freebsd-dev/sys/kern/kern_synch.c

933 lines
25 KiB
C
Raw Normal View History

1994-05-24 10:09:53 +00:00
/*-
* Copyright (c) 1982, 1986, 1990, 1991, 1993
* The Regents of the University of California. All rights reserved.
* (c) UNIX System Laboratories, Inc.
* All or some portions of this file are derived from material licensed
* to the University of California by American Telephone and Telegraph
* Co. or Unix System Laboratories, Inc. and are reproduced herein with
* the permission of UNIX System Laboratories, Inc.
*
* 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.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
*
* @(#)kern_synch.c 8.9 (Berkeley) 5/19/95
1999-08-28 01:08:13 +00:00
* $FreeBSD$
1994-05-24 10:09:53 +00:00
*/
#include "opt_ktrace.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/kernel.h>
#include <sys/signalvar.h>
#include <sys/resourcevar.h>
#include <sys/vmmeter.h>
#include <sys/sysctl.h>
#include <vm/vm.h>
#include <vm/vm_extern.h>
1994-05-24 10:09:53 +00:00
#ifdef KTRACE
#include <sys/uio.h>
1994-05-24 10:09:53 +00:00
#include <sys/ktrace.h>
#endif
#include <machine/cpu.h>
#ifdef SMP
#include <machine/smp.h>
#endif
1994-05-24 10:09:53 +00:00
static void sched_setup __P((void *dummy));
SYSINIT(sched_setup, SI_SUB_KICK_SCHEDULER, SI_ORDER_FIRST, sched_setup, NULL)
u_char curpriority;
int hogticks;
int lbolt;
int sched_quantum; /* Roundrobin scheduling quantum in ticks. */
1994-05-24 10:09:53 +00:00
1997-11-22 08:35:46 +00:00
static void endtsleep __P((void *));
1997-12-29 08:54:52 +00:00
static void roundrobin __P((void *arg));
static void schedcpu __P((void *arg));
1997-11-22 08:35:46 +00:00
static void updatepri __P((struct proc *p));
static int
sysctl_kern_quantum SYSCTL_HANDLER_ARGS
{
int error, new_val;
new_val = sched_quantum * tick;
error = sysctl_handle_int(oidp, &new_val, 0, req);
if (error != 0 || req->newptr == NULL)
return (error);
if (new_val < tick)
return (EINVAL);
sched_quantum = new_val / tick;
hogticks = 2 * sched_quantum;
return (0);
}
SYSCTL_PROC(_kern, OID_AUTO, quantum, CTLTYPE_INT|CTLFLAG_RW,
0, sizeof sched_quantum, sysctl_kern_quantum, "I", "");
/* maybe_resched: Decide if you need to reschedule or not
* taking the priorities and schedulers into account.
*/
static void maybe_resched(struct proc *chk)
{
struct proc *p = curproc; /* XXX */
/*
* Compare priorities if the new process is on the same scheduler,
* otherwise the one on the more realtimeish scheduler wins.
*
* XXX idle scheduler still broken because proccess stays on idle
* scheduler during waits (such as when getting FS locks). If a
* standard process becomes runaway cpu-bound, the system can lockup
* due to idle-scheduler processes in wakeup never getting any cpu.
*/
if (p == 0 ||
(chk->p_priority < curpriority && RTP_PRIO_BASE(p->p_rtprio.type) == RTP_PRIO_BASE(chk->p_rtprio.type)) ||
RTP_PRIO_BASE(chk->p_rtprio.type) < RTP_PRIO_BASE(p->p_rtprio.type)
) {
need_resched();
}
}
int
roundrobin_interval(void)
{
return (sched_quantum);
}
1994-05-24 10:09:53 +00:00
/*
* Force switch among equal priority processes every 100ms.
*/
/* ARGSUSED */
static void
1994-05-24 10:09:53 +00:00
roundrobin(arg)
void *arg;
{
#ifndef SMP
struct proc *p = curproc; /* XXX */
#endif
#ifdef SMP
need_resched();
forward_roundrobin();
#else
if (p == 0 || RTP_PRIO_NEED_RR(p->p_rtprio.type))
need_resched();
#endif
1994-05-24 10:09:53 +00:00
timeout(roundrobin, NULL, sched_quantum);
1994-05-24 10:09:53 +00:00
}
/*
* Constants for digital decay and forget:
* 90% of (p_estcpu) usage in 5 * loadav time
* 95% of (p_pctcpu) usage in 60 seconds (load insensitive)
* Note that, as ps(1) mentions, this can let percentages
* total over 100% (I've seen 137.9% for 3 processes).
*
* Note that schedclock() updates p_estcpu and p_cpticks asynchronously.
1994-05-24 10:09:53 +00:00
*
* We wish to decay away 90% of p_estcpu in (5 * loadavg) seconds.
* That is, the system wants to compute a value of decay such
* that the following for loop:
* for (i = 0; i < (5 * loadavg); i++)
* p_estcpu *= decay;
* will compute
* p_estcpu *= 0.1;
* for all values of loadavg:
*
* Mathematically this loop can be expressed by saying:
* decay ** (5 * loadavg) ~= .1
*
* The system computes decay as:
* decay = (2 * loadavg) / (2 * loadavg + 1)
*
* We wish to prove that the system's computation of decay
* will always fulfill the equation:
* decay ** (5 * loadavg) ~= .1
*
* If we compute b as:
* b = 2 * loadavg
* then
* decay = b / (b + 1)
*
* We now need to prove two things:
* 1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1)
* 2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg)
1995-05-30 08:16:23 +00:00
*
1994-05-24 10:09:53 +00:00
* Facts:
* For x close to zero, exp(x) =~ 1 + x, since
* exp(x) = 0! + x**1/1! + x**2/2! + ... .
* therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b.
* For x close to zero, ln(1+x) =~ x, since
* ln(1+x) = x - x**2/2 + x**3/3 - ... -1 < x < 1
* therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1).
* ln(.1) =~ -2.30
*
* Proof of (1):
* Solve (factor)**(power) =~ .1 given power (5*loadav):
* solving for factor,
* ln(factor) =~ (-2.30/5*loadav), or
* factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) =
* exp(-1/b) =~ (b-1)/b =~ b/(b+1). QED
*
* Proof of (2):
* Solve (factor)**(power) =~ .1 given factor == (b/(b+1)):
* solving for power,
* power*ln(b/(b+1)) =~ -2.30, or
* power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav. QED
*
* Actual power values for the implemented algorithm are as follows:
* loadav: 1 2 3 4
* power: 5.68 10.32 14.94 19.55
*/
/* calculations for digital decay to forget 90% of usage in 5*loadav sec */
#define loadfactor(loadav) (2 * (loadav))
#define decay_cpu(loadfac, cpu) (((loadfac) * (cpu)) / ((loadfac) + FSCALE))
/* decay 95% of `p_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */
1997-11-22 08:35:46 +00:00
static fixpt_t ccpu = 0.95122942450071400909 * FSCALE; /* exp(-1/20) */
SYSCTL_INT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0, "");
1994-05-24 10:09:53 +00:00
/* kernel uses `FSCALE', userland (SHOULD) use kern.fscale */
static int fscale __unused = FSCALE;
SYSCTL_INT(_kern, OID_AUTO, fscale, CTLFLAG_RD, 0, FSCALE, "");
1994-05-24 10:09:53 +00:00
/*
* If `ccpu' is not equal to `exp(-1/20)' and you still want to use the
* faster/more-accurate formula, you'll have to estimate CCPU_SHIFT below
* and possibly adjust FSHIFT in "param.h" so that (FSHIFT >= CCPU_SHIFT).
*
* To estimate CCPU_SHIFT for exp(-1/20), the following formula was used:
* 1 - exp(-1/20) ~= 0.0487 ~= 0.0488 == 1 (fixed pt, *11* bits).
*
* If you don't want to bother with the faster/more-accurate formula, you
1994-05-24 10:09:53 +00:00
* can set CCPU_SHIFT to (FSHIFT + 1) which will use a slower/less-accurate
* (more general) method of calculating the %age of CPU used by a process.
*/
#define CCPU_SHIFT 11
/*
* Recompute process priorities, every hz ticks.
*/
/* ARGSUSED */
static void
1994-05-24 10:09:53 +00:00
schedcpu(arg)
void *arg;
{
register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
register struct proc *p;
register int realstathz, s;
1994-05-24 10:09:53 +00:00
realstathz = stathz ? stathz : hz;
LIST_FOREACH(p, &allproc, p_list) {
1994-05-24 10:09:53 +00:00
/*
* Increment time in/out of memory and sleep time
* (if sleeping). We ignore overflow; with 16-bit int's
* (remember them?) overflow takes 45 days.
*/
p->p_swtime++;
if (p->p_stat == SSLEEP || p->p_stat == SSTOP)
p->p_slptime++;
p->p_pctcpu = (p->p_pctcpu * ccpu) >> FSHIFT;
/*
* If the process has slept the entire second,
* stop recalculating its priority until it wakes up.
*/
if (p->p_slptime > 1)
continue;
s = splhigh(); /* prevent state changes and protect run queue */
1994-05-24 10:09:53 +00:00
/*
* p_pctcpu is only for ps.
*/
#if (FSHIFT >= CCPU_SHIFT)
p->p_pctcpu += (realstathz == 100)?
1994-05-24 10:09:53 +00:00
((fixpt_t) p->p_cpticks) << (FSHIFT - CCPU_SHIFT):
100 * (((fixpt_t) p->p_cpticks)
<< (FSHIFT - CCPU_SHIFT)) / realstathz;
1994-05-24 10:09:53 +00:00
#else
p->p_pctcpu += ((FSCALE - ccpu) *
(p->p_cpticks * FSCALE / realstathz)) >> FSHIFT;
1994-05-24 10:09:53 +00:00
#endif
p->p_cpticks = 0;
Scheduler fixes equivalent to the ones logged in the following NetBSD commit to kern_synch.c: ---------------------------- revision 1.55 date: 1999/02/23 02:56:03; author: ross; state: Exp; lines: +39 -10 Scheduler bug fixes and reorganization * fix the ancient nice(1) bug, where nice +20 processes incorrectly steal 10 - 20% of the CPU, (or even more depending on load average) * provide a new schedclk() mechanism at a new clock at schedhz, so high platform hz values don't cause nice +0 processes to look like they are niced * change the algorithm slightly, and reorganize the code a lot * fix percent-CPU calculation bugs, and eliminate some no-op code === nice bug === Correctly divide the scheduler queues between niced and compute-bound processes. The current nice weight of two (sort of, see `algorithm change' below) neatly divides the USRPRI queues in half; this should have been used to clip p_estcpu, instead of UCHAR_MAX. Besides being the wrong amount, clipping an unsigned char to UCHAR_MAX is a no-op, and it was done after decay_cpu() which can only _reduce_ the value. It has to be kept <= NICE_WEIGHT * PRIO_MAX - PPQ or processes can scheduler-penalize themselves onto the same queue as nice +20 processes. (Or even a higher one.) === New schedclk() mechansism === Some platforms should be cutting down stathz before hitting the scheduler, since the scheduler algorithm only works right in the vicinity of 64 Hz. Rather than prescale hz, then scale back and forth by 4 every time p_estcpu is touched (each occurance an abstraction violation), use p_estcpu without scaling and require schedhz to be generated directly at the right frequency. Use a default stathz (well, actually, profhz) / 4, so nothing changes unless a platform defines schedhz and a new clock. Define these for alpha, where hz==1024, and nice was totally broke. === Algorithm change === The nice value used to be added to the exponentially-decayed scheduler history value p_estcpu, in _addition_ to be incorporated directly (with greater wieght) into the priority calculation. At first glance, it appears to be a pointless increase of 1/8 the nice effect (pri = p_estcpu/4 + nice*2), but it's actually at least 3x that because it will ramp up linearly but be decayed only exponentially, thus converging to an additional .75 nice for a loadaverage of one. I killed this, it makes the behavior hard to control, almost impossible to analyze, and the effect (~~nothing at for the first second, then somewhat increased niceness after three seconds or more, depending on load average) pointless. === Other bugs === hz -> profhz in the p_pctcpu = f(p_cpticks) calcuation. Collect scheduler functionality. Try to put each abstraction in just one place. ---------------------------- The details are a little different in FreeBSD: === nice bug === Fixing this is the main point of this commit. We use essentially the same clipping rule as NetBSD (our limit on p_estcpu differs by a scale factor). However, clipping at all is fundamentally bad. It gives free CPU the hoggiest hogs once they reach the limit, and reaching the limit is normal for long-running hogs. This will be fixed later. === New schedclk() mechanism === We don't use the NetBSD schedclk() (now schedclock()) mechanism. We require (real)stathz to be about 128 and scale by an extra factor of 2 compared with NetBSD's statclock(). We scale p_estcpu instead of scaling the clock. This is more accurate and flexible. === Algorithm change === Same change. === Other bugs === The p_pctcpu bug was fixed long ago. We don't try as hard to abstract functionality yet. Related changes: the new limit on p_estcpu must be exported to kern_exit.c for clipping in wait1(). Agreed with by: dufault
1999-11-28 12:12:13 +00:00
p->p_estcpu = decay_cpu(loadfac, p->p_estcpu);
1994-05-24 10:09:53 +00:00
resetpriority(p);
if (p->p_priority >= PUSER) {
if ((p != curproc) &&
#ifdef SMP
p->p_oncpu == 0xff && /* idle */
#endif
1994-05-24 10:09:53 +00:00
p->p_stat == SRUN &&
(p->p_flag & P_INMEM) &&
(p->p_priority / PPQ) != (p->p_usrpri / PPQ)) {
remrunqueue(p);
1994-05-24 10:09:53 +00:00
p->p_priority = p->p_usrpri;
setrunqueue(p);
} else
p->p_priority = p->p_usrpri;
}
splx(s);
}
vmmeter();
wakeup((caddr_t)&lbolt);
1994-05-24 10:09:53 +00:00
timeout(schedcpu, (void *)0, hz);
}
/*
* Recalculate the priority of a process after it has slept for a while.
* For all load averages >= 1 and max p_estcpu of 255, sleeping for at
* least six times the loadfactor will decay p_estcpu to zero.
*/
1997-11-22 08:35:46 +00:00
static void
1994-05-24 10:09:53 +00:00
updatepri(p)
register struct proc *p;
{
register unsigned int newcpu = p->p_estcpu;
register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
if (p->p_slptime > 5 * loadfac)
p->p_estcpu = 0;
else {
p->p_slptime--; /* the first time was done in schedcpu */
while (newcpu && --p->p_slptime)
Scheduler fixes equivalent to the ones logged in the following NetBSD commit to kern_synch.c: ---------------------------- revision 1.55 date: 1999/02/23 02:56:03; author: ross; state: Exp; lines: +39 -10 Scheduler bug fixes and reorganization * fix the ancient nice(1) bug, where nice +20 processes incorrectly steal 10 - 20% of the CPU, (or even more depending on load average) * provide a new schedclk() mechanism at a new clock at schedhz, so high platform hz values don't cause nice +0 processes to look like they are niced * change the algorithm slightly, and reorganize the code a lot * fix percent-CPU calculation bugs, and eliminate some no-op code === nice bug === Correctly divide the scheduler queues between niced and compute-bound processes. The current nice weight of two (sort of, see `algorithm change' below) neatly divides the USRPRI queues in half; this should have been used to clip p_estcpu, instead of UCHAR_MAX. Besides being the wrong amount, clipping an unsigned char to UCHAR_MAX is a no-op, and it was done after decay_cpu() which can only _reduce_ the value. It has to be kept <= NICE_WEIGHT * PRIO_MAX - PPQ or processes can scheduler-penalize themselves onto the same queue as nice +20 processes. (Or even a higher one.) === New schedclk() mechansism === Some platforms should be cutting down stathz before hitting the scheduler, since the scheduler algorithm only works right in the vicinity of 64 Hz. Rather than prescale hz, then scale back and forth by 4 every time p_estcpu is touched (each occurance an abstraction violation), use p_estcpu without scaling and require schedhz to be generated directly at the right frequency. Use a default stathz (well, actually, profhz) / 4, so nothing changes unless a platform defines schedhz and a new clock. Define these for alpha, where hz==1024, and nice was totally broke. === Algorithm change === The nice value used to be added to the exponentially-decayed scheduler history value p_estcpu, in _addition_ to be incorporated directly (with greater wieght) into the priority calculation. At first glance, it appears to be a pointless increase of 1/8 the nice effect (pri = p_estcpu/4 + nice*2), but it's actually at least 3x that because it will ramp up linearly but be decayed only exponentially, thus converging to an additional .75 nice for a loadaverage of one. I killed this, it makes the behavior hard to control, almost impossible to analyze, and the effect (~~nothing at for the first second, then somewhat increased niceness after three seconds or more, depending on load average) pointless. === Other bugs === hz -> profhz in the p_pctcpu = f(p_cpticks) calcuation. Collect scheduler functionality. Try to put each abstraction in just one place. ---------------------------- The details are a little different in FreeBSD: === nice bug === Fixing this is the main point of this commit. We use essentially the same clipping rule as NetBSD (our limit on p_estcpu differs by a scale factor). However, clipping at all is fundamentally bad. It gives free CPU the hoggiest hogs once they reach the limit, and reaching the limit is normal for long-running hogs. This will be fixed later. === New schedclk() mechanism === We don't use the NetBSD schedclk() (now schedclock()) mechanism. We require (real)stathz to be about 128 and scale by an extra factor of 2 compared with NetBSD's statclock(). We scale p_estcpu instead of scaling the clock. This is more accurate and flexible. === Algorithm change === Same change. === Other bugs === The p_pctcpu bug was fixed long ago. We don't try as hard to abstract functionality yet. Related changes: the new limit on p_estcpu must be exported to kern_exit.c for clipping in wait1(). Agreed with by: dufault
1999-11-28 12:12:13 +00:00
newcpu = decay_cpu(loadfac, newcpu);
p->p_estcpu = newcpu;
1994-05-24 10:09:53 +00:00
}
resetpriority(p);
}
/*
* We're only looking at 7 bits of the address; everything is
* aligned to 4, lots of things are aligned to greater powers
* of 2. Shift right by 8, i.e. drop the bottom 256 worth.
*/
#define TABLESIZE 128
1997-11-22 08:35:46 +00:00
static TAILQ_HEAD(slpquehead, proc) slpque[TABLESIZE];
#define LOOKUP(x) (((intptr_t)(x) >> 8) & (TABLESIZE - 1))
1994-05-24 10:09:53 +00:00
/*
* During autoconfiguration or after a panic, a sleep will simply
* lower the priority briefly to allow interrupts, then return.
* The priority to be used (safepri) is machine-dependent, thus this
* value is initialized and maintained in the machine-dependent layers.
* This priority will typically be 0, or the lowest priority
* that is safe for use on the interrupt stack; it can be made
* higher to block network software interrupts after panics.
*/
int safepri;
void
sleepinit(void)
{
int i;
sched_quantum = hz/10;
hogticks = 2 * sched_quantum;
for (i = 0; i < TABLESIZE; i++)
TAILQ_INIT(&slpque[i]);
}
1994-05-24 10:09:53 +00:00
/*
* General sleep call. Suspends the current process until a wakeup is
* performed on the specified identifier. The process will then be made
* runnable with the specified priority. Sleeps at most timo/hz seconds
* (0 means no timeout). If pri includes PCATCH flag, signals are checked
* before and after sleeping, else signals are not checked. Returns 0 if
* awakened, EWOULDBLOCK if the timeout expires. If PCATCH is set and a
* signal needs to be delivered, ERESTART is returned if the current system
* call should be restarted if possible, and EINTR is returned if the system
* call should be interrupted by the signal (return EINTR).
*/
int
tsleep(ident, priority, wmesg, timo)
void *ident;
int priority, timo;
1997-11-21 11:37:03 +00:00
const char *wmesg;
1994-05-24 10:09:53 +00:00
{
struct proc *p = curproc;
int s, sig, catch = priority & PCATCH;
init_main.c subr_autoconf.c: Add support for "interrupt driven configuration hooks". A component of the kernel can register a hook, most likely during auto-configuration, and receive a callback once interrupt services are available. This callback will occur before the root and dump devices are configured, so the configuration task can affect the selection of those two devices or complete any tasks that need to be performed prior to launching init. System boot is posponed so long as a hook is registered. The hook owner is responsible for removing the hook once their task is complete or the system boot can continue. kern_acct.c kern_clock.c kern_exit.c kern_synch.c kern_time.c: Change the interface and implementation for the kernel callout service. The new implemntaion is based on the work of Adam M. Costello and George Varghese, published in a technical report entitled "Redesigning the BSD Callout and Timer Facilities". The interface used in FreeBSD is a little different than the one outlined in the paper. The new function prototypes are: struct callout_handle timeout(void (*func)(void *), void *arg, int ticks); void untimeout(void (*func)(void *), void *arg, struct callout_handle handle); If a client wishes to remove a timeout, it must store the callout_handle returned by timeout and pass it to untimeout. The new implementation gives 0(1) insert and removal of callouts making this interface scale well even for applications that keep 100s of callouts outstanding. See the updated timeout.9 man page for more details.
1997-09-21 22:00:25 +00:00
struct callout_handle thandle;
1994-05-24 10:09:53 +00:00
#ifdef KTRACE
if (p && KTRPOINT(p, KTR_CSW))
1994-05-24 10:09:53 +00:00
ktrcsw(p->p_tracep, 1, 0);
#endif
s = splhigh();
if (cold || panicstr) {
/*
* After a panic, or during autoconfiguration,
* just give interrupts a chance, then just return;
* don't run any other procs or panic below,
* in case this is the idle process and already asleep.
*/
splx(safepri);
splx(s);
return (0);
}
KASSERT(p != NULL, ("tsleep1"));
1999-01-10 01:58:29 +00:00
KASSERT(ident != NULL && p->p_stat == SRUN, ("tsleep"));
/*
* Process may be sitting on a slpque if asleep() was called, remove
* it before re-adding.
*/
if (p->p_wchan != NULL)
unsleep(p);
1994-05-24 10:09:53 +00:00
p->p_wchan = ident;
p->p_wmesg = wmesg;
p->p_slptime = 0;
p->p_priority = priority & PRIMASK;
TAILQ_INSERT_TAIL(&slpque[LOOKUP(ident)], p, p_procq);
1994-05-24 10:09:53 +00:00
if (timo)
init_main.c subr_autoconf.c: Add support for "interrupt driven configuration hooks". A component of the kernel can register a hook, most likely during auto-configuration, and receive a callback once interrupt services are available. This callback will occur before the root and dump devices are configured, so the configuration task can affect the selection of those two devices or complete any tasks that need to be performed prior to launching init. System boot is posponed so long as a hook is registered. The hook owner is responsible for removing the hook once their task is complete or the system boot can continue. kern_acct.c kern_clock.c kern_exit.c kern_synch.c kern_time.c: Change the interface and implementation for the kernel callout service. The new implemntaion is based on the work of Adam M. Costello and George Varghese, published in a technical report entitled "Redesigning the BSD Callout and Timer Facilities". The interface used in FreeBSD is a little different than the one outlined in the paper. The new function prototypes are: struct callout_handle timeout(void (*func)(void *), void *arg, int ticks); void untimeout(void (*func)(void *), void *arg, struct callout_handle handle); If a client wishes to remove a timeout, it must store the callout_handle returned by timeout and pass it to untimeout. The new implementation gives 0(1) insert and removal of callouts making this interface scale well even for applications that keep 100s of callouts outstanding. See the updated timeout.9 man page for more details.
1997-09-21 22:00:25 +00:00
thandle = timeout(endtsleep, (void *)p, timo);
1994-05-24 10:09:53 +00:00
/*
* We put ourselves on the sleep queue and start our timeout
* before calling CURSIG, as we could stop there, and a wakeup
* or a SIGCONT (or both) could occur while we were stopped.
* A SIGCONT would cause us to be marked as SSLEEP
* without resuming us, thus we must be ready for sleep
* when CURSIG is called. If the wakeup happens while we're
* stopped, p->p_wchan will be 0 upon return from CURSIG.
*/
if (catch) {
p->p_flag |= P_SINTR;
if ((sig = CURSIG(p))) {
1994-05-24 10:09:53 +00:00
if (p->p_wchan)
unsleep(p);
p->p_stat = SRUN;
goto resume;
}
if (p->p_wchan == 0) {
catch = 0;
goto resume;
}
} else
sig = 0;
p->p_stat = SSLEEP;
p->p_stats->p_ru.ru_nvcsw++;
mi_switch();
resume:
curpriority = p->p_usrpri;
splx(s);
p->p_flag &= ~P_SINTR;
if (p->p_flag & P_TIMEOUT) {
p->p_flag &= ~P_TIMEOUT;
if (sig == 0) {
#ifdef KTRACE
if (KTRPOINT(p, KTR_CSW))
ktrcsw(p->p_tracep, 0, 0);
#endif
return (EWOULDBLOCK);
}
} else if (timo)
init_main.c subr_autoconf.c: Add support for "interrupt driven configuration hooks". A component of the kernel can register a hook, most likely during auto-configuration, and receive a callback once interrupt services are available. This callback will occur before the root and dump devices are configured, so the configuration task can affect the selection of those two devices or complete any tasks that need to be performed prior to launching init. System boot is posponed so long as a hook is registered. The hook owner is responsible for removing the hook once their task is complete or the system boot can continue. kern_acct.c kern_clock.c kern_exit.c kern_synch.c kern_time.c: Change the interface and implementation for the kernel callout service. The new implemntaion is based on the work of Adam M. Costello and George Varghese, published in a technical report entitled "Redesigning the BSD Callout and Timer Facilities". The interface used in FreeBSD is a little different than the one outlined in the paper. The new function prototypes are: struct callout_handle timeout(void (*func)(void *), void *arg, int ticks); void untimeout(void (*func)(void *), void *arg, struct callout_handle handle); If a client wishes to remove a timeout, it must store the callout_handle returned by timeout and pass it to untimeout. The new implementation gives 0(1) insert and removal of callouts making this interface scale well even for applications that keep 100s of callouts outstanding. See the updated timeout.9 man page for more details.
1997-09-21 22:00:25 +00:00
untimeout(endtsleep, (void *)p, thandle);
1994-05-24 10:09:53 +00:00
if (catch && (sig != 0 || (sig = CURSIG(p)))) {
#ifdef KTRACE
if (KTRPOINT(p, KTR_CSW))
ktrcsw(p->p_tracep, 0, 0);
#endif
if (SIGISMEMBER(p->p_sigacts->ps_sigintr, sig))
1994-05-24 10:09:53 +00:00
return (EINTR);
return (ERESTART);
}
#ifdef KTRACE
if (KTRPOINT(p, KTR_CSW))
ktrcsw(p->p_tracep, 0, 0);
#endif
return (0);
}
/*
* asleep() - async sleep call. Place process on wait queue and return
* immediately without blocking. The process stays runnable until await()
* is called. If ident is NULL, remove process from wait queue if it is still
* on one.
*
* Only the most recent sleep condition is effective when making successive
* calls to asleep() or when calling tsleep().
*
* The timeout, if any, is not initiated until await() is called. The sleep
* priority, signal, and timeout is specified in the asleep() call but may be
* overriden in the await() call.
*
* <<<<<<<< EXPERIMENTAL, UNTESTED >>>>>>>>>>
*/
int
asleep(void *ident, int priority, const char *wmesg, int timo)
{
struct proc *p = curproc;
int s;
/*
* splhigh() while manipulating sleep structures and slpque.
*
* Remove preexisting wait condition (if any) and place process
* on appropriate slpque, but do not put process to sleep.
*/
s = splhigh();
if (p->p_wchan != NULL)
unsleep(p);
if (ident) {
p->p_wchan = ident;
p->p_wmesg = wmesg;
p->p_slptime = 0;
p->p_asleep.as_priority = priority;
p->p_asleep.as_timo = timo;
TAILQ_INSERT_TAIL(&slpque[LOOKUP(ident)], p, p_procq);
}
splx(s);
return(0);
}
/*
* await() - wait for async condition to occur. The process blocks until
* wakeup() is called on the most recent asleep() address. If wakeup is called
* priority to await(), await() winds up being a NOP.
*
* If await() is called more then once (without an intervening asleep() call),
* await() is still effectively a NOP but it calls mi_switch() to give other
* processes some cpu before returning. The process is left runnable.
*
* <<<<<<<< EXPERIMENTAL, UNTESTED >>>>>>>>>>
*/
int
await(int priority, int timo)
{
struct proc *p = curproc;
int s;
s = splhigh();
if (p->p_wchan != NULL) {
struct callout_handle thandle;
int sig;
int catch;
/*
* The call to await() can override defaults specified in
* the original asleep().
*/
if (priority < 0)
priority = p->p_asleep.as_priority;
if (timo < 0)
timo = p->p_asleep.as_timo;
/*
* Install timeout
*/
if (timo)
thandle = timeout(endtsleep, (void *)p, timo);
sig = 0;
catch = priority & PCATCH;
if (catch) {
p->p_flag |= P_SINTR;
if ((sig = CURSIG(p))) {
if (p->p_wchan)
unsleep(p);
p->p_stat = SRUN;
goto resume;
}
if (p->p_wchan == NULL) {
catch = 0;
goto resume;
}
}
p->p_stat = SSLEEP;
p->p_stats->p_ru.ru_nvcsw++;
mi_switch();
resume:
curpriority = p->p_usrpri;
splx(s);
p->p_flag &= ~P_SINTR;
if (p->p_flag & P_TIMEOUT) {
p->p_flag &= ~P_TIMEOUT;
if (sig == 0) {
#ifdef KTRACE
if (KTRPOINT(p, KTR_CSW))
ktrcsw(p->p_tracep, 0, 0);
#endif
return (EWOULDBLOCK);
}
} else if (timo)
untimeout(endtsleep, (void *)p, thandle);
if (catch && (sig != 0 || (sig = CURSIG(p)))) {
#ifdef KTRACE
if (KTRPOINT(p, KTR_CSW))
ktrcsw(p->p_tracep, 0, 0);
#endif
if (SIGISMEMBER(p->p_sigacts->ps_sigintr, sig))
return (EINTR);
return (ERESTART);
}
#ifdef KTRACE
if (KTRPOINT(p, KTR_CSW))
ktrcsw(p->p_tracep, 0, 0);
#endif
} else {
/*
* If as_priority is 0, await() has been called without an
* intervening asleep(). We are still effectively a NOP,
* but we call mi_switch() for safety.
*/
if (p->p_asleep.as_priority == 0) {
p->p_stats->p_ru.ru_nvcsw++;
mi_switch();
}
splx(s);
}
/*
* clear p_asleep.as_priority as an indication that await() has been
* called. If await() is called again without an intervening asleep(),
* await() is still effectively a NOP but the above mi_switch() code
* is triggered as a safety.
*/
p->p_asleep.as_priority = 0;
return (0);
}
/*
* Implement timeout for tsleep or asleep()/await()
*
1994-05-24 10:09:53 +00:00
* If process hasn't been awakened (wchan non-zero),
* set timeout flag and undo the sleep. If proc
* is stopped, just unsleep so it will remain stopped.
*/
1997-11-22 08:35:46 +00:00
static void
1994-05-24 10:09:53 +00:00
endtsleep(arg)
void *arg;
{
register struct proc *p;
int s;
p = (struct proc *)arg;
s = splhigh();
if (p->p_wchan) {
if (p->p_stat == SSLEEP)
setrunnable(p);
else
unsleep(p);
p->p_flag |= P_TIMEOUT;
}
splx(s);
}
/*
* Remove a process from its wait queue
*/
void
unsleep(p)
register struct proc *p;
{
int s;
s = splhigh();
if (p->p_wchan) {
TAILQ_REMOVE(&slpque[LOOKUP(p->p_wchan)], p, p_procq);
1994-05-24 10:09:53 +00:00
p->p_wchan = 0;
}
splx(s);
}
/*
* Make all processes sleeping on the specified identifier runnable.
*/
void
wakeup(ident)
register void *ident;
{
register struct slpquehead *qp;
register struct proc *p;
1994-05-24 10:09:53 +00:00
int s;
s = splhigh();
qp = &slpque[LOOKUP(ident)];
restart:
TAILQ_FOREACH(p, qp, p_procq) {
1994-05-24 10:09:53 +00:00
if (p->p_wchan == ident) {
TAILQ_REMOVE(qp, p, p_procq);
1994-05-24 10:09:53 +00:00
p->p_wchan = 0;
if (p->p_stat == SSLEEP) {
/* OPTIMIZED EXPANSION OF setrunnable(p); */
if (p->p_slptime > 1)
updatepri(p);
p->p_slptime = 0;
p->p_stat = SRUN;
if (p->p_flag & P_INMEM) {
1994-05-24 10:09:53 +00:00
setrunqueue(p);
maybe_resched(p);
} else {
p->p_flag |= P_SWAPINREQ;
wakeup((caddr_t)&proc0);
}
/* END INLINE EXPANSION */
goto restart;
}
}
}
splx(s);
}
/*
* Make a process sleeping on the specified identifier runnable.
* May wake more than one process if a target prcoess is currently
* swapped out.
*/
void
wakeup_one(ident)
register void *ident;
{
register struct slpquehead *qp;
register struct proc *p;
int s;
s = splhigh();
qp = &slpque[LOOKUP(ident)];
TAILQ_FOREACH(p, qp, p_procq) {
if (p->p_wchan == ident) {
TAILQ_REMOVE(qp, p, p_procq);
p->p_wchan = 0;
if (p->p_stat == SSLEEP) {
/* OPTIMIZED EXPANSION OF setrunnable(p); */
if (p->p_slptime > 1)
updatepri(p);
p->p_slptime = 0;
p->p_stat = SRUN;
if (p->p_flag & P_INMEM) {
setrunqueue(p);
maybe_resched(p);
break;
} else {
p->p_flag |= P_SWAPINREQ;
wakeup((caddr_t)&proc0);
}
1994-05-24 10:09:53 +00:00
/* END INLINE EXPANSION */
}
}
1994-05-24 10:09:53 +00:00
}
splx(s);
}
/*
* The machine independent parts of mi_switch().
* Must be called at splstatclock() or higher.
*/
void
mi_switch()
{
struct timeval new_switchtime;
1994-05-24 10:09:53 +00:00
register struct proc *p = curproc; /* XXX */
register struct rlimit *rlim;
int x;
1994-05-24 10:09:53 +00:00
/*
* XXX this spl is almost unnecessary. It is partly to allow for
* sloppy callers that don't do it (issignal() via CURSIG() is the
* main offender). It is partly to work around a bug in the i386
* cpu_switch() (the ipl is not preserved). We ran for years
* without it. I think there was only a interrupt latency problem.
* The main caller, tsleep(), does an splx() a couple of instructions
* after calling here. The buggy caller, issignal(), usually calls
* here at spl0() and sometimes returns at splhigh(). The process
* then runs for a little too long at splhigh(). The ipl gets fixed
* when the process returns to user mode (or earlier).
*
* It would probably be better to always call here at spl0(). Callers
* are prepared to give up control to another process, so they must
* be prepared to be interrupted. The clock stuff here may not
* actually need splstatclock().
*/
x = splstatclock();
#ifdef SIMPLELOCK_DEBUG
if (p->p_simple_locks)
printf("sleep: holding simple lock\n");
#endif
1994-05-24 10:09:53 +00:00
/*
* Compute the amount of time during which the current
* process was running, and add that to its total so far.
*/
microuptime(&new_switchtime);
if (timevalcmp(&new_switchtime, &switchtime, <)) {
printf("microuptime() went backwards (%ld.%06ld -> %ld,%06ld)\n",
switchtime.tv_sec, switchtime.tv_usec,
new_switchtime.tv_sec, new_switchtime.tv_usec);
new_switchtime = switchtime;
} else {
p->p_runtime += (new_switchtime.tv_usec - switchtime.tv_usec) +
(new_switchtime.tv_sec - switchtime.tv_sec) * (int64_t)1000000;
}
1994-05-24 10:09:53 +00:00
/*
* Check if the process exceeds its cpu resource allocation.
* If over max, kill it.
1994-05-24 10:09:53 +00:00
*/
if (p->p_stat != SZOMB && p->p_limit->p_cpulimit != RLIM_INFINITY &&
p->p_runtime > p->p_limit->p_cpulimit) {
rlim = &p->p_rlimit[RLIMIT_CPU];
if (p->p_runtime / (rlim_t)1000000 >= rlim->rlim_max) {
killproc(p, "exceeded maximum CPU limit");
} else {
psignal(p, SIGXCPU);
if (rlim->rlim_cur < rlim->rlim_max) {
/* XXX: we should make a private copy */
rlim->rlim_cur += 5;
}
}
1994-05-24 10:09:53 +00:00
}
/*
* Pick a new current process and record its start time.
*/
cnt.v_swtch++;
switchtime = new_switchtime;
1994-05-24 10:09:53 +00:00
cpu_switch(p);
if (switchtime.tv_sec == 0)
microuptime(&switchtime);
switchticks = ticks;
splx(x);
1994-05-24 10:09:53 +00:00
}
/*
* Change process state to be runnable,
* placing it on the run queue if it is in memory,
* and awakening the swapper if it isn't in memory.
*/
void
setrunnable(p)
register struct proc *p;
{
register int s;
s = splhigh();
switch (p->p_stat) {
case 0:
case SRUN:
case SZOMB:
default:
panic("setrunnable");
case SSTOP:
case SSLEEP:
unsleep(p); /* e.g. when sending signals */
break;
case SIDL:
break;
}
p->p_stat = SRUN;
if (p->p_flag & P_INMEM)
setrunqueue(p);
splx(s);
if (p->p_slptime > 1)
updatepri(p);
p->p_slptime = 0;
if ((p->p_flag & P_INMEM) == 0) {
p->p_flag |= P_SWAPINREQ;
1994-05-24 10:09:53 +00:00
wakeup((caddr_t)&proc0);
}
else
maybe_resched(p);
1994-05-24 10:09:53 +00:00
}
/*
* Compute the priority of a process when running in user mode.
* Arrange to reschedule if the resulting priority is better
* than that of the current process.
*/
void
resetpriority(p)
register struct proc *p;
{
register unsigned int newpriority;
if (p->p_rtprio.type == RTP_PRIO_NORMAL) {
Scheduler fixes equivalent to the ones logged in the following NetBSD commit to kern_synch.c: ---------------------------- revision 1.55 date: 1999/02/23 02:56:03; author: ross; state: Exp; lines: +39 -10 Scheduler bug fixes and reorganization * fix the ancient nice(1) bug, where nice +20 processes incorrectly steal 10 - 20% of the CPU, (or even more depending on load average) * provide a new schedclk() mechanism at a new clock at schedhz, so high platform hz values don't cause nice +0 processes to look like they are niced * change the algorithm slightly, and reorganize the code a lot * fix percent-CPU calculation bugs, and eliminate some no-op code === nice bug === Correctly divide the scheduler queues between niced and compute-bound processes. The current nice weight of two (sort of, see `algorithm change' below) neatly divides the USRPRI queues in half; this should have been used to clip p_estcpu, instead of UCHAR_MAX. Besides being the wrong amount, clipping an unsigned char to UCHAR_MAX is a no-op, and it was done after decay_cpu() which can only _reduce_ the value. It has to be kept <= NICE_WEIGHT * PRIO_MAX - PPQ or processes can scheduler-penalize themselves onto the same queue as nice +20 processes. (Or even a higher one.) === New schedclk() mechansism === Some platforms should be cutting down stathz before hitting the scheduler, since the scheduler algorithm only works right in the vicinity of 64 Hz. Rather than prescale hz, then scale back and forth by 4 every time p_estcpu is touched (each occurance an abstraction violation), use p_estcpu without scaling and require schedhz to be generated directly at the right frequency. Use a default stathz (well, actually, profhz) / 4, so nothing changes unless a platform defines schedhz and a new clock. Define these for alpha, where hz==1024, and nice was totally broke. === Algorithm change === The nice value used to be added to the exponentially-decayed scheduler history value p_estcpu, in _addition_ to be incorporated directly (with greater wieght) into the priority calculation. At first glance, it appears to be a pointless increase of 1/8 the nice effect (pri = p_estcpu/4 + nice*2), but it's actually at least 3x that because it will ramp up linearly but be decayed only exponentially, thus converging to an additional .75 nice for a loadaverage of one. I killed this, it makes the behavior hard to control, almost impossible to analyze, and the effect (~~nothing at for the first second, then somewhat increased niceness after three seconds or more, depending on load average) pointless. === Other bugs === hz -> profhz in the p_pctcpu = f(p_cpticks) calcuation. Collect scheduler functionality. Try to put each abstraction in just one place. ---------------------------- The details are a little different in FreeBSD: === nice bug === Fixing this is the main point of this commit. We use essentially the same clipping rule as NetBSD (our limit on p_estcpu differs by a scale factor). However, clipping at all is fundamentally bad. It gives free CPU the hoggiest hogs once they reach the limit, and reaching the limit is normal for long-running hogs. This will be fixed later. === New schedclk() mechanism === We don't use the NetBSD schedclk() (now schedclock()) mechanism. We require (real)stathz to be about 128 and scale by an extra factor of 2 compared with NetBSD's statclock(). We scale p_estcpu instead of scaling the clock. This is more accurate and flexible. === Algorithm change === Same change. === Other bugs === The p_pctcpu bug was fixed long ago. We don't try as hard to abstract functionality yet. Related changes: the new limit on p_estcpu must be exported to kern_exit.c for clipping in wait1(). Agreed with by: dufault
1999-11-28 12:12:13 +00:00
newpriority = PUSER + p->p_estcpu / INVERSE_ESTCPU_WEIGHT +
NICE_WEIGHT * p->p_nice;
newpriority = min(newpriority, MAXPRI);
p->p_usrpri = newpriority;
}
maybe_resched(p);
1994-05-24 10:09:53 +00:00
}
/* ARGSUSED */
static void
sched_setup(dummy)
void *dummy;
{
/* Kick off timeout driven events by calling first time. */
roundrobin(NULL);
schedcpu(NULL);
}
/*
* We adjust the priority of the current process. The priority of
* a process gets worse as it accumulates CPU time. The cpu usage
* estimator (p_estcpu) is increased here. resetpriority() will
Scheduler fixes equivalent to the ones logged in the following NetBSD commit to kern_synch.c: ---------------------------- revision 1.55 date: 1999/02/23 02:56:03; author: ross; state: Exp; lines: +39 -10 Scheduler bug fixes and reorganization * fix the ancient nice(1) bug, where nice +20 processes incorrectly steal 10 - 20% of the CPU, (or even more depending on load average) * provide a new schedclk() mechanism at a new clock at schedhz, so high platform hz values don't cause nice +0 processes to look like they are niced * change the algorithm slightly, and reorganize the code a lot * fix percent-CPU calculation bugs, and eliminate some no-op code === nice bug === Correctly divide the scheduler queues between niced and compute-bound processes. The current nice weight of two (sort of, see `algorithm change' below) neatly divides the USRPRI queues in half; this should have been used to clip p_estcpu, instead of UCHAR_MAX. Besides being the wrong amount, clipping an unsigned char to UCHAR_MAX is a no-op, and it was done after decay_cpu() which can only _reduce_ the value. It has to be kept <= NICE_WEIGHT * PRIO_MAX - PPQ or processes can scheduler-penalize themselves onto the same queue as nice +20 processes. (Or even a higher one.) === New schedclk() mechansism === Some platforms should be cutting down stathz before hitting the scheduler, since the scheduler algorithm only works right in the vicinity of 64 Hz. Rather than prescale hz, then scale back and forth by 4 every time p_estcpu is touched (each occurance an abstraction violation), use p_estcpu without scaling and require schedhz to be generated directly at the right frequency. Use a default stathz (well, actually, profhz) / 4, so nothing changes unless a platform defines schedhz and a new clock. Define these for alpha, where hz==1024, and nice was totally broke. === Algorithm change === The nice value used to be added to the exponentially-decayed scheduler history value p_estcpu, in _addition_ to be incorporated directly (with greater wieght) into the priority calculation. At first glance, it appears to be a pointless increase of 1/8 the nice effect (pri = p_estcpu/4 + nice*2), but it's actually at least 3x that because it will ramp up linearly but be decayed only exponentially, thus converging to an additional .75 nice for a loadaverage of one. I killed this, it makes the behavior hard to control, almost impossible to analyze, and the effect (~~nothing at for the first second, then somewhat increased niceness after three seconds or more, depending on load average) pointless. === Other bugs === hz -> profhz in the p_pctcpu = f(p_cpticks) calcuation. Collect scheduler functionality. Try to put each abstraction in just one place. ---------------------------- The details are a little different in FreeBSD: === nice bug === Fixing this is the main point of this commit. We use essentially the same clipping rule as NetBSD (our limit on p_estcpu differs by a scale factor). However, clipping at all is fundamentally bad. It gives free CPU the hoggiest hogs once they reach the limit, and reaching the limit is normal for long-running hogs. This will be fixed later. === New schedclk() mechanism === We don't use the NetBSD schedclk() (now schedclock()) mechanism. We require (real)stathz to be about 128 and scale by an extra factor of 2 compared with NetBSD's statclock(). We scale p_estcpu instead of scaling the clock. This is more accurate and flexible. === Algorithm change === Same change. === Other bugs === The p_pctcpu bug was fixed long ago. We don't try as hard to abstract functionality yet. Related changes: the new limit on p_estcpu must be exported to kern_exit.c for clipping in wait1(). Agreed with by: dufault
1999-11-28 12:12:13 +00:00
* compute a different priority each time p_estcpu increases by
* INVERSE_ESTCPU_WEIGHT
* (until MAXPRI is reached). The cpu usage estimator ramps up
* quite quickly when the process is running (linearly), and decays
* away exponentially, at a rate which is proportionally slower when
* the system is busy. The basic principle is that the system will
* 90% forget that the process used a lot of CPU time in 5 * loadav
* seconds. This causes the system to favor processes which haven't
* run much recently, and to round-robin among other processes.
*/
void
schedclock(p)
struct proc *p;
{
Scheduler fixes equivalent to the ones logged in the following NetBSD commit to kern_synch.c: ---------------------------- revision 1.55 date: 1999/02/23 02:56:03; author: ross; state: Exp; lines: +39 -10 Scheduler bug fixes and reorganization * fix the ancient nice(1) bug, where nice +20 processes incorrectly steal 10 - 20% of the CPU, (or even more depending on load average) * provide a new schedclk() mechanism at a new clock at schedhz, so high platform hz values don't cause nice +0 processes to look like they are niced * change the algorithm slightly, and reorganize the code a lot * fix percent-CPU calculation bugs, and eliminate some no-op code === nice bug === Correctly divide the scheduler queues between niced and compute-bound processes. The current nice weight of two (sort of, see `algorithm change' below) neatly divides the USRPRI queues in half; this should have been used to clip p_estcpu, instead of UCHAR_MAX. Besides being the wrong amount, clipping an unsigned char to UCHAR_MAX is a no-op, and it was done after decay_cpu() which can only _reduce_ the value. It has to be kept <= NICE_WEIGHT * PRIO_MAX - PPQ or processes can scheduler-penalize themselves onto the same queue as nice +20 processes. (Or even a higher one.) === New schedclk() mechansism === Some platforms should be cutting down stathz before hitting the scheduler, since the scheduler algorithm only works right in the vicinity of 64 Hz. Rather than prescale hz, then scale back and forth by 4 every time p_estcpu is touched (each occurance an abstraction violation), use p_estcpu without scaling and require schedhz to be generated directly at the right frequency. Use a default stathz (well, actually, profhz) / 4, so nothing changes unless a platform defines schedhz and a new clock. Define these for alpha, where hz==1024, and nice was totally broke. === Algorithm change === The nice value used to be added to the exponentially-decayed scheduler history value p_estcpu, in _addition_ to be incorporated directly (with greater wieght) into the priority calculation. At first glance, it appears to be a pointless increase of 1/8 the nice effect (pri = p_estcpu/4 + nice*2), but it's actually at least 3x that because it will ramp up linearly but be decayed only exponentially, thus converging to an additional .75 nice for a loadaverage of one. I killed this, it makes the behavior hard to control, almost impossible to analyze, and the effect (~~nothing at for the first second, then somewhat increased niceness after three seconds or more, depending on load average) pointless. === Other bugs === hz -> profhz in the p_pctcpu = f(p_cpticks) calcuation. Collect scheduler functionality. Try to put each abstraction in just one place. ---------------------------- The details are a little different in FreeBSD: === nice bug === Fixing this is the main point of this commit. We use essentially the same clipping rule as NetBSD (our limit on p_estcpu differs by a scale factor). However, clipping at all is fundamentally bad. It gives free CPU the hoggiest hogs once they reach the limit, and reaching the limit is normal for long-running hogs. This will be fixed later. === New schedclk() mechanism === We don't use the NetBSD schedclk() (now schedclock()) mechanism. We require (real)stathz to be about 128 and scale by an extra factor of 2 compared with NetBSD's statclock(). We scale p_estcpu instead of scaling the clock. This is more accurate and flexible. === Algorithm change === Same change. === Other bugs === The p_pctcpu bug was fixed long ago. We don't try as hard to abstract functionality yet. Related changes: the new limit on p_estcpu must be exported to kern_exit.c for clipping in wait1(). Agreed with by: dufault
1999-11-28 12:12:13 +00:00
p->p_cpticks++;
Scheduler fixes equivalent to the ones logged in the following NetBSD commit to kern_synch.c: ---------------------------- revision 1.55 date: 1999/02/23 02:56:03; author: ross; state: Exp; lines: +39 -10 Scheduler bug fixes and reorganization * fix the ancient nice(1) bug, where nice +20 processes incorrectly steal 10 - 20% of the CPU, (or even more depending on load average) * provide a new schedclk() mechanism at a new clock at schedhz, so high platform hz values don't cause nice +0 processes to look like they are niced * change the algorithm slightly, and reorganize the code a lot * fix percent-CPU calculation bugs, and eliminate some no-op code === nice bug === Correctly divide the scheduler queues between niced and compute-bound processes. The current nice weight of two (sort of, see `algorithm change' below) neatly divides the USRPRI queues in half; this should have been used to clip p_estcpu, instead of UCHAR_MAX. Besides being the wrong amount, clipping an unsigned char to UCHAR_MAX is a no-op, and it was done after decay_cpu() which can only _reduce_ the value. It has to be kept <= NICE_WEIGHT * PRIO_MAX - PPQ or processes can scheduler-penalize themselves onto the same queue as nice +20 processes. (Or even a higher one.) === New schedclk() mechansism === Some platforms should be cutting down stathz before hitting the scheduler, since the scheduler algorithm only works right in the vicinity of 64 Hz. Rather than prescale hz, then scale back and forth by 4 every time p_estcpu is touched (each occurance an abstraction violation), use p_estcpu without scaling and require schedhz to be generated directly at the right frequency. Use a default stathz (well, actually, profhz) / 4, so nothing changes unless a platform defines schedhz and a new clock. Define these for alpha, where hz==1024, and nice was totally broke. === Algorithm change === The nice value used to be added to the exponentially-decayed scheduler history value p_estcpu, in _addition_ to be incorporated directly (with greater wieght) into the priority calculation. At first glance, it appears to be a pointless increase of 1/8 the nice effect (pri = p_estcpu/4 + nice*2), but it's actually at least 3x that because it will ramp up linearly but be decayed only exponentially, thus converging to an additional .75 nice for a loadaverage of one. I killed this, it makes the behavior hard to control, almost impossible to analyze, and the effect (~~nothing at for the first second, then somewhat increased niceness after three seconds or more, depending on load average) pointless. === Other bugs === hz -> profhz in the p_pctcpu = f(p_cpticks) calcuation. Collect scheduler functionality. Try to put each abstraction in just one place. ---------------------------- The details are a little different in FreeBSD: === nice bug === Fixing this is the main point of this commit. We use essentially the same clipping rule as NetBSD (our limit on p_estcpu differs by a scale factor). However, clipping at all is fundamentally bad. It gives free CPU the hoggiest hogs once they reach the limit, and reaching the limit is normal for long-running hogs. This will be fixed later. === New schedclk() mechanism === We don't use the NetBSD schedclk() (now schedclock()) mechanism. We require (real)stathz to be about 128 and scale by an extra factor of 2 compared with NetBSD's statclock(). We scale p_estcpu instead of scaling the clock. This is more accurate and flexible. === Algorithm change === Same change. === Other bugs === The p_pctcpu bug was fixed long ago. We don't try as hard to abstract functionality yet. Related changes: the new limit on p_estcpu must be exported to kern_exit.c for clipping in wait1(). Agreed with by: dufault
1999-11-28 12:12:13 +00:00
p->p_estcpu = ESTCPULIM(p->p_estcpu + 1);
if ((p->p_estcpu % INVERSE_ESTCPU_WEIGHT) == 0) {
resetpriority(p);
if (p->p_priority >= PUSER)
p->p_priority = p->p_usrpri;
}
}