kernel_optimize_test/arch/s390/kernel/vtime.c

397 lines
10 KiB
C
Raw Normal View History

/*
* Virtual cpu timer based timer functions.
*
* Copyright IBM Corp. 2004, 2012
* Author(s): Jan Glauber <jan.glauber@de.ibm.com>
*/
#include <linux/kernel_stat.h>
#include <linux/notifier.h>
#include <linux/kprobes.h>
#include <linux/export.h>
#include <linux/kernel.h>
#include <linux/timex.h>
#include <linux/types.h>
#include <linux/time.h>
[S390] nohz/s390: fix arch_needs_cpu() return value on offline cpus This fixes the same problem as described in the patch "nohz: fix printk_needs_cpu() return value on offline cpus" for the arch_needs_cpu() primitive: arch_needs_cpu() may return 1 if called on offline cpus. When a cpu gets offlined it schedules the idle process which, before killing its own cpu, will call tick_nohz_stop_sched_tick(). That function in turn will call arch_needs_cpu() in order to check if the local tick can be disabled. On offline cpus this function should naturally return 0 since regardless if the tick gets disabled or not the cpu will be dead short after. That is besides the fact that __cpu_disable() should already have made sure that no interrupts on the offlined cpu will be delivered anyway. In this case it prevents tick_nohz_stop_sched_tick() to call select_nohz_load_balancer(). No idea if that really is a problem. However what made me debug this is that on 2.6.32 the function get_nohz_load_balancer() is used within __mod_timer() to select a cpu on which a timer gets enqueued. If arch_needs_cpu() returns 1 then the nohz_load_balancer cpu doesn't get updated when a cpu gets offlined. It may contain the cpu number of an offline cpu. In turn timers get enqueued on an offline cpu and not very surprisingly they never expire and cause system hangs. This has been observed 2.6.32 kernels. On current kernels __mod_timer() uses get_nohz_timer_target() which doesn't have that problem. However there might be other problems because of the too early exit tick_nohz_stop_sched_tick() in case a cpu goes offline. This specific bug was indrocuded with 3c5d92a0 "nohz: Introduce arch_needs_cpu". In this case a cpu hotplug notifier is used to fix the issue in order to keep the normal/fast path small. All we need to do is to clear the condition that makes arch_needs_cpu() return 1 since it is just a performance improvement which is supposed to keep the local tick running for a short period if a cpu goes idle. Nothing special needs to be done except for clearing the condition. Cc: stable@kernel.org Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2010-12-01 17:08:01 +08:00
#include <linux/cpu.h>
#include <linux/smp.h>
#include <asm/irq_regs.h>
#include <asm/cputime.h>
#include <asm/vtimer.h>
#include <asm/irq.h>
#include "entry.h"
static void virt_timer_expire(void);
DEFINE_PER_CPU(struct s390_idle_data, s390_idle);
static LIST_HEAD(virt_timer_list);
static DEFINE_SPINLOCK(virt_timer_lock);
static atomic64_t virt_timer_current;
static atomic64_t virt_timer_elapsed;
static inline u64 get_vtimer(void)
{
u64 timer;
asm volatile("stpt %0" : "=m" (timer));
return timer;
}
static inline void set_vtimer(u64 expires)
{
u64 timer;
asm volatile(
" stpt %0\n" /* Store current cpu timer value */
" spt %1" /* Set new value imm. afterwards */
: "=m" (timer) : "m" (expires));
S390_lowcore.system_timer += S390_lowcore.last_update_timer - timer;
S390_lowcore.last_update_timer = expires;
}
static inline int virt_timer_forward(u64 elapsed)
{
BUG_ON(!irqs_disabled());
if (list_empty(&virt_timer_list))
return 0;
elapsed = atomic64_add_return(elapsed, &virt_timer_elapsed);
return elapsed >= atomic64_read(&virt_timer_current);
}
/*
* Update process times based on virtual cpu times stored by entry.S
* to the lowcore fields user_timer, system_timer & steal_clock.
*/
static int do_account_vtime(struct task_struct *tsk, int hardirq_offset)
{
struct thread_info *ti = task_thread_info(tsk);
u64 timer, clock, user, system, steal;
timer = S390_lowcore.last_update_timer;
clock = S390_lowcore.last_update_clock;
asm volatile(
" stpt %0\n" /* Store current cpu timer value */
" stck %1" /* Store current tod clock value */
: "=m" (S390_lowcore.last_update_timer),
"=m" (S390_lowcore.last_update_clock));
S390_lowcore.system_timer += timer - S390_lowcore.last_update_timer;
S390_lowcore.steal_timer += S390_lowcore.last_update_clock - clock;
user = S390_lowcore.user_timer - ti->user_timer;
S390_lowcore.steal_timer -= user;
ti->user_timer = S390_lowcore.user_timer;
account_user_time(tsk, user, user);
system = S390_lowcore.system_timer - ti->system_timer;
S390_lowcore.steal_timer -= system;
ti->system_timer = S390_lowcore.system_timer;
account_system_time(tsk, hardirq_offset, system, system);
steal = S390_lowcore.steal_timer;
if ((s64) steal > 0) {
S390_lowcore.steal_timer = 0;
account_steal_time(steal);
}
return virt_timer_forward(user + system);
}
void account_vtime(struct task_struct *prev, struct task_struct *next)
{
struct thread_info *ti;
do_account_vtime(prev, 0);
ti = task_thread_info(prev);
ti->user_timer = S390_lowcore.user_timer;
ti->system_timer = S390_lowcore.system_timer;
ti = task_thread_info(next);
S390_lowcore.user_timer = ti->user_timer;
S390_lowcore.system_timer = ti->system_timer;
}
void account_process_tick(struct task_struct *tsk, int user_tick)
{
if (do_account_vtime(tsk, HARDIRQ_OFFSET))
virt_timer_expire();
}
/*
* Update process times based on virtual cpu times stored by entry.S
* to the lowcore fields user_timer, system_timer & steal_clock.
*/
void account_system_vtime(struct task_struct *tsk)
{
struct thread_info *ti = task_thread_info(tsk);
u64 timer, system;
timer = S390_lowcore.last_update_timer;
S390_lowcore.last_update_timer = get_vtimer();
S390_lowcore.system_timer += timer - S390_lowcore.last_update_timer;
system = S390_lowcore.system_timer - ti->system_timer;
S390_lowcore.steal_timer -= system;
ti->system_timer = S390_lowcore.system_timer;
account_system_time(tsk, 0, system, system);
virt_timer_forward(system);
}
EXPORT_SYMBOL_GPL(account_system_vtime);
void __kprobes vtime_stop_cpu(void)
{
struct s390_idle_data *idle = &__get_cpu_var(s390_idle);
unsigned long long idle_time;
unsigned long psw_mask;
trace_hardirqs_on();
/* Don't trace preempt off for idle. */
stop_critical_timings();
/* Wait for external, I/O or machine check interrupt. */
psw_mask = psw_kernel_bits | PSW_MASK_WAIT | PSW_MASK_DAT |
PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
idle->nohz_delay = 0;
/* Call the assembler magic in entry.S */
psw_idle(idle, psw_mask);
/* Reenable preemption tracer. */
start_critical_timings();
/* Account time spent with enabled wait psw loaded as idle time. */
idle->sequence++;
smp_wmb();
idle_time = idle->clock_idle_exit - idle->clock_idle_enter;
idle->clock_idle_enter = idle->clock_idle_exit = 0ULL;
idle->idle_time += idle_time;
idle->idle_count++;
account_idle_time(idle_time);
smp_wmb();
idle->sequence++;
}
cputime64_t s390_get_idle_time(int cpu)
{
struct s390_idle_data *idle = &per_cpu(s390_idle, cpu);
unsigned long long now, idle_enter, idle_exit;
unsigned int sequence;
do {
now = get_clock();
sequence = ACCESS_ONCE(idle->sequence);
idle_enter = ACCESS_ONCE(idle->clock_idle_enter);
idle_exit = ACCESS_ONCE(idle->clock_idle_exit);
} while ((sequence & 1) || (idle->sequence != sequence));
return idle_enter ? ((idle_exit ?: now) - idle_enter) : 0;
}
/*
* Sorted add to a list. List is linear searched until first bigger
* element is found.
*/
static void list_add_sorted(struct vtimer_list *timer, struct list_head *head)
{
struct vtimer_list *tmp;
list_for_each_entry(tmp, head, entry) {
if (tmp->expires > timer->expires) {
list_add_tail(&timer->entry, &tmp->entry);
return;
}
}
list_add_tail(&timer->entry, head);
}
/*
* Handler for expired virtual CPU timer.
*/
static void virt_timer_expire(void)
{
struct vtimer_list *timer, *tmp;
unsigned long elapsed;
LIST_HEAD(cb_list);
/* walk timer list, fire all expired timers */
spin_lock(&virt_timer_lock);
elapsed = atomic64_read(&virt_timer_elapsed);
list_for_each_entry_safe(timer, tmp, &virt_timer_list, entry) {
if (timer->expires < elapsed)
/* move expired timer to the callback queue */
list_move_tail(&timer->entry, &cb_list);
else
timer->expires -= elapsed;
}
if (!list_empty(&virt_timer_list)) {
timer = list_first_entry(&virt_timer_list,
struct vtimer_list, entry);
atomic64_set(&virt_timer_current, timer->expires);
}
atomic64_sub(elapsed, &virt_timer_elapsed);
spin_unlock(&virt_timer_lock);
/* Do callbacks and recharge periodic timers */
list_for_each_entry_safe(timer, tmp, &cb_list, entry) {
list_del_init(&timer->entry);
timer->function(timer->data);
if (timer->interval) {
/* Recharge interval timer */
timer->expires = timer->interval +
atomic64_read(&virt_timer_elapsed);
spin_lock(&virt_timer_lock);
list_add_sorted(timer, &virt_timer_list);
spin_unlock(&virt_timer_lock);
}
}
}
void init_virt_timer(struct vtimer_list *timer)
{
timer->function = NULL;
INIT_LIST_HEAD(&timer->entry);
}
EXPORT_SYMBOL(init_virt_timer);
static inline int vtimer_pending(struct vtimer_list *timer)
{
return !list_empty(&timer->entry);
}
static void internal_add_vtimer(struct vtimer_list *timer)
{
if (list_empty(&virt_timer_list)) {
/* First timer, just program it. */
atomic64_set(&virt_timer_current, timer->expires);
atomic64_set(&virt_timer_elapsed, 0);
list_add(&timer->entry, &virt_timer_list);
} else {
/* Update timer against current base. */
timer->expires += atomic64_read(&virt_timer_elapsed);
if (likely((s64) timer->expires <
(s64) atomic64_read(&virt_timer_current)))
/* The new timer expires before the current timer. */
atomic64_set(&virt_timer_current, timer->expires);
/* Insert new timer into the list. */
list_add_sorted(timer, &virt_timer_list);
}
}
static void __add_vtimer(struct vtimer_list *timer, int periodic)
{
unsigned long flags;
timer->interval = periodic ? timer->expires : 0;
spin_lock_irqsave(&virt_timer_lock, flags);
internal_add_vtimer(timer);
spin_unlock_irqrestore(&virt_timer_lock, flags);
}
/*
* add_virt_timer - add an oneshot virtual CPU timer
*/
void add_virt_timer(struct vtimer_list *timer)
{
__add_vtimer(timer, 0);
}
EXPORT_SYMBOL(add_virt_timer);
/*
* add_virt_timer_int - add an interval virtual CPU timer
*/
void add_virt_timer_periodic(struct vtimer_list *timer)
{
__add_vtimer(timer, 1);
}
EXPORT_SYMBOL(add_virt_timer_periodic);
static int __mod_vtimer(struct vtimer_list *timer, u64 expires, int periodic)
{
unsigned long flags;
int rc;
BUG_ON(!timer->function);
if (timer->expires == expires && vtimer_pending(timer))
return 1;
spin_lock_irqsave(&virt_timer_lock, flags);
rc = vtimer_pending(timer);
if (rc)
list_del_init(&timer->entry);
timer->interval = periodic ? expires : 0;
timer->expires = expires;
internal_add_vtimer(timer);
spin_unlock_irqrestore(&virt_timer_lock, flags);
return rc;
}
/*
* returns whether it has modified a pending timer (1) or not (0)
*/
int mod_virt_timer(struct vtimer_list *timer, u64 expires)
{
return __mod_vtimer(timer, expires, 0);
}
EXPORT_SYMBOL(mod_virt_timer);
/*
* returns whether it has modified a pending timer (1) or not (0)
*/
int mod_virt_timer_periodic(struct vtimer_list *timer, u64 expires)
{
return __mod_vtimer(timer, expires, 1);
}
EXPORT_SYMBOL(mod_virt_timer_periodic);
/*
* Delete a virtual timer.
*
* returns whether the deleted timer was pending (1) or not (0)
*/
int del_virt_timer(struct vtimer_list *timer)
{
unsigned long flags;
if (!vtimer_pending(timer))
return 0;
spin_lock_irqsave(&virt_timer_lock, flags);
list_del_init(&timer->entry);
spin_unlock_irqrestore(&virt_timer_lock, flags);
return 1;
}
EXPORT_SYMBOL(del_virt_timer);
/*
* Start the virtual CPU timer on the current CPU.
*/
void __cpuinit init_cpu_vtimer(void)
{
/* set initial cpu timer */
set_vtimer(VTIMER_MAX_SLICE);
}
[S390] nohz/s390: fix arch_needs_cpu() return value on offline cpus This fixes the same problem as described in the patch "nohz: fix printk_needs_cpu() return value on offline cpus" for the arch_needs_cpu() primitive: arch_needs_cpu() may return 1 if called on offline cpus. When a cpu gets offlined it schedules the idle process which, before killing its own cpu, will call tick_nohz_stop_sched_tick(). That function in turn will call arch_needs_cpu() in order to check if the local tick can be disabled. On offline cpus this function should naturally return 0 since regardless if the tick gets disabled or not the cpu will be dead short after. That is besides the fact that __cpu_disable() should already have made sure that no interrupts on the offlined cpu will be delivered anyway. In this case it prevents tick_nohz_stop_sched_tick() to call select_nohz_load_balancer(). No idea if that really is a problem. However what made me debug this is that on 2.6.32 the function get_nohz_load_balancer() is used within __mod_timer() to select a cpu on which a timer gets enqueued. If arch_needs_cpu() returns 1 then the nohz_load_balancer cpu doesn't get updated when a cpu gets offlined. It may contain the cpu number of an offline cpu. In turn timers get enqueued on an offline cpu and not very surprisingly they never expire and cause system hangs. This has been observed 2.6.32 kernels. On current kernels __mod_timer() uses get_nohz_timer_target() which doesn't have that problem. However there might be other problems because of the too early exit tick_nohz_stop_sched_tick() in case a cpu goes offline. This specific bug was indrocuded with 3c5d92a0 "nohz: Introduce arch_needs_cpu". In this case a cpu hotplug notifier is used to fix the issue in order to keep the normal/fast path small. All we need to do is to clear the condition that makes arch_needs_cpu() return 1 since it is just a performance improvement which is supposed to keep the local tick running for a short period if a cpu goes idle. Nothing special needs to be done except for clearing the condition. Cc: stable@kernel.org Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2010-12-01 17:08:01 +08:00
static int __cpuinit s390_nohz_notify(struct notifier_block *self,
unsigned long action, void *hcpu)
{
struct s390_idle_data *idle;
long cpu = (long) hcpu;
idle = &per_cpu(s390_idle, cpu);
switch (action) {
case CPU_DYING:
case CPU_DYING_FROZEN:
idle->nohz_delay = 0;
default:
break;
}
return NOTIFY_OK;
}
void __init vtime_init(void)
{
/* Enable cpu timer interrupts on the boot cpu. */
init_cpu_vtimer();
[S390] nohz/s390: fix arch_needs_cpu() return value on offline cpus This fixes the same problem as described in the patch "nohz: fix printk_needs_cpu() return value on offline cpus" for the arch_needs_cpu() primitive: arch_needs_cpu() may return 1 if called on offline cpus. When a cpu gets offlined it schedules the idle process which, before killing its own cpu, will call tick_nohz_stop_sched_tick(). That function in turn will call arch_needs_cpu() in order to check if the local tick can be disabled. On offline cpus this function should naturally return 0 since regardless if the tick gets disabled or not the cpu will be dead short after. That is besides the fact that __cpu_disable() should already have made sure that no interrupts on the offlined cpu will be delivered anyway. In this case it prevents tick_nohz_stop_sched_tick() to call select_nohz_load_balancer(). No idea if that really is a problem. However what made me debug this is that on 2.6.32 the function get_nohz_load_balancer() is used within __mod_timer() to select a cpu on which a timer gets enqueued. If arch_needs_cpu() returns 1 then the nohz_load_balancer cpu doesn't get updated when a cpu gets offlined. It may contain the cpu number of an offline cpu. In turn timers get enqueued on an offline cpu and not very surprisingly they never expire and cause system hangs. This has been observed 2.6.32 kernels. On current kernels __mod_timer() uses get_nohz_timer_target() which doesn't have that problem. However there might be other problems because of the too early exit tick_nohz_stop_sched_tick() in case a cpu goes offline. This specific bug was indrocuded with 3c5d92a0 "nohz: Introduce arch_needs_cpu". In this case a cpu hotplug notifier is used to fix the issue in order to keep the normal/fast path small. All we need to do is to clear the condition that makes arch_needs_cpu() return 1 since it is just a performance improvement which is supposed to keep the local tick running for a short period if a cpu goes idle. Nothing special needs to be done except for clearing the condition. Cc: stable@kernel.org Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2010-12-01 17:08:01 +08:00
cpu_notifier(s390_nohz_notify, 0);
}