FreeBSD/Linux Kernel Cross Reference
sys/netinet/tcp_var.h
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 "opt_kern_tls.h"
43 #include <net/vnet.h>
44 #include <sys/mbuf.h>
45 #include <sys/ktls.h>
46 #endif
47
48 #define TCP_END_BYTE_INFO 8 /* Bytes that makeup the "end information array" */
49 /* Types of ending byte info */
50 #define TCP_EI_EMPTY_SLOT 0
51 #define TCP_EI_STATUS_CLIENT_FIN 0x1
52 #define TCP_EI_STATUS_CLIENT_RST 0x2
53 #define TCP_EI_STATUS_SERVER_FIN 0x3
54 #define TCP_EI_STATUS_SERVER_RST 0x4
55 #define TCP_EI_STATUS_RETRAN 0x5
56 #define TCP_EI_STATUS_PROGRESS 0x6
57 #define TCP_EI_STATUS_PERSIST_MAX 0x7
58 #define TCP_EI_STATUS_KEEP_MAX 0x8
59 #define TCP_EI_STATUS_DATA_A_CLOSE 0x9
60 #define TCP_EI_STATUS_RST_IN_FRONT 0xa
61 #define TCP_EI_STATUS_2MSL 0xb
62 #define TCP_EI_STATUS_MAX_VALUE 0xb
63
64 /************************************************/
65 /* Status bits we track to assure no duplicates,
66 * the bits here are not used by the code but
67 * for human representation. To check a bit we
68 * take and shift over by 1 minus the value (1-8).
69 */
70 /************************************************/
71 #define TCP_EI_BITS_CLIENT_FIN 0x001
72 #define TCP_EI_BITS_CLIENT_RST 0x002
73 #define TCP_EI_BITS_SERVER_FIN 0x004
74 #define TCP_EI_BITS_SERVER_RST 0x008
75 #define TCP_EI_BITS_RETRAN 0x010
76 #define TCP_EI_BITS_PROGRESS 0x020
77 #define TCP_EI_BITS_PRESIST_MAX 0x040
78 #define TCP_EI_BITS_KEEP_MAX 0x080
79 #define TCP_EI_BITS_DATA_A_CLO 0x100
80 #define TCP_EI_BITS_RST_IN_FR 0x200 /* a front state reset */
81 #define TCP_EI_BITS_2MS_TIMER 0x400 /* 2 MSL timer expired */
82
83 #if defined(_KERNEL) || defined(_WANT_TCPCB)
84 #include <netinet/cc/cc.h>
85
86 /* TCP segment queue entry */
87 struct tseg_qent {
88 TAILQ_ENTRY(tseg_qent) tqe_q;
89 struct mbuf *tqe_m; /* mbuf contains packet */
90 struct mbuf *tqe_last; /* last mbuf in chain */
91 tcp_seq tqe_start; /* TCP Sequence number start */
92 int tqe_len; /* TCP segment data length */
93 uint32_t tqe_flags; /* The flags from tcp_get_flags() */
94 uint32_t tqe_mbuf_cnt; /* Count of mbuf overhead */
95 };
96 TAILQ_HEAD(tsegqe_head, tseg_qent);
97
98 struct sackblk {
99 tcp_seq start; /* start seq no. of sack block */
100 tcp_seq end; /* end seq no. */
101 };
102
103 struct sackhole {
104 tcp_seq start; /* start seq no. of hole */
105 tcp_seq end; /* end seq no. */
106 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */
107 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */
108 };
109
110 struct sackhint {
111 struct sackhole *nexthole;
112 int32_t sack_bytes_rexmit;
113 tcp_seq last_sack_ack; /* Most recent/largest sacked ack */
114
115 int32_t delivered_data; /* Newly acked data from last SACK */
116
117 int32_t sacked_bytes; /* Total sacked bytes reported by the
118 * receiver via sack option
119 */
120 uint32_t recover_fs; /* Flight Size at the start of Loss recovery */
121 uint32_t prr_delivered; /* Total bytes delivered using PRR */
122 uint32_t prr_out; /* Bytes sent during IN_RECOVERY */
123 };
124
125 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
126
127 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
128
129 typedef enum {
130 TT_DELACK = 0,
131 TT_REXMT,
132 TT_PERSIST,
133 TT_KEEP,
134 TT_2MSL,
135 TT_N,
136 } tt_which;
137
138 /*
139 * Tcp control block, one per tcp connection.
140 */
141 struct tcpcb {
142 struct inpcb t_inpcb; /* embedded protocol indepenent cb */
143 #define t_start_zero t_fb
144 #define t_zero_size (sizeof(struct tcpcb) - \
145 offsetof(struct tcpcb, t_start_zero))
146 struct tcp_function_block *t_fb;/* TCP function call block */
147 void *t_fb_ptr; /* Pointer to t_fb specific data */
148
149 struct callout t_callout;
150 sbintime_t t_timers[TT_N];
151 sbintime_t t_precisions[TT_N];
152
153 uint32_t t_maxseg:24, /* maximum segment size */
154 t_logstate:8; /* State of "black box" logging */
155 uint32_t t_port:16, /* Tunneling (over udp) port */
156 t_state:4, /* state of this connection */
157 t_idle_reduce : 1,
158 t_delayed_ack: 7, /* Delayed ack variable */
159 t_fin_is_rst: 1, /* Are fin's treated as resets */
160 t_log_state_set: 1,
161 bits_spare : 2;
162 u_int t_flags;
163 tcp_seq snd_una; /* sent but unacknowledged */
164 tcp_seq snd_max; /* highest sequence number sent;
165 * used to recognize retransmits
166 */
167 tcp_seq snd_nxt; /* send next */
168 tcp_seq snd_up; /* send urgent pointer */
169 uint32_t snd_wnd; /* send window */
170 uint32_t snd_cwnd; /* congestion-controlled window */
171 uint32_t t_peakrate_thr; /* pre-calculated peak rate threshold */
172 uint32_t ts_offset; /* our timestamp offset */
173 uint32_t rfbuf_ts; /* recv buffer autoscaling timestamp */
174 int rcv_numsacks; /* # distinct sack blks present */
175 u_int t_tsomax; /* TSO total burst length limit */
176 u_int t_tsomaxsegcount; /* TSO maximum segment count */
177 u_int t_tsomaxsegsize; /* TSO maximum segment size in bytes */
178 tcp_seq rcv_nxt; /* receive next */
179 tcp_seq rcv_adv; /* advertised window */
180 uint32_t rcv_wnd; /* receive window */
181 u_int t_flags2; /* More tcpcb flags storage */
182 int t_srtt; /* smoothed round-trip time */
183 int t_rttvar; /* variance in round-trip time */
184 uint32_t ts_recent; /* timestamp echo data */
185 u_char snd_scale; /* window scaling for send window */
186 u_char rcv_scale; /* window scaling for recv window */
187 u_char snd_limited; /* segments limited transmitted */
188 u_char request_r_scale; /* pending window scaling */
189 tcp_seq last_ack_sent;
190 u_int t_rcvtime; /* inactivity time */
191 tcp_seq rcv_up; /* receive urgent pointer */
192 int t_segqlen; /* segment reassembly queue length */
193 uint32_t t_segqmbuflen; /* total reassembly queue byte length */
194 struct tsegqe_head t_segq; /* segment reassembly queue */
195 struct mbuf *t_in_pkt;
196 struct mbuf *t_tail_pkt;
197 uint32_t snd_ssthresh; /* snd_cwnd size threshold for
198 * for slow start exponential to
199 * linear switch
200 */
201 tcp_seq snd_wl1; /* window update seg seq number */
202 tcp_seq snd_wl2; /* window update seg ack number */
203
204 tcp_seq irs; /* initial receive sequence number */
205 tcp_seq iss; /* initial send sequence number */
206 u_int t_acktime; /* RACK and BBR incoming new data was acked */
207 u_int t_sndtime; /* time last data was sent */
208 u_int ts_recent_age; /* when last updated */
209 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */
210 char t_oobflags; /* have some */
211 char t_iobc; /* input character */
212 int t_rxtcur; /* current retransmit value (ticks) */
213
214 int t_rxtshift; /* log(2) of rexmt exp. backoff */
215 u_int t_rtttime; /* RTT measurement start time */
216
217 tcp_seq t_rtseq; /* sequence number being timed */
218 u_int t_starttime; /* time connection was established */
219 u_int t_fbyte_in; /* ticks time first byte queued in */
220 u_int t_fbyte_out; /* ticks time first byte queued out */
221
222 u_int t_pmtud_saved_maxseg; /* pre-blackhole MSS */
223 int t_blackhole_enter; /* when to enter blackhole detection */
224 int t_blackhole_exit; /* when to exit blackhole detection */
225 u_int t_rttmin; /* minimum rtt allowed */
226
227 int t_softerror; /* possible error not yet reported */
228 uint32_t max_sndwnd; /* largest window peer has offered */
229 uint32_t snd_cwnd_prev; /* cwnd prior to retransmit */
230 uint32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
231 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */
232 int t_sndzerowin; /* zero-window updates sent */
233 int snd_numholes; /* number of holes seen by sender */
234 u_int t_badrxtwin; /* window for retransmit recovery */
235 TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
236 /* SACK scoreboard (sorted) */
237 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/
238 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
239 struct sackhint sackhint; /* SACK scoreboard hint */
240 int t_rttlow; /* smallest observerved RTT */
241 int rfbuf_cnt; /* recv buffer autoscaling byte count */
242 struct toedev *tod; /* toedev handling this connection */
243 int t_sndrexmitpack; /* retransmit packets sent */
244 int t_rcvoopack; /* out-of-order packets received */
245 void *t_toe; /* TOE pcb pointer */
246 struct cc_algo *t_cc; /* congestion control algorithm */
247 struct cc_var t_ccv; /* congestion control specific vars */
248 int t_bytes_acked; /* # bytes acked during current RTT */
249 u_int t_maxunacktime;
250 u_int t_keepinit; /* time to establish connection */
251 u_int t_keepidle; /* time before keepalive probes begin */
252 u_int t_keepintvl; /* interval between keepalives */
253 u_int t_keepcnt; /* number of keepalives before close */
254 int t_dupacks; /* consecutive dup acks recd */
255 int t_lognum; /* Number of log entries */
256 int t_loglimit; /* Maximum number of log entries */
257 uint32_t t_rcep; /* Number of received CE marked pkts */
258 uint32_t t_scep; /* Synced number of delivered CE pkts */
259 int64_t t_pacing_rate; /* bytes / sec, -1 => unlimited */
260 struct tcp_log_stailq t_logs; /* Log buffer */
261 struct tcp_log_id_node *t_lin;
262 struct tcp_log_id_bucket *t_lib;
263 const char *t_output_caller; /* Function that called tcp_output */
264 struct statsblob *t_stats; /* Per-connection stats */
265 uint32_t t_logsn; /* Log "serial number" */
266 uint32_t gput_ts; /* Time goodput measurement started */
267 tcp_seq gput_seq; /* Outbound measurement seq */
268 tcp_seq gput_ack; /* Inbound measurement ack */
269 int32_t t_stats_gput_prev; /* XXXLAS: Prev gput measurement */
270 uint32_t t_maxpeakrate; /* max peak rate set by user, bytes/s */
271 uint32_t t_sndtlppack; /* tail loss probe packets sent */
272 uint64_t t_sndtlpbyte; /* total tail loss probe bytes sent */
273 uint64_t t_sndbytes; /* total bytes sent */
274 uint64_t t_snd_rxt_bytes; /* total bytes retransmitted */
275 uint32_t t_dsack_bytes; /* dsack bytes received */
276 uint32_t t_dsack_tlp_bytes; /* dsack bytes received for TLPs sent */
277 uint32_t t_dsack_pack; /* dsack packets we have eceived */
278 uint8_t t_rttupdated; /* number of times rtt sampled */
279 /* TCP Fast Open */
280 uint8_t t_tfo_client_cookie_len; /* TFO client cookie length */
281 uint32_t t_end_info_status; /* Status flag of end info */
282 unsigned int *t_tfo_pending; /* TFO server pending counter */
283 union {
284 uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
285 uint64_t server;
286 } t_tfo_cookie; /* TCP Fast Open cookie to send */
287 union {
288 uint8_t t_end_info_bytes[TCP_END_BYTE_INFO];
289 uint64_t t_end_info;
290 };
291 #ifdef TCPPCAP
292 struct mbufq t_inpkts; /* List of saved input packets. */
293 struct mbufq t_outpkts; /* List of saved output packets. */
294 #endif
295 #ifdef TCP_HHOOK
296 struct osd t_osd; /* storage for Khelp module data */
297 #endif
298 };
299 #endif /* _KERNEL || _WANT_TCPCB */
300
301 #ifdef _KERNEL
302 struct tcptemp {
303 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
304 struct tcphdr tt_t;
305 };
306
307 /* Enable TCP/UDP tunneling port */
308 #define TCP_TUNNELING_PORT_MIN 0
309 #define TCP_TUNNELING_PORT_MAX 65535
310 #define TCP_TUNNELING_PORT_DEFAULT 0
311
312 /* Enable TCP/UDP tunneling port */
313 #define TCP_TUNNELING_OVERHEAD_MIN sizeof(struct udphdr)
314 #define TCP_TUNNELING_OVERHEAD_MAX 1024
315 #define TCP_TUNNELING_OVERHEAD_DEFAULT TCP_TUNNELING_OVERHEAD_MIN
316
317 /* Minimum map entries limit value, if set */
318 #define TCP_MIN_MAP_ENTRIES_LIMIT 128
319
320 /*
321 * TODO: We yet need to brave plowing in
322 * to tcp_input() and the pru_usrreq() block.
323 * Right now these go to the old standards which
324 * are somewhat ok, but in the long term may
325 * need to be changed. If we do tackle tcp_input()
326 * then we need to get rid of the tcp_do_segment()
327 * function below.
328 */
329 /* Flags for tcp functions */
330 #define TCP_FUNC_BEING_REMOVED 0x01 /* Can no longer be referenced */
331 #define TCP_FUNC_OUTPUT_CANDROP 0x02 /* tfb_tcp_output may ask tcp_drop */
332
333 /*
334 * If defining the optional tcp_timers, in the
335 * tfb_tcp_timer_stop call you must use the
336 * callout_async_drain() function with the
337 * tcp_timer_discard callback. You should check
338 * the return of callout_async_drain() and if 0
339 * increment tt_draincnt. Since the timer sub-system
340 * does not know your callbacks you must provide a
341 * stop_all function that loops through and calls
342 * tcp_timer_stop() with each of your defined timers.
343 * Adding a tfb_tcp_handoff_ok function allows the socket
344 * option to change stacks to query you even if the
345 * connection is in a later stage. You return 0 to
346 * say you can take over and run your stack, you return
347 * non-zero (an error number) to say no you can't.
348 * If the function is undefined you can only change
349 * in the early states (before connect or listen).
350 * tfb_tcp_fb_fini is changed to add a flag to tell
351 * the old stack if the tcb is being destroyed or
352 * not. A one in the flag means the TCB is being
353 * destroyed, a zero indicates its transitioning to
354 * another stack (via socket option).
355 */
356 struct tcp_function_block {
357 char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
358 int (*tfb_tcp_output)(struct tcpcb *);
359 int (*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *);
360 void (*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
361 struct socket *, struct tcpcb *,
362 int, int, uint8_t);
363 int (*tfb_do_queued_segments)(struct socket *, struct tcpcb *, int);
364 int (*tfb_do_segment_nounlock)(struct mbuf *, struct tcphdr *,
365 struct socket *, struct tcpcb *,
366 int, int, uint8_t,
367 int, struct timeval *);
368 void (*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *,
369 struct socket *, struct tcpcb *,
370 int, int, uint8_t,
371 int, struct timeval *);
372 int (*tfb_tcp_ctloutput)(struct inpcb *inp, struct sockopt *sopt);
373 /* Optional memory allocation/free routine */
374 int (*tfb_tcp_fb_init)(struct tcpcb *);
375 void (*tfb_tcp_fb_fini)(struct tcpcb *, int);
376 /* Optional timers, must define all if you define one */
377 int (*tfb_tcp_timer_stop_all)(struct tcpcb *);
378 void (*tfb_tcp_rexmit_tmr)(struct tcpcb *);
379 int (*tfb_tcp_handoff_ok)(struct tcpcb *);
380 void (*tfb_tcp_mtu_chg)(struct tcpcb *);
381 int (*tfb_pru_options)(struct tcpcb *, int);
382 void (*tfb_hwtls_change)(struct tcpcb *, int);
383 int (*tfb_compute_pipe)(struct tcpcb *tp);
384 volatile uint32_t tfb_refcnt;
385 uint32_t tfb_flags;
386 uint8_t tfb_id;
387 };
388
389 struct tcp_function {
390 TAILQ_ENTRY(tcp_function) tf_next;
391 char tf_name[TCP_FUNCTION_NAME_LEN_MAX];
392 struct tcp_function_block *tf_fb;
393 };
394
395 TAILQ_HEAD(tcp_funchead, tcp_function);
396
397 struct tcpcb * tcp_drop(struct tcpcb *, int);
398
399 #ifdef _NETINET_IN_PCB_H_
400 #define intotcpcb(inp) __containerof((inp), struct tcpcb, t_inpcb)
401 #define sototcpcb(so) intotcpcb(sotoinpcb(so))
402 #define tptoinpcb(tp) (&(tp)->t_inpcb)
403 #define tptosocket(tp) (tp)->t_inpcb.inp_socket
404
405 /*
406 * tcp_output()
407 * Handles tcp_drop request from advanced stacks and reports that inpcb is
408 * gone with negative return code.
409 * Drop in replacement for the default stack.
410 */
411 static inline int
412 tcp_output(struct tcpcb *tp)
413 {
414 struct inpcb *inp = tptoinpcb(tp);
415 int rv;
416
417 INP_WLOCK_ASSERT(inp);
418
419 rv = tp->t_fb->tfb_tcp_output(tp);
420 if (rv < 0) {
421 KASSERT(tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP,
422 ("TCP stack %s requested tcp_drop(%p)",
423 tp->t_fb->tfb_tcp_block_name, tp));
424 tp = tcp_drop(tp, -rv);
425 if (tp)
426 INP_WUNLOCK(inp);
427 }
428
429 return (rv);
430 }
431
432 /*
433 * tcp_output_unlock()
434 * Always returns unlocked, handles drop request from advanced stacks.
435 * Always returns positive error code.
436 */
437 static inline int
438 tcp_output_unlock(struct tcpcb *tp)
439 {
440 struct inpcb *inp = tptoinpcb(tp);
441 int rv;
442
443 INP_WLOCK_ASSERT(inp);
444
445 rv = tp->t_fb->tfb_tcp_output(tp);
446 if (rv < 0) {
447 KASSERT(tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP,
448 ("TCP stack %s requested tcp_drop(%p)",
449 tp->t_fb->tfb_tcp_block_name, tp));
450 rv = -rv;
451 tp = tcp_drop(tp, rv);
452 if (tp)
453 INP_WUNLOCK(inp);
454 } else
455 INP_WUNLOCK(inp);
456
457 return (rv);
458 }
459
460 /*
461 * tcp_output_nodrop()
462 * Always returns locked. It is caller's responsibility to run tcp_drop()!
463 * Useful in syscall implementations, when we want to perform some logging
464 * and/or tracing with tcpcb before calling tcp_drop(). To be used with
465 * tcp_unlock_or_drop() later.
466 *
467 * XXXGL: maybe don't allow stacks to return a drop request at certain
468 * TCP states? Why would it do in connect(2)? In recv(2)?
469 */
470 static inline int
471 tcp_output_nodrop(struct tcpcb *tp)
472 {
473 int rv;
474
475 INP_WLOCK_ASSERT(tptoinpcb(tp));
476
477 rv = tp->t_fb->tfb_tcp_output(tp);
478 KASSERT(rv >= 0 || tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP,
479 ("TCP stack %s requested tcp_drop(%p)",
480 tp->t_fb->tfb_tcp_block_name, tp));
481 return (rv);
482 }
483
484 /*
485 * tcp_unlock_or_drop()
486 * Handle return code from tfb_tcp_output() after we have logged/traced,
487 * to be used with tcp_output_nodrop().
488 */
489 static inline int
490 tcp_unlock_or_drop(struct tcpcb *tp, int tcp_output_retval)
491 {
492 struct inpcb *inp = tptoinpcb(tp);
493
494 INP_WLOCK_ASSERT(inp);
495
496 if (tcp_output_retval < 0) {
497 tcp_output_retval = -tcp_output_retval;
498 if (tcp_drop(tp, tcp_output_retval) != NULL)
499 INP_WUNLOCK(inp);
500 } else
501 INP_WUNLOCK(inp);
502
503 return (tcp_output_retval);
504 }
505 #endif /* _NETINET_IN_PCB_H_ */
506
507 static int inline
508 tcp_packets_this_ack(struct tcpcb *tp, tcp_seq ack)
509 {
510 return ((ack - tp->snd_una) / tp->t_maxseg +
511 ((((ack - tp->snd_una) % tp->t_maxseg) != 0) ? 1 : 0));
512 }
513 #endif /* _KERNEL */
514
515 /*
516 * Flags and utility macros for the t_flags field.
517 */
518 #define TF_ACKNOW 0x00000001 /* ack peer immediately */
519 #define TF_DELACK 0x00000002 /* ack, but try to delay it */
520 #define TF_NODELAY 0x00000004 /* don't delay packets to coalesce */
521 #define TF_NOOPT 0x00000008 /* don't use tcp options */
522 #define TF_SENTFIN 0x00000010 /* have sent FIN */
523 #define TF_REQ_SCALE 0x00000020 /* have/will request window scaling */
524 #define TF_RCVD_SCALE 0x00000040 /* other side has requested scaling */
525 #define TF_REQ_TSTMP 0x00000080 /* have/will request timestamps */
526 #define TF_RCVD_TSTMP 0x00000100 /* a timestamp was received in SYN */
527 #define TF_SACK_PERMIT 0x00000200 /* other side said I could SACK */
528 #define TF_NEEDSYN 0x00000400 /* send SYN (implicit state) */
529 #define TF_NEEDFIN 0x00000800 /* send FIN (implicit state) */
530 #define TF_NOPUSH 0x00001000 /* don't push */
531 #define TF_PREVVALID 0x00002000 /* saved values for bad rxmit valid
532 * Note: accessing and restoring from
533 * these may only be done in the 1st
534 * RTO recovery round (t_rxtshift == 1)
535 */
536 #define TF_WAKESOR 0x00004000 /* wake up receive socket */
537 #define TF_GPUTINPROG 0x00008000 /* Goodput measurement in progress */
538 #define TF_MORETOCOME 0x00010000 /* More data to be appended to sock */
539 #define TF_SONOTCONN 0x00020000 /* needs soisconnected() on ESTAB */
540 #define TF_LASTIDLE 0x00040000 /* connection was previously idle */
541 #define TF_RXWIN0SENT 0x00080000 /* sent a receiver win 0 in response */
542 #define TF_FASTRECOVERY 0x00100000 /* in NewReno Fast Recovery */
543 #define TF_WASFRECOVERY 0x00200000 /* was in NewReno Fast Recovery */
544 #define TF_SIGNATURE 0x00400000 /* require MD5 digests (RFC2385) */
545 #define TF_FORCEDATA 0x00800000 /* force out a byte */
546 #define TF_TSO 0x01000000 /* TSO enabled on this connection */
547 #define TF_TOE 0x02000000 /* this connection is offloaded */
548 #define TF_CLOSED 0x04000000 /* close(2) called on socket */
549 #define TF_UNUSED1 0x08000000 /* unused */
550 #define TF_LRD 0x10000000 /* Lost Retransmission Detection */
551 #define TF_CONGRECOVERY 0x20000000 /* congestion recovery mode */
552 #define TF_WASCRECOVERY 0x40000000 /* was in congestion recovery */
553 #define TF_FASTOPEN 0x80000000 /* TCP Fast Open indication */
554
555 #define IN_FASTRECOVERY(t_flags) (t_flags & TF_FASTRECOVERY)
556 #define ENTER_FASTRECOVERY(t_flags) t_flags |= TF_FASTRECOVERY
557 #define EXIT_FASTRECOVERY(t_flags) t_flags &= ~TF_FASTRECOVERY
558
559 #define IN_CONGRECOVERY(t_flags) (t_flags & TF_CONGRECOVERY)
560 #define ENTER_CONGRECOVERY(t_flags) t_flags |= TF_CONGRECOVERY
561 #define EXIT_CONGRECOVERY(t_flags) t_flags &= ~TF_CONGRECOVERY
562
563 #define IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
564 #define ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
565 #define EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
566
567 #if defined(_KERNEL) && !defined(TCP_RFC7413)
568 #define IS_FASTOPEN(t_flags) (false)
569 #else
570 #define IS_FASTOPEN(t_flags) (t_flags & TF_FASTOPEN)
571 #endif
572
573 #define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una)
574
575 /*
576 * Flags for the t_oobflags field.
577 */
578 #define TCPOOB_HAVEDATA 0x01
579 #define TCPOOB_HADDATA 0x02
580
581 /*
582 * Flags for the extended TCP flags field, t_flags2
583 */
584 #define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */
585 #define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */
586 #define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */
587 #define TF2_LOG_AUTO 0x00000008 /* Session is auto-logging. */
588 #define TF2_DROP_AF_DATA 0x00000010 /* Drop after all data ack'd */
589 #define TF2_ECN_PERMIT 0x00000020 /* connection ECN-ready */
590 #define TF2_ECN_SND_CWR 0x00000040 /* ECN CWR in queue */
591 #define TF2_ECN_SND_ECE 0x00000080 /* ECN ECE in queue */
592 #define TF2_ACE_PERMIT 0x00000100 /* Accurate ECN mode */
593 #define TF2_FBYTES_COMPLETE 0x00000400 /* We have first bytes in and out */
594 #define TF2_ECN_USE_ECT1 0x00000800 /* Use ECT(1) marking on session */
595
596 /*
597 * Structure to hold TCP options that are only used during segment
598 * processing (in tcp_input), but not held in the tcpcb.
599 * It's basically used to reduce the number of parameters
600 * to tcp_dooptions and tcp_addoptions.
601 * The binary order of the to_flags is relevant for packing of the
602 * options in tcp_addoptions.
603 */
604 struct tcpopt {
605 u_int32_t to_flags; /* which options are present */
606 #define TOF_MSS 0x0001 /* maximum segment size */
607 #define TOF_SCALE 0x0002 /* window scaling */
608 #define TOF_SACKPERM 0x0004 /* SACK permitted */
609 #define TOF_TS 0x0010 /* timestamp */
610 #define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */
611 #define TOF_SACK 0x0080 /* Peer sent SACK option */
612 #define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */
613 #define TOF_MAXOPT 0x0200
614 u_int32_t to_tsval; /* new timestamp */
615 u_int32_t to_tsecr; /* reflected timestamp */
616 u_char *to_sacks; /* pointer to the first SACK blocks */
617 u_char *to_signature; /* pointer to the TCP-MD5 signature */
618 u_int8_t *to_tfo_cookie; /* pointer to the TFO cookie */
619 u_int16_t to_mss; /* maximum segment size */
620 u_int8_t to_wscale; /* window scaling */
621 u_int8_t to_nsacks; /* number of SACK blocks */
622 u_int8_t to_tfo_len; /* TFO cookie length */
623 u_int32_t to_spare; /* UTO */
624 };
625
626 /*
627 * Flags for tcp_dooptions.
628 */
629 #define TO_SYN 0x01 /* parse SYN-only options */
630
631 struct hc_metrics_lite { /* must stay in sync with hc_metrics */
632 uint32_t rmx_mtu; /* MTU for this path */
633 uint32_t rmx_ssthresh; /* outbound gateway buffer limit */
634 uint32_t rmx_rtt; /* estimated round trip time */
635 uint32_t rmx_rttvar; /* estimated rtt variance */
636 uint32_t rmx_cwnd; /* congestion window */
637 uint32_t rmx_sendpipe; /* outbound delay-bandwidth product */
638 uint32_t rmx_recvpipe; /* inbound delay-bandwidth product */
639 };
640
641 /*
642 * Used by tcp_maxmtu() to communicate interface specific features
643 * and limits at the time of connection setup.
644 */
645 struct tcp_ifcap {
646 int ifcap;
647 u_int tsomax;
648 u_int tsomaxsegcount;
649 u_int tsomaxsegsize;
650 };
651
652 #ifndef _NETINET_IN_PCB_H_
653 struct in_conninfo;
654 #endif /* _NETINET_IN_PCB_H_ */
655
656 /*
657 * The smoothed round-trip time and estimated variance
658 * are stored as fixed point numbers scaled by the values below.
659 * For convenience, these scales are also used in smoothing the average
660 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
661 * With these scales, srtt has 3 bits to the right of the binary point,
662 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the
663 * binary point, and is smoothed with an ALPHA of 0.75.
664 */
665 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */
666 #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */
667 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */
668 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */
669 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
670
671 /*
672 * The initial retransmission should happen at rtt + 4 * rttvar.
673 * Because of the way we do the smoothing, srtt and rttvar
674 * will each average +1/2 tick of bias. When we compute
675 * the retransmit timer, we want 1/2 tick of rounding and
676 * 1 extra tick because of +-1/2 tick uncertainty in the
677 * firing of the timer. The bias will give us exactly the
678 * 1.5 tick we need. But, because the bias is
679 * statistical, we have to test that we don't drop below
680 * the minimum feasible timer (which is 2 ticks).
681 * This version of the macro adapted from a paper by Lawrence
682 * Brakmo and Larry Peterson which outlines a problem caused
683 * by insufficient precision in the original implementation,
684 * which results in inappropriately large RTO values for very
685 * fast networks.
686 */
687 #define TCP_REXMTVAL(tp) \
688 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
689 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
690
691 /*
692 * TCP statistics.
693 * Many of these should be kept per connection,
694 * but that's inconvenient at the moment.
695 */
696 struct tcpstat {
697 uint64_t tcps_connattempt; /* connections initiated */
698 uint64_t tcps_accepts; /* connections accepted */
699 uint64_t tcps_connects; /* connections established */
700 uint64_t tcps_drops; /* connections dropped */
701 uint64_t tcps_conndrops; /* embryonic connections dropped */
702 uint64_t tcps_minmssdrops; /* average minmss too low drops */
703 uint64_t tcps_closed; /* conn. closed (includes drops) */
704 uint64_t tcps_segstimed; /* segs where we tried to get rtt */
705 uint64_t tcps_rttupdated; /* times we succeeded */
706 uint64_t tcps_delack; /* delayed acks sent */
707 uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
708 uint64_t tcps_rexmttimeo; /* retransmit timeouts */
709 uint64_t tcps_persisttimeo; /* persist timeouts */
710 uint64_t tcps_keeptimeo; /* keepalive timeouts */
711 uint64_t tcps_keepprobe; /* keepalive probes sent */
712 uint64_t tcps_keepdrops; /* connections dropped in keepalive */
713 uint64_t tcps_progdrops; /* drops due to no progress */
714
715 uint64_t tcps_sndtotal; /* total packets sent */
716 uint64_t tcps_sndpack; /* data packets sent */
717 uint64_t tcps_sndbyte; /* data bytes sent */
718 uint64_t tcps_sndrexmitpack; /* data packets retransmitted */
719 uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */
720 uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */
721 uint64_t tcps_sndacks; /* ack-only packets sent */
722 uint64_t tcps_sndprobe; /* window probes sent */
723 uint64_t tcps_sndurg; /* packets sent with URG only */
724 uint64_t tcps_sndwinup; /* window update-only packets sent */
725 uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
726
727 uint64_t tcps_rcvtotal; /* total packets received */
728 uint64_t tcps_rcvpack; /* packets received in sequence */
729 uint64_t tcps_rcvbyte; /* bytes received in sequence */
730 uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */
731 uint64_t tcps_rcvbadoff; /* packets received with bad offset */
732 uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */
733 uint64_t tcps_rcvshort; /* packets received too short */
734 uint64_t tcps_rcvduppack; /* duplicate-only packets received */
735 uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */
736 uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */
737 uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
738 uint64_t tcps_rcvoopack; /* out-of-order packets received */
739 uint64_t tcps_rcvoobyte; /* out-of-order bytes received */
740 uint64_t tcps_rcvpackafterwin; /* packets with data after window */
741 uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */
742 uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */
743 uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */
744 uint64_t tcps_rcvdupack; /* rcvd duplicate acks */
745 uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */
746 uint64_t tcps_rcvackpack; /* rcvd ack packets */
747 uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */
748 uint64_t tcps_rcvwinupd; /* rcvd window update packets */
749 uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */
750 uint64_t tcps_predack; /* times hdr predict ok for acks */
751 uint64_t tcps_preddat; /* times hdr predict ok for data pkts */
752 uint64_t tcps_pcbcachemiss;
753 uint64_t tcps_cachedrtt; /* times cached RTT in route updated */
754 uint64_t tcps_cachedrttvar; /* times cached rttvar updated */
755 uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */
756 uint64_t tcps_usedrtt; /* times RTT initialized from route */
757 uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */
758 uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/
759 uint64_t tcps_persistdrop; /* timeout in persist state */
760 uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */
761 uint64_t tcps_mturesent; /* resends due to MTU discovery */
762 uint64_t tcps_listendrop; /* listen queue overflows */
763 uint64_t tcps_badrst; /* ignored RSTs in the window */
764
765 uint64_t tcps_sc_added; /* entry added to syncache */
766 uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */
767 uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */
768 uint64_t tcps_sc_dropped; /* could not reply to packet */
769 uint64_t tcps_sc_completed; /* successful extraction of entry */
770 uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
771 uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */
772 uint64_t tcps_sc_reset; /* RST removed entry from syncache */
773 uint64_t tcps_sc_stale; /* timed out or listen socket gone */
774 uint64_t tcps_sc_aborted; /* syncache entry aborted */
775 uint64_t tcps_sc_badack; /* removed due to bad ACK */
776 uint64_t tcps_sc_unreach; /* ICMP unreachable received */
777 uint64_t tcps_sc_zonefail; /* zalloc() failed */
778 uint64_t tcps_sc_sendcookie; /* SYN cookie sent */
779 uint64_t tcps_sc_recvcookie; /* SYN cookie received */
780
781 uint64_t tcps_hc_added; /* entry added to hostcache */
782 uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
783
784 uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */
785
786 /* SACK related stats */
787 uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
788 uint64_t tcps_sack_rexmits; /* SACK rexmit segments */
789 uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */
790 uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */
791 uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */
792 uint64_t tcps_sack_lostrexmt; /* SACK lost retransmission recovered */
793 uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */
794
795 /* ECN related stats */
796 uint64_t tcps_ecn_rcvce; /* ECN Congestion Experienced */
797 uint64_t tcps_ecn_rcvect0; /* ECN Capable Transport */
798 uint64_t tcps_ecn_rcvect1; /* ECN Capable Transport */
799 uint64_t tcps_ecn_shs; /* ECN successful handshakes */
800 uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */
801
802 /* TCP_SIGNATURE related stats */
803 uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */
804 uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */
805 uint64_t tcps_sig_err_buildsig; /* Failed to make signature */
806 uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */
807 uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */
808
809 /* Path MTU Discovery Black Hole Detection related stats */
810 uint64_t tcps_pmtud_blackhole_activated; /* Black Hole Count */
811 uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
812 uint64_t tcps_pmtud_blackhole_failed; /* Black Hole Failure Count */
813
814 uint64_t tcps_tunneled_pkts; /* Packets encap's in UDP received */
815 uint64_t tcps_tunneled_errs; /* Packets that had errors that were UDP encaped */
816
817 /* Dsack related stats */
818 uint64_t tcps_dsack_count; /* Number of ACKs arriving with DSACKs */
819 uint64_t tcps_dsack_bytes; /* Number of bytes DSACK'ed no TLP */
820 uint64_t tcps_dsack_tlp_bytes; /* Number of bytes DSACK'ed due to TLPs */
821
822 /* TCPS_TIME_WAIT usage stats */
823 uint64_t tcps_tw_recycles; /* Times time-wait was recycled. */
824 uint64_t tcps_tw_resets; /* Times time-wait sent a reset. */
825 uint64_t tcps_tw_responds; /* Times time-wait sent a valid ack. */
826
827 /* Accurate ECN Handshake stats */
828 uint64_t tcps_ace_nect; /* ACE SYN packet with Non-ECT */
829 uint64_t tcps_ace_ect1; /* ACE SYN packet with ECT1 */
830 uint64_t tcps_ace_ect0; /* ACE SYN packet with ECT0 */
831 uint64_t tcps_ace_ce; /* ACE SYN packet with CE */
832
833 /* ECN related stats */
834 uint64_t tcps_ecn_sndect0; /* ECN Capable Transport */
835 uint64_t tcps_ecn_sndect1; /* ECN Capable Transport */
836
837 uint64_t _pad[4]; /* 4 TBD placeholder for STABLE */
838 };
839
840 #define tcps_rcvmemdrop tcps_rcvreassfull /* compat */
841
842 #ifdef _KERNEL
843 #define TI_UNLOCKED 1
844 #define TI_RLOCKED 2
845 #include <sys/counter.h>
846
847 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */
848 /*
849 * In-kernel consumers can use these accessor macros directly to update
850 * stats.
851 */
852 #define TCPSTAT_ADD(name, val) \
853 VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
854 #define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1)
855
856 /*
857 * Kernel module consumers must use this accessor macro.
858 */
859 void kmod_tcpstat_add(int statnum, int val);
860 #define KMOD_TCPSTAT_ADD(name, val) \
861 kmod_tcpstat_add(offsetof(struct tcpstat, name) / sizeof(uint64_t), val)
862 #define KMOD_TCPSTAT_INC(name) KMOD_TCPSTAT_ADD(name, 1)
863
864 /*
865 * Running TCP connection count by state.
866 */
867 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
868 #define V_tcps_states VNET(tcps_states)
869 #define TCPSTATES_INC(state) counter_u64_add(V_tcps_states[state], 1)
870 #define TCPSTATES_DEC(state) counter_u64_add(V_tcps_states[state], -1)
871
872 /*
873 * TCP specific helper hook point identifiers.
874 */
875 #define HHOOK_TCP_EST_IN 0
876 #define HHOOK_TCP_EST_OUT 1
877 #define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT
878
879 struct tcp_hhook_data {
880 struct tcpcb *tp;
881 struct tcphdr *th;
882 struct tcpopt *to;
883 uint32_t len;
884 int tso;
885 tcp_seq curack;
886 };
887 #ifdef TCP_HHOOK
888 void hhook_run_tcp_est_out(struct tcpcb *tp,
889 struct tcphdr *th, struct tcpopt *to,
890 uint32_t len, int tso);
891 #endif
892 #endif
893
894 /*
895 * TCB structure exported to user-land via sysctl(3).
896 *
897 * Fields prefixed with "xt_" are unique to the export structure, and fields
898 * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
899 *
900 * Legend:
901 * (s) - used by userland utilities in src
902 * (p) - used by utilities in ports
903 * (3) - is known to be used by third party software not in ports
904 * (n) - no known usage
905 *
906 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
907 * included. Not all of our clients do.
908 */
909 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
910 struct xtcpcb {
911 ksize_t xt_len; /* length of this structure */
912 struct xinpcb xt_inp;
913 char xt_stack[TCP_FUNCTION_NAME_LEN_MAX]; /* (s) */
914 char xt_logid[TCP_LOG_ID_LEN]; /* (s) */
915 char xt_cc[TCP_CA_NAME_MAX]; /* (s) */
916 int64_t spare64[6];
917 int32_t t_state; /* (s,p) */
918 uint32_t t_flags; /* (s,p) */
919 int32_t t_sndzerowin; /* (s) */
920 int32_t t_sndrexmitpack; /* (s) */
921 int32_t t_rcvoopack; /* (s) */
922 int32_t t_rcvtime; /* (s) */
923 int32_t tt_rexmt; /* (s) */
924 int32_t tt_persist; /* (s) */
925 int32_t tt_keep; /* (s) */
926 int32_t tt_2msl; /* (s) */
927 int32_t tt_delack; /* (s) */
928 int32_t t_logstate; /* (3) */
929 uint32_t t_snd_cwnd; /* (s) */
930 uint32_t t_snd_ssthresh; /* (s) */
931 uint32_t t_maxseg; /* (s) */
932 uint32_t t_rcv_wnd; /* (s) */
933 uint32_t t_snd_wnd; /* (s) */
934 uint32_t xt_ecn; /* (s) */
935 uint32_t t_dsack_bytes; /* (n) */
936 uint32_t t_dsack_tlp_bytes; /* (n) */
937 uint32_t t_dsack_pack; /* (n) */
938 uint16_t xt_encaps_port; /* (s) */
939 int16_t spare16;
940 int32_t spare32[22];
941 } __aligned(8);
942
943 #ifdef _KERNEL
944 void tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
945 #endif
946 #endif
947
948 /*
949 * TCP function information (name-to-id mapping, aliases, and refcnt)
950 * exported to user-land via sysctl(3).
951 */
952 struct tcp_function_info {
953 uint32_t tfi_refcnt;
954 uint8_t tfi_id;
955 char tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
956 char tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
957 };
958
959 /*
960 * Identifiers for TCP sysctl nodes
961 */
962 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
963 #define TCPCTL_MSSDFLT 3 /* MSS default */
964 #define TCPCTL_STATS 4 /* statistics */
965 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */
966 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
967 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
968 #define TCPCTL_SENDSPACE 8 /* send buffer space */
969 #define TCPCTL_RECVSPACE 9 /* receive buffer space */
970 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */
971 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */
972 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */
973 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */
974 #define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */
975 #define TCPCTL_DROP 15 /* drop tcp connection */
976 #define TCPCTL_STATES 16 /* connection counts by TCP state */
977
978 #ifdef _KERNEL
979 #ifdef SYSCTL_DECL
980 SYSCTL_DECL(_net_inet_tcp);
981 SYSCTL_DECL(_net_inet_tcp_sack);
982 MALLOC_DECLARE(M_TCPLOG);
983 #endif
984
985 VNET_DECLARE(int, tcp_log_in_vain);
986 #define V_tcp_log_in_vain VNET(tcp_log_in_vain)
987
988 /*
989 * Global TCP tunables shared between different stacks.
990 * Please keep the list sorted.
991 */
992 VNET_DECLARE(int, drop_synfin);
993 VNET_DECLARE(int, path_mtu_discovery);
994 VNET_DECLARE(int, tcp_abc_l_var);
995 VNET_DECLARE(int, tcp_autorcvbuf_max);
996 VNET_DECLARE(int, tcp_autosndbuf_inc);
997 VNET_DECLARE(int, tcp_autosndbuf_max);
998 VNET_DECLARE(int, tcp_delack_enabled);
999 VNET_DECLARE(int, tcp_do_autorcvbuf);
1000 VNET_DECLARE(int, tcp_do_autosndbuf);
1001 VNET_DECLARE(int, tcp_do_ecn);
1002 VNET_DECLARE(int, tcp_do_lrd);
1003 VNET_DECLARE(int, tcp_do_prr);
1004 VNET_DECLARE(int, tcp_do_prr_conservative);
1005 VNET_DECLARE(int, tcp_do_newcwv);
1006 VNET_DECLARE(int, tcp_do_rfc1323);
1007 VNET_DECLARE(int, tcp_tolerate_missing_ts);
1008 VNET_DECLARE(int, tcp_do_rfc3042);
1009 VNET_DECLARE(int, tcp_do_rfc3390);
1010 VNET_DECLARE(int, tcp_do_rfc3465);
1011 VNET_DECLARE(int, tcp_do_newsack);
1012 VNET_DECLARE(int, tcp_do_sack);
1013 VNET_DECLARE(int, tcp_do_tso);
1014 VNET_DECLARE(int, tcp_ecn_maxretries);
1015 VNET_DECLARE(int, tcp_initcwnd_segments);
1016 VNET_DECLARE(int, tcp_insecure_rst);
1017 VNET_DECLARE(int, tcp_insecure_syn);
1018 VNET_DECLARE(uint32_t, tcp_map_entries_limit);
1019 VNET_DECLARE(uint32_t, tcp_map_split_limit);
1020 VNET_DECLARE(int, tcp_minmss);
1021 VNET_DECLARE(int, tcp_mssdflt);
1022 #ifdef STATS
1023 VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl);
1024 VNET_DECLARE(int, tcp_perconn_stats_enable);
1025 #endif /* STATS */
1026 VNET_DECLARE(int, tcp_recvspace);
1027 VNET_DECLARE(int, tcp_sack_globalholes);
1028 VNET_DECLARE(int, tcp_sack_globalmaxholes);
1029 VNET_DECLARE(int, tcp_sack_maxholes);
1030 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
1031 VNET_DECLARE(int, tcp_sendspace);
1032 VNET_DECLARE(int, tcp_udp_tunneling_overhead);
1033 VNET_DECLARE(int, tcp_udp_tunneling_port);
1034 VNET_DECLARE(struct inpcbinfo, tcbinfo);
1035
1036 #define V_tcp_do_lrd VNET(tcp_do_lrd)
1037 #define V_tcp_do_prr VNET(tcp_do_prr)
1038 #define V_tcp_do_prr_conservative VNET(tcp_do_prr_conservative)
1039 #define V_tcp_do_newcwv VNET(tcp_do_newcwv)
1040 #define V_drop_synfin VNET(drop_synfin)
1041 #define V_path_mtu_discovery VNET(path_mtu_discovery)
1042 #define V_tcbinfo VNET(tcbinfo)
1043 #define V_tcp_abc_l_var VNET(tcp_abc_l_var)
1044 #define V_tcp_autorcvbuf_max VNET(tcp_autorcvbuf_max)
1045 #define V_tcp_autosndbuf_inc VNET(tcp_autosndbuf_inc)
1046 #define V_tcp_autosndbuf_max VNET(tcp_autosndbuf_max)
1047 #define V_tcp_delack_enabled VNET(tcp_delack_enabled)
1048 #define V_tcp_do_autorcvbuf VNET(tcp_do_autorcvbuf)
1049 #define V_tcp_do_autosndbuf VNET(tcp_do_autosndbuf)
1050 #define V_tcp_do_ecn VNET(tcp_do_ecn)
1051 #define V_tcp_do_rfc1323 VNET(tcp_do_rfc1323)
1052 #define V_tcp_tolerate_missing_ts VNET(tcp_tolerate_missing_ts)
1053 #define V_tcp_ts_offset_per_conn VNET(tcp_ts_offset_per_conn)
1054 #define V_tcp_do_rfc3042 VNET(tcp_do_rfc3042)
1055 #define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390)
1056 #define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465)
1057 #define V_tcp_do_newsack VNET(tcp_do_newsack)
1058 #define V_tcp_do_sack VNET(tcp_do_sack)
1059 #define V_tcp_do_tso VNET(tcp_do_tso)
1060 #define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries)
1061 #define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments)
1062 #define V_tcp_insecure_rst VNET(tcp_insecure_rst)
1063 #define V_tcp_insecure_syn VNET(tcp_insecure_syn)
1064 #define V_tcp_map_entries_limit VNET(tcp_map_entries_limit)
1065 #define V_tcp_map_split_limit VNET(tcp_map_split_limit)
1066 #define V_tcp_minmss VNET(tcp_minmss)
1067 #define V_tcp_mssdflt VNET(tcp_mssdflt)
1068 #ifdef STATS
1069 #define V_tcp_perconn_stats_dflt_tpl VNET(tcp_perconn_stats_dflt_tpl)
1070 #define V_tcp_perconn_stats_enable VNET(tcp_perconn_stats_enable)
1071 #endif /* STATS */
1072 #define V_tcp_recvspace VNET(tcp_recvspace)
1073 #define V_tcp_sack_globalholes VNET(tcp_sack_globalholes)
1074 #define V_tcp_sack_globalmaxholes VNET(tcp_sack_globalmaxholes)
1075 #define V_tcp_sack_maxholes VNET(tcp_sack_maxholes)
1076 #define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail)
1077 #define V_tcp_sendspace VNET(tcp_sendspace)
1078 #define V_tcp_udp_tunneling_overhead VNET(tcp_udp_tunneling_overhead)
1079 #define V_tcp_udp_tunneling_port VNET(tcp_udp_tunneling_port)
1080
1081 #ifdef TCP_HHOOK
1082 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
1083 #define V_tcp_hhh VNET(tcp_hhh)
1084 #endif
1085
1086 int tcp_addoptions(struct tcpopt *, u_char *);
1087 struct tcpcb *
1088 tcp_close(struct tcpcb *);
1089 void tcp_discardcb(struct tcpcb *);
1090 void tcp_twstart(struct tcpcb *);
1091 int tcp_ctloutput(struct socket *, struct sockopt *);
1092 void tcp_fini(void *);
1093 char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, const void *,
1094 const void *);
1095 char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, const void *,
1096 const void *);
1097 int tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *,
1098 struct mbuf *);
1099 void tcp_reass_global_init(void);
1100 void tcp_reass_flush(struct tcpcb *);
1101 void tcp_dooptions(struct tcpopt *, u_char *, int, int);
1102 void tcp_dropwithreset(struct mbuf *, struct tcphdr *,
1103 struct tcpcb *, int, int);
1104 void tcp_pulloutofband(struct socket *,
1105 struct tcphdr *, struct mbuf *, int);
1106 void tcp_xmit_timer(struct tcpcb *, int);
1107 void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
1108 void cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
1109 uint16_t nsegs, uint16_t type);
1110 void cc_conn_init(struct tcpcb *tp);
1111 void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
1112 void cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos);
1113 void cc_ecnpkt_handler_flags(struct tcpcb *tp, uint16_t flags, uint8_t iptos);
1114 void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
1115 #ifdef TCP_HHOOK
1116 void hhook_run_tcp_est_in(struct tcpcb *tp,
1117 struct tcphdr *th, struct tcpopt *to);
1118 #endif
1119
1120 int tcp_input(struct mbuf **, int *, int);
1121 int tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
1122 struct tcpcb *, int);
1123 int tcp_input_with_port(struct mbuf **, int *, int, uint16_t);
1124 void tcp_do_segment(struct mbuf *, struct tcphdr *,
1125 struct socket *, struct tcpcb *, int, int, uint8_t);
1126
1127 int register_tcp_functions(struct tcp_function_block *blk, int wait);
1128 int register_tcp_functions_as_names(struct tcp_function_block *blk,
1129 int wait, const char *names[], int *num_names);
1130 int register_tcp_functions_as_name(struct tcp_function_block *blk,
1131 const char *name, int wait);
1132 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
1133 bool force);
1134 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
1135 int find_tcp_function_alias(struct tcp_function_block *blk, struct tcp_function_set *fs);
1136 void tcp_switch_back_to_default(struct tcpcb *tp);
1137 struct tcp_function_block *
1138 find_and_ref_tcp_fb(struct tcp_function_block *fs);
1139 int tcp_default_ctloutput(struct inpcb *inp, struct sockopt *sopt);
1140 int tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt);
1141
1142 extern counter_u64_t tcp_inp_lro_direct_queue;
1143 extern counter_u64_t tcp_inp_lro_wokeup_queue;
1144 extern counter_u64_t tcp_inp_lro_compressed;
1145 extern counter_u64_t tcp_inp_lro_locks_taken;
1146 extern counter_u64_t tcp_extra_mbuf;
1147 extern counter_u64_t tcp_would_have_but;
1148 extern counter_u64_t tcp_comp_total;
1149 extern counter_u64_t tcp_uncomp_total;
1150 extern counter_u64_t tcp_bad_csums;
1151
1152 #ifdef NETFLIX_EXP_DETECTION
1153 /* Various SACK attack thresholds */
1154 extern int32_t tcp_force_detection;
1155 extern int32_t tcp_sack_to_ack_thresh;
1156 extern int32_t tcp_sack_to_move_thresh;
1157 extern int32_t tcp_restoral_thresh;
1158 extern int32_t tcp_sad_decay_val;
1159 extern int32_t tcp_sad_pacing_interval;
1160 extern int32_t tcp_sad_low_pps;
1161 extern int32_t tcp_map_minimum;
1162 extern int32_t tcp_attack_on_turns_on_logging;
1163 #endif
1164 extern uint32_t tcp_ack_war_time_window;
1165 extern uint32_t tcp_ack_war_cnt;
1166
1167 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
1168 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
1169 void tcp6_use_min_mtu(struct tcpcb *);
1170 u_int tcp_maxseg(const struct tcpcb *);
1171 u_int tcp_fixed_maxseg(const struct tcpcb *);
1172 void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
1173 struct tcp_ifcap *);
1174 void tcp_mss(struct tcpcb *, int);
1175 int tcp_mssopt(struct in_conninfo *);
1176 struct tcpcb *
1177 tcp_newtcpcb(struct inpcb *);
1178 int tcp_default_output(struct tcpcb *);
1179 void tcp_state_change(struct tcpcb *, int);
1180 void tcp_respond(struct tcpcb *, void *,
1181 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, uint16_t);
1182 bool tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
1183 struct mbuf *, int);
1184 void tcp_setpersist(struct tcpcb *);
1185 void tcp_record_dsack(struct tcpcb *tp, tcp_seq start, tcp_seq end, int tlp);
1186 struct tcptemp *
1187 tcpip_maketemplate(struct inpcb *);
1188 void tcpip_fillheaders(struct inpcb *, uint16_t, void *, void *);
1189 void tcp_timer_activate(struct tcpcb *, tt_which, u_int);
1190 bool tcp_timer_active(struct tcpcb *, tt_which);
1191 void tcp_timer_stop(struct tcpcb *);
1192 void tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
1193 int inp_to_cpuid(struct inpcb *inp);
1194 /*
1195 * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
1196 */
1197 void tcp_hc_init(void);
1198 #ifdef VIMAGE
1199 void tcp_hc_destroy(void);
1200 #endif
1201 void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
1202 uint32_t tcp_hc_getmtu(struct in_conninfo *);
1203 void tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
1204 void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
1205 void cc_after_idle(struct tcpcb *tp);
1206
1207 extern struct protosw tcp_protosw; /* shared for TOE */
1208 extern struct protosw tcp6_protosw; /* shared for TOE */
1209
1210 uint32_t tcp_new_ts_offset(struct in_conninfo *);
1211 tcp_seq tcp_new_isn(struct in_conninfo *);
1212
1213 int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
1214 int tcp_dsack_block_exists(struct tcpcb *);
1215 void tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq);
1216 void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1217 void tcp_clean_dsack_blocks(struct tcpcb *tp);
1218 void tcp_clean_sackreport(struct tcpcb *tp);
1219 void tcp_sack_adjust(struct tcpcb *tp);
1220 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1221 void tcp_do_prr_ack(struct tcpcb *, struct tcphdr *, struct tcpopt *);
1222 void tcp_lost_retransmission(struct tcpcb *, struct tcphdr *);
1223 void tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1224 void tcp_free_sackholes(struct tcpcb *tp);
1225 void tcp_sack_lost_retransmission(struct tcpcb *, struct tcphdr *);
1226 int tcp_newreno(struct tcpcb *, struct tcphdr *);
1227 int tcp_compute_pipe(struct tcpcb *);
1228 uint32_t tcp_compute_initwnd(uint32_t);
1229 void tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
1230 int tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes,
1231 size_t seed_len);
1232 int tcp_can_enable_pacing(void);
1233 void tcp_decrement_paced_conn(void);
1234
1235 struct mbuf *
1236 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen,
1237 int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls);
1238
1239 int tcp_stats_init(void);
1240 void tcp_log_end_status(struct tcpcb *tp, uint8_t status);
1241
1242 static inline void
1243 tcp_fields_to_host(struct tcphdr *th)
1244 {
1245
1246 th->th_seq = ntohl(th->th_seq);
1247 th->th_ack = ntohl(th->th_ack);
1248 th->th_win = ntohs(th->th_win);
1249 th->th_urp = ntohs(th->th_urp);
1250 }
1251
1252 static inline void
1253 tcp_fields_to_net(struct tcphdr *th)
1254 {
1255
1256 th->th_seq = htonl(th->th_seq);
1257 th->th_ack = htonl(th->th_ack);
1258 th->th_win = htons(th->th_win);
1259 th->th_urp = htons(th->th_urp);
1260 }
1261
1262 static inline uint16_t
1263 tcp_get_flags(const struct tcphdr *th)
1264 {
1265 return (((uint16_t)th->th_x2 << 8) | th->th_flags);
1266 }
1267
1268 static inline void
1269 tcp_set_flags(struct tcphdr *th, uint16_t flags)
1270 {
1271 th->th_x2 = (flags >> 8) & 0x0f;
1272 th->th_flags = flags & 0xff;
1273 }
1274
1275 static inline void
1276 tcp_account_for_send(struct tcpcb *tp, uint32_t len, uint8_t is_rxt,
1277 uint8_t is_tlp, int hw_tls)
1278 {
1279 if (is_tlp) {
1280 tp->t_sndtlppack++;
1281 tp->t_sndtlpbyte += len;
1282 }
1283 /* To get total bytes sent you must add t_snd_rxt_bytes to t_sndbytes */
1284 if (is_rxt)
1285 tp->t_snd_rxt_bytes += len;
1286 else
1287 tp->t_sndbytes += len;
1288
1289 #ifdef KERN_TLS
1290 if (hw_tls && is_rxt && len != 0) {
1291 uint64_t rexmit_percent = (1000ULL * tp->t_snd_rxt_bytes) / (10ULL * (tp->t_snd_rxt_bytes + tp->t_sndbytes));
1292 if (rexmit_percent > ktls_ifnet_max_rexmit_pct)
1293 ktls_disable_ifnet(tp);
1294 }
1295 #endif
1296
1297 }
1298 #endif /* _KERNEL */
1299
1300 #endif /* _NETINET_TCP_VAR_H_ */
Cache object: 0faf61bc6555914a467087a8a0f4ce28
|