xref: /netbsd-src/sys/netinet/tcp_congctl.c (revision 946379e7b37692fc43f68eb0d1c10daa0a7f3b6c)
1 /*	$NetBSD: tcp_congctl.c,v 1.20 2015/08/24 22:21:26 pooka Exp $	*/
2 
3 /*-
4  * Copyright (c) 1997, 1998, 1999, 2001, 2005, 2006 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation
9  * Facility, NASA Ames Research Center.
10  * This code is derived from software contributed to The NetBSD Foundation
11  * by Charles M. Hannum.
12  * This code is derived from software contributed to The NetBSD Foundation
13  * by Rui Paulo.
14  *
15  * Redistribution and use in source and binary forms, with or without
16  * modification, are permitted provided that the following conditions
17  * are met:
18  * 1. Redistributions of source code must retain the above copyright
19  *    notice, this list of conditions and the following disclaimer.
20  * 2. Redistributions in binary form must reproduce the above copyright
21  *    notice, this list of conditions and the following disclaimer in the
22  *    documentation and/or other materials provided with the distribution.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
25  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
26  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
27  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
28  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 /*
38  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
39  * All rights reserved.
40  *
41  * Redistribution and use in source and binary forms, with or without
42  * modification, are permitted provided that the following conditions
43  * are met:
44  * 1. Redistributions of source code must retain the above copyright
45  *    notice, this list of conditions and the following disclaimer.
46  * 2. Redistributions in binary form must reproduce the above copyright
47  *    notice, this list of conditions and the following disclaimer in the
48  *    documentation and/or other materials provided with the distribution.
49  * 3. Neither the name of the project nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  */
65 
66 /*
67  *      @(#)COPYRIGHT   1.1 (NRL) 17 January 1995
68  *
69  * NRL grants permission for redistribution and use in source and binary
70  * forms, with or without modification, of the software and documentation
71  * created at NRL provided that the following conditions are met:
72  *
73  * 1. Redistributions of source code must retain the above copyright
74  *    notice, this list of conditions and the following disclaimer.
75  * 2. Redistributions in binary form must reproduce the above copyright
76  *    notice, this list of conditions and the following disclaimer in the
77  *    documentation and/or other materials provided with the distribution.
78  * 3. All advertising materials mentioning features or use of this software
79  *    must display the following acknowledgements:
80  *      This product includes software developed by the University of
81  *      California, Berkeley and its contributors.
82  *      This product includes software developed at the Information
83  *      Technology Division, US Naval Research Laboratory.
84  * 4. Neither the name of the NRL nor the names of its contributors
85  *    may be used to endorse or promote products derived from this software
86  *    without specific prior written permission.
87  *
88  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
89  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
90  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
91  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
92  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
93  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
94  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
95  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
96  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
97  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
98  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
99  *
100  * The views and conclusions contained in the software and documentation
101  * are those of the authors and should not be interpreted as representing
102  * official policies, either expressed or implied, of the US Naval
103  * Research Laboratory (NRL).
104  */
105 
106 /*
107  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995
108  *	The Regents of the University of California.  All rights reserved.
109  *
110  * Redistribution and use in source and binary forms, with or without
111  * modification, are permitted provided that the following conditions
112  * are met:
113  * 1. Redistributions of source code must retain the above copyright
114  *    notice, this list of conditions and the following disclaimer.
115  * 2. Redistributions in binary form must reproduce the above copyright
116  *    notice, this list of conditions and the following disclaimer in the
117  *    documentation and/or other materials provided with the distribution.
118  * 3. Neither the name of the University nor the names of its contributors
119  *    may be used to endorse or promote products derived from this software
120  *    without specific prior written permission.
121  *
122  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
123  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
124  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
125  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
126  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
127  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
128  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
129  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
130  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
131  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
132  * SUCH DAMAGE.
133  *
134  *	@(#)tcp_input.c	8.12 (Berkeley) 5/24/95
135  */
136 
137 #include <sys/cdefs.h>
138 __KERNEL_RCSID(0, "$NetBSD: tcp_congctl.c,v 1.20 2015/08/24 22:21:26 pooka Exp $");
139 
140 #ifdef _KERNEL_OPT
141 #include "opt_inet.h"
142 #include "opt_tcp_debug.h"
143 #include "opt_tcp_congctl.h"
144 #endif
145 
146 #include <sys/param.h>
147 #include <sys/systm.h>
148 #include <sys/malloc.h>
149 #include <sys/mbuf.h>
150 #include <sys/protosw.h>
151 #include <sys/socket.h>
152 #include <sys/socketvar.h>
153 #include <sys/errno.h>
154 #include <sys/syslog.h>
155 #include <sys/pool.h>
156 #include <sys/domain.h>
157 #include <sys/kernel.h>
158 #include <sys/mutex.h>
159 
160 #include <net/if.h>
161 #include <net/route.h>
162 
163 #include <netinet/in.h>
164 #include <netinet/in_systm.h>
165 #include <netinet/ip.h>
166 #include <netinet/in_pcb.h>
167 #include <netinet/in_var.h>
168 #include <netinet/ip_var.h>
169 
170 #ifdef INET6
171 #ifndef INET
172 #include <netinet/in.h>
173 #endif
174 #include <netinet/ip6.h>
175 #include <netinet6/ip6_var.h>
176 #include <netinet6/in6_pcb.h>
177 #include <netinet6/ip6_var.h>
178 #include <netinet6/in6_var.h>
179 #include <netinet/icmp6.h>
180 #include <netinet6/nd6.h>
181 #endif
182 
183 #include <netinet/tcp.h>
184 #include <netinet/tcp_fsm.h>
185 #include <netinet/tcp_seq.h>
186 #include <netinet/tcp_timer.h>
187 #include <netinet/tcp_var.h>
188 #include <netinet/tcpip.h>
189 #include <netinet/tcp_congctl.h>
190 #ifdef TCP_DEBUG
191 #include <netinet/tcp_debug.h>
192 #endif
193 
194 /*
195  * TODO:
196  *   consider separating the actual implementations in another file.
197  */
198 
199 static void tcp_common_congestion_exp(struct tcpcb *, int, int);
200 
201 static int  tcp_reno_do_fast_retransmit(struct tcpcb *, const struct tcphdr *);
202 static int  tcp_reno_fast_retransmit(struct tcpcb *, const struct tcphdr *);
203 static void tcp_reno_slow_retransmit(struct tcpcb *);
204 static void tcp_reno_fast_retransmit_newack(struct tcpcb *,
205     const struct tcphdr *);
206 static void tcp_reno_newack(struct tcpcb *, const struct tcphdr *);
207 static void tcp_reno_congestion_exp(struct tcpcb *tp);
208 
209 static int  tcp_newreno_fast_retransmit(struct tcpcb *, const struct tcphdr *);
210 static void tcp_newreno_fast_retransmit_newack(struct tcpcb *,
211 	const struct tcphdr *);
212 static void tcp_newreno_newack(struct tcpcb *, const struct tcphdr *);
213 
214 static int tcp_cubic_fast_retransmit(struct tcpcb *, const struct tcphdr *);
215 static void tcp_cubic_slow_retransmit(struct tcpcb *tp);
216 static void tcp_cubic_newack(struct tcpcb *, const struct tcphdr *);
217 static void tcp_cubic_congestion_exp(struct tcpcb *);
218 
219 static void tcp_congctl_fillnames(void);
220 
221 extern int tcprexmtthresh;
222 
223 MALLOC_DEFINE(M_TCPCONGCTL, "tcpcongctl", "TCP congestion control structures");
224 
225 /* currently selected global congestion control */
226 char tcp_congctl_global_name[TCPCC_MAXLEN];
227 
228 /* available global congestion control algorithms */
229 char tcp_congctl_avail[10 * TCPCC_MAXLEN];
230 
231 /*
232  * Used to list the available congestion control algorithms.
233  */
234 TAILQ_HEAD(, tcp_congctlent) tcp_congctlhd =
235     TAILQ_HEAD_INITIALIZER(tcp_congctlhd);
236 
237 static struct tcp_congctlent * tcp_congctl_global;
238 
239 static kmutex_t tcp_congctl_mtx;
240 
241 void
242 tcp_congctl_init(void)
243 {
244 	int r __diagused;
245 
246 	mutex_init(&tcp_congctl_mtx, MUTEX_DEFAULT, IPL_NONE);
247 
248 	/* Base algorithms. */
249 	r = tcp_congctl_register("reno", &tcp_reno_ctl);
250 	KASSERT(r == 0);
251 	r = tcp_congctl_register("newreno", &tcp_newreno_ctl);
252 	KASSERT(r == 0);
253 	r = tcp_congctl_register("cubic", &tcp_cubic_ctl);
254 	KASSERT(r == 0);
255 
256 	/* NewReno is the default. */
257 #ifndef TCP_CONGCTL_DEFAULT
258 #define TCP_CONGCTL_DEFAULT "newreno"
259 #endif
260 
261 	r = tcp_congctl_select(NULL, TCP_CONGCTL_DEFAULT);
262 	KASSERT(r == 0);
263 }
264 
265 /*
266  * Register a congestion algorithm and select it if we have none.
267  */
268 int
269 tcp_congctl_register(const char *name, const struct tcp_congctl *tcc)
270 {
271 	struct tcp_congctlent *ntcc, *tccp;
272 
273 	TAILQ_FOREACH(tccp, &tcp_congctlhd, congctl_ent)
274 		if (!strcmp(name, tccp->congctl_name)) {
275 			/* name already registered */
276 			return EEXIST;
277 		}
278 
279 	ntcc = malloc(sizeof(*ntcc), M_TCPCONGCTL, M_WAITOK|M_ZERO);
280 
281 	strlcpy(ntcc->congctl_name, name, sizeof(ntcc->congctl_name) - 1);
282 	ntcc->congctl_ctl = tcc;
283 
284 	TAILQ_INSERT_TAIL(&tcp_congctlhd, ntcc, congctl_ent);
285 	tcp_congctl_fillnames();
286 
287 	if (TAILQ_FIRST(&tcp_congctlhd) == ntcc)
288 		tcp_congctl_select(NULL, name);
289 
290 	return 0;
291 }
292 
293 int
294 tcp_congctl_unregister(const char *name)
295 {
296 	struct tcp_congctlent *tccp, *rtccp;
297 	unsigned int size;
298 
299 	rtccp = NULL;
300 	size = 0;
301 	TAILQ_FOREACH(tccp, &tcp_congctlhd, congctl_ent) {
302 		if (!strcmp(name, tccp->congctl_name))
303 			rtccp = tccp;
304 		size++;
305 	}
306 
307 	if (!rtccp)
308 		return ENOENT;
309 
310 	if (size <= 1 || tcp_congctl_global == rtccp || rtccp->congctl_refcnt)
311 		return EBUSY;
312 
313 	TAILQ_REMOVE(&tcp_congctlhd, rtccp, congctl_ent);
314 	free(rtccp, M_TCPCONGCTL);
315 	tcp_congctl_fillnames();
316 
317 	return 0;
318 }
319 
320 /*
321  * Select a congestion algorithm by name.
322  */
323 int
324 tcp_congctl_select(struct tcpcb *tp, const char *name)
325 {
326 	struct tcp_congctlent *tccp, *old_tccp, *new_tccp;
327 	bool old_found, new_found;
328 
329 	KASSERT(name);
330 
331 	old_found = (tp == NULL || tp->t_congctl == NULL);
332 	old_tccp = NULL;
333 	new_found = false;
334 	new_tccp = NULL;
335 
336 	TAILQ_FOREACH(tccp, &tcp_congctlhd, congctl_ent) {
337 		if (!old_found && tccp->congctl_ctl == tp->t_congctl) {
338 			old_tccp = tccp;
339 			old_found = true;
340 		}
341 
342 		if (!new_found && !strcmp(name, tccp->congctl_name)) {
343 			new_tccp = tccp;
344 			new_found = true;
345 		}
346 
347 		if (new_found && old_found) {
348 			if (tp) {
349 				mutex_enter(&tcp_congctl_mtx);
350 				if (old_tccp)
351 					old_tccp->congctl_refcnt--;
352 				tp->t_congctl = new_tccp->congctl_ctl;
353 				new_tccp->congctl_refcnt++;
354 				mutex_exit(&tcp_congctl_mtx);
355 			} else {
356 				tcp_congctl_global = new_tccp;
357 				strlcpy(tcp_congctl_global_name,
358 				    new_tccp->congctl_name,
359 				    sizeof(tcp_congctl_global_name) - 1);
360 			}
361 			return 0;
362 		}
363 	}
364 
365 	return EINVAL;
366 }
367 
368 void
369 tcp_congctl_release(struct tcpcb *tp)
370 {
371 	struct tcp_congctlent *tccp;
372 
373 	KASSERT(tp->t_congctl);
374 
375 	TAILQ_FOREACH(tccp, &tcp_congctlhd, congctl_ent) {
376 		if (tccp->congctl_ctl == tp->t_congctl) {
377 			tccp->congctl_refcnt--;
378 			return;
379 		}
380 	}
381 }
382 
383 /*
384  * Returns the name of a congestion algorithm.
385  */
386 const char *
387 tcp_congctl_bystruct(const struct tcp_congctl *tcc)
388 {
389 	struct tcp_congctlent *tccp;
390 
391 	KASSERT(tcc);
392 
393 	TAILQ_FOREACH(tccp, &tcp_congctlhd, congctl_ent)
394 		if (tccp->congctl_ctl == tcc)
395 			return tccp->congctl_name;
396 
397 	return NULL;
398 }
399 
400 static void
401 tcp_congctl_fillnames(void)
402 {
403 	struct tcp_congctlent *tccp;
404 	const char *delim = " ";
405 
406 	tcp_congctl_avail[0] = '\0';
407 	TAILQ_FOREACH(tccp, &tcp_congctlhd, congctl_ent) {
408 		strlcat(tcp_congctl_avail, tccp->congctl_name,
409 		    sizeof(tcp_congctl_avail) - 1);
410 		if (TAILQ_NEXT(tccp, congctl_ent))
411 			strlcat(tcp_congctl_avail, delim,
412 			    sizeof(tcp_congctl_avail) - 1);
413 	}
414 
415 }
416 
417 /* ------------------------------------------------------------------------ */
418 
419 /*
420  * Common stuff
421  */
422 
423 /* Window reduction (1-beta) for [New]Reno: 0.5 */
424 #define RENO_BETAA 1
425 #define RENO_BETAB 2
426 /* Window reduction (1-beta) for Cubic: 0.8 */
427 #define CUBIC_BETAA 4
428 #define CUBIC_BETAB 5
429 /* Draft Rhee Section 4.1 */
430 #define CUBIC_CA 4
431 #define CUBIC_CB 10
432 
433 static void
434 tcp_common_congestion_exp(struct tcpcb *tp, int betaa, int betab)
435 {
436 	u_int win;
437 
438 	/*
439 	 * Reduce the congestion window and the slow start threshold.
440 	 */
441 	win = min(tp->snd_wnd, tp->snd_cwnd) * betaa / betab / tp->t_segsz;
442 	if (win < 2)
443 		win = 2;
444 
445 	tp->snd_ssthresh = win * tp->t_segsz;
446 	tp->snd_recover = tp->snd_max;
447 	tp->snd_cwnd = tp->snd_ssthresh;
448 
449 	/*
450 	 * When using TCP ECN, notify the peer that
451 	 * we reduced the cwnd.
452 	 */
453 	if (TCP_ECN_ALLOWED(tp))
454 		tp->t_flags |= TF_ECN_SND_CWR;
455 }
456 
457 
458 /* ------------------------------------------------------------------------ */
459 
460 /*
461  * TCP/Reno congestion control.
462  */
463 static void
464 tcp_reno_congestion_exp(struct tcpcb *tp)
465 {
466 
467 	tcp_common_congestion_exp(tp, RENO_BETAA, RENO_BETAB);
468 }
469 
470 static int
471 tcp_reno_do_fast_retransmit(struct tcpcb *tp, const struct tcphdr *th)
472 {
473 	/*
474 	 * Dup acks mean that packets have left the
475 	 * network (they're now cached at the receiver)
476 	 * so bump cwnd by the amount in the receiver
477 	 * to keep a constant cwnd packets in the
478 	 * network.
479 	 *
480 	 * If we are using TCP/SACK, then enter
481 	 * Fast Recovery if the receiver SACKs
482 	 * data that is tcprexmtthresh * MSS
483 	 * bytes past the last ACKed segment,
484 	 * irrespective of the number of DupAcks.
485 	 */
486 
487 	tcp_seq onxt = tp->snd_nxt;
488 
489 	tp->t_partialacks = 0;
490 	TCP_TIMER_DISARM(tp, TCPT_REXMT);
491 	tp->t_rtttime = 0;
492 	if (TCP_SACK_ENABLED(tp)) {
493 		tp->t_dupacks = tcprexmtthresh;
494 		tp->sack_newdata = tp->snd_nxt;
495 		tp->snd_cwnd = tp->t_segsz;
496 		(void) tcp_output(tp);
497 		return 0;
498 	}
499 	tp->snd_nxt = th->th_ack;
500 	tp->snd_cwnd = tp->t_segsz;
501 	(void) tcp_output(tp);
502 	tp->snd_cwnd = tp->snd_ssthresh + tp->t_segsz * tp->t_dupacks;
503 	if (SEQ_GT(onxt, tp->snd_nxt))
504 		tp->snd_nxt = onxt;
505 
506 	return 0;
507 }
508 
509 static int
510 tcp_reno_fast_retransmit(struct tcpcb *tp, const struct tcphdr *th)
511 {
512 
513 	/*
514 	 * We know we're losing at the current
515 	 * window size so do congestion avoidance
516 	 * (set ssthresh to half the current window
517 	 * and pull our congestion window back to
518 	 * the new ssthresh).
519 	 */
520 
521 	tcp_reno_congestion_exp(tp);
522 	return tcp_reno_do_fast_retransmit(tp, th);
523 }
524 
525 static void
526 tcp_reno_slow_retransmit(struct tcpcb *tp)
527 {
528 	u_int win;
529 
530 	/*
531 	 * Close the congestion window down to one segment
532 	 * (we'll open it by one segment for each ack we get).
533 	 * Since we probably have a window's worth of unacked
534 	 * data accumulated, this "slow start" keeps us from
535 	 * dumping all that data as back-to-back packets (which
536 	 * might overwhelm an intermediate gateway).
537 	 *
538 	 * There are two phases to the opening: Initially we
539 	 * open by one mss on each ack.  This makes the window
540 	 * size increase exponentially with time.  If the
541 	 * window is larger than the path can handle, this
542 	 * exponential growth results in dropped packet(s)
543 	 * almost immediately.  To get more time between
544 	 * drops but still "push" the network to take advantage
545 	 * of improving conditions, we switch from exponential
546 	 * to linear window opening at some threshhold size.
547 	 * For a threshhold, we use half the current window
548 	 * size, truncated to a multiple of the mss.
549 	 *
550 	 * (the minimum cwnd that will give us exponential
551 	 * growth is 2 mss.  We don't allow the threshhold
552 	 * to go below this.)
553 	 */
554 
555 	win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_segsz;
556 	if (win < 2)
557 		win = 2;
558 	/* Loss Window MUST be one segment. */
559 	tp->snd_cwnd = tp->t_segsz;
560 	tp->snd_ssthresh = win * tp->t_segsz;
561 	tp->t_partialacks = -1;
562 	tp->t_dupacks = 0;
563 	tp->t_bytes_acked = 0;
564 
565 	if (TCP_ECN_ALLOWED(tp))
566 		tp->t_flags |= TF_ECN_SND_CWR;
567 }
568 
569 static void
570 tcp_reno_fast_retransmit_newack(struct tcpcb *tp,
571     const struct tcphdr *th)
572 {
573 	if (tp->t_partialacks < 0) {
574 		/*
575 		 * We were not in fast recovery.  Reset the duplicate ack
576 		 * counter.
577 		 */
578 		tp->t_dupacks = 0;
579 	} else {
580 		/*
581 		 * Clamp the congestion window to the crossover point and
582 		 * exit fast recovery.
583 		 */
584 		if (tp->snd_cwnd > tp->snd_ssthresh)
585 			tp->snd_cwnd = tp->snd_ssthresh;
586 		tp->t_partialacks = -1;
587 		tp->t_dupacks = 0;
588 		tp->t_bytes_acked = 0;
589 		if (TCP_SACK_ENABLED(tp) && SEQ_GT(th->th_ack, tp->snd_fack))
590 			tp->snd_fack = th->th_ack;
591 	}
592 }
593 
594 static void
595 tcp_reno_newack(struct tcpcb *tp, const struct tcphdr *th)
596 {
597 	/*
598 	 * When new data is acked, open the congestion window.
599 	 */
600 
601 	u_int cw = tp->snd_cwnd;
602 	u_int incr = tp->t_segsz;
603 
604 	if (tcp_do_abc) {
605 
606 		/*
607 		 * RFC 3465 Appropriate Byte Counting (ABC)
608 		 */
609 
610 		int acked = th->th_ack - tp->snd_una;
611 
612 		if (cw >= tp->snd_ssthresh) {
613 			tp->t_bytes_acked += acked;
614 			if (tp->t_bytes_acked >= cw) {
615 				/* Time to increase the window. */
616 				tp->t_bytes_acked -= cw;
617 			} else {
618 				/* No need to increase yet. */
619 				incr = 0;
620 			}
621 		} else {
622 			/*
623 			 * use 2*SMSS or 1*SMSS for the "L" param,
624 			 * depending on sysctl setting.
625 			 *
626 			 * (See RFC 3465 2.3 Choosing the Limit)
627 			 */
628 			u_int abc_lim;
629 
630 			abc_lim = (tcp_abc_aggressive == 0 ||
631 			    tp->snd_nxt != tp->snd_max) ? incr : incr * 2;
632 			incr = min(acked, abc_lim);
633 		}
634 	} else {
635 
636 		/*
637 		 * If the window gives us less than ssthresh packets
638 		 * in flight, open exponentially (segsz per packet).
639 		 * Otherwise open linearly: segsz per window
640 		 * (segsz^2 / cwnd per packet).
641 		 */
642 
643 		if (cw >= tp->snd_ssthresh) {
644 			incr = incr * incr / cw;
645 		}
646 	}
647 
648 	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN << tp->snd_scale);
649 }
650 
651 const struct tcp_congctl tcp_reno_ctl = {
652 	.fast_retransmit = tcp_reno_fast_retransmit,
653 	.slow_retransmit = tcp_reno_slow_retransmit,
654 	.fast_retransmit_newack = tcp_reno_fast_retransmit_newack,
655 	.newack = tcp_reno_newack,
656 	.cong_exp = tcp_reno_congestion_exp,
657 };
658 
659 /*
660  * TCP/NewReno Congestion control.
661  */
662 static int
663 tcp_newreno_fast_retransmit(struct tcpcb *tp, const struct tcphdr *th)
664 {
665 
666 	if (SEQ_LT(th->th_ack, tp->snd_high)) {
667 		/*
668 		 * False fast retransmit after timeout.
669 		 * Do not enter fast recovery
670 		 */
671 		tp->t_dupacks = 0;
672 		return 1;
673 	}
674 	/*
675 	 * Fast retransmit is same as reno.
676 	 */
677 	return tcp_reno_fast_retransmit(tp, th);
678 }
679 
680 /*
681  * Implement the NewReno response to a new ack, checking for partial acks in
682  * fast recovery.
683  */
684 static void
685 tcp_newreno_fast_retransmit_newack(struct tcpcb *tp, const struct tcphdr *th)
686 {
687 	if (tp->t_partialacks < 0) {
688 		/*
689 		 * We were not in fast recovery.  Reset the duplicate ack
690 		 * counter.
691 		 */
692 		tp->t_dupacks = 0;
693 	} else if (SEQ_LT(th->th_ack, tp->snd_recover)) {
694 		/*
695 		 * This is a partial ack.  Retransmit the first unacknowledged
696 		 * segment and deflate the congestion window by the amount of
697 		 * acknowledged data.  Do not exit fast recovery.
698 		 */
699 		tcp_seq onxt = tp->snd_nxt;
700 		u_long ocwnd = tp->snd_cwnd;
701 		int sack_num_segs = 1, sack_bytes_rxmt = 0;
702 
703 		/*
704 		 * snd_una has not yet been updated and the socket's send
705 		 * buffer has not yet drained off the ACK'd data, so we
706 		 * have to leave snd_una as it was to get the correct data
707 		 * offset in tcp_output().
708 		 */
709 		tp->t_partialacks++;
710 		TCP_TIMER_DISARM(tp, TCPT_REXMT);
711 		tp->t_rtttime = 0;
712 		tp->snd_nxt = th->th_ack;
713 
714 		if (TCP_SACK_ENABLED(tp)) {
715 			/*
716 			 * Partial ack handling within a sack recovery episode.
717 			 * Keeping this very simple for now. When a partial ack
718 			 * is received, force snd_cwnd to a value that will
719 			 * allow the sender to transmit no more than 2 segments.
720 			 * If necessary, a fancier scheme can be adopted at a
721 			 * later point, but for now, the goal is to prevent the
722 			 * sender from bursting a large amount of data in the
723 			 * midst of sack recovery.
724 		 	 */
725 
726 			/*
727 			 * send one or 2 segments based on how much
728 			 * new data was acked
729 			 */
730 			if (((th->th_ack - tp->snd_una) / tp->t_segsz) > 2)
731 				sack_num_segs = 2;
732 			(void)tcp_sack_output(tp, &sack_bytes_rxmt);
733 			tp->snd_cwnd = sack_bytes_rxmt +
734 			    (tp->snd_nxt - tp->sack_newdata) +
735 			    sack_num_segs * tp->t_segsz;
736 			tp->t_flags |= TF_ACKNOW;
737 			(void) tcp_output(tp);
738 		} else {
739 			/*
740 			 * Set snd_cwnd to one segment beyond ACK'd offset
741 			 * snd_una is not yet updated when we're called
742 			 */
743 			tp->snd_cwnd = tp->t_segsz + (th->th_ack - tp->snd_una);
744 			(void) tcp_output(tp);
745 			tp->snd_cwnd = ocwnd;
746 			if (SEQ_GT(onxt, tp->snd_nxt))
747 				tp->snd_nxt = onxt;
748 			/*
749 			 * Partial window deflation.  Relies on fact that
750 			 * tp->snd_una not updated yet.
751 		 	 */
752 			tp->snd_cwnd -= (th->th_ack - tp->snd_una -
753 			    tp->t_segsz);
754 		}
755 	} else {
756 		/*
757 		 * Complete ack.  Inflate the congestion window to ssthresh
758 		 * and exit fast recovery.
759 		 *
760 		 * Window inflation should have left us with approx.
761 		 * snd_ssthresh outstanding data.  But in case we
762 		 * would be inclined to send a burst, better to do
763 		 * it via the slow start mechanism.
764 		 */
765 		if (SEQ_SUB(tp->snd_max, th->th_ack) < tp->snd_ssthresh)
766 			tp->snd_cwnd = SEQ_SUB(tp->snd_max, th->th_ack)
767 			    + tp->t_segsz;
768 		else
769 			tp->snd_cwnd = tp->snd_ssthresh;
770 		tp->t_partialacks = -1;
771 		tp->t_dupacks = 0;
772 		tp->t_bytes_acked = 0;
773 		if (TCP_SACK_ENABLED(tp) && SEQ_GT(th->th_ack, tp->snd_fack))
774 			tp->snd_fack = th->th_ack;
775 	}
776 }
777 
778 static void
779 tcp_newreno_newack(struct tcpcb *tp, const struct tcphdr *th)
780 {
781 	/*
782 	 * If we are still in fast recovery (meaning we are using
783 	 * NewReno and we have only received partial acks), do not
784 	 * inflate the window yet.
785 	 */
786 	if (tp->t_partialacks < 0)
787 		tcp_reno_newack(tp, th);
788 }
789 
790 
791 const struct tcp_congctl tcp_newreno_ctl = {
792 	.fast_retransmit = tcp_newreno_fast_retransmit,
793 	.slow_retransmit = tcp_reno_slow_retransmit,
794 	.fast_retransmit_newack = tcp_newreno_fast_retransmit_newack,
795 	.newack = tcp_newreno_newack,
796 	.cong_exp = tcp_reno_congestion_exp,
797 };
798 
799 /*
800  * CUBIC - http://tools.ietf.org/html/draft-rhee-tcpm-cubic-02
801  */
802 
803 /* Cubic prototypes */
804 static void	tcp_cubic_update_ctime(struct tcpcb *tp);
805 static uint32_t	tcp_cubic_diff_ctime(struct tcpcb *);
806 static uint32_t	tcp_cubic_cbrt(uint32_t);
807 static ulong	tcp_cubic_getW(struct tcpcb *, uint32_t, uint32_t);
808 
809 /* Cubic TIME functions - XXX I don't like using timevals and microuptime */
810 /*
811  * Set congestion timer to now
812  */
813 static void
814 tcp_cubic_update_ctime(struct tcpcb *tp)
815 {
816 	struct timeval now_timeval;
817 
818 	getmicrouptime(&now_timeval);
819 	tp->snd_cubic_ctime = now_timeval.tv_sec * 1000 +
820 	    now_timeval.tv_usec / 1000;
821 }
822 
823 /*
824  * miliseconds from last congestion
825  */
826 static uint32_t
827 tcp_cubic_diff_ctime(struct tcpcb *tp)
828 {
829 	struct timeval now_timeval;
830 
831 	getmicrouptime(&now_timeval);
832 	return now_timeval.tv_sec * 1000 + now_timeval.tv_usec / 1000 -
833 	    tp->snd_cubic_ctime;
834 }
835 
836 /*
837  * Approximate cubic root
838  */
839 #define CBRT_ROUNDS 30
840 static uint32_t
841 tcp_cubic_cbrt(uint32_t v)
842 {
843 	int i, rounds = CBRT_ROUNDS;
844 	uint64_t x = v / 3;
845 
846 	/* We fail to calculate correct for small numbers */
847 	if (v == 0)
848 		return 0;
849 	else if (v < 4)
850 		return 1;
851 
852 	/*
853 	 * largest x that 2*x^3+3*x fits 64bit
854 	 * Avoid overflow for a time cost
855 	 */
856 	if (x > 2097151)
857 		rounds += 10;
858 
859 	for (i = 0; i < rounds; i++)
860 		if (rounds == CBRT_ROUNDS)
861 			x = (v + 2 * x * x * x) / (3 * x * x);
862 		else
863 			/* Avoid overflow */
864 			x = v / (3 * x * x) + 2 * x / 3;
865 
866 	return (uint32_t)x;
867 }
868 
869 /* Draft Rhee Section 3.1 - get W(t+rtt) - Eq. 1 */
870 static ulong
871 tcp_cubic_getW(struct tcpcb *tp, uint32_t ms_elapsed, uint32_t rtt)
872 {
873 	uint32_t K;
874 	long tK3;
875 
876 	/* Section 3.1 Eq. 2 */
877 	K = tcp_cubic_cbrt(tp->snd_cubic_wmax / CUBIC_BETAB *
878 	    CUBIC_CB / CUBIC_CA);
879 	/*  (t-K)^3 - not clear why is the measure unit mattering */
880 	tK3 = (long)(ms_elapsed + rtt) - (long)K;
881 	tK3 = tK3 * tK3 * tK3;
882 
883 	return CUBIC_CA * tK3 / CUBIC_CB + tp->snd_cubic_wmax;
884 }
885 
886 static void
887 tcp_cubic_congestion_exp(struct tcpcb *tp)
888 {
889 
890 	/*
891 	 * Congestion - Set WMax and shrink cwnd
892 	 */
893 	tcp_cubic_update_ctime(tp);
894 
895 	/* Section 3.6 - Fast Convergence */
896 	if (tp->snd_cubic_wmax < tp->snd_cubic_wmax_last) {
897 		tp->snd_cubic_wmax_last = tp->snd_cubic_wmax;
898 		tp->snd_cubic_wmax = tp->snd_cubic_wmax / 2 +
899 		    tp->snd_cubic_wmax * CUBIC_BETAA / CUBIC_BETAB / 2;
900 	} else {
901 		tp->snd_cubic_wmax_last = tp->snd_cubic_wmax;
902 		tp->snd_cubic_wmax = tp->snd_cwnd;
903 	}
904 
905 	tp->snd_cubic_wmax = max(tp->t_segsz, tp->snd_cubic_wmax);
906 
907 	/* Shrink CWND */
908 	tcp_common_congestion_exp(tp, CUBIC_BETAA, CUBIC_BETAB);
909 }
910 
911 static int
912 tcp_cubic_fast_retransmit(struct tcpcb *tp, const struct tcphdr *th)
913 {
914 
915 	if (SEQ_LT(th->th_ack, tp->snd_high)) {
916 		/* See newreno */
917 		tp->t_dupacks = 0;
918 		return 1;
919 	}
920 
921 	/*
922 	 * mark WMax
923 	 */
924 	tcp_cubic_congestion_exp(tp);
925 
926 	/* Do fast retransmit */
927 	return tcp_reno_do_fast_retransmit(tp, th);
928 }
929 
930 static void
931 tcp_cubic_newack(struct tcpcb *tp, const struct tcphdr *th)
932 {
933 	uint32_t ms_elapsed, rtt;
934 	u_long w_tcp;
935 
936 	/* Congestion avoidance and not in fast recovery and usable rtt */
937 	if (tp->snd_cwnd > tp->snd_ssthresh && tp->t_partialacks < 0 &&
938 	    /*
939 	     * t_srtt is 1/32 units of slow ticks
940 	     * converting it in ms would be equal to
941 	     * (t_srtt >> 5) * 1000 / PR_SLOWHZ ~= (t_srtt << 5) / PR_SLOWHZ
942 	     */
943 	    (rtt = (tp->t_srtt << 5) / PR_SLOWHZ) > 0) {
944 		ms_elapsed = tcp_cubic_diff_ctime(tp);
945 
946 		/* Compute W_tcp(t) */
947 		w_tcp = tp->snd_cubic_wmax * CUBIC_BETAA / CUBIC_BETAB +
948 		    ms_elapsed / rtt / 3;
949 
950 		if (tp->snd_cwnd > w_tcp) {
951 			/* Not in TCP friendly mode */
952 			tp->snd_cwnd += (tcp_cubic_getW(tp, ms_elapsed, rtt) -
953 			    tp->snd_cwnd) / tp->snd_cwnd;
954 		} else {
955 			/* friendly TCP mode */
956 			tp->snd_cwnd = w_tcp;
957 		}
958 
959 		/* Make sure we are within limits */
960 		tp->snd_cwnd = max(tp->snd_cwnd, tp->t_segsz);
961 		tp->snd_cwnd = min(tp->snd_cwnd, TCP_MAXWIN << tp->snd_scale);
962 	} else {
963 		/* Use New Reno */
964 		tcp_newreno_newack(tp, th);
965 	}
966 }
967 
968 static void
969 tcp_cubic_slow_retransmit(struct tcpcb *tp)
970 {
971 
972 	/* Timeout - Mark new congestion */
973 	tcp_cubic_congestion_exp(tp);
974 
975 	/* Loss Window MUST be one segment. */
976 	tp->snd_cwnd = tp->t_segsz;
977 	tp->t_partialacks = -1;
978 	tp->t_dupacks = 0;
979 	tp->t_bytes_acked = 0;
980 
981 	if (TCP_ECN_ALLOWED(tp))
982 		tp->t_flags |= TF_ECN_SND_CWR;
983 }
984 
985 const struct tcp_congctl tcp_cubic_ctl = {
986 	.fast_retransmit = tcp_cubic_fast_retransmit,
987 	.slow_retransmit = tcp_cubic_slow_retransmit,
988 	.fast_retransmit_newack = tcp_newreno_fast_retransmit_newack,
989 	.newack = tcp_cubic_newack,
990 	.cong_exp = tcp_cubic_congestion_exp,
991 };
992