mirror of
https://github.com/torvalds/linux.git
synced 2024-12-12 22:23:55 +00:00
c83e448420
The ever-vigilant users of linode.com noticed that an idle 2.6 UML has a persistent load average of ~.4. It turns out that because the UML timer handler processed softirqs before actually delivering the tick, the tick was counted in the context of the idle thread about half the time. Signed-off-by: Jeff Dike <jdike@linux.intel.com> Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
190 lines
4.2 KiB
C
190 lines
4.2 KiB
C
/*
|
|
* Copyright (C) 2000 Jeff Dike (jdike@karaya.com)
|
|
* Licensed under the GPL
|
|
*/
|
|
|
|
#include "linux/kernel.h"
|
|
#include "linux/module.h"
|
|
#include "linux/unistd.h"
|
|
#include "linux/stddef.h"
|
|
#include "linux/spinlock.h"
|
|
#include "linux/time.h"
|
|
#include "linux/sched.h"
|
|
#include "linux/interrupt.h"
|
|
#include "linux/init.h"
|
|
#include "linux/delay.h"
|
|
#include "linux/hrtimer.h"
|
|
#include "asm/irq.h"
|
|
#include "asm/param.h"
|
|
#include "asm/current.h"
|
|
#include "kern_util.h"
|
|
#include "mode.h"
|
|
#include "os.h"
|
|
|
|
int hz(void)
|
|
{
|
|
return(HZ);
|
|
}
|
|
|
|
/*
|
|
* Scheduler clock - returns current time in nanosec units.
|
|
*/
|
|
unsigned long long sched_clock(void)
|
|
{
|
|
return (unsigned long long)jiffies_64 * (1000000000 / HZ);
|
|
}
|
|
|
|
#ifdef CONFIG_UML_REAL_TIME_CLOCK
|
|
static unsigned long long prev_nsecs[NR_CPUS];
|
|
static long long delta[NR_CPUS]; /* Deviation per interval */
|
|
#endif
|
|
|
|
void timer_irq(union uml_pt_regs *regs)
|
|
{
|
|
unsigned long long ticks = 0;
|
|
#ifdef CONFIG_UML_REAL_TIME_CLOCK
|
|
int c = cpu();
|
|
if(prev_nsecs[c]){
|
|
/* We've had 1 tick */
|
|
unsigned long long nsecs = os_nsecs();
|
|
|
|
delta[c] += nsecs - prev_nsecs[c];
|
|
prev_nsecs[c] = nsecs;
|
|
|
|
/* Protect against the host clock being set backwards */
|
|
if(delta[c] < 0)
|
|
delta[c] = 0;
|
|
|
|
ticks += (delta[c] * HZ) / BILLION;
|
|
delta[c] -= (ticks * BILLION) / HZ;
|
|
}
|
|
else prev_nsecs[c] = os_nsecs();
|
|
#else
|
|
ticks = 1;
|
|
#endif
|
|
while(ticks > 0){
|
|
do_IRQ(TIMER_IRQ, regs);
|
|
ticks--;
|
|
}
|
|
}
|
|
|
|
/* Protects local_offset */
|
|
static DEFINE_SPINLOCK(timer_spinlock);
|
|
static unsigned long long local_offset = 0;
|
|
|
|
static inline unsigned long long get_time(void)
|
|
{
|
|
unsigned long long nsecs;
|
|
unsigned long flags;
|
|
|
|
spin_lock_irqsave(&timer_spinlock, flags);
|
|
nsecs = os_nsecs();
|
|
nsecs += local_offset;
|
|
spin_unlock_irqrestore(&timer_spinlock, flags);
|
|
|
|
return nsecs;
|
|
}
|
|
|
|
irqreturn_t um_timer(int irq, void *dev)
|
|
{
|
|
unsigned long long nsecs;
|
|
unsigned long flags;
|
|
|
|
write_seqlock_irqsave(&xtime_lock, flags);
|
|
|
|
do_timer(1);
|
|
|
|
#ifdef CONFIG_UML_REAL_TIME_CLOCK
|
|
nsecs = get_time();
|
|
#else
|
|
nsecs = (unsigned long long) xtime.tv_sec * BILLION + xtime.tv_nsec +
|
|
BILLION / HZ;
|
|
#endif
|
|
xtime.tv_sec = nsecs / NSEC_PER_SEC;
|
|
xtime.tv_nsec = nsecs - xtime.tv_sec * NSEC_PER_SEC;
|
|
|
|
write_sequnlock_irqrestore(&xtime_lock, flags);
|
|
|
|
return IRQ_HANDLED;
|
|
}
|
|
|
|
static void register_timer(void)
|
|
{
|
|
int err;
|
|
|
|
err = request_irq(TIMER_IRQ, um_timer, IRQF_DISABLED, "timer", NULL);
|
|
if(err != 0)
|
|
printk(KERN_ERR "register_timer : request_irq failed - "
|
|
"errno = %d\n", -err);
|
|
|
|
err = set_interval(1);
|
|
if(err != 0)
|
|
printk(KERN_ERR "register_timer : set_interval failed - "
|
|
"errno = %d\n", -err);
|
|
}
|
|
|
|
extern void (*late_time_init)(void);
|
|
|
|
void time_init(void)
|
|
{
|
|
long long nsecs;
|
|
|
|
nsecs = os_nsecs();
|
|
set_normalized_timespec(&wall_to_monotonic, -nsecs / BILLION,
|
|
-nsecs % BILLION);
|
|
set_normalized_timespec(&xtime, nsecs / BILLION, nsecs % BILLION);
|
|
late_time_init = register_timer;
|
|
}
|
|
|
|
void do_gettimeofday(struct timeval *tv)
|
|
{
|
|
#ifdef CONFIG_UML_REAL_TIME_CLOCK
|
|
unsigned long long nsecs = get_time();
|
|
#else
|
|
unsigned long long nsecs = (unsigned long long) xtime.tv_sec * BILLION +
|
|
xtime.tv_nsec;
|
|
#endif
|
|
tv->tv_sec = nsecs / NSEC_PER_SEC;
|
|
/* Careful about calculations here - this was originally done as
|
|
* (nsecs - tv->tv_sec * NSEC_PER_SEC) / NSEC_PER_USEC
|
|
* which gave bogus (> 1000000) values. Dunno why, suspect gcc
|
|
* (4.0.0) miscompiled it, or there's a subtle 64/32-bit conversion
|
|
* problem that I missed.
|
|
*/
|
|
nsecs -= tv->tv_sec * NSEC_PER_SEC;
|
|
tv->tv_usec = (unsigned long) nsecs / NSEC_PER_USEC;
|
|
}
|
|
|
|
static inline void set_time(unsigned long long nsecs)
|
|
{
|
|
unsigned long long now;
|
|
unsigned long flags;
|
|
|
|
spin_lock_irqsave(&timer_spinlock, flags);
|
|
now = os_nsecs();
|
|
local_offset = nsecs - now;
|
|
spin_unlock_irqrestore(&timer_spinlock, flags);
|
|
|
|
clock_was_set();
|
|
}
|
|
|
|
int do_settimeofday(struct timespec *tv)
|
|
{
|
|
set_time((unsigned long long) tv->tv_sec * NSEC_PER_SEC + tv->tv_nsec);
|
|
|
|
return 0;
|
|
}
|
|
|
|
void timer_handler(int sig, union uml_pt_regs *regs)
|
|
{
|
|
if(current_thread->cpu == 0)
|
|
timer_irq(regs);
|
|
local_irq_disable();
|
|
irq_enter();
|
|
update_process_times(CHOOSE_MODE(
|
|
(UPT_SC(regs) && user_context(UPT_SP(regs))),
|
|
(regs)->skas.is_user));
|
|
irq_exit();
|
|
local_irq_enable();
|
|
}
|