_len; /* TCP segment data length */
88 uint32_t tqe_flags; /* The flags from the th->th_flags */
89 uint32_t tqe_mbuf_cnt; /* Count of mbuf overhead */
90 };
91 TAILQ_HEAD(tsegqe_head, tseg_qent);
92
93 struct sackblk {
94 tcp_seq start; /* start seq no. of sack block */
95 tcp_seq end; /* end seq no. */
96 };
97
98 struct sackhole {
99 tcp_seq start; /* start seq no. of hole */
100 tcp_seq end; /* end seq no. */
101 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */
102 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */
103 };
104
105 struct sackhint {
106 struct sackhole *nexthole;
107 int32_t sack_bytes_rexmit;
108 tcp_seq last_sack_ack; /* Most recent/largest sacked ack */
109
110 int32_t delivered_data; /* Newly acked data from last SACK */
111
112 int32_t sacked_bytes; /* Total sacked bytes reported by the
113 * receiver via sack option
114 */
115 uint32_t recover_fs; /* Flight Size at the start of Loss recovery */
116 uint32_t prr_delivered; /* Total bytes delivered using PRR */
117 uint32_t prr_out; /* Bytes sent during IN_RECOVERY */
118 };
119
120 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
121
122 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
123
124 /*
125 * Tcp control block, one per tcp; fields:
126 * Organized for 64 byte cacheline efficiency based
127 * on common tcp_input/tcp_output processing.
128 */
129 struct tcpcb {
130 /* Cache line 1 */
131 struct inpcb *t_inpcb; /* back pointer to internet pcb */
132 struct tcp_function_block *t_fb;/* TCP function call block */
133 void *t_fb_ptr; /* Pointer to t_fb specific data */
134 uint32_t t_maxseg:24, /* maximum segment size */
135 t_logstate:8; /* State of "black box" logging */
136 uint32_t t_port:16, /* Tunneling (over udp) port */
137 t_state:4, /* state of this connection */
138 t_idle_reduce : 1,
139 t_delayed_ack: 7, /* Delayed ack variable */
140 t_fin_is_rst: 1, /* Are fin's treated as resets */
141 t_log_state_set: 1,
142 bits_spare : 2;
143 u_int t_flags;
144 tcp_seq snd_una; /* sent but unacknowledged */
145 tcp_seq snd_max; /* highest sequence number sent;
146 * used to recognize retransmits
147 */
148 tcp_seq snd_nxt; /* send next */
149 tcp_seq snd_up; /* send urgent pointer */
150 uint32_t snd_wnd; /* send window */
151 uint32_t snd_cwnd; /* congestion-controlled window */
152 uint32_t t_peakrate_thr; /* pre-calculated peak rate threshold */
153 /* Cache line 2 */
154 u_int32_t ts_offset; /* our timestamp offset */
155 u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */
156 int rcv_numsacks; /* # distinct sack blks present */
157 u_int t_tsomax; /* TSO total burst length limit in bytes */
158 u_int t_tsomaxsegcount; /* TSO maximum segment count */
159 u_int t_tsomaxsegsize; /* TSO maximum segment size in bytes */
160 tcp_seq rcv_nxt; /* receive next */
161 tcp_seq rcv_adv; /* advertised window */
162 uint32_t rcv_wnd; /* receive window */
163 u_int t_flags2; /* More tcpcb flags storage */
164 int t_srtt; /* smoothed round-trip time */
165 int t_rttvar; /* variance in round-trip time */
166 u_int32_t ts_recent; /* timestamp echo data */
167 u_char snd_scale; /* window scaling for send window */
168 u_char rcv_scale; /* window scaling for recv window */
169 u_char snd_limited; /* segments limited transmitted */
170 u_char request_r_scale; /* pending window scaling */
171 tcp_seq last_ack_sent;
172 u_int t_rcvtime; /* inactivity time */
173 /* Cache line 3 */
174 tcp_seq rcv_up; /* receive urgent pointer */
175 int t_segqlen; /* segment reassembly queue length */
176 uint32_t t_segqmbuflen; /* Count of bytes mbufs on all entries */
177 struct tsegqe_head t_segq; /* segment reassembly queue */
178 struct mbuf *t_in_pkt;
179 struct mbuf *t_tail_pkt;
180 struct tcp_timer *t_timers; /* All the TCP timers in one struct */
181 struct vnet *t_vnet; /* back pointer to parent vnet */
182 uint32_t snd_ssthresh; /* snd_cwnd size threshold for
183 * for slow start exponential to
184 * linear switch
185 */
186 tcp_seq snd_wl1; /* window update seg seq number */
187 /* Cache line 4 */
188 tcp_seq snd_wl2; /* window update seg ack number */
189
190 tcp_seq irs; /* initial receive sequence number */
191 tcp_seq iss; /* initial send sequence number */
192 u_int t_acktime; /* RACK and BBR incoming new data was acked */
193 u_int t_sndtime; /* time last data was sent */
194 u_int ts_recent_age; /* when last updated */
195 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */
196 uint16_t cl4_spare; /* Spare to adjust CL 4 */
197 char t_oobflags; /* have some */
198 char t_iobc; /* input character */
199 int t_rxtcur; /* current retransmit value (ticks) */
200
201 int t_rxtshift; /* log(2) of rexmt exp. backoff */
202 u_int t_rtttime; /* RTT measurement start time */
203
204 tcp_seq t_rtseq; /* sequence number being timed */
205 u_int t_starttime; /* time connection was established */
206 u_int t_fbyte_in; /* ticks time when first byte queued in */
207 u_int t_fbyte_out; /* ticks time when first byte queued out */
208
209 u_int t_pmtud_saved_maxseg; /* pre-blackhole MSS */
210 int t_blackhole_enter; /* when to enter blackhole detection */
211 int t_blackhole_exit; /* when to exit blackhole detection */
212 u_int t_rttmin; /* minimum rtt allowed */
213
214 int t_softerror; /* possible error not yet reported */
215 uint32_t max_sndwnd; /* largest window peer has offered */
216 /* Cache line 5 */
217 uint32_t snd_cwnd_prev; /* cwnd prior to retransmit */
218 uint32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
219 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */
220 int t_sndzerowin; /* zero-window updates sent */
221 u_long t_rttupdated; /* number of times rtt sampled */
222 int snd_numholes; /* number of holes seen by sender */
223 u_int t_badrxtwin; /* window for retransmit recovery */
224 TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
225 /* SACK scoreboard (sorted) */
226 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/
227 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
228 struct sackhint sackhint; /* SACK scoreboard hint */
229 int t_rttlow; /* smallest observerved RTT */
230 int rfbuf_cnt; /* recv buffer autoscaling byte count */
231 struct toedev *tod; /* toedev handling this connection */
232 int t_sndrexmitpack; /* retransmit packets sent */
233 int t_rcvoopack; /* out-of-order packets received */
234 void *t_toe; /* TOE pcb pointer */
235 struct cc_algo *cc_algo; /* congestion control algorithm */
236 struct cc_var *ccv; /* congestion control specific vars */
237 struct osd *osd; /* storage for Khelp module data */
238 int t_bytes_acked; /* # bytes acked during current RTT */
239 u_int t_maxunacktime;
240 u_int t_keepinit; /* time to establish connection */
241 u_int t_keepidle; /* time before keepalive probes begin */
242 u_int t_keepintvl; /* interval between keepalives */
243 u_int t_keepcnt; /* number of keepalives before close */
244 int t_dupacks; /* consecutive dup acks recd */
245 int t_lognum; /* Number of log entries */
246 int t_loglimit; /* Maximum number of log entries */
247 int64_t t_pacing_rate; /* bytes / sec, -1 => unlimited */
248 struct tcp_log_stailq t_logs; /* Log buffer */
249 struct tcp_log_id_node *t_lin;
250 struct tcp_log_id_bucket *t_lib;
251 const char *t_output_caller; /* Function that called tcp_output */
252 struct statsblob *t_stats; /* Per-connection stats */
253 uint32_t t_logsn; /* Log "serial number" */
254 uint32_t gput_ts; /* Time goodput measurement started */
255 tcp_seq gput_seq; /* Outbound measurement seq */
256 tcp_seq gput_ack; /* Inbound measurement ack */
257 int32_t t_stats_gput_prev; /* XXXLAS: Prev gput measurement */
258 uint32_t t_maxpeakrate; /* max peak rate set by user, in bytes/s */
259 uint32_t t_sndtlppack; /* tail loss probe packets sent */
260 uint64_t t_sndtlpbyte; /* total tail loss probe bytes sent */
261 uint64_t t_sndbytes; /* total bytes sent */
262 uint64_t t_snd_rxt_bytes; /* total bytes retransmitted */
263
264 uint8_t t_tfo_client_cookie_len; /* TCP Fast Open client cookie length */
265 uint32_t t_end_info_status; /* Status flag of end info */
266 unsigned int *t_tfo_pending; /* TCP Fast Open server pending counter */
267 union {
268 uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
269 uint64_t server;
270 } t_tfo_cookie; /* TCP Fast Open cookie to send */
271 union {
272 uint8_t t_end_info_bytes[TCP_END_BYTE_INFO];
273 uint64_t t_end_info;
274 };
275 #ifdef TCPPCAP
276 struct mbufq t_inpkts; /* List of saved input packets. */
277 struct mbufq t_outpkts; /* List of saved output packets. */
278 #endif
279 };
280 #endif /* _KERNEL || _WANT_TCPCB */
281
282 #ifdef _KERNEL
283 struct tcptemp {
284 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
285 struct tcphdr tt_t;
286 };
287
288 /* Enable TCP/UDP tunneling port */
289 #define TCP_TUNNELING_PORT_MIN 0
290 #define TCP_TUNNELING_PORT_MAX 65535
291 #define TCP_TUNNELING_PORT_DEFAULT 0
292
293 /* Enable TCP/UDP tunneling port */
294 #define TCP_TUNNELING_OVERHEAD_MIN sizeof(struct udphdr)
295 #define TCP_TUNNELING_OVERHEAD_MAX 1024
296 #define TCP_TUNNELING_OVERHEAD_DEFAULT TCP_TUNNELING_OVERHEAD_MIN
297
298 /* Minimum map entries limit value, if set */
299 #define TCP_MIN_MAP_ENTRIES_LIMIT 128
300
301 /*
302 * TODO: We yet need to brave plowing in
303 * to tcp_input() and the pru_usrreq() block.
304 * Right now these go to the old standards which
305 * are somewhat ok, but in the long term may
306 * need to be changed. If we do tackle tcp_input()
307 * then we need to get rid of the tcp_do_segment()
308 * function below.
309 */
310 /* Flags for tcp functions */
311 #define TCP_FUNC_BEING_REMOVED 0x01 /* Can no longer be referenced */
312
313 /*
314 * If defining the optional tcp_timers, in the
315 * tfb_tcp_timer_stop call you must use the
316 * callout_async_drain() function with the
317 * tcp_timer_discard callback. You should check
318 * the return of callout_async_drain() and if 0
319 * increment tt_draincnt. Since the timer sub-system
320 * does not know your callbacks you must provide a
321 * stop_all function that loops through and calls
322 * tcp_timer_stop() with each of your defined timers.
323 * Adding a tfb_tcp_handoff_ok function allows the socket
324 * option to change stacks to query you even if the
325 * connection is in a later stage. You return 0 to
326 * say you can take over and run your stack, you return
327 * non-zero (an error number) to say no you can't.
328 * If the function is undefined you can only change
329 * in the early states (before connect or listen).
330 * tfb_tcp_fb_fini is changed to add a flag to tell
331 * the old stack if the tcb is being destroyed or
332 * not. A one in the flag means the TCB is being
333 * destroyed, a zero indicates its transitioning to
334 * another stack (via socket option).
335 */
336 struct tcp_function_block {
337 char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
338 int (*tfb_tcp_output)(struct tcpcb *);
339 int (*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *);
340 void (*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
341 struct socket *, struct tcpcb *,
342 int, int, uint8_t);
343 int (*tfb_do_queued_segments)(struct socket *, struct tcpcb *, int);
344 int (*tfb_do_segment_nounlock)(struct mbuf *, struct tcphdr *,
345 struct socket *, struct tcpcb *,
346 int, int, uint8_t,
347 int, struct timeval *);
348 void (*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *,
349 struct socket *, struct tcpcb *,
350 int, int, uint8_t,
351 int, struct timeval *);
352 int (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
353 struct inpcb *inp, struct tcpcb *tp);
354 /* Optional memory allocation/free routine */
355 int (*tfb_tcp_fb_init)(struct tcpcb *);
356 void (*tfb_tcp_fb_fini)(struct tcpcb *, int);
357 /* Optional timers, must define all if you define one */
358 int (*tfb_tcp_timer_stop_all)(struct tcpcb *);
359 void (*tfb_tcp_timer_activate)(struct tcpcb *,
360 uint32_t, u_int);
361 int (*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
362 void (*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
363 void (*tfb_tcp_rexmit_tmr)(struct tcpcb *);
364 int (*tfb_tcp_handoff_ok)(struct tcpcb *);
365 void (*tfb_tcp_mtu_chg)(struct tcpcb *);
366 int (*tfb_pru_options)(struct tcpcb *, int);
367 void (*tfb_hwtls_change)(struct tcpcb *, int);
368 volatile uint32_t tfb_refcnt;
369 uint32_t tfb_flags;
370 uint8_t tfb_id;
371 };
372
373 struct tcp_function {
374 TAILQ_ENTRY(tcp_function) tf_next;
375 char tf_name[TCP_FUNCTION_NAME_LEN_MAX];
376 struct tcp_function_block *tf_fb;
377 };
378
379 TAILQ_HEAD(tcp_funchead, tcp_function);
380 #endif /* _KERNEL */
381
382 /*
383 * Flags and utility macros for the t_flags field.
384 */
385 #define TF_ACKNOW 0x00000001 /* ack peer immediately */
386 #define TF_DELACK 0x00000002 /* ack, but try to delay it */
387 #define TF_NODELAY 0x00000004 /* don't delay packets to coalesce */
388 #define TF_NOOPT 0x00000008 /* don't use tcp options */
389 #define TF_SENTFIN 0x00000010 /* have sent FIN */
390 #define TF_REQ_SCALE 0x00000020 /* have/will request window scaling */
391 #define TF_RCVD_SCALE 0x00000040 /* other side has requested scaling */
392 #define TF_REQ_TSTMP 0x00000080 /* have/will request timestamps */
393 #define TF_RCVD_TSTMP 0x00000100 /* a timestamp was received in SYN */
394 #define TF_SACK_PERMIT 0x00000200 /* other side said I could SACK */
395 #define TF_NEEDSYN 0x00000400 /* send SYN (implicit state) */
396 #define TF_NEEDFIN 0x00000800 /* send FIN (implicit state) */
397 #define TF_NOPUSH 0x00001000 /* don't push */
398 #define TF_PREVVALID 0x00002000 /* saved values for bad rxmit valid
399 * Note: accessing and restoring from
400 * these may only be done in the 1st
401 * RTO recovery round (t_rxtshift == 1)
402 */
403 #define TF_WAKESOR 0x00004000 /* wake up receive socket */
404 #define TF_GPUTINPROG 0x00008000 /* Goodput measurement in progress */
405 #define TF_MORETOCOME 0x00010000 /* More data to be appended to sock */
406 #define TF_LQ_OVERFLOW 0x00020000 /* listen queue overflow */
407 #define TF_LASTIDLE 0x00040000 /* connection was previously idle */
408 #define TF_RXWIN0SENT 0x00080000 /* sent a receiver win 0 in response */
409 #define TF_FASTRECOVERY 0x00100000 /* in NewReno Fast Recovery */
410 #define TF_WASFRECOVERY 0x00200000 /* was in NewReno Fast Recovery */
411 #define TF_SIGNATURE 0x00400000 /* require MD5 digests (RFC2385) */
412 #define TF_FORCEDATA 0x00800000 /* force out a byte */
413 #define TF_TSO 0x01000000 /* TSO enabled on this connection */
414 #define TF_TOE 0x02000000 /* this connection is offloaded */
415 #define TF_UNUSED0 0x04000000 /* unused */
416 #define TF_UNUSED1 0x08000000 /* unused */
417 #define TF_UNUSED2 0x10000000 /* unused */
418 #define TF_CONGRECOVERY 0x20000000 /* congestion recovery mode */
419 #define TF_WASCRECOVERY 0x40000000 /* was in congestion recovery */
420 #define TF_FASTOPEN 0x80000000 /* TCP Fast Open indication */
421
422 #define IN_FASTRECOVERY(t_flags) (t_flags & TF_FASTRECOVERY)
423 #define ENTER_FASTRECOVERY(t_flags) t_flags |= TF_FASTRECOVERY
424 #define EXIT_FASTRECOVERY(t_flags) t_flags &= ~TF_FASTRECOVERY
425
426 #define IN_CONGRECOVERY(t_flags) (t_flags & TF_CONGRECOVERY)
427 #define ENTER_CONGRECOVERY(t_flags) t_flags |= TF_CONGRECOVERY
428 #define EXIT_CONGRECOVERY(t_flags) t_flags &= ~TF_CONGRECOVERY
429
430 #define IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
431 #define ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
432 #define EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
433
434 #if defined(_KERNEL)
435 #if !defined(TCP_RFC7413)
436 #define IS_FASTOPEN(t_flags) (false)
437 #else
438 #define IS_FASTOPEN(t_flags) (t_flags & TF_FASTOPEN)
439 #endif
440 #endif
441
442 #define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una)
443
444 /*
445 * Flags for the t_oobflags field.
446 */
447 #define TCPOOB_HAVEDATA 0x01
448 #define TCPOOB_HADDATA 0x02
449
450 /*
451 * Flags for the extended TCP flags field, t_flags2
452 */
453 #define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */
454 #define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */
455 #define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */
456 #define TF2_LOG_AUTO 0x00000008 /* Session is auto-logging. */
457 #define TF2_DROP_AF_DATA 0x00000010 /* Drop after all data ack'd */
458 #define TF2_ECN_PERMIT 0x00000020 /* connection ECN-ready */
459 #define TF2_ECN_SND_CWR 0x00000040 /* ECN CWR in queue */
460 #define TF2_ECN_SND_ECE 0x00000080 /* ECN ECE in queue */
461 #define TF2_ACE_PERMIT 0x00000100 /* Accurate ECN mode */
462 #define TF2_FBYTES_COMPLETE 0x00000400 /* We have first bytes in and out */
463 /*
464 * Structure to hold TCP options that are only used during segment
465 * processing (in tcp_input), but not held in the tcpcb.
466 * It's basically used to reduce the number of parameters
467 * to tcp_dooptions and tcp_addoptions.
468 * The binary order of the to_flags is relevant for packing of the
469 * options in tcp_addoptions.
470 */
471 struct tcpopt {
472 u_int32_t to_flags; /* which options are present */
473 #define TOF_MSS 0x0001 /* maximum segment size */
474 #define TOF_SCALE 0x0002 /* window scaling */
475 #define TOF_SACKPERM 0x0004 /* SACK permitted */
476 #define TOF_TS 0x0010 /* timestamp */
477 #define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */
478 #define TOF_SACK 0x0080 /* Peer sent SACK option */
479 #define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */
480 #define TOF_MAXOPT 0x0200
481 u_int32_t to_tsval; /* new timestamp */
482 u_int32_t to_tsecr; /* reflected timestamp */
483 u_char *to_sacks; /* pointer to the first SACK blocks */
484 u_char *to_signature; /* pointer to the TCP-MD5 signature */
485 u_int8_t *to_tfo_cookie; /* pointer to the TFO cookie */
486 u_int16_t to_mss; /* maximum segment size */
487 u_int8_t to_wscale; /* window scaling */
488 u_int8_t to_nsacks; /* number of SACK blocks */
489 u_int8_t to_tfo_len; /* TFO cookie length */
490 u_int32_t to_spare; /* UTO */
491 };
492
493 /*
494 * Flags for tcp_dooptions.
495 */
496 #define TO_SYN 0x01 /* parse SYN-only options */
497
498 struct hc_metrics_lite { /* must stay in sync with hc_metrics */
499 uint32_t rmx_mtu; /* MTU for this path */
500 uint32_t rmx_ssthresh; /* outbound gateway buffer limit */
501 uint32_t rmx_rtt; /* estimated round trip time */
502 uint32_t rmx_rttvar; /* estimated rtt variance */
503 uint32_t rmx_cwnd; /* congestion window */
504 uint32_t rmx_sendpipe; /* outbound delay-bandwidth product */
505 uint32_t rmx_recvpipe; /* inbound delay-bandwidth product */
506 };
507
508 /*
509 * Used by tcp_maxmtu() to communicate interface specific features
510 * and limits at the time of connection setup.
511 */
512 struct tcp_ifcap {
513 int ifcap;
514 u_int tsomax;
515 u_int tsomaxsegcount;
516 u_int tsomaxsegsize;
517 };
518
519 #ifndef _NETINET_IN_PCB_H_
520 struct in_conninfo;
521 #endif /* _NETINET_IN_PCB_H_ */
522
523 struct tcptw {
524 struct inpcb *tw_inpcb; /* XXX back pointer to internet pcb */
525 uint32_t t_port:16, /* UDP port number if TCPoUDP */
526 t_unused:16;
527 tcp_seq snd_nxt;
528 tcp_seq rcv_nxt;
529 tcp_seq iss;
530 tcp_seq irs;
531 u_short last_win; /* cached window value */
532 short tw_so_options; /* copy of so_options */
533 struct ucred *tw_cred; /* user credentials */
534 u_int32_t t_recent;
535 u_int32_t ts_offset; /* our timestamp offset */
536 u_int t_starttime;
537 int tw_time;
538 TAILQ_ENTRY(tcptw) tw_2msl;
539 void *tw_pspare; /* TCP_SIGNATURE */
540 u_int tw_flags; /* tcpcb t_flags */
541 };
542
543 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb)
544 #define intotw(ip) ((struct tcptw *)(ip)->inp_ppcb)
545 #define sototcpcb(so) (intotcpcb(sotoinpcb(so)))
546
547 /*
548 * The smoothed round-trip time and estimated variance
549 * are stored as fixed point numbers scaled by the values below.
550 * For convenience, these scales are also used in smoothing the average
551 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
552 * With these scales, srtt has 3 bits to the right of the binary point,
553 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the
554 * binary point, and is smoothed with an ALPHA of 0.75.
555 */
556 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 5 bits frac. */
557 #define TCP_RTT_SHIFT 5 /* shift for srtt; 5 bits frac. */
558 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 4 bits */
559 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 4 bits */
560 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
561
562 /*
563 * The initial retransmission should happen at rtt + 4 * rttvar.
564 * Because of the way we do the smoothing, srtt and rttvar
565 * will each average +1/2 tick of bias. When we compute
566 * the retransmit timer, we want 1/2 tick of rounding and
567 * 1 extra tick because of +-1/2 tick uncertainty in the
568 * firing of the timer. The bias will give us exactly the
569 * 1.5 tick we need. But, because the bias is
570 * statistical, we have to test that we don't drop below
571 * the minimum feasible timer (which is 2 ticks).
572 * This version of the macro adapted from a paper by Lawrence
573 * Brakmo and Larry Peterson which outlines a problem caused
574 * by insufficient precision in the original implementation,
575 * which results in inappropriately large RTO values for very
576 * fast networks.
577 */
578 #define TCP_REXMTVAL(tp) \
579 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
580 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
581
582 /*
583 * TCP statistics.
584 * Many of these should be kept per connection,
585 * but that's inconvenient at the moment.
586 */
587 struct tcpstat {
588 uint64_t tcps_connattempt; /* connections initiated */
589 uint64_t tcps_accepts; /* connections accepted */
590 uint64_t tcps_connects; /* connections established */
591 uint64_t tcps_drops; /* connections dropped */
592 uint64_t tcps_conndrops; /* embryonic connections dropped */
593 uint64_t tcps_minmssdrops; /* average minmss too low drops */
594 uint64_t tcps_closed; /* conn. closed (includes drops) */
595 uint64_t tcps_segstimed; /* segs where we tried to get rtt */
596 uint64_t tcps_rttupdated; /* times we succeeded */
597 uint64_t tcps_delack; /* delayed acks sent */
598 uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
599 uint64_t tcps_rexmttimeo; /* retransmit timeouts */
600 uint64_t tcps_persisttimeo; /* persist timeouts */
601 uint64_t tcps_keeptimeo; /* keepalive timeouts */
602 uint64_t tcps_keepprobe; /* keepalive probes sent */
603 uint64_t tcps_keepdrops; /* connections dropped in keepalive */
604
605 uint64_t tcps_sndtotal; /* total packets sent */
606 uint64_t tcps_sndpack; /* data packets sent */
607 uint64_t tcps_sndbyte; /* data bytes sent */
608 uint64_t tcps_sndrexmitpack; /* data packets retransmitted */
609 uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */
610 uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */
611 uint64_t tcps_sndacks; /* ack-only packets sent */
612 uint64_t tcps_sndprobe; /* window probes sent */
613 uint64_t tcps_sndurg; /* packets sent with URG only */
614 uint64_t tcps_sndwinup; /* window update-only packets sent */
615 uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
616
617 uint64_t tcps_rcvtotal; /* total packets received */
618 uint64_t tcps_rcvpack; /* packets received in sequence */
619 uint64_t tcps_rcvbyte; /* bytes received in sequence */
620 uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */
621 uint64_t tcps_rcvbadoff; /* packets received with bad offset */
622 uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */
623 uint64_t tcps_rcvshort; /* packets received too short */
624 uint64_t tcps_rcvduppack; /* duplicate-only packets received */
625 uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */
626 uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */
627 uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
628 uint64_t tcps_rcvoopack; /* out-of-order packets received */
629 uint64_t tcps_rcvoobyte; /* out-of-order bytes received */
630 uint64_t tcps_rcvpackafterwin; /* packets with data after window */
631 uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */
632 uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */
633 uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */
634 uint64_t tcps_rcvdupack; /* rcvd duplicate acks */
635 uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */
636 uint64_t tcps_rcvackpack; /* rcvd ack packets */
637 uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */
638 uint64_t tcps_rcvwinupd; /* rcvd window update packets */
639 uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */
640 uint64_t tcps_predack; /* times hdr predict ok for acks */
641 uint64_t tcps_preddat; /* times hdr predict ok for data pkts */
642 uint64_t tcps_pcbcachemiss;
643 uint64_t tcps_cachedrtt; /* times cached RTT in route updated */
644 uint64_t tcps_cachedrttvar; /* times cached rttvar updated */
645 uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */
646 uint64_t tcps_usedrtt; /* times RTT initialized from route */
647 uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */
648 uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/
649 uint64_t tcps_persistdrop; /* timeout in persist state */
650 uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */
651 uint64_t tcps_mturesent; /* resends due to MTU discovery */
652 uint64_t tcps_listendrop; /* listen queue overflows */
653 uint64_t tcps_badrst; /* ignored RSTs in the window */
654
655 uint64_t tcps_sc_added; /* entry added to syncache */
656 uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */
657 uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */
658 uint64_t tcps_sc_dropped; /* could not reply to packet */
659 uint64_t tcps_sc_completed; /* successful extraction of entry */
660 uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
661 uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */
662 uint64_t tcps_sc_reset; /* RST removed entry from syncache */
663 uint64_t tcps_sc_stale; /* timed out or listen socket gone */
664 uint64_t tcps_sc_aborted; /* syncache entry aborted */
665 uint64_t tcps_sc_badack; /* removed due to bad ACK */
666 uint64_t tcps_sc_unreach; /* ICMP unreachable received */
667 uint64_t tcps_sc_zonefail; /* zalloc() failed */
668 uint64_t tcps_sc_sendcookie; /* SYN cookie sent */
669 uint64_t tcps_sc_recvcookie; /* SYN cookie received */
670
671 uint64_t tcps_hc_added; /* entry added to hostcache */
672 uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
673
674 uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */
675
676 /* SACK related stats */
677 uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
678 uint64_t tcps_sack_rexmits; /* SACK rexmit segments */
679 uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */
680 uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */
681 uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */
682 uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */
683
684 /* ECN related stats */
685 uint64_t tcps_ecn_ce; /* ECN Congestion Experienced */
686 uint64_t tcps_ecn_ect0; /* ECN Capable Transport */
687 uint64_t tcps_ecn_ect1; /* ECN Capable Transport */
688 uint64_t tcps_ecn_shs; /* ECN successful handshakes */
689 uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */
690
691 /* TCP_SIGNATURE related stats */
692 uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */
693 uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */
694 uint64_t tcps_sig_err_buildsig; /* Failed to make signature */
695 uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */
696 uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */
697
698 /* Path MTU Discovery Black Hole Detection related stats */
699 uint64_t tcps_pmtud_blackhole_activated; /* Black Hole Count */
700 uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
701 uint64_t tcps_pmtud_blackhole_failed; /* Black Hole Failure Count */
702
703 uint64_t tcps_tunneled_pkts; /* Packets encap's in UDP received */
704 uint64_t tcps_tunneled_errs; /* Packets that had errors that were UDP encaped */
705
706 /* TCPS_TIME_WAIT usage stats */
707 uint64_t tcps_tw_recycles; /* Times time-wait was recycled. */
708 uint64_t tcps_tw_resets; /* Times time-wait sent a reset. */
709 uint64_t tcps_tw_responds; /* Times time-wait sent a valid ack. */
710
711 uint64_t _pad[7];
712 };
713
714 #define tcps_rcvmemdrop tcps_rcvreassfull /* compat */
715
716 #ifdef _KERNEL
717 #define TI_UNLOCKED 1
718 #define TI_RLOCKED 2
719 #include <sys/counter.h>
720
721 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */
722 /*
723 * In-kernel consumers can use these accessor macros directly to update
724 * stats.
725 */
726 #define TCPSTAT_ADD(name, val) \
727 VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
728 #define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1)
729
730 /*
731 * Kernel module consumers must use this accessor macro.
732 */
733 void kmod_tcpstat_add(int statnum, int val);
734 #define KMOD_TCPSTAT_ADD(name, val) \
735 kmod_tcpstat_add(offsetof(struct tcpstat, name) / sizeof(uint64_t), val)
736 #define KMOD_TCPSTAT_INC(name) KMOD_TCPSTAT_ADD(name, 1)
737
738 /*
739 * Running TCP connection count by state.
740 */
741 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
742 #define V_tcps_states VNET(tcps_states)
743 #define TCPSTATES_INC(state) counter_u64_add(V_tcps_states[state], 1)
744 #define TCPSTATES_DEC(state) counter_u64_add(V_tcps_states[state], -1)
745
746 /*
747 * TCP specific helper hook point identifiers.
748 */
749 #define HHOOK_TCP_EST_IN 0
750 #define HHOOK_TCP_EST_OUT 1
751 #define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT
752
753 struct tcp_hhook_data {
754 struct tcpcb *tp;
755 struct tcphdr *th;
756 struct tcpopt *to;
757 uint32_t len;
758 int tso;
759 tcp_seq curack;
760 };
761 #ifdef TCP_HHOOK
762 void hhook_run_tcp_est_out(struct tcpcb *tp,
763 struct tcphdr *th, struct tcpopt *to,
764 uint32_t len, int tso);
765 #endif
766 #endif
767
768 /*
769 * TCB structure exported to user-land via sysctl(3).
770 *
771 * Fields prefixed with "xt_" are unique to the export structure, and fields
772 * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
773 *
774 * Legend:
775 * (s) - used by userland utilities in src
776 * (p) - used by utilities in ports
777 * (3) - is known to be used by third party software not in ports
778 * (n) - no known usage
779 *
780 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
781 * included. Not all of our clients do.
782 */
783 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
784 struct xtcpcb {
785 ksize_t xt_len; /* length of this structure */
786 struct xinpcb xt_inp;
787 char xt_stack[TCP_FUNCTION_NAME_LEN_MAX]; /* (s) */
788 char xt_logid[TCP_LOG_ID_LEN]; /* (s) */
789 char xt_cc[TCP_CA_NAME_MAX]; /* (s) */
790 int64_t spare64[6];
791 int32_t t_state; /* (s,p) */
792 uint32_t t_flags; /* (s,p) */
793 int32_t t_sndzerowin; /* (s) */
794 int32_t t_sndrexmitpack; /* (s) */
795 int32_t t_rcvoopack; /* (s) */
796 int32_t t_rcvtime; /* (s) */
797 int32_t tt_rexmt; /* (s) */
798 int32_t tt_persist; /* (s) */
799 int32_t tt_keep; /* (s) */
800 int32_t tt_2msl; /* (s) */
801 int32_t tt_delack; /* (s) */
802 int32_t t_logstate; /* (3) */
803 uint32_t t_snd_cwnd; /* (s) */
804 uint32_t t_snd_ssthresh; /* (s) */
805 uint32_t t_maxseg; /* (s) */
806 uint32_t t_rcv_wnd; /* (s) */
807 uint32_t t_snd_wnd; /* (s) */
808 uint32_t xt_ecn; /* (s) */
809 uint16_t xt_encaps_port; /* (s) */
810 int16_t spare16;
811 int32_t spare32[25];
812 } __aligned(8);
813
814 #ifdef _KERNEL
815 void tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
816 #endif
817 #endif
818
819 /*
820 * TCP function information (name-to-id mapping, aliases, and refcnt)
821 * exported to user-land via sysctl(3).
822 */
823 struct tcp_function_info {
824 uint32_t tfi_refcnt;
825 uint8_t tfi_id;
826 char tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
827 char tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
828 };
829
830 /*
831 * Identifiers for TCP sysctl nodes
832 */
833 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
834 #define TCPCTL_MSSDFLT 3 /* MSS default */
835 #define TCPCTL_STATS 4 /* statistics */
836 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */
837 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
838 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
839 #define TCPCTL_SENDSPACE 8 /* send buffer space */
840 #define TCPCTL_RECVSPACE 9 /* receive buffer space */
841 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */
842 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */
843 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */
844 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */
845 #define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */
846 #define TCPCTL_DROP 15 /* drop tcp connection */
847 #define TCPCTL_STATES 16 /* connection counts by TCP state */
848
849 #ifdef _KERNEL
850 #ifdef SYSCTL_DECL
851 SYSCTL_DECL(_net_inet_tcp);
852 SYSCTL_DECL(_net_inet_tcp_sack);
853 MALLOC_DECLARE(M_TCPLOG);
854 #endif
855
856 VNET_DECLARE(int, tcp_log_in_vain);
857 #define V_tcp_log_in_vain VNET(tcp_log_in_vain)
858
859 /*
860 * Global TCP tunables shared between different stacks.
861 * Please keep the list sorted.
862 */
863 VNET_DECLARE(int, drop_synfin);
864 VNET_DECLARE(int, path_mtu_discovery);
865 VNET_DECLARE(int, tcp_abc_l_var);
866 VNET_DECLARE(int, tcp_autorcvbuf_max);
867 VNET_DECLARE(int, tcp_autosndbuf_inc);
868 VNET_DECLARE(int, tcp_autosndbuf_max);
869 VNET_DECLARE(int, tcp_delack_enabled);
870 VNET_DECLARE(int, tcp_do_autorcvbuf);
871 VNET_DECLARE(int, tcp_do_autosndbuf);
872 VNET_DECLARE(int, tcp_do_ecn);
873 VNET_DECLARE(int, tcp_do_prr);
874 VNET_DECLARE(int, tcp_do_prr_conservative);
875 VNET_DECLARE(int, tcp_do_newcwv);
876 VNET_DECLARE(int, tcp_do_rfc1323);
877 VNET_DECLARE(int, tcp_tolerate_missing_ts);
878 VNET_DECLARE(int, tcp_do_rfc3042);
879 VNET_DECLARE(int, tcp_do_rfc3390);
880 VNET_DECLARE(int, tcp_do_rfc3465);
881 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
882 VNET_DECLARE(int, tcp_do_sack);
883 VNET_DECLARE(int, tcp_do_tso);
884 VNET_DECLARE(int, tcp_ecn_maxretries);
885 VNET_DECLARE(int, tcp_initcwnd_segments);
886 VNET_DECLARE(int, tcp_insecure_rst);
887 VNET_DECLARE(int, tcp_insecure_syn);
888 VNET_DECLARE(uint32_t, tcp_map_entries_limit);
889 VNET_DECLARE(uint32_t, tcp_map_split_limit);
890 VNET_DECLARE(int, tcp_minmss);
891 VNET_DECLARE(int, tcp_mssdflt);
892 #ifdef STATS
893 VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl);
894 VNET_DECLARE(int, tcp_perconn_stats_enable);
895 #endif /* STATS */
896 VNET_DECLARE(int, tcp_recvspace);
897 VNET_DECLARE(int, tcp_sack_globalholes);
898 VNET_DECLARE(int, tcp_sack_globalmaxholes);
899 VNET_DECLARE(int, tcp_sack_maxholes);
900 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
901 VNET_DECLARE(int, tcp_sendspace);
902 VNET_DECLARE(int, tcp_udp_tunneling_overhead);
903 VNET_DECLARE(int, tcp_udp_tunneling_port);
904 VNET_DECLARE(struct inpcbhead, tcb);
905 VNET_DECLARE(struct inpcbinfo, tcbinfo);
906
907 #define V_tcp_do_prr VNET(tcp_do_prr)
908 #define V_tcp_do_prr_conservative VNET(tcp_do_prr_conservative)
909 #define V_tcp_do_newcwv VNET(tcp_do_newcwv)
910 #define V_drop_synfin VNET(drop_synfin)
911 #define V_path_mtu_discovery VNET(path_mtu_discovery)
912 #define V_tcb VNET(tcb)
913 #define V_tcbinfo VNET(tcbinfo)
914 #define V_tcp_abc_l_var VNET(tcp_abc_l_var)
915 #define V_tcp_autorcvbuf_max VNET(tcp_autorcvbuf_max)
916 #define V_tcp_autosndbuf_inc VNET(tcp_autosndbuf_inc)
917 #define V_tcp_autosndbuf_max VNET(tcp_autosndbuf_max)
918 #define V_tcp_delack_enabled VNET(tcp_delack_enabled)
919 #define V_tcp_do_autorcvbuf VNET(tcp_do_autorcvbuf)
920 #define V_tcp_do_autosndbuf VNET(tcp_do_autosndbuf)
921 #define V_tcp_do_ecn VNET(tcp_do_ecn)
922 #define V_tcp_do_rfc1323 VNET(tcp_do_rfc1323)
923 #define V_tcp_tolerate_missing_ts VNET(tcp_tolerate_missing_ts)
924 #define V_tcp_ts_offset_per_conn VNET(tcp_ts_offset_per_conn)
925 #define V_tcp_do_rfc3042 VNET(tcp_do_rfc3042)
926 #define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390)
927 #define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465)
928 #define V_tcp_do_rfc6675_pipe VNET(tcp_do_rfc6675_pipe)
929 #define V_tcp_do_sack VNET(tcp_do_sack)
930 #define V_tcp_do_tso VNET(tcp_do_tso)
931 #define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries)
932 #define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments)
933 #define V_tcp_insecure_rst VNET(tcp_insecure_rst)
934 #define V_tcp_insecure_syn VNET(tcp_insecure_syn)
935 #define V_tcp_map_entries_limit VNET(tcp_map_entries_limit)
936 #define V_tcp_map_split_limit VNET(tcp_map_split_limit)
937 #define V_tcp_minmss VNET(tcp_minmss)
938 #define V_tcp_mssdflt VNET(tcp_mssdflt)
939 #ifdef STATS
940 #define V_tcp_perconn_stats_dflt_tpl VNET(tcp_perconn_stats_dflt_tpl)
941 #define V_tcp_perconn_stats_enable VNET(tcp_perconn_stats_enable)
942 #endif /* STATS */
943 #define V_tcp_recvspace VNET(tcp_recvspace)
944 #define V_tcp_sack_globalholes VNET(tcp_sack_globalholes)
945 #define V_tcp_sack_globalmaxholes VNET(tcp_sack_globalmaxholes)
946 #define V_tcp_sack_maxholes VNET(tcp_sack_maxholes)
947 #define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail)
948 #define V_tcp_sendspace VNET(tcp_sendspace)
949 #define V_tcp_udp_tunneling_overhead VNET(tcp_udp_tunneling_overhead)
950 #define V_tcp_udp_tunneling_port VNET(tcp_udp_tunneling_port)
951
952 #ifdef TCP_HHOOK
953 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
954 #define V_tcp_hhh VNET(tcp_hhh)
955 #endif
956
957 int tcp_addoptions(struct tcpopt *, u_char *);
958 int tcp_ccalgounload(struct cc_algo *unload_algo);
959 struct tcpcb *
960 tcp_close(struct tcpcb *);
961 void tcp_discardcb(struct tcpcb *);
962 void tcp_twstart(struct tcpcb *);
963 void tcp_twclose(struct tcptw *, int);
964 void tcp_ctlinput(int, struct sockaddr *, void *);
965 int tcp_ctloutput(struct socket *, struct sockopt *);
966 void tcp_ctlinput_viaudp(int, struct sockaddr *, void *, void *);
967 struct tcpcb *
968 tcp_drop(struct tcpcb *, int);
969 void tcp_drain(void);
970 void tcp_init(void);
971 void tcp_fini(void *);
972 char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
973 const void *);
974 char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
975 const void *);
976 int tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *,
977 struct mbuf *);
978 void tcp_reass_global_init(void);
979 void tcp_reass_flush(struct tcpcb *);
980 void tcp_dooptions(struct tcpopt *, u_char *, int, int);
981 void tcp_dropwithreset(struct mbuf *, struct tcphdr *,
982 struct tcpcb *, int, int);
983 void tcp_pulloutofband(struct socket *,
984 struct tcphdr *, struct mbuf *, int);
985 void tcp_xmit_timer(struct tcpcb *, int);
986 void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
987 void cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
988 uint16_t nsegs, uint16_t type);
989 void cc_conn_init(struct tcpcb *tp);
990 void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
991 void cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos);
992 void cc_ecnpkt_handler_flags(struct tcpcb *tp, uint16_t flags, uint8_t iptos);
993 void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
994 #ifdef TCP_HHOOK
995 void hhook_run_tcp_est_in(struct tcpcb *tp,
996 struct tcphdr *th, struct tcpopt *to);
997 #endif
998
999 int tcp_input(struct mbuf **, int *, int);
1000 int tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
1001 struct tcpcb *, int);
1002 int tcp_input_with_port(struct mbuf **, int *, int, uint16_t);
1003 void tcp_handle_wakeup(struct tcpcb *, struct socket *);
1004 void tcp_do_segment(struct mbuf *, struct tcphdr *,
1005 struct socket *, struct tcpcb *, int, int, uint8_t);
1006
1007 int register_tcp_functions(struct tcp_function_block *blk, int wait);
1008 int register_tcp_functions_as_names(struct tcp_function_block *blk,
1009 int wait, const char *names[], int *num_names);
1010 int register_tcp_functions_as_name(struct tcp_function_block *blk,
1011 const char *name, int wait);
1012 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
1013 bool force);
1014 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
1015 void tcp_switch_back_to_default(struct tcpcb *tp);
1016 struct tcp_function_block *
1017 find_and_ref_tcp_fb(struct tcp_function_block *fs);
1018 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
1019
1020 extern counter_u64_t tcp_inp_lro_direct_queue;
1021 extern counter_u64_t tcp_inp_lro_wokeup_queue;
1022 extern counter_u64_t tcp_inp_lro_compressed;
1023 extern counter_u64_t tcp_inp_lro_locks_taken;
1024 extern counter_u64_t tcp_extra_mbuf;
1025 extern counter_u64_t tcp_would_have_but;
1026 extern counter_u64_t tcp_comp_total;
1027 extern counter_u64_t tcp_uncomp_total;
1028
1029 #ifdef NETFLIX_EXP_DETECTION
1030 /* Various SACK attack thresholds */
1031 extern int32_t tcp_force_detection;
1032 extern int32_t tcp_sack_to_ack_thresh;
1033 extern int32_t tcp_sack_to_move_thresh;
1034 extern int32_t tcp_restoral_thresh;
1035 extern int32_t tcp_sad_decay_val;
1036 extern int32_t tcp_sad_pacing_interval;
1037 extern int32_t tcp_sad_low_pps;
1038 extern int32_t tcp_map_minimum;
1039 extern int32_t tcp_attack_on_turns_on_logging;
1040 #endif
1041 extern uint32_t tcp_ack_war_time_window;
1042 extern uint32_t tcp_ack_war_cnt;
1043
1044 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
1045 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
1046 u_int tcp_maxseg(const struct tcpcb *);
1047 u_int tcp_fixed_maxseg(const struct tcpcb *);
1048 void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
1049 struct tcp_ifcap *);
1050 void tcp_mss(struct tcpcb *, int);
1051 int tcp_mssopt(struct in_conninfo *);
1052 struct inpcb *
1053 tcp_drop_syn_sent(struct inpcb *, int);
1054 struct tcpcb *
1055 tcp_newtcpcb(struct inpcb *);
1056 int tcp_output(struct tcpcb *);
1057 void tcp_state_change(struct tcpcb *, int);
1058 void tcp_respond(struct tcpcb *, void *,
1059 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
1060 void tcp_tw_init(void);
1061 #ifdef VIMAGE
1062 void tcp_tw_destroy(void);
1063 #endif
1064 void tcp_tw_zone_change(void);
1065 int tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
1066 struct mbuf *, int);
1067 void tcp_setpersist(struct tcpcb *);
1068 void tcp_slowtimo(void);
1069 struct tcptemp *
1070 tcpip_maketemplate(struct inpcb *);
1071 void tcpip_fillheaders(struct inpcb *, uint16_t, void *, void *);
1072 void tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
1073 int tcp_timer_suspend(struct tcpcb *, uint32_t);
1074 void tcp_timers_unsuspend(struct tcpcb *, uint32_t);
1075 int tcp_timer_active(struct tcpcb *, uint32_t);
1076 void tcp_timer_stop(struct tcpcb *, uint32_t);
1077 void tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
1078 int inp_to_cpuid(struct inpcb *inp);
1079 /*
1080 * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
1081 */
1082 void tcp_hc_init(void);
1083 #ifdef VIMAGE
1084 void tcp_hc_destroy(void);
1085 #endif
1086 void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
1087 uint32_t tcp_hc_getmtu(struct in_conninfo *);
1088 void tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
1089 void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
1090
1091 extern struct pr_usrreqs tcp_usrreqs;
1092
1093 uint32_t tcp_new_ts_offset(struct in_conninfo *);
1094 tcp_seq tcp_new_isn(struct in_conninfo *);
1095
1096 int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
1097 int tcp_dsack_block_exists(struct tcpcb *);
1098 void tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq);
1099 void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1100 void tcp_clean_dsack_blocks(struct tcpcb *tp);
1101 void tcp_clean_sackreport(struct tcpcb *tp);
1102 void tcp_sack_adjust(struct tcpcb *tp);
1103 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1104 void tcp_do_prr_ack(struct tcpcb *, struct tcphdr *, struct tcpopt *);
1105 void tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1106 void tcp_free_sackholes(struct tcpcb *tp);
1107 int tcp_newreno(struct tcpcb *, struct tcphdr *);
1108 int tcp_compute_pipe(struct tcpcb *);
1109 uint32_t tcp_compute_initwnd(uint32_t);
1110 void tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
1111 int tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes,
1112 size_t seed_len);
1113 int tcp_can_enable_pacing(void);
1114 void tcp_decrement_paced_conn(void);
1115
1116 struct mbuf *
1117 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen,
1118 int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls);
1119
1120 int tcp_stats_init(void);
1121 void tcp_log_end_status(struct tcpcb *tp, uint8_t status);
1122
1123 static inline void
tcp_fields_to_host(struct tcphdr * th)1124 tcp_fields_to_host(struct tcphdr *th)
1125 {
1126
1127 th->th_seq = ntohl(th->th_seq);
1128 th->th_ack = ntohl(th->th_ack);
1129 th->th_win = ntohs(th->th_win);
1130 th->th_urp = ntohs(th->th_urp);
1131 }
1132
1133 static inline void
tcp_fields_to_net(struct tcphdr * th)1134 tcp_fields_to_net(struct tcphdr *th)
1135 {
1136
1137 th->th_seq = htonl(th->th_seq);
1138 th->th_ack = htonl(th->th_ack);
1139 th->th_win = htons(th->th_win);
1140 th->th_urp = htons(th->th_urp);
1141 }
1142
1143 static inline void
tcp_account_for_send(struct tcpcb * tp,uint32_t len,uint8_t is_rxt,uint8_t is_tlp,int hw_tls __unused)1144 tcp_account_for_send(struct tcpcb *tp, uint32_t len, uint8_t is_rxt,
1145 uint8_t is_tlp, int hw_tls __unused)
1146 {
1147 if (is_tlp) {
1148 tp->t_sndtlppack++;
1149 tp->t_sndtlpbyte += len;
1150 }
1151 /* To get total bytes sent you must add t_snd_rxt_bytes to t_sndbytes */
1152 if (is_rxt)
1153 tp->t_snd_rxt_bytes += len;
1154 else
1155 tp->t_sndbytes += len;
1156 }
1157 #endif /* _KERNEL */
1158
1159 #endif /* _NETINET_TCP_VAR_H_ */
1160