2007-02-16 17:28:19 +08:00
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/clocksource.h>
|
|
|
|
#include <linux/time.h>
|
|
|
|
#include <linux/acpi.h>
|
|
|
|
#include <linux/cpufreq.h>
|
2007-10-13 05:04:06 +08:00
|
|
|
#include <linux/acpi_pmtmr.h>
|
2007-02-16 17:28:19 +08:00
|
|
|
|
2007-10-13 05:04:06 +08:00
|
|
|
#include <asm/hpet.h>
|
2007-02-16 17:28:19 +08:00
|
|
|
#include <asm/timex.h>
|
2008-01-30 20:30:06 +08:00
|
|
|
#include <asm/timer.h>
|
2008-04-02 01:45:18 +08:00
|
|
|
#include <asm/vgtod.h>
|
2007-02-16 17:28:19 +08:00
|
|
|
|
2007-02-16 17:28:20 +08:00
|
|
|
static int notsc __initdata = 0;
|
2007-02-16 17:28:19 +08:00
|
|
|
|
|
|
|
unsigned int cpu_khz; /* TSC clocks / usec, not used here */
|
|
|
|
EXPORT_SYMBOL(cpu_khz);
|
2007-05-03 01:27:06 +08:00
|
|
|
unsigned int tsc_khz;
|
|
|
|
EXPORT_SYMBOL(tsc_khz);
|
2007-02-16 17:28:19 +08:00
|
|
|
|
2008-01-30 20:30:06 +08:00
|
|
|
/* Accelerators for sched_clock()
|
|
|
|
* convert from cycles(64bits) => nanoseconds (64bits)
|
|
|
|
* basic equation:
|
|
|
|
* ns = cycles / (freq / ns_per_sec)
|
|
|
|
* ns = cycles * (ns_per_sec / freq)
|
|
|
|
* ns = cycles * (10^9 / (cpu_khz * 10^3))
|
|
|
|
* ns = cycles * (10^6 / cpu_khz)
|
|
|
|
*
|
|
|
|
* Then we use scaling math (suggested by george@mvista.com) to get:
|
|
|
|
* ns = cycles * (10^6 * SC / cpu_khz) / SC
|
|
|
|
* ns = cycles * cyc2ns_scale / SC
|
|
|
|
*
|
|
|
|
* And since SC is a constant power of two, we can convert the div
|
|
|
|
* into a shift.
|
|
|
|
*
|
|
|
|
* We can use khz divisor instead of mhz to keep a better precision, since
|
|
|
|
* cyc2ns_scale is limited to 10^6 * 2^10, which fits in 32 bits.
|
|
|
|
* (mathieu.desnoyers@polymtl.ca)
|
|
|
|
*
|
|
|
|
* -johnstul@us.ibm.com "math is hard, lets go shopping!"
|
|
|
|
*/
|
|
|
|
DEFINE_PER_CPU(unsigned long, cyc2ns);
|
2007-02-16 17:28:19 +08:00
|
|
|
|
2008-01-30 20:30:06 +08:00
|
|
|
static void set_cyc2ns_scale(unsigned long cpu_khz, int cpu)
|
2007-02-16 17:28:19 +08:00
|
|
|
{
|
2008-01-30 20:30:06 +08:00
|
|
|
unsigned long long tsc_now, ns_now;
|
2008-03-31 20:52:15 +08:00
|
|
|
unsigned long flags, *scale;
|
2007-02-16 17:28:19 +08:00
|
|
|
|
2008-01-30 20:30:06 +08:00
|
|
|
local_irq_save(flags);
|
|
|
|
sched_clock_idle_sleep_event();
|
|
|
|
|
|
|
|
scale = &per_cpu(cyc2ns, cpu);
|
|
|
|
|
|
|
|
rdtscll(tsc_now);
|
|
|
|
ns_now = __cycles_2_ns(tsc_now);
|
|
|
|
|
|
|
|
if (cpu_khz)
|
|
|
|
*scale = (NSEC_PER_MSEC << CYC2NS_SCALE_FACTOR)/cpu_khz;
|
|
|
|
|
|
|
|
sched_clock_idle_wakeup_event(0);
|
|
|
|
local_irq_restore(flags);
|
2007-02-16 17:28:19 +08:00
|
|
|
}
|
|
|
|
|
2008-01-30 20:31:03 +08:00
|
|
|
unsigned long long native_sched_clock(void)
|
2007-02-16 17:28:19 +08:00
|
|
|
{
|
|
|
|
unsigned long a = 0;
|
|
|
|
|
|
|
|
/* Could do CPU core sync here. Opteron can execute rdtsc speculatively,
|
|
|
|
* which means it is not completely exact and may not be monotonous
|
|
|
|
* between CPUs. But the errors should be too small to matter for
|
|
|
|
* scheduling purposes.
|
|
|
|
*/
|
|
|
|
|
|
|
|
rdtscll(a);
|
|
|
|
return cycles_2_ns(a);
|
|
|
|
}
|
|
|
|
|
2008-01-30 20:31:03 +08:00
|
|
|
/* We need to define a real function for sched_clock, to override the
|
|
|
|
weak default version */
|
|
|
|
#ifdef CONFIG_PARAVIRT
|
|
|
|
unsigned long long sched_clock(void)
|
|
|
|
{
|
|
|
|
return paravirt_sched_clock();
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
unsigned long long
|
|
|
|
sched_clock(void) __attribute__((alias("native_sched_clock")));
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
2007-02-16 17:28:20 +08:00
|
|
|
static int tsc_unstable;
|
|
|
|
|
2008-01-30 20:33:24 +08:00
|
|
|
int check_tsc_unstable(void)
|
2007-02-16 17:28:20 +08:00
|
|
|
{
|
|
|
|
return tsc_unstable;
|
|
|
|
}
|
2008-01-30 20:33:24 +08:00
|
|
|
EXPORT_SYMBOL_GPL(check_tsc_unstable);
|
|
|
|
|
2007-02-16 17:28:19 +08:00
|
|
|
#ifdef CONFIG_CPU_FREQ
|
|
|
|
|
|
|
|
/* Frequency scaling support. Adjust the TSC based timer when the cpu frequency
|
|
|
|
* changes.
|
|
|
|
*
|
|
|
|
* RED-PEN: On SMP we assume all CPUs run with the same frequency. It's
|
|
|
|
* not that important because current Opteron setups do not support
|
|
|
|
* scaling on SMP anyroads.
|
|
|
|
*
|
|
|
|
* Should fix up last_tsc too. Currently gettimeofday in the
|
|
|
|
* first tick after the change will be slightly wrong.
|
|
|
|
*/
|
|
|
|
|
2007-07-21 23:10:13 +08:00
|
|
|
static unsigned int ref_freq;
|
|
|
|
static unsigned long loops_per_jiffy_ref;
|
|
|
|
static unsigned long tsc_khz_ref;
|
2007-02-16 17:28:19 +08:00
|
|
|
|
|
|
|
static int time_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
struct cpufreq_freqs *freq = data;
|
|
|
|
unsigned long *lpj, dummy;
|
|
|
|
|
2007-10-20 02:35:04 +08:00
|
|
|
if (cpu_has(&cpu_data(freq->cpu), X86_FEATURE_CONSTANT_TSC))
|
2007-02-16 17:28:19 +08:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
lpj = &dummy;
|
|
|
|
if (!(freq->flags & CPUFREQ_CONST_LOOPS))
|
|
|
|
#ifdef CONFIG_SMP
|
2007-10-20 02:35:04 +08:00
|
|
|
lpj = &cpu_data(freq->cpu).loops_per_jiffy;
|
2007-02-16 17:28:19 +08:00
|
|
|
#else
|
|
|
|
lpj = &boot_cpu_data.loops_per_jiffy;
|
|
|
|
#endif
|
|
|
|
|
|
|
|
if (!ref_freq) {
|
|
|
|
ref_freq = freq->old;
|
|
|
|
loops_per_jiffy_ref = *lpj;
|
2007-05-03 01:27:06 +08:00
|
|
|
tsc_khz_ref = tsc_khz;
|
2007-02-16 17:28:19 +08:00
|
|
|
}
|
|
|
|
if ((val == CPUFREQ_PRECHANGE && freq->old < freq->new) ||
|
|
|
|
(val == CPUFREQ_POSTCHANGE && freq->old > freq->new) ||
|
|
|
|
(val == CPUFREQ_RESUMECHANGE)) {
|
|
|
|
*lpj =
|
|
|
|
cpufreq_scale(loops_per_jiffy_ref, ref_freq, freq->new);
|
|
|
|
|
2007-05-03 01:27:06 +08:00
|
|
|
tsc_khz = cpufreq_scale(tsc_khz_ref, ref_freq, freq->new);
|
2007-02-16 17:28:19 +08:00
|
|
|
if (!(freq->flags & CPUFREQ_CONST_LOOPS))
|
2007-05-03 01:27:08 +08:00
|
|
|
mark_tsc_unstable("cpufreq changes");
|
2007-02-16 17:28:19 +08:00
|
|
|
}
|
|
|
|
|
2008-04-07 18:14:45 +08:00
|
|
|
set_cyc2ns_scale(tsc_khz_ref, freq->cpu);
|
2007-02-16 17:28:19 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct notifier_block time_cpufreq_notifier_block = {
|
|
|
|
.notifier_call = time_cpufreq_notifier
|
|
|
|
};
|
|
|
|
|
|
|
|
static int __init cpufreq_tsc(void)
|
|
|
|
{
|
2007-07-21 23:10:13 +08:00
|
|
|
cpufreq_register_notifier(&time_cpufreq_notifier_block,
|
|
|
|
CPUFREQ_TRANSITION_NOTIFIER);
|
2007-02-16 17:28:19 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
core_initcall(cpufreq_tsc);
|
|
|
|
|
|
|
|
#endif
|
|
|
|
|
2007-10-13 05:04:06 +08:00
|
|
|
#define MAX_RETRIES 5
|
|
|
|
#define SMI_TRESHOLD 50000
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Read TSC and the reference counters. Take care of SMI disturbance
|
|
|
|
*/
|
|
|
|
static unsigned long __init tsc_read_refs(unsigned long *pm,
|
|
|
|
unsigned long *hpet)
|
|
|
|
{
|
|
|
|
unsigned long t1, t2;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < MAX_RETRIES; i++) {
|
2008-01-30 20:32:39 +08:00
|
|
|
t1 = get_cycles();
|
2007-10-13 05:04:06 +08:00
|
|
|
if (hpet)
|
|
|
|
*hpet = hpet_readl(HPET_COUNTER) & 0xFFFFFFFF;
|
|
|
|
else
|
|
|
|
*pm = acpi_pm_read_early();
|
2008-01-30 20:32:39 +08:00
|
|
|
t2 = get_cycles();
|
2007-10-13 05:04:06 +08:00
|
|
|
if ((t2 - t1) < SMI_TRESHOLD)
|
|
|
|
return t2;
|
|
|
|
}
|
|
|
|
return ULONG_MAX;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* tsc_calibrate - calibrate the tsc on boot
|
|
|
|
*/
|
|
|
|
void __init tsc_calibrate(void)
|
|
|
|
{
|
|
|
|
unsigned long flags, tsc1, tsc2, tr1, tr2, pm1, pm2, hpet1, hpet2;
|
2008-01-30 20:30:06 +08:00
|
|
|
int hpet = is_hpet_enabled(), cpu;
|
2007-10-13 05:04:06 +08:00
|
|
|
|
|
|
|
local_irq_save(flags);
|
|
|
|
|
|
|
|
tsc1 = tsc_read_refs(&pm1, hpet ? &hpet1 : NULL);
|
|
|
|
|
|
|
|
outb((inb(0x61) & ~0x02) | 0x01, 0x61);
|
|
|
|
|
|
|
|
outb(0xb0, 0x43);
|
|
|
|
outb((CLOCK_TICK_RATE / (1000 / 50)) & 0xff, 0x42);
|
|
|
|
outb((CLOCK_TICK_RATE / (1000 / 50)) >> 8, 0x42);
|
2008-01-30 20:32:39 +08:00
|
|
|
tr1 = get_cycles();
|
2007-10-13 05:04:06 +08:00
|
|
|
while ((inb(0x61) & 0x20) == 0);
|
2008-01-30 20:32:39 +08:00
|
|
|
tr2 = get_cycles();
|
2007-10-13 05:04:06 +08:00
|
|
|
|
|
|
|
tsc2 = tsc_read_refs(&pm2, hpet ? &hpet2 : NULL);
|
|
|
|
|
|
|
|
local_irq_restore(flags);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Preset the result with the raw and inaccurate PIT
|
|
|
|
* calibration value
|
|
|
|
*/
|
|
|
|
tsc_khz = (tr2 - tr1) / 50;
|
|
|
|
|
|
|
|
/* hpet or pmtimer available ? */
|
|
|
|
if (!hpet && !pm1 && !pm2) {
|
|
|
|
printk(KERN_INFO "TSC calibrated against PIT\n");
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Check, whether the sampling was disturbed by an SMI */
|
|
|
|
if (tsc1 == ULONG_MAX || tsc2 == ULONG_MAX) {
|
|
|
|
printk(KERN_WARNING "TSC calibration disturbed by SMI, "
|
|
|
|
"using PIT calibration result\n");
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
tsc2 = (tsc2 - tsc1) * 1000000L;
|
|
|
|
|
|
|
|
if (hpet) {
|
|
|
|
printk(KERN_INFO "TSC calibrated against HPET\n");
|
|
|
|
if (hpet2 < hpet1)
|
|
|
|
hpet2 += 0x100000000;
|
|
|
|
hpet2 -= hpet1;
|
|
|
|
tsc1 = (hpet2 * hpet_readl(HPET_PERIOD)) / 1000000;
|
|
|
|
} else {
|
|
|
|
printk(KERN_INFO "TSC calibrated against PM_TIMER\n");
|
|
|
|
if (pm2 < pm1)
|
|
|
|
pm2 += ACPI_PM_OVRRUN;
|
|
|
|
pm2 -= pm1;
|
|
|
|
tsc1 = (pm2 * 1000000000) / PMTMR_TICKS_PER_SEC;
|
|
|
|
}
|
|
|
|
|
|
|
|
tsc_khz = tsc2 / tsc1;
|
2008-01-30 20:30:06 +08:00
|
|
|
|
|
|
|
for_each_possible_cpu(cpu)
|
|
|
|
set_cyc2ns_scale(tsc_khz, cpu);
|
2007-10-13 05:04:06 +08:00
|
|
|
}
|
|
|
|
|
2007-02-16 17:28:19 +08:00
|
|
|
/*
|
|
|
|
* Make an educated guess if the TSC is trustworthy and synchronized
|
|
|
|
* over all CPUs.
|
|
|
|
*/
|
|
|
|
__cpuinit int unsynchronized_tsc(void)
|
|
|
|
{
|
|
|
|
if (tsc_unstable)
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
#ifdef CONFIG_SMP
|
|
|
|
if (apic_is_clustered_box())
|
|
|
|
return 1;
|
|
|
|
#endif
|
2008-01-30 20:32:40 +08:00
|
|
|
|
2008-01-30 20:32:41 +08:00
|
|
|
if (boot_cpu_has(X86_FEATURE_CONSTANT_TSC))
|
2007-07-21 23:10:13 +08:00
|
|
|
return 0;
|
2007-02-16 17:28:19 +08:00
|
|
|
|
2007-07-21 23:10:13 +08:00
|
|
|
/* Assume multi socket systems are not synchronized */
|
|
|
|
return num_present_cpus() > 1;
|
2007-02-16 17:28:19 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
int __init notsc_setup(char *s)
|
|
|
|
{
|
|
|
|
notsc = 1;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
__setup("notsc", notsc_setup);
|
2007-02-16 17:28:20 +08:00
|
|
|
|
2008-04-02 01:45:18 +08:00
|
|
|
static struct clocksource clocksource_tsc;
|
2007-02-16 17:28:20 +08:00
|
|
|
|
2008-04-02 01:45:18 +08:00
|
|
|
/*
|
|
|
|
* We compare the TSC to the cycle_last value in the clocksource
|
|
|
|
* structure to avoid a nasty time-warp. This can be observed in a
|
|
|
|
* very small window right after one CPU updated cycle_last under
|
|
|
|
* xtime/vsyscall_gtod lock and the other CPU reads a TSC value which
|
|
|
|
* is smaller than the cycle_last reference value due to a TSC which
|
|
|
|
* is slighty behind. This delta is nowhere else observable, but in
|
|
|
|
* that case it results in a forward time jump in the range of hours
|
|
|
|
* due to the unsigned delta calculation of the time keeping core
|
|
|
|
* code, which is necessary to support wrapping clocksources like pm
|
|
|
|
* timer.
|
|
|
|
*/
|
2007-02-16 17:28:20 +08:00
|
|
|
static cycle_t read_tsc(void)
|
|
|
|
{
|
2008-01-30 20:32:39 +08:00
|
|
|
cycle_t ret = (cycle_t)get_cycles();
|
2008-04-02 01:45:18 +08:00
|
|
|
|
|
|
|
return ret >= clocksource_tsc.cycle_last ?
|
|
|
|
ret : clocksource_tsc.cycle_last;
|
2007-02-16 17:28:20 +08:00
|
|
|
}
|
|
|
|
|
2007-02-16 17:28:21 +08:00
|
|
|
static cycle_t __vsyscall_fn vread_tsc(void)
|
|
|
|
{
|
2008-01-30 20:32:39 +08:00
|
|
|
cycle_t ret = (cycle_t)vget_cycles();
|
2008-04-02 01:45:18 +08:00
|
|
|
|
|
|
|
return ret >= __vsyscall_gtod_data.clock.cycle_last ?
|
|
|
|
ret : __vsyscall_gtod_data.clock.cycle_last;
|
2007-02-16 17:28:21 +08:00
|
|
|
}
|
|
|
|
|
2007-02-16 17:28:20 +08:00
|
|
|
static struct clocksource clocksource_tsc = {
|
|
|
|
.name = "tsc",
|
|
|
|
.rating = 300,
|
|
|
|
.read = read_tsc,
|
|
|
|
.mask = CLOCKSOURCE_MASK(64),
|
|
|
|
.shift = 22,
|
|
|
|
.flags = CLOCK_SOURCE_IS_CONTINUOUS |
|
|
|
|
CLOCK_SOURCE_MUST_VERIFY,
|
2007-02-16 17:28:21 +08:00
|
|
|
.vread = vread_tsc,
|
2007-02-16 17:28:20 +08:00
|
|
|
};
|
|
|
|
|
2007-05-03 01:27:08 +08:00
|
|
|
void mark_tsc_unstable(char *reason)
|
2007-02-16 17:28:20 +08:00
|
|
|
{
|
|
|
|
if (!tsc_unstable) {
|
|
|
|
tsc_unstable = 1;
|
2007-05-03 01:27:08 +08:00
|
|
|
printk("Marking TSC unstable due to %s\n", reason);
|
2007-02-16 17:28:20 +08:00
|
|
|
/* Change only the rating, when not registered */
|
|
|
|
if (clocksource_tsc.mult)
|
|
|
|
clocksource_change_rating(&clocksource_tsc, 0);
|
|
|
|
else
|
|
|
|
clocksource_tsc.rating = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(mark_tsc_unstable);
|
|
|
|
|
2007-03-05 16:30:50 +08:00
|
|
|
void __init init_tsc_clocksource(void)
|
2007-02-16 17:28:20 +08:00
|
|
|
{
|
|
|
|
if (!notsc) {
|
2007-05-03 01:27:06 +08:00
|
|
|
clocksource_tsc.mult = clocksource_khz2mult(tsc_khz,
|
2007-02-16 17:28:20 +08:00
|
|
|
clocksource_tsc.shift);
|
|
|
|
if (check_tsc_unstable())
|
|
|
|
clocksource_tsc.rating = 0;
|
|
|
|
|
2007-03-05 16:30:50 +08:00
|
|
|
clocksource_register(&clocksource_tsc);
|
2007-02-16 17:28:20 +08:00
|
|
|
}
|
|
|
|
}
|