OpenCloudOS-Kernel/arch/arm/mach-davinci/time.c

462 lines
12 KiB
C
Raw Normal View History

/*
* DaVinci timer subsystem
*
* Author: Kevin Hilman, MontaVista Software, Inc. <source@mvista.com>
*
* 2007 (c) MontaVista Software, Inc. This file is licensed under
* the terms of the GNU General Public License version 2. This program
* is licensed "as is" without any warranty of any kind, whether express
* or implied.
*/
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/types.h>
#include <linux/interrupt.h>
#include <linux/clocksource.h>
#include <linux/clockchips.h>
#include <linux/io.h>
#include <linux/clk.h>
#include <linux/err.h>
#include <linux/platform_device.h>
#include <linux/sched_clock.h>
#include <asm/mach/irq.h>
#include <asm/mach/time.h>
#include <mach/cputype.h>
#include <mach/hardware.h>
#include <mach/time.h>
#include "clock.h"
static struct clock_event_device clockevent_davinci;
static unsigned int davinci_clock_tick_rate;
/*
* This driver configures the 2 64-bit count-up timers as 4 independent
* 32-bit count-up timers used as follows:
*/
enum {
TID_CLOCKEVENT,
TID_CLOCKSOURCE,
};
/* Timer register offsets */
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
#define PID12 0x0
#define TIM12 0x10
#define TIM34 0x14
#define PRD12 0x18
#define PRD34 0x1c
#define TCR 0x20
#define TGCR 0x24
#define WDTCR 0x28
/* Offsets of the 8 compare registers */
#define CMP12_0 0x60
#define CMP12_1 0x64
#define CMP12_2 0x68
#define CMP12_3 0x6c
#define CMP12_4 0x70
#define CMP12_5 0x74
#define CMP12_6 0x78
#define CMP12_7 0x7c
/* Timer register bitfields */
#define TCR_ENAMODE_DISABLE 0x0
#define TCR_ENAMODE_ONESHOT 0x1
#define TCR_ENAMODE_PERIODIC 0x2
#define TCR_ENAMODE_MASK 0x3
#define TGCR_TIMMODE_SHIFT 2
#define TGCR_TIMMODE_64BIT_GP 0x0
#define TGCR_TIMMODE_32BIT_UNCHAINED 0x1
#define TGCR_TIMMODE_64BIT_WDOG 0x2
#define TGCR_TIMMODE_32BIT_CHAINED 0x3
#define TGCR_TIM12RS_SHIFT 0
#define TGCR_TIM34RS_SHIFT 1
#define TGCR_RESET 0x0
#define TGCR_UNRESET 0x1
#define TGCR_RESET_MASK 0x3
#define WDTCR_WDEN_SHIFT 14
#define WDTCR_WDEN_DISABLE 0x0
#define WDTCR_WDEN_ENABLE 0x1
#define WDTCR_WDKEY_SHIFT 16
#define WDTCR_WDKEY_SEQ0 0xa5c6
#define WDTCR_WDKEY_SEQ1 0xda7e
struct timer_s {
char *name;
unsigned int id;
unsigned long period;
unsigned long opts;
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
unsigned long flags;
void __iomem *base;
unsigned long tim_off;
unsigned long prd_off;
unsigned long enamode_shift;
struct irqaction irqaction;
};
static struct timer_s timers[];
/* values for 'opts' field of struct timer_s */
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
#define TIMER_OPTS_DISABLED 0x01
#define TIMER_OPTS_ONESHOT 0x02
#define TIMER_OPTS_PERIODIC 0x04
#define TIMER_OPTS_STATE_MASK 0x07
#define TIMER_OPTS_USE_COMPARE 0x80000000
#define USING_COMPARE(t) ((t)->opts & TIMER_OPTS_USE_COMPARE)
static char *id_to_name[] = {
[T0_BOT] = "timer0_0",
[T0_TOP] = "timer0_1",
[T1_BOT] = "timer1_0",
[T1_TOP] = "timer1_1",
};
static int timer32_config(struct timer_s *t)
{
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
u32 tcr;
struct davinci_soc_info *soc_info = &davinci_soc_info;
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
if (USING_COMPARE(t)) {
struct davinci_timer_instance *dtip =
soc_info->timer_info->timers;
int event_timer = ID_TO_TIMER(timers[TID_CLOCKEVENT].id);
/*
* Next interrupt should be the current time reg value plus
* the new period (using 32-bit unsigned addition/wrapping
* to 0 on overflow). This assumes that the clocksource
* is setup to count to 2^32-1 before wrapping around to 0.
*/
__raw_writel(__raw_readl(t->base + t->tim_off) + t->period,
t->base + dtip[event_timer].cmp_off);
} else {
tcr = __raw_readl(t->base + TCR);
/* disable timer */
tcr &= ~(TCR_ENAMODE_MASK << t->enamode_shift);
__raw_writel(tcr, t->base + TCR);
/* reset counter to zero, set new period */
__raw_writel(0, t->base + t->tim_off);
__raw_writel(t->period, t->base + t->prd_off);
/* Set enable mode */
if (t->opts & TIMER_OPTS_ONESHOT)
tcr |= TCR_ENAMODE_ONESHOT << t->enamode_shift;
else if (t->opts & TIMER_OPTS_PERIODIC)
tcr |= TCR_ENAMODE_PERIODIC << t->enamode_shift;
__raw_writel(tcr, t->base + TCR);
}
return 0;
}
static inline u32 timer32_read(struct timer_s *t)
{
return __raw_readl(t->base + t->tim_off);
}
static irqreturn_t timer_interrupt(int irq, void *dev_id)
{
struct clock_event_device *evt = &clockevent_davinci;
evt->event_handler(evt);
return IRQ_HANDLED;
}
/* called when 32-bit counter wraps */
static irqreturn_t freerun_interrupt(int irq, void *dev_id)
{
return IRQ_HANDLED;
}
static struct timer_s timers[] = {
[TID_CLOCKEVENT] = {
.name = "clockevent",
.opts = TIMER_OPTS_DISABLED,
.irqaction = {
.flags = IRQF_TIMER,
.handler = timer_interrupt,
}
},
[TID_CLOCKSOURCE] = {
.name = "free-run counter",
.period = ~0,
.opts = TIMER_OPTS_PERIODIC,
.irqaction = {
.flags = IRQF_TIMER,
.handler = freerun_interrupt,
}
},
};
static void __init timer_init(void)
{
struct davinci_soc_info *soc_info = &davinci_soc_info;
struct davinci_timer_instance *dtip = soc_info->timer_info->timers;
void __iomem *base[2];
int i;
/* Global init of each 64-bit timer as a whole */
for(i=0; i<2; i++) {
u32 tgcr;
base[i] = ioremap(dtip[i].base, SZ_4K);
if (WARN_ON(!base[i]))
continue;
/* Disabled, Internal clock source */
__raw_writel(0, base[i] + TCR);
/* reset both timers, no pre-scaler for timer34 */
tgcr = 0;
__raw_writel(tgcr, base[i] + TGCR);
/* Set both timers to unchained 32-bit */
tgcr = TGCR_TIMMODE_32BIT_UNCHAINED << TGCR_TIMMODE_SHIFT;
__raw_writel(tgcr, base[i] + TGCR);
/* Unreset timers */
tgcr |= (TGCR_UNRESET << TGCR_TIM12RS_SHIFT) |
(TGCR_UNRESET << TGCR_TIM34RS_SHIFT);
__raw_writel(tgcr, base[i] + TGCR);
/* Init both counters to zero */
__raw_writel(0, base[i] + TIM12);
__raw_writel(0, base[i] + TIM34);
}
/* Init of each timer as a 32-bit timer */
for (i=0; i< ARRAY_SIZE(timers); i++) {
struct timer_s *t = &timers[i];
int timer = ID_TO_TIMER(t->id);
u32 irq;
t->base = base[timer];
if (!t->base)
continue;
if (IS_TIMER_BOT(t->id)) {
t->enamode_shift = 6;
t->tim_off = TIM12;
t->prd_off = PRD12;
irq = dtip[timer].bottom_irq;
} else {
t->enamode_shift = 22;
t->tim_off = TIM34;
t->prd_off = PRD34;
irq = dtip[timer].top_irq;
}
/* Register interrupt */
t->irqaction.name = t->name;
t->irqaction.dev_id = (void *)t;
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
if (t->irqaction.handler != NULL) {
irq = USING_COMPARE(t) ? dtip[i].cmp_irq : irq;
setup_irq(irq, &t->irqaction);
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
}
}
}
/*
* clocksource
*/
static u64 read_cycles(struct clocksource *cs)
{
struct timer_s *t = &timers[TID_CLOCKSOURCE];
return (cycles_t)timer32_read(t);
}
static struct clocksource clocksource_davinci = {
.rating = 300,
.read = read_cycles,
.mask = CLOCKSOURCE_MASK(32),
.flags = CLOCK_SOURCE_IS_CONTINUOUS,
};
/*
* Overwrite weak default sched_clock with something more precise
*/
static u64 notrace davinci_read_sched_clock(void)
{
return timer32_read(&timers[TID_CLOCKSOURCE]);
}
/*
* clockevent
*/
static int davinci_set_next_event(unsigned long cycles,
struct clock_event_device *evt)
{
struct timer_s *t = &timers[TID_CLOCKEVENT];
t->period = cycles;
timer32_config(t);
return 0;
}
static int davinci_shutdown(struct clock_event_device *evt)
{
struct timer_s *t = &timers[TID_CLOCKEVENT];
t->opts &= ~TIMER_OPTS_STATE_MASK;
t->opts |= TIMER_OPTS_DISABLED;
return 0;
}
static int davinci_set_oneshot(struct clock_event_device *evt)
{
struct timer_s *t = &timers[TID_CLOCKEVENT];
t->opts &= ~TIMER_OPTS_STATE_MASK;
t->opts |= TIMER_OPTS_ONESHOT;
return 0;
}
static int davinci_set_periodic(struct clock_event_device *evt)
{
struct timer_s *t = &timers[TID_CLOCKEVENT];
t->period = davinci_clock_tick_rate / (HZ);
t->opts &= ~TIMER_OPTS_STATE_MASK;
t->opts |= TIMER_OPTS_PERIODIC;
timer32_config(t);
return 0;
}
static struct clock_event_device clockevent_davinci = {
.features = CLOCK_EVT_FEAT_PERIODIC |
CLOCK_EVT_FEAT_ONESHOT,
.set_next_event = davinci_set_next_event,
.set_state_shutdown = davinci_shutdown,
.set_state_periodic = davinci_set_periodic,
.set_state_oneshot = davinci_set_oneshot,
};
void __init davinci_timer_init(void)
{
struct clk *timer_clk;
struct davinci_soc_info *soc_info = &davinci_soc_info;
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
unsigned int clockevent_id;
unsigned int clocksource_id;
int i;
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
clockevent_id = soc_info->timer_info->clockevent_id;
clocksource_id = soc_info->timer_info->clocksource_id;
timers[TID_CLOCKEVENT].id = clockevent_id;
timers[TID_CLOCKSOURCE].id = clocksource_id;
/*
* If using same timer for both clock events & clocksource,
* a compare register must be used to generate an event interrupt.
* This is equivalent to a oneshot timer only (not periodic).
*/
if (clockevent_id == clocksource_id) {
struct davinci_timer_instance *dtip =
soc_info->timer_info->timers;
int event_timer = ID_TO_TIMER(clockevent_id);
/* Only bottom timers can use compare regs */
if (IS_TIMER_TOP(clockevent_id))
pr_warn("%s: Invalid use of system timers. Results unpredictable.\n",
__func__);
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
else if ((dtip[event_timer].cmp_off == 0)
|| (dtip[event_timer].cmp_irq == 0))
pr_warn("%s: Invalid timer instance setup. Results unpredictable.\n",
__func__);
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
else {
timers[TID_CLOCKEVENT].opts |= TIMER_OPTS_USE_COMPARE;
clockevent_davinci.features = CLOCK_EVT_FEAT_ONESHOT;
}
}
timer_clk = clk_get(NULL, "timer0");
BUG_ON(IS_ERR(timer_clk));
clk_prepare_enable(timer_clk);
/* init timer hw */
timer_init();
davinci_clock_tick_rate = clk_get_rate(timer_clk);
/* setup clocksource */
davinci: Add compare register support to timer code The Timer64p timer has 8 compare registers that can be used to generate interrupts when the timer value matches the compare reg's value. They do not disturb the timer itself. This can be useful when there is only one timer available for both clock events and clocksource. When enabled, the clocksource remains a continuous 32-bit counter but the clock event will no longer support periodic interrupts. Instead only oneshot timers will be supported and implemented by setting the compare register to the current timer value plus the period that the clock event subsystem is requesting. Compare registers support is enabled automatically when the following conditions are met: 1) The same timer is being used for clock events and clocksource. 2) The timer is the bottom half (32 bits) of the 64-bit timer (hardware limitation). 3) The the compare register offset and irq are not zero. Since the timer is always running, there is a hardware race in timer32_config() between reading the current timer value, and adding the period to the current timer value and writing the compare register. Testing on a da830 evm board with the timer clocked at 24 MHz and the processor clocked at 300 MHz, showed the number of counter ticks to do this ranged from 20-53 (~1-2.2 usecs) but usually around 41 ticks. This includes some artifacts from collecting the information. So, the minimum period should be at least 5 usecs to be safe. There is also an non-critical lower limit that the period should be since there is no point in setting an event that is much shorter than the time it takes to set the event, and get & handle the timer interrupt for that event. There can also be all sorts of delays from activities occuring elsewhere in the system (including hardware activitis like cache & TLB management). These are virtually impossible to quantify so a minimum period of 50 usecs was chosen. That will certianly be enough to avoid the actual hardware race but hopefully not large enough to cause unreasonably course-grained timers. Signed-off-by: Mark A. Greer <mgreer@mvista.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
2009-04-16 03:41:54 +08:00
clocksource_davinci.name = id_to_name[clocksource_id];
if (clocksource_register_hz(&clocksource_davinci,
davinci_clock_tick_rate))
pr_err("%s: can't register clocksource!\n",
clocksource_davinci.name);
sched_clock_register(davinci_read_sched_clock, 32,
davinci_clock_tick_rate);
/* setup clockevent */
clockevent_davinci.name = id_to_name[timers[TID_CLOCKEVENT].id];
clockevent_davinci.cpumask = cpumask_of(0);
clockevents_config_and_register(&clockevent_davinci,
davinci_clock_tick_rate, 1, 0xfffffffe);
for (i=0; i< ARRAY_SIZE(timers); i++)
timer32_config(&timers[i]);
}
/* reset board using watchdog timer */
void davinci_watchdog_reset(struct platform_device *pdev)
{
u32 tgcr, wdtcr;
void __iomem *base;
struct clk *wd_clk;
base = ioremap(pdev->resource[0].start, SZ_4K);
if (WARN_ON(!base))
return;
wd_clk = clk_get(&pdev->dev, NULL);
if (WARN_ON(IS_ERR(wd_clk)))
return;
clk_prepare_enable(wd_clk);
/* disable, internal clock source */
__raw_writel(0, base + TCR);
/* reset timer, set mode to 64-bit watchdog, and unreset */
tgcr = 0;
__raw_writel(tgcr, base + TGCR);
tgcr = TGCR_TIMMODE_64BIT_WDOG << TGCR_TIMMODE_SHIFT;
tgcr |= (TGCR_UNRESET << TGCR_TIM12RS_SHIFT) |
(TGCR_UNRESET << TGCR_TIM34RS_SHIFT);
__raw_writel(tgcr, base + TGCR);
/* clear counter and period regs */
__raw_writel(0, base + TIM12);
__raw_writel(0, base + TIM34);
__raw_writel(0, base + PRD12);
__raw_writel(0, base + PRD34);
/* put watchdog in pre-active state */
wdtcr = __raw_readl(base + WDTCR);
wdtcr = (WDTCR_WDKEY_SEQ0 << WDTCR_WDKEY_SHIFT) |
(WDTCR_WDEN_ENABLE << WDTCR_WDEN_SHIFT);
__raw_writel(wdtcr, base + WDTCR);
/* put watchdog in active state */
wdtcr = (WDTCR_WDKEY_SEQ1 << WDTCR_WDKEY_SHIFT) |
(WDTCR_WDEN_ENABLE << WDTCR_WDEN_SHIFT);
__raw_writel(wdtcr, base + WDTCR);
/* write an invalid value to the WDKEY field to trigger
* a watchdog reset */
wdtcr = 0x00004000;
__raw_writel(wdtcr, base + WDTCR);
}