freebsd-skq/sys/netinet/cc/cc_newreno.c
Hiren Panchasara e04310d59b Set slow start threshold more accurately on loss to be flightsize/2 instead of
cwnd/2 as recommended by RFC5681. (spotted by mmacy at nextbsd dot org)

Restore pre-r307901 behavior of aligning ssthresh/cwnd on mss boundary. (spotted
by slawa at zxy dot spb dot ru)

Tested by:	    dim, Slawa <slawa at zxy dot spb dot ru>
MFC after:	    1 month
X-MFC with:	    r307901
Sponsored by:	    Limelight Networks
Differential Revision:	https://reviews.freebsd.org/D8349
2016-11-01 21:08:37 +00:00

257 lines
8.1 KiB
C

/*-
* Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995
* The Regents of the University of California.
* Copyright (c) 2007-2008,2010
* Swinburne University of Technology, Melbourne, Australia.
* Copyright (c) 2009-2010 Lawrence Stewart <lstewart@freebsd.org>
* Copyright (c) 2010 The FreeBSD Foundation
* All rights reserved.
*
* This software was developed at the Centre for Advanced Internet
* Architectures, Swinburne University of Technology, by Lawrence Stewart, James
* Healy and David Hayes, made possible in part by a grant from the Cisco
* University Research Program Fund at Community Foundation Silicon Valley.
*
* Portions of this software were developed at the Centre for Advanced
* Internet Architectures, Swinburne University of Technology, Melbourne,
* Australia by David Hayes under sponsorship from the FreeBSD Foundation.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/*
* This software was first released in 2007 by James Healy and Lawrence Stewart
* whilst working on the NewTCP research project at Swinburne University of
* Technology's Centre for Advanced Internet Architectures, Melbourne,
* Australia, which was made possible in part by a grant from the Cisco
* University Research Program Fund at Community Foundation Silicon Valley.
* More details are available at:
* http://caia.swin.edu.au/urp/newtcp/
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/sysctl.h>
#include <sys/systm.h>
#include <net/vnet.h>
#include <netinet/tcp.h>
#include <netinet/tcp_seq.h>
#include <netinet/tcp_var.h>
#include <netinet/cc/cc.h>
#include <netinet/cc/cc_module.h>
static void newreno_ack_received(struct cc_var *ccv, uint16_t type);
static void newreno_after_idle(struct cc_var *ccv);
static void newreno_cong_signal(struct cc_var *ccv, uint32_t type);
static void newreno_post_recovery(struct cc_var *ccv);
struct cc_algo newreno_cc_algo = {
.name = "newreno",
.ack_received = newreno_ack_received,
.after_idle = newreno_after_idle,
.cong_signal = newreno_cong_signal,
.post_recovery = newreno_post_recovery,
};
static void
newreno_ack_received(struct cc_var *ccv, uint16_t type)
{
if (type == CC_ACK && !IN_RECOVERY(CCV(ccv, t_flags)) &&
(ccv->flags & CCF_CWND_LIMITED)) {
u_int cw = CCV(ccv, snd_cwnd);
u_int incr = CCV(ccv, t_maxseg);
/*
* Regular in-order ACK, open the congestion window.
* Method depends on which congestion control state we're
* in (slow start or cong avoid) and if ABC (RFC 3465) is
* enabled.
*
* slow start: cwnd <= ssthresh
* cong avoid: cwnd > ssthresh
*
* slow start and ABC (RFC 3465):
* Grow cwnd exponentially by the amount of data
* ACKed capping the max increment per ACK to
* (abc_l_var * maxseg) bytes.
*
* slow start without ABC (RFC 5681):
* Grow cwnd exponentially by maxseg per ACK.
*
* cong avoid and ABC (RFC 3465):
* Grow cwnd linearly by maxseg per RTT for each
* cwnd worth of ACKed data.
*
* cong avoid without ABC (RFC 5681):
* Grow cwnd linearly by approximately maxseg per RTT using
* maxseg^2 / cwnd per ACK as the increment.
* If cwnd > maxseg^2, fix the cwnd increment at 1 byte to
* avoid capping cwnd.
*/
if (cw > CCV(ccv, snd_ssthresh)) {
if (V_tcp_do_rfc3465) {
if (ccv->flags & CCF_ABC_SENTAWND)
ccv->flags &= ~CCF_ABC_SENTAWND;
else
incr = 0;
} else
incr = max((incr * incr / cw), 1);
} else if (V_tcp_do_rfc3465) {
/*
* In slow-start with ABC enabled and no RTO in sight?
* (Must not use abc_l_var > 1 if slow starting after
* an RTO. On RTO, snd_nxt = snd_una, so the
* snd_nxt == snd_max check is sufficient to
* handle this).
*
* XXXLAS: Find a way to signal SS after RTO that
* doesn't rely on tcpcb vars.
*/
if (CCV(ccv, snd_nxt) == CCV(ccv, snd_max))
incr = min(ccv->bytes_this_ack,
ccv->nsegs * V_tcp_abc_l_var *
CCV(ccv, t_maxseg));
else
incr = min(ccv->bytes_this_ack, CCV(ccv, t_maxseg));
}
/* ABC is on by default, so incr equals 0 frequently. */
if (incr > 0)
CCV(ccv, snd_cwnd) = min(cw + incr,
TCP_MAXWIN << CCV(ccv, snd_scale));
}
}
static void
newreno_after_idle(struct cc_var *ccv)
{
int rw;
/*
* If we've been idle for more than one retransmit timeout the old
* congestion window is no longer current and we have to reduce it to
* the restart window before we can transmit again.
*
* The restart window is the initial window or the last CWND, whichever
* is smaller.
*
* This is done to prevent us from flooding the path with a full CWND at
* wirespeed, overloading router and switch buffers along the way.
*
* See RFC5681 Section 4.1. "Restarting Idle Connections".
*/
if (V_tcp_do_rfc3390)
rw = min(4 * CCV(ccv, t_maxseg),
max(2 * CCV(ccv, t_maxseg), 4380));
else
rw = CCV(ccv, t_maxseg) * 2;
CCV(ccv, snd_cwnd) = min(rw, CCV(ccv, snd_cwnd));
}
/*
* Perform any necessary tasks before we enter congestion recovery.
*/
static void
newreno_cong_signal(struct cc_var *ccv, uint32_t type)
{
uint32_t cwin, ssthresh_on_loss;
u_int mss;
cwin = CCV(ccv, snd_cwnd);
mss = CCV(ccv, t_maxseg);
ssthresh_on_loss =
max((CCV(ccv, snd_max) - CCV(ccv, snd_una)) / 2 / mss, 2)
* mss;
/* Catch algos which mistakenly leak private signal types. */
KASSERT((type & CC_SIGPRIVMASK) == 0,
("%s: congestion signal type 0x%08x is private\n", __func__, type));
cwin = max(cwin / 2 / mss, 2) * mss;
switch (type) {
case CC_NDUPACK:
if (!IN_FASTRECOVERY(CCV(ccv, t_flags))) {
if (!IN_CONGRECOVERY(CCV(ccv, t_flags))) {
CCV(ccv, snd_ssthresh) = ssthresh_on_loss;
CCV(ccv, snd_cwnd) = cwin;
}
ENTER_RECOVERY(CCV(ccv, t_flags));
}
break;
case CC_ECN:
if (!IN_CONGRECOVERY(CCV(ccv, t_flags))) {
CCV(ccv, snd_ssthresh) = ssthresh_on_loss;
CCV(ccv, snd_cwnd) = cwin;
ENTER_CONGRECOVERY(CCV(ccv, t_flags));
}
break;
case CC_RTO:
CCV(ccv, snd_ssthresh) = ssthresh_on_loss;
CCV(ccv, snd_cwnd) = mss;
break;
}
}
/*
* Perform any necessary tasks before we exit congestion recovery.
*/
static void
newreno_post_recovery(struct cc_var *ccv)
{
int pipe;
pipe = 0;
if (IN_FASTRECOVERY(CCV(ccv, t_flags))) {
/*
* Fast recovery will conclude after returning from this
* function. Window inflation should have left us with
* approximately snd_ssthresh outstanding data. But in case we
* would be inclined to send a burst, better to do it via the
* slow start mechanism.
*
* XXXLAS: Find a way to do this without needing curack
*/
if (V_tcp_do_rfc6675_pipe)
pipe = tcp_compute_pipe(ccv->ccvc.tcp);
else
pipe = CCV(ccv, snd_max) - ccv->curack;
if (pipe < CCV(ccv, snd_ssthresh))
CCV(ccv, snd_cwnd) = pipe + CCV(ccv, t_maxseg);
else
CCV(ccv, snd_cwnd) = CCV(ccv, snd_ssthresh);
}
}
DECLARE_CC_MODULE(newreno, &newreno_cc_algo);