freebsd-nq/sys/kern/kern_et.c
Alexander Motin a157e42516 Refactor timer management code with priority to one-shot operation mode.
The main goal of this is to generate timer interrupts only when there is
some work to do. When CPU is busy interrupts are generating at full rate
of hz + stathz to fullfill scheduler and timekeeping requirements. But
when CPU is idle, only minimum set of interrupts (down to 8 interrupts per
second per CPU now), needed to handle scheduled callouts is executed.
This allows significantly increase idle CPU sleep time, increasing effect
of static power-saving technologies. Also it should reduce host CPU load
on virtualized systems, when guest system is idle.

There is set of tunables, also available as writable sysctls, allowing to
control wanted event timer subsystem behavior:
  kern.eventtimer.timer - allows to choose event timer hardware to use.
On x86 there is up to 4 different kinds of timers. Depending on whether
chosen timer is per-CPU, behavior of other options slightly differs.
  kern.eventtimer.periodic - allows to choose periodic and one-shot
operation mode. In periodic mode, current timer hardware taken as the only
source of time for time events. This mode is quite alike to previous kernel
behavior. One-shot mode instead uses currently selected time counter
hardware to schedule all needed events one by one and program timer to
generate interrupt exactly in specified time. Default value depends of
chosen timer capabilities, but one-shot mode is preferred, until other is
forced by user or hardware.
  kern.eventtimer.singlemul - in periodic mode specifies how much times
higher timer frequency should be, to not strictly alias hardclock() and
statclock() events. Default values are 2 and 4, but could be reduced to 1
if extra interrupts are unwanted.
  kern.eventtimer.idletick - makes each CPU to receive every timer interrupt
independently of whether they busy or not. By default this options is
disabled. If chosen timer is per-CPU and runs in periodic mode, this option
has no effect - all interrupts are generating.

As soon as this patch modifies cpu_idle() on some platforms, I have also
refactored one on x86. Now it makes use of MONITOR/MWAIT instrunctions
(if supported) under high sleep/wakeup rate, as fast alternative to other
methods. It allows SMP scheduler to wake up sleeping CPUs much faster
without using IPI, significantly increasing performance on some highly
task-switching loads.

Tested by:	many (on i386, amd64, sparc64 and powerc)
H/W donated by:	Gheorghe Ardelean
Sponsored by:	iXsystems, Inc.
2010-09-13 07:25:35 +00:00

260 lines
6.9 KiB
C

/*-
* Copyright (c) 2010 Alexander Motin <mav@FreeBSD.org>
* All rights reserved.
*
* 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,
* without modification, immediately at the beginning of the file.
* 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 ``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 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/systm.h>
#include <sys/queue.h>
#include <sys/timeet.h>
SLIST_HEAD(et_eventtimers_list, eventtimer);
static struct et_eventtimers_list eventtimers = SLIST_HEAD_INITIALIZER(et_eventtimers);
struct mtx et_eventtimers_mtx;
MTX_SYSINIT(et_eventtimers_init, &et_eventtimers_mtx, "et_mtx", MTX_DEF);
SYSCTL_NODE(_kern, OID_AUTO, eventtimer, CTLFLAG_RW, 0, "Event timers");
SYSCTL_NODE(_kern_eventtimer, OID_AUTO, et, CTLFLAG_RW, 0, "");
/*
* Register a new event timer hardware.
*/
int
et_register(struct eventtimer *et)
{
struct eventtimer *tmp, *next;
if (et->et_quality >= 0 || bootverbose) {
if (et->et_frequency == 0) {
printf("Event timer \"%s\" quality %d\n",
et->et_name, et->et_quality);
} else {
printf("Event timer \"%s\" "
"frequency %ju Hz quality %d\n",
et->et_name, (uintmax_t)et->et_frequency,
et->et_quality);
}
}
et->et_sysctl = SYSCTL_ADD_NODE(NULL,
SYSCTL_STATIC_CHILDREN(_kern_eventtimer_et), OID_AUTO, et->et_name,
CTLFLAG_RW, 0, "event timer description");
SYSCTL_ADD_UINT(NULL, SYSCTL_CHILDREN(et->et_sysctl), OID_AUTO,
"flags", CTLFLAG_RD, &(et->et_flags), 0,
"Event timer capabilities");
SYSCTL_ADD_QUAD(NULL, SYSCTL_CHILDREN(et->et_sysctl), OID_AUTO,
"frequency", CTLFLAG_RD, &(et->et_frequency),
"Event timer base frequency");
SYSCTL_ADD_INT(NULL, SYSCTL_CHILDREN(et->et_sysctl), OID_AUTO,
"quality", CTLFLAG_RD, &(et->et_quality), 0,
"Goodness of event timer");
ET_LOCK();
if (SLIST_EMPTY(&eventtimers) ||
SLIST_FIRST(&eventtimers)->et_quality < et->et_quality) {
SLIST_INSERT_HEAD(&eventtimers, et, et_all);
} else {
SLIST_FOREACH(tmp, &eventtimers, et_all) {
next = SLIST_NEXT(tmp, et_all);
if (next == NULL || next->et_quality < et->et_quality) {
SLIST_INSERT_AFTER(tmp, et, et_all);
break;
}
}
}
ET_UNLOCK();
return (0);
}
/*
* Deregister event timer hardware.
*/
int
et_deregister(struct eventtimer *et)
{
int err = 0;
if (et->et_deregister_cb != NULL) {
if ((err = et->et_deregister_cb(et, et->et_arg)) != 0)
return (err);
}
ET_LOCK();
SLIST_REMOVE(&eventtimers, et, eventtimer, et_all);
ET_UNLOCK();
sysctl_remove_oid(et->et_sysctl, 1, 1);
return (0);
}
/*
* Find free event timer hardware with specified parameters.
*/
struct eventtimer *
et_find(const char *name, int check, int want)
{
struct eventtimer *et = NULL;
SLIST_FOREACH(et, &eventtimers, et_all) {
if (et->et_active)
continue;
if (name != NULL && strcasecmp(et->et_name, name) != 0)
continue;
if (name == NULL && et->et_quality < 0)
continue;
if ((et->et_flags & check) != want)
continue;
break;
}
return (et);
}
/*
* Initialize event timer hardware. Set callbacks.
*/
int
et_init(struct eventtimer *et, et_event_cb_t *event,
et_deregister_cb_t *deregister, void *arg)
{
if (event == NULL)
return (EINVAL);
if (et->et_active)
return (EBUSY);
et->et_active = 1;
et->et_event_cb = event;
et->et_deregister_cb = deregister;
et->et_arg = arg;
return (0);
}
/*
* Start event timer hardware.
* first - delay before first tick.
* period - period of subsequent periodic ticks.
*/
int
et_start(struct eventtimer *et,
struct bintime *first, struct bintime *period)
{
if (!et->et_active)
return (ENXIO);
if (first == NULL && period == NULL)
return (EINVAL);
if ((et->et_flags & ET_FLAGS_PERIODIC) == 0 &&
period != NULL)
return (ENODEV);
if ((et->et_flags & ET_FLAGS_ONESHOT) == 0 &&
period == NULL)
return (ENODEV);
if (first != NULL) {
if (first->sec < et->et_min_period.sec ||
(first->sec == et->et_min_period.sec &&
first->frac < et->et_min_period.frac))
first = &et->et_min_period;
if (first->sec > et->et_max_period.sec ||
(first->sec == et->et_max_period.sec &&
first->frac > et->et_max_period.frac))
first = &et->et_max_period;
}
if (period != NULL) {
if (period->sec < et->et_min_period.sec ||
(period->sec == et->et_min_period.sec &&
period->frac < et->et_min_period.frac))
period = &et->et_min_period;
if (period->sec > et->et_max_period.sec ||
(period->sec == et->et_max_period.sec &&
period->frac > et->et_max_period.frac))
period = &et->et_max_period;
}
if (et->et_start)
return (et->et_start(et, first, period));
return (0);
}
/* Stop event timer hardware. */
int
et_stop(struct eventtimer *et)
{
if (!et->et_active)
return (ENXIO);
if (et->et_stop)
return (et->et_stop(et));
return (0);
}
/* Mark event timer hardware as broken. */
int
et_ban(struct eventtimer *et)
{
et->et_flags &= ~(ET_FLAGS_PERIODIC | ET_FLAGS_ONESHOT);
return (0);
}
/* Free event timer hardware. */
int
et_free(struct eventtimer *et)
{
if (!et->et_active)
return (ENXIO);
et->et_active = 0;
return (0);
}
/* Report list of supported event timers hardware via sysctl. */
static int
sysctl_kern_eventtimer_choice(SYSCTL_HANDLER_ARGS)
{
char buf[512], *spc;
struct eventtimer *et;
int error, off;
spc = "";
error = 0;
buf[0] = 0;
off = 0;
ET_LOCK();
SLIST_FOREACH(et, &eventtimers, et_all) {
off += snprintf(buf + off, sizeof(buf) - off, "%s%s(%d)",
spc, et->et_name, et->et_quality);
spc = " ";
}
ET_UNLOCK();
error = SYSCTL_OUT(req, buf, strlen(buf));
return (error);
}
SYSCTL_PROC(_kern_eventtimer, OID_AUTO, choice,
CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE,
0, 0, sysctl_kern_eventtimer_choice, "A", "Present event timers");