linux_dsm_epyc7002/arch/s390/kernel/vtime.c

603 lines
15 KiB
C
Raw Normal View History

/*
* arch/s390/kernel/vtime.c
* Virtual cpu timer based timer functions.
*
* S390 version
* Copyright (C) 2004 IBM Deutschland Entwicklung GmbH, IBM Corporation
* Author(s): Jan Glauber <jan.glauber@de.ibm.com>
*/
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/time.h>
#include <linux/delay.h>
#include <linux/init.h>
#include <linux/smp.h>
#include <linux/types.h>
#include <linux/timex.h>
#include <linux/notifier.h>
#include <linux/kernel_stat.h>
#include <linux/rcupdate.h>
#include <linux/posix-timers.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 16:08:01 +07:00
#include <linux/cpu.h>
#include <linux/kprobes.h>
#include <asm/timer.h>
#include <asm/irq_regs.h>
#include <asm/cputime.h>
#include <asm/irq.h>
static DEFINE_PER_CPU(struct vtimer_queue, virt_cpu_timer);
DEFINE_PER_CPU(struct s390_idle_data, s390_idle);
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 immediately afterwards */
: "=m" (timer) : "m" (expires) );
S390_lowcore.system_timer += S390_lowcore.last_update_timer - timer;
S390_lowcore.last_update_timer = expires;
}
/*
* Update process times based on virtual cpu times stored by entry.S
* to the lowcore fields user_timer, system_timer & steal_clock.
*/
static void 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);
}
}
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)
{
do_account_vtime(tsk, HARDIRQ_OFFSET);
}
/*
* 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);
}
EXPORT_SYMBOL_GPL(account_system_vtime);
void __kprobes vtime_start_cpu(__u64 int_clock, __u64 enter_timer)
{
struct s390_idle_data *idle = &__get_cpu_var(s390_idle);
struct vtimer_queue *vq = &__get_cpu_var(virt_cpu_timer);
__u64 idle_time, expires;
if (idle->idle_enter == 0ULL)
return;
/* Account time spent with enabled wait psw loaded as idle time. */
idle_time = int_clock - idle->idle_enter;
account_idle_time(idle_time);
S390_lowcore.steal_timer +=
idle->idle_enter - S390_lowcore.last_update_clock;
S390_lowcore.last_update_clock = int_clock;
/* Account system time spent going idle. */
S390_lowcore.system_timer += S390_lowcore.last_update_timer - vq->idle;
S390_lowcore.last_update_timer = enter_timer;
/* Restart vtime CPU timer */
if (vq->do_spt) {
/* Program old expire value but first save progress. */
expires = vq->idle - enter_timer;
expires += get_vtimer();
set_vtimer(expires);
} else {
/* Don't account the CPU timer delta while the cpu was idle. */
vq->elapsed -= vq->idle - enter_timer;
}
idle->sequence++;
smp_wmb();
idle->idle_time += idle_time;
idle->idle_enter = 0ULL;
idle->idle_count++;
smp_wmb();
idle->sequence++;
}
void __kprobes vtime_stop_cpu(void)
{
struct s390_idle_data *idle = &__get_cpu_var(s390_idle);
struct vtimer_queue *vq = &__get_cpu_var(virt_cpu_timer);
psw_t psw;
/* 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;
/* Check if the CPU timer needs to be reprogrammed. */
if (vq->do_spt) {
__u64 vmax = VTIMER_MAX_SLICE;
/*
* The inline assembly is equivalent to
* vq->idle = get_cpu_timer();
* set_cpu_timer(VTIMER_MAX_SLICE);
* idle->idle_enter = get_clock();
* __load_psw_mask(psw_kernel_bits | PSW_MASK_WAIT |
* PSW_MASK_DAT | PSW_MASK_IO |
* PSW_MASK_EXT | PSW_MASK_MCHECK);
* The difference is that the inline assembly makes sure that
* the last three instruction are stpt, stck and lpsw in that
* order. This is done to increase the precision.
*/
asm volatile(
#ifndef CONFIG_64BIT
" basr 1,0\n"
"0: ahi 1,1f-0b\n"
" st 1,4(%2)\n"
#else /* CONFIG_64BIT */
" larl 1,1f\n"
" stg 1,8(%2)\n"
#endif /* CONFIG_64BIT */
" stpt 0(%4)\n"
" spt 0(%5)\n"
" stck 0(%3)\n"
#ifndef CONFIG_64BIT
" lpsw 0(%2)\n"
#else /* CONFIG_64BIT */
" lpswe 0(%2)\n"
#endif /* CONFIG_64BIT */
"1:"
: "=m" (idle->idle_enter), "=m" (vq->idle)
: "a" (&psw), "a" (&idle->idle_enter),
"a" (&vq->idle), "a" (&vmax), "m" (vmax), "m" (psw)
: "memory", "cc", "1");
} else {
/*
* The inline assembly is equivalent to
* vq->idle = get_cpu_timer();
* idle->idle_enter = get_clock();
* __load_psw_mask(psw_kernel_bits | PSW_MASK_WAIT |
* PSW_MASK_DAT | PSW_MASK_IO |
* PSW_MASK_EXT | PSW_MASK_MCHECK);
* The difference is that the inline assembly makes sure that
* the last three instruction are stpt, stck and lpsw in that
* order. This is done to increase the precision.
*/
asm volatile(
#ifndef CONFIG_64BIT
" basr 1,0\n"
"0: ahi 1,1f-0b\n"
" st 1,4(%2)\n"
#else /* CONFIG_64BIT */
" larl 1,1f\n"
" stg 1,8(%2)\n"
#endif /* CONFIG_64BIT */
" stpt 0(%4)\n"
" stck 0(%3)\n"
#ifndef CONFIG_64BIT
" lpsw 0(%2)\n"
#else /* CONFIG_64BIT */
" lpswe 0(%2)\n"
#endif /* CONFIG_64BIT */
"1:"
: "=m" (idle->idle_enter), "=m" (vq->idle)
: "a" (&psw), "a" (&idle->idle_enter),
"a" (&vq->idle), "m" (psw)
: "memory", "cc", "1");
}
}
cputime64_t s390_get_idle_time(int cpu)
{
struct s390_idle_data *idle;
unsigned long long now, idle_time, idle_enter;
unsigned int sequence;
idle = &per_cpu(s390_idle, cpu);
now = get_clock();
repeat:
sequence = idle->sequence;
smp_rmb();
if (sequence & 1)
goto repeat;
idle_time = 0;
idle_enter = idle->idle_enter;
if (idle_enter != 0ULL && idle_enter < now)
idle_time = now - idle_enter;
smp_rmb();
if (idle->sequence != sequence)
goto repeat;
return idle_time;
}
/*
* 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 *event;
list_for_each_entry(event, head, entry) {
if (event->expires > timer->expires) {
list_add_tail(&timer->entry, &event->entry);
return;
}
}
list_add_tail(&timer->entry, head);
}
/*
* Do the callback functions of expired vtimer events.
* Called from within the interrupt handler.
*/
static void do_callbacks(struct list_head *cb_list)
{
struct vtimer_queue *vq;
struct vtimer_list *event, *tmp;
if (list_empty(cb_list))
return;
vq = &__get_cpu_var(virt_cpu_timer);
list_for_each_entry_safe(event, tmp, cb_list, entry) {
list_del_init(&event->entry);
(event->function)(event->data);
if (event->interval) {
/* Recharge interval timer */
event->expires = event->interval + vq->elapsed;
spin_lock(&vq->lock);
list_add_sorted(event, &vq->list);
spin_unlock(&vq->lock);
}
}
}
/*
* Handler for the virtual CPU timer.
*/
static void do_cpu_timer_interrupt(unsigned int ext_int_code,
unsigned int param32, unsigned long param64)
{
struct vtimer_queue *vq;
struct vtimer_list *event, *tmp;
struct list_head cb_list; /* the callback queue */
__u64 elapsed, next;
kstat_cpu(smp_processor_id()).irqs[EXTINT_TMR]++;
INIT_LIST_HEAD(&cb_list);
vq = &__get_cpu_var(virt_cpu_timer);
/* walk timer list, fire all expired events */
spin_lock(&vq->lock);
elapsed = vq->elapsed + (vq->timer - S390_lowcore.async_enter_timer);
BUG_ON((s64) elapsed < 0);
vq->elapsed = 0;
list_for_each_entry_safe(event, tmp, &vq->list, entry) {
if (event->expires < elapsed)
/* move expired timer to the callback queue */
list_move_tail(&event->entry, &cb_list);
else
event->expires -= elapsed;
}
spin_unlock(&vq->lock);
vq->do_spt = list_empty(&cb_list);
do_callbacks(&cb_list);
/* next event is first in list */
next = VTIMER_MAX_SLICE;
spin_lock(&vq->lock);
if (!list_empty(&vq->list)) {
event = list_first_entry(&vq->list, struct vtimer_list, entry);
next = event->expires;
} else
vq->do_spt = 0;
spin_unlock(&vq->lock);
/*
* To improve precision add the time spent by the
* interrupt handler to the elapsed time.
* Note: CPU timer counts down and we got an interrupt,
* the current content is negative
*/
elapsed = S390_lowcore.async_enter_timer - get_vtimer();
set_vtimer(next - elapsed);
vq->timer = next - elapsed;
vq->elapsed = elapsed;
}
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));
}
/*
* this function should only run on the specified CPU
*/
static void internal_add_vtimer(struct vtimer_list *timer)
{
struct vtimer_queue *vq;
unsigned long flags;
__u64 left, expires;
vq = &per_cpu(virt_cpu_timer, timer->cpu);
spin_lock_irqsave(&vq->lock, flags);
BUG_ON(timer->cpu != smp_processor_id());
if (list_empty(&vq->list)) {
/* First timer on this cpu, just program it. */
list_add(&timer->entry, &vq->list);
set_vtimer(timer->expires);
vq->timer = timer->expires;
vq->elapsed = 0;
} else {
/* Check progress of old timers. */
expires = timer->expires;
left = get_vtimer();
if (likely((s64) expires < (s64) left)) {
/* The new timer expires before the current timer. */
set_vtimer(expires);
vq->elapsed += vq->timer - left;
vq->timer = expires;
} else {
vq->elapsed += vq->timer - left;
vq->timer = left;
}
/* Insert new timer into per cpu list. */
timer->expires += vq->elapsed;
list_add_sorted(timer, &vq->list);
}
spin_unlock_irqrestore(&vq->lock, flags);
/* release CPU acquired in prepare_vtimer or mod_virt_timer() */
put_cpu();
}
static inline void prepare_vtimer(struct vtimer_list *timer)
{
BUG_ON(!timer->function);
BUG_ON(!timer->expires || timer->expires > VTIMER_MAX_SLICE);
BUG_ON(vtimer_pending(timer));
timer->cpu = get_cpu();
}
/*
* add_virt_timer - add an oneshot virtual CPU timer
*/
void add_virt_timer(void *new)
{
struct vtimer_list *timer;
timer = (struct vtimer_list *)new;
prepare_vtimer(timer);
timer->interval = 0;
internal_add_vtimer(timer);
}
EXPORT_SYMBOL(add_virt_timer);
/*
* add_virt_timer_int - add an interval virtual CPU timer
*/
void add_virt_timer_periodic(void *new)
{
struct vtimer_list *timer;
timer = (struct vtimer_list *)new;
prepare_vtimer(timer);
timer->interval = timer->expires;
internal_add_vtimer(timer);
}
EXPORT_SYMBOL(add_virt_timer_periodic);
int __mod_vtimer(struct vtimer_list *timer, __u64 expires, int periodic)
{
struct vtimer_queue *vq;
unsigned long flags;
int cpu;
BUG_ON(!timer->function);
BUG_ON(!expires || expires > VTIMER_MAX_SLICE);
if (timer->expires == expires && vtimer_pending(timer))
return 1;
cpu = get_cpu();
vq = &per_cpu(virt_cpu_timer, cpu);
/* disable interrupts before test if timer is pending */
spin_lock_irqsave(&vq->lock, flags);
/* if timer isn't pending add it on the current CPU */
if (!vtimer_pending(timer)) {
spin_unlock_irqrestore(&vq->lock, flags);
if (periodic)
timer->interval = expires;
else
timer->interval = 0;
timer->expires = expires;
timer->cpu = cpu;
internal_add_vtimer(timer);
return 0;
}
/* check if we run on the right CPU */
BUG_ON(timer->cpu != cpu);
list_del_init(&timer->entry);
timer->expires = expires;
if (periodic)
timer->interval = expires;
/* the timer can't expire anymore so we can release the lock */
spin_unlock_irqrestore(&vq->lock, flags);
internal_add_vtimer(timer);
return 1;
}
/*
* If we change a pending timer the function must be called on the CPU
* where the timer is running on.
*
* 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);
/*
* If we change a pending timer the function must be called on the CPU
* where the timer is running on.
*
* 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;
struct vtimer_queue *vq;
/* check if timer is pending */
if (!vtimer_pending(timer))
return 0;
vq = &per_cpu(virt_cpu_timer, timer->cpu);
spin_lock_irqsave(&vq->lock, flags);
/* we don't interrupt a running timer, just let it expire! */
list_del_init(&timer->entry);
spin_unlock_irqrestore(&vq->lock, flags);
return 1;
}
EXPORT_SYMBOL(del_virt_timer);
/*
* Start the virtual CPU timer on the current CPU.
*/
void init_cpu_vtimer(void)
{
struct vtimer_queue *vq;
/* initialize per cpu vtimer structure */
vq = &__get_cpu_var(virt_cpu_timer);
INIT_LIST_HEAD(&vq->list);
spin_lock_init(&vq->lock);
/* enable cpu timer interrupts */
__ctl_set_bit(0,10);
}
[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 16:08:01 +07:00
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)
{
/* request the cpu timer external interrupt */
if (register_external_interrupt(0x1005, do_cpu_timer_interrupt))
panic("Couldn't request external interrupt 0x1005");
/* 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 16:08:01 +07:00
cpu_notifier(s390_nohz_notify, 0);
}