android_kernel_samsung_msm8976/arch/i386/kernel/timers/timer_hpet.c

196 lines
5.4 KiB
C
Raw Normal View History

/*
* This code largely moved from arch/i386/kernel/time.c.
* See comments there for proper credits.
*/
#include <linux/spinlock.h>
#include <linux/init.h>
#include <linux/timex.h>
#include <linux/errno.h>
#include <linux/string.h>
#include <linux/jiffies.h>
#include <asm/timer.h>
#include <asm/io.h>
#include <asm/processor.h>
#include "io_ports.h"
#include "mach_timer.h"
#include <asm/hpet.h>
static unsigned long hpet_usec_quotient; /* convert hpet clks to usec */
static unsigned long tsc_hpet_quotient; /* convert tsc to hpet clks */
static unsigned long hpet_last; /* hpet counter value at last tick*/
static unsigned long last_tsc_low; /* lsb 32 bits of Time Stamp Counter */
static unsigned long last_tsc_high; /* msb 32 bits of Time Stamp Counter */
static unsigned long long monotonic_base;
static seqlock_t monotonic_lock = SEQLOCK_UNLOCKED;
/* 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_mhz * 10^6))
* ns = cycles * (10^3 / cpu_mhz)
*
* Then we use scaling math (suggested by george@mvista.com) to get:
* ns = cycles * (10^3 * SC / cpu_mhz) / SC
* ns = cycles * cyc2ns_scale / SC
*
* And since SC is a constant power of two, we can convert the div
* into a shift.
* -johnstul@us.ibm.com "math is hard, lets go shopping!"
*/
static unsigned long cyc2ns_scale;
#define CYC2NS_SCALE_FACTOR 10 /* 2^10, carefully chosen */
static inline void set_cyc2ns_scale(unsigned long cpu_mhz)
{
cyc2ns_scale = (1000 << CYC2NS_SCALE_FACTOR)/cpu_mhz;
}
static inline unsigned long long cycles_2_ns(unsigned long long cyc)
{
return (cyc * cyc2ns_scale) >> CYC2NS_SCALE_FACTOR;
}
static unsigned long long monotonic_clock_hpet(void)
{
unsigned long long last_offset, this_offset, base;
unsigned seq;
/* atomically read monotonic base & last_offset */
do {
seq = read_seqbegin(&monotonic_lock);
last_offset = ((unsigned long long)last_tsc_high<<32)|last_tsc_low;
base = monotonic_base;
} while (read_seqretry(&monotonic_lock, seq));
/* Read the Time Stamp Counter */
rdtscll(this_offset);
/* return the value in ns */
return base + cycles_2_ns(this_offset - last_offset);
}
static unsigned long get_offset_hpet(void)
{
register unsigned long eax, edx;
eax = hpet_readl(HPET_COUNTER);
eax -= hpet_last; /* hpet delta */
eax = min(hpet_tick, eax);
/*
* Time offset = (hpet delta) * ( usecs per HPET clock )
* = (hpet delta) * ( usecs per tick / HPET clocks per tick)
* = (hpet delta) * ( hpet_usec_quotient ) / (2^32)
*
* Where,
* hpet_usec_quotient = (2^32 * usecs per tick)/HPET clocks per tick
*
* Using a mull instead of a divl saves some cycles in critical path.
*/
ASM_MUL64_REG(eax, edx, hpet_usec_quotient, eax);
/* our adjusted time offset in microseconds */
return edx;
}
static void mark_offset_hpet(void)
{
unsigned long long this_offset, last_offset;
unsigned long offset;
write_seqlock(&monotonic_lock);
last_offset = ((unsigned long long)last_tsc_high<<32)|last_tsc_low;
rdtsc(last_tsc_low, last_tsc_high);
if (hpet_use_timer)
offset = hpet_readl(HPET_T0_CMP) - hpet_tick;
else
offset = hpet_readl(HPET_COUNTER);
if (unlikely(((offset - hpet_last) >= (2*hpet_tick)) && (hpet_last != 0))) {
int lost_ticks = ((offset - hpet_last) / hpet_tick) - 1;
jiffies_64 += lost_ticks;
}
hpet_last = offset;
/* update the monotonic base value */
this_offset = ((unsigned long long)last_tsc_high<<32)|last_tsc_low;
monotonic_base += cycles_2_ns(this_offset - last_offset);
write_sequnlock(&monotonic_lock);
}
static void delay_hpet(unsigned long loops)
{
unsigned long hpet_start, hpet_end;
unsigned long eax;
/* loops is the number of cpu cycles. Convert it to hpet clocks */
ASM_MUL64_REG(eax, loops, tsc_hpet_quotient, loops);
hpet_start = hpet_readl(HPET_COUNTER);
do {
rep_nop();
hpet_end = hpet_readl(HPET_COUNTER);
} while ((hpet_end - hpet_start) < (loops));
}
static int __init init_hpet(char* override)
{
unsigned long result, remain;
/* check clock override */
if (override[0] && strncmp(override,"hpet",4))
return -ENODEV;
if (!is_hpet_enabled())
return -ENODEV;
printk("Using HPET for gettimeofday\n");
if (cpu_has_tsc) {
unsigned long tsc_quotient = calibrate_tsc_hpet(&tsc_hpet_quotient);
if (tsc_quotient) {
/* report CPU clock rate in Hz.
* The formula is (10^6 * 2^32) / (2^32 * 1 / (clocks/us)) =
* clock/second. Our precision is about 100 ppm.
*/
{ unsigned long eax=0, edx=1000;
ASM_DIV64_REG(cpu_khz, edx, tsc_quotient,
eax, edx);
printk("Detected %lu.%03lu MHz processor.\n",
cpu_khz / 1000, cpu_khz % 1000);
}
set_cyc2ns_scale(cpu_khz/1000);
}
}
/*
* Math to calculate hpet to usec multiplier
* Look for the comments at get_offset_hpet()
*/
ASM_DIV64_REG(result, remain, hpet_tick, 0, KERNEL_TICK_USEC);
if (remain > (hpet_tick >> 1))
result++; /* rounding the result */
hpet_usec_quotient = result;
return 0;
}
/************************************************************/
/* tsc timer_opts struct */
static struct timer_opts timer_hpet = {
.name = "hpet",
.mark_offset = mark_offset_hpet,
.get_offset = get_offset_hpet,
.monotonic_clock = monotonic_clock_hpet,
.delay = delay_hpet,
[PATCH] Platform SMIs and their interferance with tsc based delay calibration Issue: Current tsc based delay_calibration can result in significant errors in loops_per_jiffy count when the platform events like SMIs (System Management Interrupts that are non-maskable) are present. This could lead to potential kernel panic(). This issue is becoming more visible with 2.6 kernel (as default HZ is 1000) and on platforms with higher SMI handling latencies. During the boot time, SMIs are mostly used by BIOS (for things like legacy keyboard emulation). Description: The psuedocode for current delay calibration with tsc based delay looks like (0) Estimate a value for loops_per_jiffy (1) While (loops_per_jiffy estimate is accurate enough) (2) wait for jiffy transition (jiffy1) (3) Note down current tsc (tsc1) (4) loop until tsc becomes tsc1 + loops_per_jiffy (5) check whether jiffy changed since jiffy1 or not and refine loops_per_jiffy estimate Consider the following cases Case 1: If SMIs happen between (2) and (3) above, we can end up with a loops_per_jiffy value that is too low. This results in shorted delays and kernel can panic () during boot (Mostly at IOAPIC timer initialization timer_irq_works() as we don't have enough timer interrupts in a specified interval). Case 2: If SMIs happen between (3) and (4) above, then we can end up with a loops_per_jiffy value that is too high. And with current i386 code, too high lpj value (greater than 17M) can result in a overflow in delay.c:__const_udelay() again resulting in shorter delay and panic(). Solution: The patch below makes the calibration routine aware of asynchronous events like SMIs. We increase the delay calibration time and also identify any significant errors (greater than 12.5%) in the calibration and notify it to user. Patch below changes both i386 and x86-64 architectures to use this new and improved calibrate_delay_direct() routine. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 07:08:13 +00:00
.read_timer = read_timer_tsc,
};
struct init_timer_opts __initdata timer_hpet_init = {
.init = init_hpet,
.opts = &timer_hpet,
};