NetBSD/sys/kern/kern_synch.c

1114 lines
28 KiB
C
Raw Normal View History

2002-09-30 01:11:36 +04:00
/* $NetBSD: kern_synch.c,v 1.114 2002/09/29 21:11:36 gmcgarry Exp $ */
/*-
* Copyright (c) 1999, 2000 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
* NASA Ames Research Center.
*
* 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 NetBSD
* Foundation, Inc. and its contributors.
* 4. Neither the name of The NetBSD Foundation 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 NETBSD FOUNDATION, INC. 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 FOUNDATION 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.
*/
/*-
* 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.
*
1998-03-01 05:20:01 +03:00
* @(#)kern_synch.c 8.9 (Berkeley) 5/19/95
*/
2001-11-12 18:25:01 +03:00
#include <sys/cdefs.h>
2002-09-30 01:11:36 +04:00
__KERNEL_RCSID(0, "$NetBSD: kern_synch.c,v 1.114 2002/09/29 21:11:36 gmcgarry Exp $");
2001-11-12 18:25:01 +03:00
1998-07-05 02:18:13 +04:00
#include "opt_ddb.h"
1998-06-26 01:17:15 +04:00
#include "opt_ktrace.h"
#include "opt_kstack.h"
#include "opt_lockdebug.h"
#include "opt_multiprocessor.h"
#include "opt_perfctrs.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/callout.h>
#include <sys/proc.h>
#include <sys/kernel.h>
#include <sys/buf.h>
#if defined(PERFCTRS)
#include <sys/pmc.h>
#endif
#include <sys/signalvar.h>
#include <sys/resourcevar.h>
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.
1999-02-23 05:56:03 +03:00
#include <sys/sched.h>
#include <uvm/uvm_extern.h>
#ifdef KTRACE
#include <sys/ktrace.h>
#endif
#include <machine/cpu.h>
int lbolt; /* once a second sleep address */
int rrticks; /* number of hardclock ticks per roundrobin() */
/*
* The global scheduler state.
*/
struct prochd sched_qs[RUNQUE_NQS]; /* run queues */
__volatile u_int32_t sched_whichqs; /* bitmap of non-empty queues */
struct slpque sched_slpque[SLPQUE_TABLESIZE]; /* sleep queues */
struct simplelock sched_lock = SIMPLELOCK_INITIALIZER;
void schedcpu(void *);
void updatepri(struct proc *);
void endtsleep(void *);
1996-02-04 05:15:01 +03:00
__inline void awaken(struct proc *);
struct callout schedcpu_ch = CALLOUT_INITIALIZER;
/*
* Force switch among equal priority processes every 100ms.
* Called from hardclock every hz/10 == rrticks hardclock ticks.
*/
/* ARGSUSED */
void
roundrobin(struct cpu_info *ci)
{
struct schedstate_percpu *spc = &ci->ci_schedstate;
spc->spc_rrticks = rrticks;
if (curproc != NULL) {
if (spc->spc_flags & SPCF_SEENRR) {
/*
* The process has already been through a roundrobin
* without switching and may be hogging the CPU.
* Indicate that the process should yield.
*/
spc->spc_flags |= SPCF_SHOULDYIELD;
} else
spc->spc_flags |= SPCF_SEENRR;
}
need_resched(curcpu());
}
/*
* 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 hardclock updates p_estcpu and p_cpticks independently.
*
* 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)
*
* 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 */
fixpt_t ccpu = 0.95122942450071400909 * FSCALE; /* exp(-1/20) */
/*
* 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 dont want to bother with the faster/more-accurate formula, you
* 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 */
void
schedcpu(void *arg)
{
2000-03-30 13:27:11 +04:00
fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
struct proc *p;
int s, s1;
2000-03-30 13:27:11 +04:00
unsigned int newcpu;
int clkhz;
proclist_lock_read();
LIST_FOREACH(p, &allproc, p_list) {
/*
* 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 = splstatclock(); /* prevent state changes */
/*
* p_pctcpu is only for ps.
*/
clkhz = stathz != 0 ? stathz : hz;
#if (FSHIFT >= CCPU_SHIFT)
p->p_pctcpu += (clkhz == 100)?
((fixpt_t) p->p_cpticks) << (FSHIFT - CCPU_SHIFT):
100 * (((fixpt_t) p->p_cpticks)
<< (FSHIFT - CCPU_SHIFT)) / clkhz;
#else
p->p_pctcpu += ((FSCALE - ccpu) *
(p->p_cpticks * FSCALE / clkhz)) >> FSHIFT;
#endif
p->p_cpticks = 0;
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.
1999-02-23 05:56:03 +03:00
newcpu = (u_int)decay_cpu(loadfac, p->p_estcpu);
p->p_estcpu = newcpu;
SCHED_LOCK(s1);
resetpriority(p);
if (p->p_priority >= PUSER) {
if (p->p_stat == SRUN &&
(p->p_flag & P_INMEM) &&
(p->p_priority / PPQ) != (p->p_usrpri / PPQ)) {
remrunqueue(p);
p->p_priority = p->p_usrpri;
setrunqueue(p);
} else
p->p_priority = p->p_usrpri;
}
SCHED_UNLOCK(s1);
splx(s);
}
proclist_unlock_read();
uvm_meter();
wakeup((caddr_t)&lbolt);
callout_reset(&schedcpu_ch, hz, schedcpu, NULL);
}
/*
* 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.
*/
void
updatepri(struct proc *p)
{
unsigned int newcpu;
fixpt_t loadfac;
SCHED_ASSERT_LOCKED();
newcpu = p->p_estcpu;
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)
newcpu = (int) decay_cpu(loadfac, newcpu);
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.
1999-02-23 05:56:03 +03:00
p->p_estcpu = newcpu;
}
resetpriority(p);
}
/*
* 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;
/*
* 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).
*
* The interlock is held until the scheduler_slock is acquired. The
* interlock will be locked before returning back to the caller
* unless the PNORELOCK flag is specified, in which case the
* interlock will always be unlocked upon return.
*/
int
ltsleep(void *ident, int priority, const char *wmesg, int timo,
__volatile struct simplelock *interlock)
{
2000-03-30 13:27:11 +04:00
struct proc *p = curproc;
struct slpque *qp;
int sig, s;
int catch = priority & PCATCH;
int relock = (priority & PNORELOCK) == 0;
/*
* XXXSMP
* This is probably bogus. Figure out what the right
* thing to do here really is.
* Note that not sleeping if ltsleep is called with curproc == NULL
* in the shutdown case is disgusting but partly necessary given
* how shutdown (barely) works.
*/
if (cold || (doing_shutdown && (panicstr || (p == NULL)))) {
/*
* 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.
*/
s = splhigh();
splx(safepri);
splx(s);
if (interlock != NULL && relock == 0)
simple_unlock(interlock);
return (0);
}
KASSERT(p != NULL);
LOCK_ASSERT(interlock == NULL || simple_lock_held(interlock));
#ifdef KTRACE
if (KTRPOINT(p, KTR_CSW))
ktrcsw(p, 1, 0);
#endif
SCHED_LOCK(s);
#ifdef DIAGNOSTIC
if (ident == NULL)
panic("ltsleep: ident == NULL");
if (p->p_stat != SONPROC)
panic("ltsleep: p_stat %d != SONPROC", p->p_stat);
if (p->p_back != NULL)
panic("ltsleep: p_back != NULL");
#endif
p->p_wchan = ident;
p->p_wmesg = wmesg;
p->p_slptime = 0;
p->p_priority = priority & PRIMASK;
qp = SLPQUE(ident);
if (qp->sq_head == 0)
qp->sq_head = p;
else
*qp->sq_tailp = p;
*(qp->sq_tailp = &p->p_forw) = 0;
if (timo)
callout_reset(&p->p_tsleep_ch, timo, endtsleep, p);
/*
* We can now release the interlock; the scheduler_slock
* is held, so a thread can't get in to do wakeup() before
* we do the switch.
*
* XXX We leave the code block here, after inserting ourselves
* on the sleep queue, because we might want a more clever
* data structure for the sleep queues at some point.
*/
if (interlock != NULL)
simple_unlock(interlock);
/*
* 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;
1996-02-04 05:15:01 +03:00
if ((sig = CURSIG(p)) != 0) {
if (p->p_wchan != NULL)
unsleep(p);
p->p_stat = SONPROC;
SCHED_UNLOCK(s);
goto resume;
}
if (p->p_wchan == NULL) {
catch = 0;
SCHED_UNLOCK(s);
goto resume;
}
} else
sig = 0;
p->p_stat = SSLEEP;
p->p_stats->p_ru.ru_nvcsw++;
SCHED_ASSERT_LOCKED();
mi_switch(p, NULL);
#if defined(DDB) && !defined(GPROF)
/* handy breakpoint location after process "wakes" */
2001-11-30 19:21:16 +03:00
__asm(".globl bpendtsleep ; bpendtsleep:");
#endif
SCHED_ASSERT_UNLOCKED();
splx(s);
resume:
KDASSERT(p->p_cpu != NULL);
KDASSERT(p->p_cpu == curcpu());
p->p_cpu->ci_schedstate.spc_curpriority = p->p_usrpri;
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, 0, 0);
#endif
if (relock && interlock != NULL)
simple_lock(interlock);
return (EWOULDBLOCK);
}
} else if (timo)
callout_stop(&p->p_tsleep_ch);
1996-02-04 05:15:01 +03:00
if (catch && (sig != 0 || (sig = CURSIG(p)) != 0)) {
#ifdef KTRACE
if (KTRPOINT(p, KTR_CSW))
ktrcsw(p, 0, 0);
#endif
if (relock && interlock != NULL)
simple_lock(interlock);
if ((SIGACTION(p, sig).sa_flags & SA_RESTART) == 0)
return (EINTR);
return (ERESTART);
}
#ifdef KTRACE
if (KTRPOINT(p, KTR_CSW))
ktrcsw(p, 0, 0);
#endif
if (relock && interlock != NULL)
simple_lock(interlock);
return (0);
}
/*
* Implement timeout for tsleep.
* 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.
*/
void
endtsleep(void *arg)
{
2000-03-30 13:27:11 +04:00
struct proc *p;
int s;
p = (struct proc *)arg;
SCHED_LOCK(s);
if (p->p_wchan) {
if (p->p_stat == SSLEEP)
setrunnable(p);
else
unsleep(p);
p->p_flag |= P_TIMEOUT;
}
SCHED_UNLOCK(s);
}
/*
* Remove a process from its wait queue
*/
void
unsleep(struct proc *p)
{
2000-03-30 13:27:11 +04:00
struct slpque *qp;
struct proc **hp;
SCHED_ASSERT_LOCKED();
if (p->p_wchan) {
hp = &(qp = SLPQUE(p->p_wchan))->sq_head;
while (*hp != p)
hp = &(*hp)->p_forw;
*hp = p->p_forw;
if (qp->sq_tailp == &p->p_forw)
qp->sq_tailp = hp;
p->p_wchan = 0;
}
}
/*
* Optimized-for-wakeup() version of setrunnable().
*/
__inline void
awaken(struct proc *p)
{
SCHED_ASSERT_LOCKED();
if (p->p_slptime > 1)
updatepri(p);
p->p_slptime = 0;
p->p_stat = SRUN;
/*
* Since curpriority is a user priority, p->p_priority
* is always better than curpriority.
*/
if (p->p_flag & P_INMEM) {
setrunqueue(p);
KASSERT(p->p_cpu != NULL);
need_resched(p->p_cpu);
} else
sched_wakeup(&proc0);
}
#if defined(MULTIPROCESSOR) || defined(LOCKDEBUG)
void
sched_unlock_idle(void)
{
simple_unlock(&sched_lock);
}
void
sched_lock_idle(void)
{
simple_lock(&sched_lock);
}
#endif /* MULTIPROCESSOR || LOCKDEBUG */
/*
* Make all processes sleeping on the specified identifier runnable.
*/
void
wakeup(void *ident)
{
int s;
SCHED_ASSERT_UNLOCKED();
SCHED_LOCK(s);
sched_wakeup(ident);
SCHED_UNLOCK(s);
}
void
sched_wakeup(void *ident)
{
2000-03-30 13:27:11 +04:00
struct slpque *qp;
struct proc *p, **q;
SCHED_ASSERT_LOCKED();
qp = SLPQUE(ident);
restart:
1996-02-04 05:15:01 +03:00
for (q = &qp->sq_head; (p = *q) != NULL; ) {
#ifdef DIAGNOSTIC
1996-02-04 05:15:01 +03:00
if (p->p_back || (p->p_stat != SSLEEP && p->p_stat != SSTOP))
panic("wakeup");
#endif
if (p->p_wchan == ident) {
p->p_wchan = 0;
*q = p->p_forw;
if (qp->sq_tailp == &p->p_forw)
qp->sq_tailp = q;
if (p->p_stat == SSLEEP) {
awaken(p);
goto restart;
}
} else
q = &p->p_forw;
}
}
/*
* Make the highest priority process first in line on the specified
* identifier runnable.
*/
void
wakeup_one(void *ident)
{
struct slpque *qp;
struct proc *p, **q;
struct proc *best_sleepp, **best_sleepq;
struct proc *best_stopp, **best_stopq;
int s;
best_sleepp = best_stopp = NULL;
best_sleepq = best_stopq = NULL;
SCHED_LOCK(s);
qp = SLPQUE(ident);
for (q = &qp->sq_head; (p = *q) != NULL; q = &p->p_forw) {
#ifdef DIAGNOSTIC
if (p->p_back || (p->p_stat != SSLEEP && p->p_stat != SSTOP))
panic("wakeup_one");
#endif
if (p->p_wchan == ident) {
if (p->p_stat == SSLEEP) {
if (best_sleepp == NULL ||
p->p_priority < best_sleepp->p_priority) {
best_sleepp = p;
best_sleepq = q;
}
} else {
if (best_stopp == NULL ||
p->p_priority < best_stopp->p_priority) {
best_stopp = p;
best_stopq = q;
}
}
}
}
/*
* Consider any SSLEEP process higher than the highest priority SSTOP
* process.
*/
if (best_sleepp != NULL) {
p = best_sleepp;
q = best_sleepq;
} else {
p = best_stopp;
q = best_stopq;
}
if (p != NULL) {
p->p_wchan = NULL;
*q = p->p_forw;
if (qp->sq_tailp == &p->p_forw)
qp->sq_tailp = q;
if (p->p_stat == SSLEEP)
awaken(p);
}
SCHED_UNLOCK(s);
}
/*
* General yield call. Puts the current process back on its run queue and
* performs a voluntary context switch.
*/
void
yield(void)
{
struct proc *p = curproc;
int s;
SCHED_LOCK(s);
p->p_priority = p->p_usrpri;
p->p_stat = SRUN;
setrunqueue(p);
p->p_stats->p_ru.ru_nvcsw++;
mi_switch(p, NULL);
SCHED_ASSERT_UNLOCKED();
splx(s);
}
/*
* General preemption call. Puts the current process back on its run queue
* and performs an involuntary context switch. If a process is supplied,
* we switch to that process. Otherwise, we use the normal process selection
* criteria.
*/
void
preempt(struct proc *newp)
{
struct proc *p = curproc;
int s;
SCHED_LOCK(s);
p->p_priority = p->p_usrpri;
p->p_stat = SRUN;
setrunqueue(p);
p->p_stats->p_ru.ru_nivcsw++;
mi_switch(p, newp);
SCHED_ASSERT_UNLOCKED();
splx(s);
}
/*
* The machine independent parts of context switch.
2000-08-24 10:14:34 +04:00
* Must be called at splsched() (no higher!) and with
* the sched_lock held.
*/
void
mi_switch(struct proc *p, struct proc *newp)
{
struct schedstate_percpu *spc;
2000-03-30 13:27:11 +04:00
struct rlimit *rlim;
long s, u;
struct timeval tv;
#if defined(MULTIPROCESSOR)
int hold_count;
#endif
SCHED_ASSERT_LOCKED();
#if defined(MULTIPROCESSOR)
/*
* Release the kernel_lock, as we are about to yield the CPU.
* The scheduler lock is still held until cpu_switch()
* selects a new process and removes it from the run queue.
*/
if (p->p_flag & P_BIGLOCK)
hold_count = spinlock_release_all(&kernel_lock);
#endif
KDASSERT(p->p_cpu != NULL);
KDASSERT(p->p_cpu == curcpu());
KDASSERT(newp == NULL);
spc = &p->p_cpu->ci_schedstate;
#if defined(LOCKDEBUG) || defined(DIAGNOSTIC)
spinlock_switchcheck();
#endif
#ifdef LOCKDEBUG
simple_lock_switchcheck();
1998-03-01 05:20:01 +03:00
#endif
/*
* Compute the amount of time during which the current
* process was running.
*/
microtime(&tv);
u = p->p_rtime.tv_usec + (tv.tv_usec - spc->spc_runtime.tv_usec);
s = p->p_rtime.tv_sec + (tv.tv_sec - spc->spc_runtime.tv_sec);
if (u < 0) {
u += 1000000;
s--;
} else if (u >= 1000000) {
u -= 1000000;
s++;
}
2002-09-30 01:11:36 +04:00
p->p_rtime.tv_usec = u;
p->p_rtime.tv_sec = s;
/*
* Check if the process exceeds its cpu resource allocation.
* If over max, kill it. In any case, if it has run for more
* than 10 minutes, reduce priority to give others a chance.
*/
rlim = &p->p_rlimit[RLIMIT_CPU];
if (s >= rlim->rlim_cur) {
/*
* XXXSMP: we're inside the scheduler lock perimeter;
* use sched_psignal.
*/
if (s >= rlim->rlim_max)
sched_psignal(p, SIGKILL);
else {
sched_psignal(p, SIGXCPU);
if (rlim->rlim_cur < rlim->rlim_max)
rlim->rlim_cur += 5;
}
}
if (autonicetime && s > autonicetime && p->p_ucred->cr_uid &&
p->p_nice == NZERO) {
p->p_nice = autoniceval + NZERO;
resetpriority(p);
}
/*
* Process is about to yield the CPU; clear the appropriate
* scheduling flags.
*/
spc->spc_flags &= ~SPCF_SWITCHCLEAR;
#ifdef KSTACK_CHECK_MAGIC
kstack_check_magic(p);
#endif
/*
2002-09-30 01:11:36 +04:00
* If we are using h/w performance counters, save context.
*/
2002-09-30 01:11:36 +04:00
#if PERFCTRS
if (PMC_ENABLED(p))
pmc_save_context(p);
#endif
/*
2002-09-30 01:11:36 +04:00
* Switch to the new current process. When we
* run again, we'll return back here.
*/
2002-09-30 01:11:36 +04:00
uvmexp.swtch++;
cpu_switch(p, NULL);
/*
2002-09-30 01:11:36 +04:00
* If we are using h/w performance counters, restore context.
*/
#if PERFCTRS
2002-09-30 01:11:36 +04:00
if (PMC_ENABLED(p))
pmc_restore_context(p);
#endif
2002-09-30 01:11:36 +04:00
/*
* Make sure that MD code released the scheduler lock before
* resuming us.
*/
SCHED_ASSERT_UNLOCKED();
/*
* We're running again; record our new start time. We might
* be running on a new CPU now, so don't use the cache'd
* schedstate_percpu pointer.
*/
KDASSERT(p->p_cpu != NULL);
KDASSERT(p->p_cpu == curcpu());
microtime(&p->p_cpu->ci_schedstate.spc_runtime);
#if defined(MULTIPROCESSOR)
/*
* Reacquire the kernel_lock now. We do this after we've
* released the scheduler lock to avoid deadlock, and before
* we reacquire the interlock.
*/
if (p->p_flag & P_BIGLOCK)
spinlock_acquire_count(&kernel_lock, hold_count);
#endif
}
/*
* Initialize the (doubly-linked) run queues
* to be empty.
*/
void
rqinit()
{
2000-03-30 13:27:11 +04:00
int i;
for (i = 0; i < RUNQUE_NQS; i++)
sched_qs[i].ph_link = sched_qs[i].ph_rlink =
(struct proc *)&sched_qs[i];
}
/*
* 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(struct proc *p)
{
SCHED_ASSERT_LOCKED();
switch (p->p_stat) {
case 0:
case SRUN:
case SONPROC:
case SZOMB:
case SDEAD:
default:
panic("setrunnable");
case SSTOP:
/*
* If we're being traced (possibly because someone attached us
* while we were stopped), check for a signal from the debugger.
*/
if ((p->p_flag & P_TRACED) != 0 && p->p_xstat != 0) {
sigaddset(&p->p_sigctx.ps_siglist, p->p_xstat);
CHECKSIGS(p);
}
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);
if (p->p_slptime > 1)
updatepri(p);
p->p_slptime = 0;
if ((p->p_flag & P_INMEM) == 0)
sched_wakeup((caddr_t)&proc0);
else if (p->p_priority < curcpu()->ci_schedstate.spc_curpriority) {
/*
* XXXSMP
* This is not exactly right. Since p->p_cpu persists
* across a context switch, this gives us some sort
* of processor affinity. But we need to figure out
* at what point it's better to reschedule on a different
* CPU than the last one.
*/
need_resched((p->p_cpu != NULL) ? p->p_cpu : curcpu());
}
}
/*
* 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(struct proc *p)
{
2000-03-30 13:27:11 +04:00
unsigned int newpriority;
SCHED_ASSERT_LOCKED();
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.
1999-02-23 05:56:03 +03:00
newpriority = PUSER + p->p_estcpu + NICE_WEIGHT * (p->p_nice - NZERO);
newpriority = min(newpriority, MAXPRI);
p->p_usrpri = newpriority;
if (newpriority < curcpu()->ci_schedstate.spc_curpriority) {
/*
* XXXSMP
* Same applies as in setrunnable() above.
*/
need_resched((p->p_cpu != NULL) ? p->p_cpu : curcpu());
}
}
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.
1999-02-23 05:56:03 +03:00
/*
* 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. The formula for computing priorities (in kern_synch.c)
* will compute a different value each time p_estcpu increases. This can
* cause a switch, but unless the priority crosses a PPQ boundary the actual
* queue will not change. 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
2000-08-03 00:21:36 +04:00
* 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.
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.
1999-02-23 05:56:03 +03:00
*/
void
schedclock(struct proc *p)
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.
1999-02-23 05:56:03 +03:00
{
int s;
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.
1999-02-23 05:56:03 +03:00
p->p_estcpu = ESTCPULIM(p->p_estcpu + 1);
SCHED_LOCK(s);
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.
1999-02-23 05:56:03 +03:00
resetpriority(p);
SCHED_UNLOCK(s);
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.
1999-02-23 05:56:03 +03:00
if (p->p_priority >= PUSER)
p->p_priority = p->p_usrpri;
}
void
suspendsched()
{
struct proc *p;
int s;
/*
* Convert all non-P_SYSTEM SSLEEP or SRUN processes to SSTOP.
*/
proclist_lock_read();
SCHED_LOCK(s);
LIST_FOREACH(p, &allproc, p_list) {
if ((p->p_flag & P_SYSTEM) != 0)
continue;
switch (p->p_stat) {
case SRUN:
if ((p->p_flag & P_INMEM) != 0)
remrunqueue(p);
/* FALLTHROUGH */
case SSLEEP:
p->p_stat = SSTOP;
break;
case SONPROC:
/*
* XXX SMP: we need to deal with processes on
* others CPU !
*/
break;
default:
break;
}
}
SCHED_UNLOCK(s);
proclist_unlock_read();
}
/*
* Low-level routines to access the run queue. Optimised assembler
* routines can override these.
*/
#ifndef __HAVE_MD_RUNQUEUE
void
setrunqueue(struct proc *p)
{
struct prochd *rq;
struct proc *prev;
int whichq;
#ifdef DIAGNOSTIC
if (p->p_back != NULL || p->p_wchan != NULL || p->p_stat != SRUN)
panic("setrunqueue");
#endif
whichq = p->p_priority / 4;
sched_whichqs |= (1<<whichq);
rq = &sched_qs[whichq];
prev = rq->ph_rlink;
p->p_forw = (struct proc *)rq;
rq->ph_rlink = p;
prev->p_forw = p;
p->p_back = prev;
}
void
remrunqueue(struct proc *p)
{
struct proc *prev, *next;
int whichq;
whichq = p->p_priority / 4;
#ifdef DIAGNOSTIC
if (((sched_whichqs & (1<<whichq)) == 0))
panic("remrunqueue");
#endif
prev = p->p_back;
p->p_back = NULL;
next = p->p_forw;
prev->p_forw = next;
next->p_back = prev;
if (prev == next)
sched_whichqs &= ~(1<<whichq);
}
#endif