NetBSD/sys/kern/kern_clock.c

1432 lines
42 KiB
C

/* $NetBSD: kern_clock.c,v 1.92 2004/09/15 04:56:14 tls Exp $ */
/*-
* Copyright (c) 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, 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. 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_clock.c 8.5 (Berkeley) 1/21/94
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: kern_clock.c,v 1.92 2004/09/15 04:56:14 tls Exp $");
#include "opt_ntp.h"
#include "opt_multiprocessor.h"
#include "opt_perfctrs.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/callout.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/resourcevar.h>
#include <sys/signalvar.h>
#include <sys/sysctl.h>
#include <sys/timex.h>
#include <sys/sched.h>
#include <sys/time.h>
#include <machine/cpu.h>
#ifdef __HAVE_GENERIC_SOFT_INTERRUPTS
#include <machine/intr.h>
#endif
#ifdef GPROF
#include <sys/gmon.h>
#endif
/*
* Clock handling routines.
*
* This code is written to operate with two timers that run independently of
* each other. The main clock, running hz times per second, is used to keep
* track of real time. The second timer handles kernel and user profiling,
* and does resource use estimation. If the second timer is programmable,
* it is randomized to avoid aliasing between the two clocks. For example,
* the randomization prevents an adversary from always giving up the CPU
* just before its quantum expires. Otherwise, it would never accumulate
* CPU ticks. The mean frequency of the second timer is stathz.
*
* If no second timer exists, stathz will be zero; in this case we drive
* profiling and statistics off the main clock. This WILL NOT be accurate;
* do not do it unless absolutely necessary.
*
* The statistics clock may (or may not) be run at a higher rate while
* profiling. This profile clock runs at profhz. We require that profhz
* be an integral multiple of stathz.
*
* If the statistics clock is running fast, it must be divided by the ratio
* profhz/stathz for statistics. (For profiling, every tick counts.)
*/
#ifdef NTP /* NTP phase-locked loop in kernel */
/*
* Phase/frequency-lock loop (PLL/FLL) definitions
*
* The following variables are read and set by the ntp_adjtime() system
* call.
*
* time_state shows the state of the system clock, with values defined
* in the timex.h header file.
*
* time_status shows the status of the system clock, with bits defined
* in the timex.h header file.
*
* time_offset is used by the PLL/FLL to adjust the system time in small
* increments.
*
* time_constant determines the bandwidth or "stiffness" of the PLL.
*
* time_tolerance determines maximum frequency error or tolerance of the
* CPU clock oscillator and is a property of the architecture; however,
* in principle it could change as result of the presence of external
* discipline signals, for instance.
*
* time_precision is usually equal to the kernel tick variable; however,
* in cases where a precision clock counter or external clock is
* available, the resolution can be much less than this and depend on
* whether the external clock is working or not.
*
* time_maxerror is initialized by a ntp_adjtime() call and increased by
* the kernel once each second to reflect the maximum error bound
* growth.
*
* time_esterror is set and read by the ntp_adjtime() call, but
* otherwise not used by the kernel.
*/
int time_state = TIME_OK; /* clock state */
int time_status = STA_UNSYNC; /* clock status bits */
long time_offset = 0; /* time offset (us) */
long time_constant = 0; /* pll time constant */
long time_tolerance = MAXFREQ; /* frequency tolerance (scaled ppm) */
long time_precision = 1; /* clock precision (us) */
long time_maxerror = MAXPHASE; /* maximum error (us) */
long time_esterror = MAXPHASE; /* estimated error (us) */
/*
* The following variables establish the state of the PLL/FLL and the
* residual time and frequency offset of the local clock. The scale
* factors are defined in the timex.h header file.
*
* time_phase and time_freq are the phase increment and the frequency
* increment, respectively, of the kernel time variable.
*
* time_freq is set via ntp_adjtime() from a value stored in a file when
* the synchronization daemon is first started. Its value is retrieved
* via ntp_adjtime() and written to the file about once per hour by the
* daemon.
*
* time_adj is the adjustment added to the value of tick at each timer
* interrupt and is recomputed from time_phase and time_freq at each
* seconds rollover.
*
* time_reftime is the second's portion of the system time at the last
* call to ntp_adjtime(). It is used to adjust the time_freq variable
* and to increase the time_maxerror as the time since last update
* increases.
*/
long time_phase = 0; /* phase offset (scaled us) */
long time_freq = 0; /* frequency offset (scaled ppm) */
long time_adj = 0; /* tick adjust (scaled 1 / hz) */
long time_reftime = 0; /* time at last adjustment (s) */
#ifdef PPS_SYNC
/*
* The following variables are used only if the kernel PPS discipline
* code is configured (PPS_SYNC). The scale factors are defined in the
* timex.h header file.
*
* pps_time contains the time at each calibration interval, as read by
* microtime(). pps_count counts the seconds of the calibration
* interval, the duration of which is nominally pps_shift in powers of
* two.
*
* pps_offset is the time offset produced by the time median filter
* pps_tf[], while pps_jitter is the dispersion (jitter) measured by
* this filter.
*
* pps_freq is the frequency offset produced by the frequency median
* filter pps_ff[], while pps_stabil is the dispersion (wander) measured
* by this filter.
*
* pps_usec is latched from a high resolution counter or external clock
* at pps_time. Here we want the hardware counter contents only, not the
* contents plus the time_tv.usec as usual.
*
* pps_valid counts the number of seconds since the last PPS update. It
* is used as a watchdog timer to disable the PPS discipline should the
* PPS signal be lost.
*
* pps_glitch counts the number of seconds since the beginning of an
* offset burst more than tick/2 from current nominal offset. It is used
* mainly to suppress error bursts due to priority conflicts between the
* PPS interrupt and timer interrupt.
*
* pps_intcnt counts the calibration intervals for use in the interval-
* adaptation algorithm. It's just too complicated for words.
*
* pps_kc_hardpps_source contains an arbitrary value that uniquely
* identifies the currently bound source of the PPS signal, or NULL
* if no source is bound.
*
* pps_kc_hardpps_mode indicates which transitions, if any, of the PPS
* signal should be reported.
*/
struct timeval pps_time; /* kernel time at last interval */
long pps_tf[] = {0, 0, 0}; /* pps time offset median filter (us) */
long pps_offset = 0; /* pps time offset (us) */
long pps_jitter = MAXTIME; /* time dispersion (jitter) (us) */
long pps_ff[] = {0, 0, 0}; /* pps frequency offset median filter */
long pps_freq = 0; /* frequency offset (scaled ppm) */
long pps_stabil = MAXFREQ; /* frequency dispersion (scaled ppm) */
long pps_usec = 0; /* microsec counter at last interval */
long pps_valid = PPS_VALID; /* pps signal watchdog counter */
int pps_glitch = 0; /* pps signal glitch counter */
int pps_count = 0; /* calibration interval counter (s) */
int pps_shift = PPS_SHIFT; /* interval duration (s) (shift) */
int pps_intcnt = 0; /* intervals at current duration */
void *pps_kc_hardpps_source = NULL; /* current PPS supplier's identifier */
int pps_kc_hardpps_mode = 0; /* interesting edges of PPS signal */
/*
* PPS signal quality monitors
*
* pps_jitcnt counts the seconds that have been discarded because the
* jitter measured by the time median filter exceeds the limit MAXTIME
* (100 us).
*
* pps_calcnt counts the frequency calibration intervals, which are
* variable from 4 s to 256 s.
*
* pps_errcnt counts the calibration intervals which have been discarded
* because the wander exceeds the limit MAXFREQ (100 ppm) or where the
* calibration interval jitter exceeds two ticks.
*
* pps_stbcnt counts the calibration intervals that have been discarded
* because the frequency wander exceeds the limit MAXFREQ / 4 (25 us).
*/
long pps_jitcnt = 0; /* jitter limit exceeded */
long pps_calcnt = 0; /* calibration intervals */
long pps_errcnt = 0; /* calibration errors */
long pps_stbcnt = 0; /* stability limit exceeded */
#endif /* PPS_SYNC */
#ifdef EXT_CLOCK
/*
* External clock definitions
*
* The following definitions and declarations are used only if an
* external clock is configured on the system.
*/
#define CLOCK_INTERVAL 30 /* CPU clock update interval (s) */
/*
* The clock_count variable is set to CLOCK_INTERVAL at each PPS
* interrupt and decremented once each second.
*/
int clock_count = 0; /* CPU clock counter */
#ifdef HIGHBALL
/*
* The clock_offset and clock_cpu variables are used by the HIGHBALL
* interface. The clock_offset variable defines the offset between
* system time and the HIGBALL counters. The clock_cpu variable contains
* the offset between the system clock and the HIGHBALL clock for use in
* disciplining the kernel time variable.
*/
extern struct timeval clock_offset; /* Highball clock offset */
long clock_cpu = 0; /* CPU clock adjust */
#endif /* HIGHBALL */
#endif /* EXT_CLOCK */
#endif /* NTP */
/*
* Bump a timeval by a small number of usec's.
*/
#define BUMPTIME(t, usec) { \
volatile struct timeval *tp = (t); \
long us; \
\
tp->tv_usec = us = tp->tv_usec + (usec); \
if (us >= 1000000) { \
tp->tv_usec = us - 1000000; \
tp->tv_sec++; \
} \
}
int stathz;
int profhz;
int profsrc;
int schedhz;
int profprocs;
int hardclock_ticks;
static int statscheddiv; /* stat => sched divider (used if schedhz == 0) */
static int psdiv; /* prof => stat divider */
int psratio; /* ratio: prof / stat */
int tickfix, tickfixinterval; /* used if tick not really integral */
#ifndef NTP
static int tickfixcnt; /* accumulated fractional error */
#else
int fixtick; /* used by NTP for same */
int shifthz;
#endif
/*
* We might want ldd to load the both words from time at once.
* To succeed we need to be quadword aligned.
* The sparc already does that, and that it has worked so far is a fluke.
*/
volatile struct timeval time __attribute__((__aligned__(__alignof__(quad_t))));
volatile struct timeval mono_time;
void *softclock_si;
/*
* Initialize clock frequencies and start both clocks running.
*/
void
initclocks(void)
{
int i;
#ifdef __HAVE_GENERIC_SOFT_INTERRUPTS
softclock_si = softintr_establish(IPL_SOFTCLOCK, softclock, NULL);
if (softclock_si == NULL)
panic("initclocks: unable to register softclock intr");
#endif
/*
* Set divisors to 1 (normal case) and let the machine-specific
* code do its bit.
*/
psdiv = 1;
cpu_initclocks();
/*
* Compute profhz/stathz/rrticks, and fix profhz if needed.
*/
i = stathz ? stathz : hz;
if (profhz == 0)
profhz = i;
psratio = profhz / i;
rrticks = hz / 10;
if (schedhz == 0) {
/* 16Hz is best */
statscheddiv = i / 16;
if (statscheddiv <= 0)
panic("statscheddiv");
}
#ifdef NTP
switch (hz) {
case 1:
shifthz = SHIFT_SCALE - 0;
break;
case 2:
shifthz = SHIFT_SCALE - 1;
break;
case 4:
shifthz = SHIFT_SCALE - 2;
break;
case 8:
shifthz = SHIFT_SCALE - 3;
break;
case 16:
shifthz = SHIFT_SCALE - 4;
break;
case 32:
shifthz = SHIFT_SCALE - 5;
break;
case 50:
case 60:
case 64:
shifthz = SHIFT_SCALE - 6;
break;
case 96:
case 100:
case 128:
shifthz = SHIFT_SCALE - 7;
break;
case 256:
shifthz = SHIFT_SCALE - 8;
break;
case 512:
shifthz = SHIFT_SCALE - 9;
break;
case 1000:
case 1024:
shifthz = SHIFT_SCALE - 10;
break;
case 1200:
case 2048:
shifthz = SHIFT_SCALE - 11;
break;
case 4096:
shifthz = SHIFT_SCALE - 12;
break;
case 8192:
shifthz = SHIFT_SCALE - 13;
break;
case 16384:
shifthz = SHIFT_SCALE - 14;
break;
case 32768:
shifthz = SHIFT_SCALE - 15;
break;
case 65536:
shifthz = SHIFT_SCALE - 16;
break;
default:
panic("weird hz");
}
if (fixtick == 0) {
/*
* Give MD code a chance to set this to a better
* value; but, if it doesn't, we should.
*/
fixtick = (1000000 - (hz*tick));
}
#endif
}
/*
* The real-time timer, interrupting hz times per second.
*/
void
hardclock(struct clockframe *frame)
{
struct lwp *l;
struct proc *p;
int delta;
extern int tickdelta;
extern long timedelta;
struct cpu_info *ci = curcpu();
struct ptimer *pt;
#ifdef NTP
int time_update;
int ltemp;
#endif
l = curlwp;
if (l) {
p = l->l_proc;
/*
* Run current process's virtual and profile time, as needed.
*/
if (CLKF_USERMODE(frame) && p->p_timers &&
(pt = LIST_FIRST(&p->p_timers->pts_virtual)) != NULL)
if (itimerdecr(pt, tick) == 0)
itimerfire(pt);
if (p->p_timers &&
(pt = LIST_FIRST(&p->p_timers->pts_prof)) != NULL)
if (itimerdecr(pt, tick) == 0)
itimerfire(pt);
}
/*
* If no separate statistics clock is available, run it from here.
*/
if (stathz == 0)
statclock(frame);
if ((--ci->ci_schedstate.spc_rrticks) <= 0)
roundrobin(ci);
#if defined(MULTIPROCESSOR)
/*
* If we are not the primary CPU, we're not allowed to do
* any more work.
*/
if (CPU_IS_PRIMARY(ci) == 0)
return;
#endif
/*
* Increment the time-of-day. The increment is normally just
* ``tick''. If the machine is one which has a clock frequency
* such that ``hz'' would not divide the second evenly into
* milliseconds, a periodic adjustment must be applied. Finally,
* if we are still adjusting the time (see adjtime()),
* ``tickdelta'' may also be added in.
*/
hardclock_ticks++;
delta = tick;
#ifndef NTP
if (tickfix) {
tickfixcnt += tickfix;
if (tickfixcnt >= tickfixinterval) {
delta++;
tickfixcnt -= tickfixinterval;
}
}
#endif /* !NTP */
/* Imprecise 4bsd adjtime() handling */
if (timedelta != 0) {
delta += tickdelta;
timedelta -= tickdelta;
}
#ifdef notyet
microset();
#endif
#ifndef NTP
BUMPTIME(&time, delta); /* XXX Now done using NTP code below */
#endif
BUMPTIME(&mono_time, delta);
#ifdef NTP
time_update = delta;
/*
* Compute the phase adjustment. If the low-order bits
* (time_phase) of the update overflow, bump the high-order bits
* (time_update).
*/
time_phase += time_adj;
if (time_phase <= -FINEUSEC) {
ltemp = -time_phase >> SHIFT_SCALE;
time_phase += ltemp << SHIFT_SCALE;
time_update -= ltemp;
} else if (time_phase >= FINEUSEC) {
ltemp = time_phase >> SHIFT_SCALE;
time_phase -= ltemp << SHIFT_SCALE;
time_update += ltemp;
}
#ifdef HIGHBALL
/*
* If the HIGHBALL board is installed, we need to adjust the
* external clock offset in order to close the hardware feedback
* loop. This will adjust the external clock phase and frequency
* in small amounts. The additional phase noise and frequency
* wander this causes should be minimal. We also need to
* discipline the kernel time variable, since the PLL is used to
* discipline the external clock. If the Highball board is not
* present, we discipline kernel time with the PLL as usual. We
* assume that the external clock phase adjustment (time_update)
* and kernel phase adjustment (clock_cpu) are less than the
* value of tick.
*/
clock_offset.tv_usec += time_update;
if (clock_offset.tv_usec >= 1000000) {
clock_offset.tv_sec++;
clock_offset.tv_usec -= 1000000;
}
if (clock_offset.tv_usec < 0) {
clock_offset.tv_sec--;
clock_offset.tv_usec += 1000000;
}
time.tv_usec += clock_cpu;
clock_cpu = 0;
#else
time.tv_usec += time_update;
#endif /* HIGHBALL */
/*
* On rollover of the second the phase adjustment to be used for
* the next second is calculated. Also, the maximum error is
* increased by the tolerance. If the PPS frequency discipline
* code is present, the phase is increased to compensate for the
* CPU clock oscillator frequency error.
*
* On a 32-bit machine and given parameters in the timex.h
* header file, the maximum phase adjustment is +-512 ms and
* maximum frequency offset is a tad less than) +-512 ppm. On a
* 64-bit machine, you shouldn't need to ask.
*/
if (time.tv_usec >= 1000000) {
time.tv_usec -= 1000000;
time.tv_sec++;
time_maxerror += time_tolerance >> SHIFT_USEC;
/*
* Leap second processing. If in leap-insert state at
* the end of the day, the system clock is set back one
* second; if in leap-delete state, the system clock is
* set ahead one second. The microtime() routine or
* external clock driver will insure that reported time
* is always monotonic. The ugly divides should be
* replaced.
*/
switch (time_state) {
case TIME_OK:
if (time_status & STA_INS)
time_state = TIME_INS;
else if (time_status & STA_DEL)
time_state = TIME_DEL;
break;
case TIME_INS:
if (time.tv_sec % 86400 == 0) {
time.tv_sec--;
time_state = TIME_OOP;
}
break;
case TIME_DEL:
if ((time.tv_sec + 1) % 86400 == 0) {
time.tv_sec++;
time_state = TIME_WAIT;
}
break;
case TIME_OOP:
time_state = TIME_WAIT;
break;
case TIME_WAIT:
if (!(time_status & (STA_INS | STA_DEL)))
time_state = TIME_OK;
break;
}
/*
* Compute the phase adjustment for the next second. In
* PLL mode, the offset is reduced by a fixed factor
* times the time constant. In FLL mode the offset is
* used directly. In either mode, the maximum phase
* adjustment for each second is clamped so as to spread
* the adjustment over not more than the number of
* seconds between updates.
*/
if (time_offset < 0) {
ltemp = -time_offset;
if (!(time_status & STA_FLL))
ltemp >>= SHIFT_KG + time_constant;
if (ltemp > (MAXPHASE / MINSEC) << SHIFT_UPDATE)
ltemp = (MAXPHASE / MINSEC) <<
SHIFT_UPDATE;
time_offset += ltemp;
time_adj = -ltemp << (shifthz - SHIFT_UPDATE);
} else if (time_offset > 0) {
ltemp = time_offset;
if (!(time_status & STA_FLL))
ltemp >>= SHIFT_KG + time_constant;
if (ltemp > (MAXPHASE / MINSEC) << SHIFT_UPDATE)
ltemp = (MAXPHASE / MINSEC) <<
SHIFT_UPDATE;
time_offset -= ltemp;
time_adj = ltemp << (shifthz - SHIFT_UPDATE);
} else
time_adj = 0;
/*
* Compute the frequency estimate and additional phase
* adjustment due to frequency error for the next
* second. When the PPS signal is engaged, gnaw on the
* watchdog counter and update the frequency computed by
* the pll and the PPS signal.
*/
#ifdef PPS_SYNC
pps_valid++;
if (pps_valid == PPS_VALID) {
pps_jitter = MAXTIME;
pps_stabil = MAXFREQ;
time_status &= ~(STA_PPSSIGNAL | STA_PPSJITTER |
STA_PPSWANDER | STA_PPSERROR);
}
ltemp = time_freq + pps_freq;
#else
ltemp = time_freq;
#endif /* PPS_SYNC */
if (ltemp < 0)
time_adj -= -ltemp >> (SHIFT_USEC - shifthz);
else
time_adj += ltemp >> (SHIFT_USEC - shifthz);
time_adj += (long)fixtick << shifthz;
/*
* When the CPU clock oscillator frequency is not a
* power of 2 in Hz, shifthz is only an approximate
* scale factor.
*
* To determine the adjustment, you can do the following:
* bc -q
* scale=24
* obase=2
* idealhz/realhz
* where `idealhz' is the next higher power of 2, and `realhz'
* is the actual value. You may need to factor this result
* into a sequence of 2 multipliers to get better precision.
*
* Likewise, the error can be calculated with (e.g. for 100Hz):
* bc -q
* scale=24
* ((1+2^-2+2^-5)*(1-2^-10)*realhz-idealhz)/idealhz
* (and then multiply by 1000000 to get ppm).
*/
switch (hz) {
case 60:
/* A factor of 1.000100010001 gives about 15ppm
error. */
if (time_adj < 0) {
time_adj -= (-time_adj >> 4);
time_adj -= (-time_adj >> 8);
} else {
time_adj += (time_adj >> 4);
time_adj += (time_adj >> 8);
}
break;
case 96:
/* A factor of 1.0101010101 gives about 244ppm error. */
if (time_adj < 0) {
time_adj -= (-time_adj >> 2);
time_adj -= (-time_adj >> 4) + (-time_adj >> 8);
} else {
time_adj += (time_adj >> 2);
time_adj += (time_adj >> 4) + (time_adj >> 8);
}
break;
case 50:
case 100:
/* A factor of 1.010001111010111 gives about 1ppm
error. */
if (time_adj < 0) {
time_adj -= (-time_adj >> 2) + (-time_adj >> 5);
time_adj += (-time_adj >> 10);
} else {
time_adj += (time_adj >> 2) + (time_adj >> 5);
time_adj -= (time_adj >> 10);
}
break;
case 1000:
/* A factor of 1.000001100010100001 gives about 50ppm
error. */
if (time_adj < 0) {
time_adj -= (-time_adj >> 6) + (-time_adj >> 11);
time_adj -= (-time_adj >> 7);
} else {
time_adj += (time_adj >> 6) + (time_adj >> 11);
time_adj += (time_adj >> 7);
}
break;
case 1200:
/* A factor of 1.1011010011100001 gives about 64ppm
error. */
if (time_adj < 0) {
time_adj -= (-time_adj >> 1) + (-time_adj >> 6);
time_adj -= (-time_adj >> 3) + (-time_adj >> 10);
} else {
time_adj += (time_adj >> 1) + (time_adj >> 6);
time_adj += (time_adj >> 3) + (time_adj >> 10);
}
break;
}
#ifdef EXT_CLOCK
/*
* If an external clock is present, it is necessary to
* discipline the kernel time variable anyway, since not
* all system components use the microtime() interface.
* Here, the time offset between the external clock and
* kernel time variable is computed every so often.
*/
clock_count++;
if (clock_count > CLOCK_INTERVAL) {
clock_count = 0;
microtime(&clock_ext);
delta.tv_sec = clock_ext.tv_sec - time.tv_sec;
delta.tv_usec = clock_ext.tv_usec -
time.tv_usec;
if (delta.tv_usec < 0)
delta.tv_sec--;
if (delta.tv_usec >= 500000) {
delta.tv_usec -= 1000000;
delta.tv_sec++;
}
if (delta.tv_usec < -500000) {
delta.tv_usec += 1000000;
delta.tv_sec--;
}
if (delta.tv_sec > 0 || (delta.tv_sec == 0 &&
delta.tv_usec > MAXPHASE) ||
delta.tv_sec < -1 || (delta.tv_sec == -1 &&
delta.tv_usec < -MAXPHASE)) {
time = clock_ext;
delta.tv_sec = 0;
delta.tv_usec = 0;
}
#ifdef HIGHBALL
clock_cpu = delta.tv_usec;
#else /* HIGHBALL */
hardupdate(delta.tv_usec);
#endif /* HIGHBALL */
}
#endif /* EXT_CLOCK */
}
#endif /* NTP */
/*
* Update real-time timeout queue.
* Process callouts at a very low CPU priority, so we don't keep the
* relatively high clock interrupt priority any longer than necessary.
*/
if (callout_hardclock()) {
if (CLKF_BASEPRI(frame)) {
/*
* Save the overhead of a software interrupt;
* it will happen as soon as we return, so do
* it now.
*/
spllowersoftclock();
KERNEL_LOCK(LK_CANRECURSE|LK_EXCLUSIVE);
softclock(NULL);
KERNEL_UNLOCK();
} else {
#ifdef __HAVE_GENERIC_SOFT_INTERRUPTS
softintr_schedule(softclock_si);
#else
setsoftclock();
#endif
}
}
}
/*
* Compute number of hz until specified time. Used to compute second
* argument to callout_reset() from an absolute time.
*/
int
hzto(struct timeval *tv)
{
unsigned long ticks;
long sec, usec;
int s;
/*
* If the number of usecs in the whole seconds part of the time
* difference fits in a long, then the total number of usecs will
* fit in an unsigned long. Compute the total and convert it to
* ticks, rounding up and adding 1 to allow for the current tick
* to expire. Rounding also depends on unsigned long arithmetic
* to avoid overflow.
*
* Otherwise, if the number of ticks in the whole seconds part of
* the time difference fits in a long, then convert the parts to
* ticks separately and add, using similar rounding methods and
* overflow avoidance. This method would work in the previous
* case, but it is slightly slower and assume that hz is integral.
*
* Otherwise, round the time difference down to the maximum
* representable value.
*
* If ints are 32-bit, then the maximum value for any timeout in
* 10ms ticks is 248 days.
*/
s = splclock();
sec = tv->tv_sec - time.tv_sec;
usec = tv->tv_usec - time.tv_usec;
splx(s);
if (usec < 0) {
sec--;
usec += 1000000;
}
if (sec < 0 || (sec == 0 && usec <= 0)) {
/*
* Would expire now or in the past. Return 0 ticks.
* This is different from the legacy hzto() interface,
* and callers need to check for it.
*/
ticks = 0;
} else if (sec <= (LONG_MAX / 1000000))
ticks = (((sec * 1000000) + (unsigned long)usec + (tick - 1))
/ tick) + 1;
else if (sec <= (LONG_MAX / hz))
ticks = (sec * hz) +
(((unsigned long)usec + (tick - 1)) / tick) + 1;
else
ticks = LONG_MAX;
if (ticks > INT_MAX)
ticks = INT_MAX;
return ((int)ticks);
}
/*
* Start profiling on a process.
*
* Kernel profiling passes proc0 which never exits and hence
* keeps the profile clock running constantly.
*/
void
startprofclock(struct proc *p)
{
if ((p->p_flag & P_PROFIL) == 0) {
p->p_flag |= P_PROFIL;
/*
* This is only necessary if using the clock as the
* profiling source.
*/
if (++profprocs == 1 && stathz != 0)
psdiv = psratio;
}
}
/*
* Stop profiling on a process.
*/
void
stopprofclock(struct proc *p)
{
if (p->p_flag & P_PROFIL) {
p->p_flag &= ~P_PROFIL;
/*
* This is only necessary if using the clock as the
* profiling source.
*/
if (--profprocs == 0 && stathz != 0)
psdiv = 1;
}
}
#if defined(PERFCTRS)
/*
* Independent profiling "tick" in case we're using a separate
* clock or profiling event source. Currently, that's just
* performance counters--hence the wrapper.
*/
void
proftick(struct clockframe *frame)
{
#ifdef GPROF
struct gmonparam *g;
intptr_t i;
#endif
struct proc *p;
p = curproc;
if (CLKF_USERMODE(frame)) {
if (p->p_flag & P_PROFIL)
addupc_intr(p, CLKF_PC(frame));
} else {
#ifdef GPROF
g = &_gmonparam;
if (g->state == GMON_PROF_ON) {
i = CLKF_PC(frame) - g->lowpc;
if (i < g->textsize) {
i /= HISTFRACTION * sizeof(*g->kcount);
g->kcount[i]++;
}
}
#endif
#ifdef PROC_PC
if (p && p->p_flag & P_PROFIL)
addupc_intr(p, PROC_PC(p));
#endif
}
}
#endif
/*
* Statistics clock. Grab profile sample, and if divider reaches 0,
* do process and kernel statistics.
*/
void
statclock(struct clockframe *frame)
{
#ifdef GPROF
struct gmonparam *g;
intptr_t i;
#endif
struct cpu_info *ci = curcpu();
struct schedstate_percpu *spc = &ci->ci_schedstate;
struct lwp *l;
struct proc *p;
/*
* Notice changes in divisor frequency, and adjust clock
* frequency accordingly.
*/
if (spc->spc_psdiv != psdiv) {
spc->spc_psdiv = psdiv;
spc->spc_pscnt = psdiv;
if (psdiv == 1) {
setstatclockrate(stathz);
} else {
setstatclockrate(profhz);
}
}
l = curlwp;
p = (l ? l->l_proc : 0);
if (CLKF_USERMODE(frame)) {
if (p->p_flag & P_PROFIL && profsrc == PROFSRC_CLOCK)
addupc_intr(p, CLKF_PC(frame));
if (--spc->spc_pscnt > 0)
return;
/*
* Came from user mode; CPU was in user state.
* If this process is being profiled record the tick.
*/
p->p_uticks++;
if (p->p_nice > NZERO)
spc->spc_cp_time[CP_NICE]++;
else
spc->spc_cp_time[CP_USER]++;
} else {
#ifdef GPROF
/*
* Kernel statistics are just like addupc_intr, only easier.
*/
g = &_gmonparam;
if (profsrc == PROFSRC_CLOCK && g->state == GMON_PROF_ON) {
i = CLKF_PC(frame) - g->lowpc;
if (i < g->textsize) {
i /= HISTFRACTION * sizeof(*g->kcount);
g->kcount[i]++;
}
}
#endif
#ifdef LWP_PC
if (p && profsrc == PROFSRC_CLOCK && p->p_flag & P_PROFIL)
addupc_intr(p, LWP_PC(l));
#endif
if (--spc->spc_pscnt > 0)
return;
/*
* Came from kernel mode, so we were:
* - handling an interrupt,
* - doing syscall or trap work on behalf of the current
* user process, or
* - spinning in the idle loop.
* Whichever it is, charge the time as appropriate.
* Note that we charge interrupts to the current process,
* regardless of whether they are ``for'' that process,
* so that we know how much of its real time was spent
* in ``non-process'' (i.e., interrupt) work.
*/
if (CLKF_INTR(frame)) {
if (p != NULL)
p->p_iticks++;
spc->spc_cp_time[CP_INTR]++;
} else if (p != NULL) {
p->p_sticks++;
spc->spc_cp_time[CP_SYS]++;
} else
spc->spc_cp_time[CP_IDLE]++;
}
spc->spc_pscnt = psdiv;
if (l != NULL) {
++p->p_cpticks;
/*
* If no separate schedclock is provided, call it here
* at about 16 Hz.
*/
if (schedhz == 0)
if ((int)(--ci->ci_schedstate.spc_schedticks) <= 0) {
schedclock(l);
ci->ci_schedstate.spc_schedticks = statscheddiv;
}
}
}
#ifdef NTP /* NTP phase-locked loop in kernel */
/*
* hardupdate() - local clock update
*
* This routine is called by ntp_adjtime() to update the local clock
* phase and frequency. The implementation is of an adaptive-parameter,
* hybrid phase/frequency-lock loop (PLL/FLL). The routine computes new
* time and frequency offset estimates for each call. If the kernel PPS
* discipline code is configured (PPS_SYNC), the PPS signal itself
* determines the new time offset, instead of the calling argument.
* Presumably, calls to ntp_adjtime() occur only when the caller
* believes the local clock is valid within some bound (+-128 ms with
* NTP). If the caller's time is far different than the PPS time, an
* argument will ensue, and it's not clear who will lose.
*
* For uncompensated quartz crystal oscillatores and nominal update
* intervals less than 1024 s, operation should be in phase-lock mode
* (STA_FLL = 0), where the loop is disciplined to phase. For update
* intervals greater than thiss, operation should be in frequency-lock
* mode (STA_FLL = 1), where the loop is disciplined to frequency.
*
* Note: splclock() is in effect.
*/
void
hardupdate(long offset)
{
long ltemp, mtemp;
if (!(time_status & STA_PLL) && !(time_status & STA_PPSTIME))
return;
ltemp = offset;
#ifdef PPS_SYNC
if (time_status & STA_PPSTIME && time_status & STA_PPSSIGNAL)
ltemp = pps_offset;
#endif /* PPS_SYNC */
/*
* Scale the phase adjustment and clamp to the operating range.
*/
if (ltemp > MAXPHASE)
time_offset = MAXPHASE << SHIFT_UPDATE;
else if (ltemp < -MAXPHASE)
time_offset = -(MAXPHASE << SHIFT_UPDATE);
else
time_offset = ltemp << SHIFT_UPDATE;
/*
* Select whether the frequency is to be controlled and in which
* mode (PLL or FLL). Clamp to the operating range. Ugly
* multiply/divide should be replaced someday.
*/
if (time_status & STA_FREQHOLD || time_reftime == 0)
time_reftime = time.tv_sec;
mtemp = time.tv_sec - time_reftime;
time_reftime = time.tv_sec;
if (time_status & STA_FLL) {
if (mtemp >= MINSEC) {
ltemp = ((time_offset / mtemp) << (SHIFT_USEC -
SHIFT_UPDATE));
if (ltemp < 0)
time_freq -= -ltemp >> SHIFT_KH;
else
time_freq += ltemp >> SHIFT_KH;
}
} else {
if (mtemp < MAXSEC) {
ltemp *= mtemp;
if (ltemp < 0)
time_freq -= -ltemp >> (time_constant +
time_constant + SHIFT_KF -
SHIFT_USEC);
else
time_freq += ltemp >> (time_constant +
time_constant + SHIFT_KF -
SHIFT_USEC);
}
}
if (time_freq > time_tolerance)
time_freq = time_tolerance;
else if (time_freq < -time_tolerance)
time_freq = -time_tolerance;
}
#ifdef PPS_SYNC
/*
* hardpps() - discipline CPU clock oscillator to external PPS signal
*
* This routine is called at each PPS interrupt in order to discipline
* the CPU clock oscillator to the PPS signal. It measures the PPS phase
* and leaves it in a handy spot for the hardclock() routine. It
* integrates successive PPS phase differences and calculates the
* frequency offset. This is used in hardclock() to discipline the CPU
* clock oscillator so that intrinsic frequency error is cancelled out.
* The code requires the caller to capture the time and hardware counter
* value at the on-time PPS signal transition.
*
* Note that, on some Unix systems, this routine runs at an interrupt
* priority level higher than the timer interrupt routine hardclock().
* Therefore, the variables used are distinct from the hardclock()
* variables, except for certain exceptions: The PPS frequency pps_freq
* and phase pps_offset variables are determined by this routine and
* updated atomically. The time_tolerance variable can be considered a
* constant, since it is infrequently changed, and then only when the
* PPS signal is disabled. The watchdog counter pps_valid is updated
* once per second by hardclock() and is atomically cleared in this
* routine.
*/
void
hardpps(struct timeval *tvp, /* time at PPS */
long usec /* hardware counter at PPS */)
{
long u_usec, v_usec, bigtick;
long cal_sec, cal_usec;
/*
* An occasional glitch can be produced when the PPS interrupt
* occurs in the hardclock() routine before the time variable is
* updated. Here the offset is discarded when the difference
* between it and the last one is greater than tick/2, but not
* if the interval since the first discard exceeds 30 s.
*/
time_status |= STA_PPSSIGNAL;
time_status &= ~(STA_PPSJITTER | STA_PPSWANDER | STA_PPSERROR);
pps_valid = 0;
u_usec = -tvp->tv_usec;
if (u_usec < -500000)
u_usec += 1000000;
v_usec = pps_offset - u_usec;
if (v_usec < 0)
v_usec = -v_usec;
if (v_usec > (tick >> 1)) {
if (pps_glitch > MAXGLITCH) {
pps_glitch = 0;
pps_tf[2] = u_usec;
pps_tf[1] = u_usec;
} else {
pps_glitch++;
u_usec = pps_offset;
}
} else
pps_glitch = 0;
/*
* A three-stage median filter is used to help deglitch the pps
* time. The median sample becomes the time offset estimate; the
* difference between the other two samples becomes the time
* dispersion (jitter) estimate.
*/
pps_tf[2] = pps_tf[1];
pps_tf[1] = pps_tf[0];
pps_tf[0] = u_usec;
if (pps_tf[0] > pps_tf[1]) {
if (pps_tf[1] > pps_tf[2]) {
pps_offset = pps_tf[1]; /* 0 1 2 */
v_usec = pps_tf[0] - pps_tf[2];
} else if (pps_tf[2] > pps_tf[0]) {
pps_offset = pps_tf[0]; /* 2 0 1 */
v_usec = pps_tf[2] - pps_tf[1];
} else {
pps_offset = pps_tf[2]; /* 0 2 1 */
v_usec = pps_tf[0] - pps_tf[1];
}
} else {
if (pps_tf[1] < pps_tf[2]) {
pps_offset = pps_tf[1]; /* 2 1 0 */
v_usec = pps_tf[2] - pps_tf[0];
} else if (pps_tf[2] < pps_tf[0]) {
pps_offset = pps_tf[0]; /* 1 0 2 */
v_usec = pps_tf[1] - pps_tf[2];
} else {
pps_offset = pps_tf[2]; /* 1 2 0 */
v_usec = pps_tf[1] - pps_tf[0];
}
}
if (v_usec > MAXTIME)
pps_jitcnt++;
v_usec = (v_usec << PPS_AVG) - pps_jitter;
if (v_usec < 0)
pps_jitter -= -v_usec >> PPS_AVG;
else
pps_jitter += v_usec >> PPS_AVG;
if (pps_jitter > (MAXTIME >> 1))
time_status |= STA_PPSJITTER;
/*
* During the calibration interval adjust the starting time when
* the tick overflows. At the end of the interval compute the
* duration of the interval and the difference of the hardware
* counters at the beginning and end of the interval. This code
* is deliciously complicated by the fact valid differences may
* exceed the value of tick when using long calibration
* intervals and small ticks. Note that the counter can be
* greater than tick if caught at just the wrong instant, but
* the values returned and used here are correct.
*/
bigtick = (long)tick << SHIFT_USEC;
pps_usec -= pps_freq;
if (pps_usec >= bigtick)
pps_usec -= bigtick;
if (pps_usec < 0)
pps_usec += bigtick;
pps_time.tv_sec++;
pps_count++;
if (pps_count < (1 << pps_shift))
return;
pps_count = 0;
pps_calcnt++;
u_usec = usec << SHIFT_USEC;
v_usec = pps_usec - u_usec;
if (v_usec >= bigtick >> 1)
v_usec -= bigtick;
if (v_usec < -(bigtick >> 1))
v_usec += bigtick;
if (v_usec < 0)
v_usec = -(-v_usec >> pps_shift);
else
v_usec = v_usec >> pps_shift;
pps_usec = u_usec;
cal_sec = tvp->tv_sec;
cal_usec = tvp->tv_usec;
cal_sec -= pps_time.tv_sec;
cal_usec -= pps_time.tv_usec;
if (cal_usec < 0) {
cal_usec += 1000000;
cal_sec--;
}
pps_time = *tvp;
/*
* Check for lost interrupts, noise, excessive jitter and
* excessive frequency error. The number of timer ticks during
* the interval may vary +-1 tick. Add to this a margin of one
* tick for the PPS signal jitter and maximum frequency
* deviation. If the limits are exceeded, the calibration
* interval is reset to the minimum and we start over.
*/
u_usec = (long)tick << 1;
if (!((cal_sec == -1 && cal_usec > (1000000 - u_usec))
|| (cal_sec == 0 && cal_usec < u_usec))
|| v_usec > time_tolerance || v_usec < -time_tolerance) {
pps_errcnt++;
pps_shift = PPS_SHIFT;
pps_intcnt = 0;
time_status |= STA_PPSERROR;
return;
}
/*
* A three-stage median filter is used to help deglitch the pps
* frequency. The median sample becomes the frequency offset
* estimate; the difference between the other two samples
* becomes the frequency dispersion (stability) estimate.
*/
pps_ff[2] = pps_ff[1];
pps_ff[1] = pps_ff[0];
pps_ff[0] = v_usec;
if (pps_ff[0] > pps_ff[1]) {
if (pps_ff[1] > pps_ff[2]) {
u_usec = pps_ff[1]; /* 0 1 2 */
v_usec = pps_ff[0] - pps_ff[2];
} else if (pps_ff[2] > pps_ff[0]) {
u_usec = pps_ff[0]; /* 2 0 1 */
v_usec = pps_ff[2] - pps_ff[1];
} else {
u_usec = pps_ff[2]; /* 0 2 1 */
v_usec = pps_ff[0] - pps_ff[1];
}
} else {
if (pps_ff[1] < pps_ff[2]) {
u_usec = pps_ff[1]; /* 2 1 0 */
v_usec = pps_ff[2] - pps_ff[0];
} else if (pps_ff[2] < pps_ff[0]) {
u_usec = pps_ff[0]; /* 1 0 2 */
v_usec = pps_ff[1] - pps_ff[2];
} else {
u_usec = pps_ff[2]; /* 1 2 0 */
v_usec = pps_ff[1] - pps_ff[0];
}
}
/*
* Here the frequency dispersion (stability) is updated. If it
* is less than one-fourth the maximum (MAXFREQ), the frequency
* offset is updated as well, but clamped to the tolerance. It
* will be processed later by the hardclock() routine.
*/
v_usec = (v_usec >> 1) - pps_stabil;
if (v_usec < 0)
pps_stabil -= -v_usec >> PPS_AVG;
else
pps_stabil += v_usec >> PPS_AVG;
if (pps_stabil > MAXFREQ >> 2) {
pps_stbcnt++;
time_status |= STA_PPSWANDER;
return;
}
if (time_status & STA_PPSFREQ) {
if (u_usec < 0) {
pps_freq -= -u_usec >> PPS_AVG;
if (pps_freq < -time_tolerance)
pps_freq = -time_tolerance;
u_usec = -u_usec;
} else {
pps_freq += u_usec >> PPS_AVG;
if (pps_freq > time_tolerance)
pps_freq = time_tolerance;
}
}
/*
* Here the calibration interval is adjusted. If the maximum
* time difference is greater than tick / 4, reduce the interval
* by half. If this is not the case for four consecutive
* intervals, double the interval.
*/
if (u_usec << pps_shift > bigtick >> 2) {
pps_intcnt = 0;
if (pps_shift > PPS_SHIFT)
pps_shift--;
} else if (pps_intcnt >= 4) {
pps_intcnt = 0;
if (pps_shift < PPS_SHIFTMAX)
pps_shift++;
} else
pps_intcnt++;
}
#endif /* PPS_SYNC */
#endif /* NTP */