WSL2-Linux-Kernel/kernel/sched/cputime.c

839 строки
21 KiB
C
Исходник Обычный вид История

#include <linux/export.h>
#include <linux/sched.h>
#include <linux/tsacct_kern.h>
#include <linux/kernel_stat.h>
#include <linux/static_key.h>
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
#include <linux/context_tracking.h>
#include "sched.h"
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
/*
* There are no locks covering percpu hardirq/softirq time.
* They are only modified in vtime_account, on corresponding CPU
* with interrupts disabled. So, writes are safe.
* They are read and saved off onto struct rq in update_rq_clock().
* This may result in other CPU reading this CPU's irq time and can
* race with irq/vtime_account on this CPU. We would either get old
* or new value with a side effect of accounting a slice of irq time to wrong
* task when irq is in progress while we read rq->clock. That is a worthy
* compromise in place of having locks on each irq in account_system_time.
*/
DEFINE_PER_CPU(u64, cpu_hardirq_time);
DEFINE_PER_CPU(u64, cpu_softirq_time);
static DEFINE_PER_CPU(u64, irq_start_time);
static int sched_clock_irqtime;
void enable_sched_clock_irqtime(void)
{
sched_clock_irqtime = 1;
}
void disable_sched_clock_irqtime(void)
{
sched_clock_irqtime = 0;
}
#ifndef CONFIG_64BIT
DEFINE_PER_CPU(seqcount_t, irq_time_seq);
#endif /* CONFIG_64BIT */
/*
* Called before incrementing preempt_count on {soft,}irq_enter
* and before decrementing preempt_count on {soft,}irq_exit.
*/
void irqtime_account_irq(struct task_struct *curr)
{
unsigned long flags;
s64 delta;
int cpu;
if (!sched_clock_irqtime)
return;
local_irq_save(flags);
cpu = smp_processor_id();
delta = sched_clock_cpu(cpu) - __this_cpu_read(irq_start_time);
__this_cpu_add(irq_start_time, delta);
irq_time_write_begin();
/*
* We do not account for softirq time from ksoftirqd here.
* We want to continue accounting softirq time to ksoftirqd thread
* in that case, so as not to confuse scheduler with a special task
* that do not consume any time, but still wants to run.
*/
if (hardirq_count())
__this_cpu_add(cpu_hardirq_time, delta);
else if (in_serving_softirq() && curr != this_cpu_ksoftirqd())
__this_cpu_add(cpu_softirq_time, delta);
irq_time_write_end();
local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(irqtime_account_irq);
static int irqtime_account_hi_update(void)
{
u64 *cpustat = kcpustat_this_cpu->cpustat;
unsigned long flags;
u64 latest_ns;
int ret = 0;
local_irq_save(flags);
latest_ns = this_cpu_read(cpu_hardirq_time);
if (nsecs_to_cputime64(latest_ns) > cpustat[CPUTIME_IRQ])
ret = 1;
local_irq_restore(flags);
return ret;
}
static int irqtime_account_si_update(void)
{
u64 *cpustat = kcpustat_this_cpu->cpustat;
unsigned long flags;
u64 latest_ns;
int ret = 0;
local_irq_save(flags);
latest_ns = this_cpu_read(cpu_softirq_time);
if (nsecs_to_cputime64(latest_ns) > cpustat[CPUTIME_SOFTIRQ])
ret = 1;
local_irq_restore(flags);
return ret;
}
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
#define sched_clock_irqtime (0)
#endif /* !CONFIG_IRQ_TIME_ACCOUNTING */
static inline void task_group_account_field(struct task_struct *p, int index,
u64 tmp)
{
/*
* Since all updates are sure to touch the root cgroup, we
* get ourselves ahead and touch it first. If the root cgroup
* is the only cgroup, then nothing else should be necessary.
*
*/
__this_cpu_add(kernel_cpustat.cpustat[index], tmp);
cpuacct_account_field(p, index, tmp);
}
/*
* Account user cpu time to a process.
* @p: the process that the cpu time gets accounted to
* @cputime: the cpu time spent in user space since the last update
* @cputime_scaled: cputime scaled by cpu frequency
*/
void account_user_time(struct task_struct *p, cputime_t cputime,
cputime_t cputime_scaled)
{
int index;
/* Add user time to process. */
p->utime += cputime;
p->utimescaled += cputime_scaled;
account_group_user_time(p, cputime);
index = (task_nice(p) > 0) ? CPUTIME_NICE : CPUTIME_USER;
/* Add user time to cpustat. */
task_group_account_field(p, index, (__force u64) cputime);
/* Account for user time used */
acct_account_cputime(p);
}
/*
* Account guest cpu time to a process.
* @p: the process that the cpu time gets accounted to
* @cputime: the cpu time spent in virtual machine since the last update
* @cputime_scaled: cputime scaled by cpu frequency
*/
static void account_guest_time(struct task_struct *p, cputime_t cputime,
cputime_t cputime_scaled)
{
u64 *cpustat = kcpustat_this_cpu->cpustat;
/* Add guest time to process. */
p->utime += cputime;
p->utimescaled += cputime_scaled;
account_group_user_time(p, cputime);
p->gtime += cputime;
/* Add guest time to cpustat. */
if (task_nice(p) > 0) {
cpustat[CPUTIME_NICE] += (__force u64) cputime;
cpustat[CPUTIME_GUEST_NICE] += (__force u64) cputime;
} else {
cpustat[CPUTIME_USER] += (__force u64) cputime;
cpustat[CPUTIME_GUEST] += (__force u64) cputime;
}
}
/*
* Account system cpu time to a process and desired cpustat field
* @p: the process that the cpu time gets accounted to
* @cputime: the cpu time spent in kernel space since the last update
* @cputime_scaled: cputime scaled by cpu frequency
* @target_cputime64: pointer to cpustat field that has to be updated
*/
static inline
void __account_system_time(struct task_struct *p, cputime_t cputime,
cputime_t cputime_scaled, int index)
{
/* Add system time to process. */
p->stime += cputime;
p->stimescaled += cputime_scaled;
account_group_system_time(p, cputime);
/* Add system time to cpustat. */
task_group_account_field(p, index, (__force u64) cputime);
/* Account for system time used */
acct_account_cputime(p);
}
/*
* Account system cpu time to a process.
* @p: the process that the cpu time gets accounted to
* @hardirq_offset: the offset to subtract from hardirq_count()
* @cputime: the cpu time spent in kernel space since the last update
* @cputime_scaled: cputime scaled by cpu frequency
*/
void account_system_time(struct task_struct *p, int hardirq_offset,
cputime_t cputime, cputime_t cputime_scaled)
{
int index;
if ((p->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0)) {
account_guest_time(p, cputime, cputime_scaled);
return;
}
if (hardirq_count() - hardirq_offset)
index = CPUTIME_IRQ;
else if (in_serving_softirq())
index = CPUTIME_SOFTIRQ;
else
index = CPUTIME_SYSTEM;
__account_system_time(p, cputime, cputime_scaled, index);
}
/*
* Account for involuntary wait time.
* @cputime: the cpu time spent in involuntary wait
*/
void account_steal_time(cputime_t cputime)
{
u64 *cpustat = kcpustat_this_cpu->cpustat;
cpustat[CPUTIME_STEAL] += (__force u64) cputime;
}
/*
* Account for idle time.
* @cputime: the cpu time spent in idle wait
*/
void account_idle_time(cputime_t cputime)
{
u64 *cpustat = kcpustat_this_cpu->cpustat;
struct rq *rq = this_rq();
if (atomic_read(&rq->nr_iowait) > 0)
cpustat[CPUTIME_IOWAIT] += (__force u64) cputime;
else
cpustat[CPUTIME_IDLE] += (__force u64) cputime;
}
static __always_inline bool steal_account_process_tick(void)
{
#ifdef CONFIG_PARAVIRT
if (static_key_false(&paravirt_steal_enabled)) {
u64 steal;
cputime_t steal_ct;
steal = paravirt_steal_clock(smp_processor_id());
steal -= this_rq()->prev_steal_time;
/*
* cputime_t may be less precise than nsecs (eg: if it's
* based on jiffies). Lets cast the result to cputime
* granularity and account the rest on the next rounds.
*/
steal_ct = nsecs_to_cputime(steal);
this_rq()->prev_steal_time += cputime_to_nsecs(steal_ct);
account_steal_time(steal_ct);
return steal_ct;
}
#endif
return false;
}
/*
* Accumulate raw cputime values of dead tasks (sig->[us]time) and live
* tasks (sum on group iteration) belonging to @tsk's group.
*/
void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times)
{
struct signal_struct *sig = tsk->signal;
cputime_t utime, stime;
struct task_struct *t;
times->utime = sig->utime;
times->stime = sig->stime;
times->sum_exec_runtime = sig->sum_sched_runtime;
rcu_read_lock();
/* make sure we can trust tsk->thread_group list */
if (!likely(pid_alive(tsk)))
goto out;
t = tsk;
do {
task_cputime(t, &utime, &stime);
times->utime += utime;
times->stime += stime;
times->sum_exec_runtime += task_sched_runtime(t);
} while_each_thread(tsk, t);
out:
rcu_read_unlock();
}
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
/*
* Account a tick to a process and cpustat
* @p: the process that the cpu time gets accounted to
* @user_tick: is the tick from userspace
* @rq: the pointer to rq
*
* Tick demultiplexing follows the order
* - pending hardirq update
* - pending softirq update
* - user_time
* - idle_time
* - system time
* - check for guest_time
* - else account as system_time
*
* Check for hardirq is done both for system and user time as there is
* no timer going off while we are on hardirq and hence we may never get an
* opportunity to update it solely in system time.
* p->stime and friends are only updated on system time and not on irq
* softirq as those do not count in task exec_runtime any more.
*/
static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
struct rq *rq)
{
cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
u64 *cpustat = kcpustat_this_cpu->cpustat;
if (steal_account_process_tick())
return;
if (irqtime_account_hi_update()) {
cpustat[CPUTIME_IRQ] += (__force u64) cputime_one_jiffy;
} else if (irqtime_account_si_update()) {
cpustat[CPUTIME_SOFTIRQ] += (__force u64) cputime_one_jiffy;
} else if (this_cpu_ksoftirqd() == p) {
/*
* ksoftirqd time do not get accounted in cpu_softirq_time.
* So, we have to handle it separately here.
* Also, p->stime needs to be updated for ksoftirqd.
*/
__account_system_time(p, cputime_one_jiffy, one_jiffy_scaled,
CPUTIME_SOFTIRQ);
} else if (user_tick) {
account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
} else if (p == rq->idle) {
account_idle_time(cputime_one_jiffy);
} else if (p->flags & PF_VCPU) { /* System time or guest time */
account_guest_time(p, cputime_one_jiffy, one_jiffy_scaled);
} else {
__account_system_time(p, cputime_one_jiffy, one_jiffy_scaled,
CPUTIME_SYSTEM);
}
}
static void irqtime_account_idle_ticks(int ticks)
{
int i;
struct rq *rq = this_rq();
for (i = 0; i < ticks; i++)
irqtime_account_process_tick(current, 0, rq);
}
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
static inline void irqtime_account_idle_ticks(int ticks) {}
static inline void irqtime_account_process_tick(struct task_struct *p, int user_tick,
struct rq *rq) {}
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
/*
* Use precise platform statistics if available:
*/
#ifdef CONFIG_VIRT_CPU_ACCOUNTING
#ifndef __ARCH_HAS_VTIME_TASK_SWITCH
void vtime_common_task_switch(struct task_struct *prev)
{
if (is_idle_task(prev))
vtime_account_idle(prev);
else
vtime_account_system(prev);
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
#ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
vtime_account_user(prev);
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
#endif
arch_vtime_task_switch(prev);
}
#endif
/*
* Archs that account the whole time spent in the idle task
* (outside irq) as idle time can rely on this and just implement
* vtime_account_system() and vtime_account_idle(). Archs that
* have other meaning of the idle time (s390 only includes the
* time spent by the CPU when it's in low power mode) must override
* vtime_account().
*/
#ifndef __ARCH_HAS_VTIME_ACCOUNT
void vtime_common_account_irq_enter(struct task_struct *tsk)
{
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
if (!in_interrupt()) {
/*
* If we interrupted user, context_tracking_in_user()
* is 1 because the context tracking don't hook
* on irq entry/exit. This way we know if
* we need to flush user time on kernel entry.
*/
if (context_tracking_in_user()) {
vtime_account_user(tsk);
return;
}
if (is_idle_task(tsk)) {
vtime_account_idle(tsk);
return;
}
}
vtime_account_system(tsk);
}
EXPORT_SYMBOL_GPL(vtime_common_account_irq_enter);
#endif /* __ARCH_HAS_VTIME_ACCOUNT */
#endif /* CONFIG_VIRT_CPU_ACCOUNTING */
#ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st)
{
*ut = p->utime;
*st = p->stime;
}
void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st)
{
struct task_cputime cputime;
thread_group_cputime(p, &cputime);
*ut = cputime.utime;
*st = cputime.stime;
}
#else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
/*
* Account a single tick of cpu time.
* @p: the process that the cpu time gets accounted to
* @user_tick: indicates if the tick is a user or a system tick
*/
void account_process_tick(struct task_struct *p, int user_tick)
{
cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
struct rq *rq = this_rq();
if (vtime_accounting_enabled())
return;
if (sched_clock_irqtime) {
irqtime_account_process_tick(p, user_tick, rq);
return;
}
if (steal_account_process_tick())
return;
if (user_tick)
account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET))
account_system_time(p, HARDIRQ_OFFSET, cputime_one_jiffy,
one_jiffy_scaled);
else
account_idle_time(cputime_one_jiffy);
}
/*
* Account multiple ticks of steal time.
* @p: the process from which the cpu time has been stolen
* @ticks: number of stolen ticks
*/
void account_steal_ticks(unsigned long ticks)
{
account_steal_time(jiffies_to_cputime(ticks));
}
/*
* Account multiple ticks of idle time.
* @ticks: number of stolen ticks
*/
void account_idle_ticks(unsigned long ticks)
{
if (sched_clock_irqtime) {
irqtime_account_idle_ticks(ticks);
return;
}
account_idle_time(jiffies_to_cputime(ticks));
}
sched: Lower chances of cputime scaling overflow Some users have reported that after running a process with hundreds of threads on intensive CPU-bound loads, the cputime of the group started to freeze after a few days. This is due to how we scale the tick-based cputime against the scheduler precise execution time value. We add the values of all threads in the group and we multiply that against the sum of the scheduler exec runtime of the whole group. This easily overflows after a few days/weeks of execution. A proposed solution to solve this was to compute that multiplication on stime instead of utime: 62188451f0d63add7ad0cd2a1ae269d600c1663d ("cputime: Avoid multiplication overflow on utime scaling") The rationale behind that was that it's easy for a thread to spend most of its time in userspace under intensive CPU-bound workload but it's much harder to do CPU-bound intensive long run in the kernel. This postulate got defeated when a user recently reported he was still seeing cputime freezes after the above patch. The workload that triggers this issue relates to intensive networking workloads where most of the cputime is consumed in the kernel. To reduce much more the opportunities for multiplication overflow, lets reduce the multiplication factors to the remainders of the division between sched exec runtime and cputime. Assuming the difference between these shouldn't ever be that large, it could work on many situations. This gets the same results as in the upstream scaling code except for a small difference: the upstream code always rounds the results to the nearest integer not greater to what would be the precise result. The new code rounds to the nearest integer either greater or not greater. In practice this difference probably shouldn't matter but it's worth mentioning. If this solution appears not to be enough in the end, we'll need to partly revert back to the behaviour prior to commit 0cf55e1ec08bb5a22e068309e2d8ba1180ab4239 ("sched, cputime: Introduce thread_group_times()") Back then, the scaling was done on exit() time before adding the cputime of an exiting thread to the signal struct. And then we'll need to scale one-by-one the live threads cputime in thread_group_cputime(). The drawback may be a slightly slower code on exit time. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Stanislaw Gruszka <sgruszka@redhat.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org>
2013-02-20 21:54:55 +04:00
/*
* Perform (stime * rtime) / total, but avoid multiplication overflow by
* loosing precision when the numbers are big.
sched: Lower chances of cputime scaling overflow Some users have reported that after running a process with hundreds of threads on intensive CPU-bound loads, the cputime of the group started to freeze after a few days. This is due to how we scale the tick-based cputime against the scheduler precise execution time value. We add the values of all threads in the group and we multiply that against the sum of the scheduler exec runtime of the whole group. This easily overflows after a few days/weeks of execution. A proposed solution to solve this was to compute that multiplication on stime instead of utime: 62188451f0d63add7ad0cd2a1ae269d600c1663d ("cputime: Avoid multiplication overflow on utime scaling") The rationale behind that was that it's easy for a thread to spend most of its time in userspace under intensive CPU-bound workload but it's much harder to do CPU-bound intensive long run in the kernel. This postulate got defeated when a user recently reported he was still seeing cputime freezes after the above patch. The workload that triggers this issue relates to intensive networking workloads where most of the cputime is consumed in the kernel. To reduce much more the opportunities for multiplication overflow, lets reduce the multiplication factors to the remainders of the division between sched exec runtime and cputime. Assuming the difference between these shouldn't ever be that large, it could work on many situations. This gets the same results as in the upstream scaling code except for a small difference: the upstream code always rounds the results to the nearest integer not greater to what would be the precise result. The new code rounds to the nearest integer either greater or not greater. In practice this difference probably shouldn't matter but it's worth mentioning. If this solution appears not to be enough in the end, we'll need to partly revert back to the behaviour prior to commit 0cf55e1ec08bb5a22e068309e2d8ba1180ab4239 ("sched, cputime: Introduce thread_group_times()") Back then, the scaling was done on exit() time before adding the cputime of an exiting thread to the signal struct. And then we'll need to scale one-by-one the live threads cputime in thread_group_cputime(). The drawback may be a slightly slower code on exit time. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Stanislaw Gruszka <sgruszka@redhat.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org>
2013-02-20 21:54:55 +04:00
*/
static cputime_t scale_stime(u64 stime, u64 rtime, u64 total)
{
u64 scaled;
for (;;) {
/* Make sure "rtime" is the bigger of stime/rtime */
if (stime > rtime)
swap(rtime, stime);
/* Make sure 'total' fits in 32 bits */
if (total >> 32)
goto drop_precision;
/* Does rtime (and thus stime) fit in 32 bits? */
if (!(rtime >> 32))
break;
/* Can we just balance rtime/stime rather than dropping bits? */
if (stime >> 31)
goto drop_precision;
/* We can grow stime and shrink rtime and try to make them both fit */
stime <<= 1;
rtime >>= 1;
continue;
drop_precision:
/* We drop from rtime, it has more bits than stime */
rtime >>= 1;
total >>= 1;
sched: Lower chances of cputime scaling overflow Some users have reported that after running a process with hundreds of threads on intensive CPU-bound loads, the cputime of the group started to freeze after a few days. This is due to how we scale the tick-based cputime against the scheduler precise execution time value. We add the values of all threads in the group and we multiply that against the sum of the scheduler exec runtime of the whole group. This easily overflows after a few days/weeks of execution. A proposed solution to solve this was to compute that multiplication on stime instead of utime: 62188451f0d63add7ad0cd2a1ae269d600c1663d ("cputime: Avoid multiplication overflow on utime scaling") The rationale behind that was that it's easy for a thread to spend most of its time in userspace under intensive CPU-bound workload but it's much harder to do CPU-bound intensive long run in the kernel. This postulate got defeated when a user recently reported he was still seeing cputime freezes after the above patch. The workload that triggers this issue relates to intensive networking workloads where most of the cputime is consumed in the kernel. To reduce much more the opportunities for multiplication overflow, lets reduce the multiplication factors to the remainders of the division between sched exec runtime and cputime. Assuming the difference between these shouldn't ever be that large, it could work on many situations. This gets the same results as in the upstream scaling code except for a small difference: the upstream code always rounds the results to the nearest integer not greater to what would be the precise result. The new code rounds to the nearest integer either greater or not greater. In practice this difference probably shouldn't matter but it's worth mentioning. If this solution appears not to be enough in the end, we'll need to partly revert back to the behaviour prior to commit 0cf55e1ec08bb5a22e068309e2d8ba1180ab4239 ("sched, cputime: Introduce thread_group_times()") Back then, the scaling was done on exit() time before adding the cputime of an exiting thread to the signal struct. And then we'll need to scale one-by-one the live threads cputime in thread_group_cputime(). The drawback may be a slightly slower code on exit time. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Stanislaw Gruszka <sgruszka@redhat.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org>
2013-02-20 21:54:55 +04:00
}
/*
* Make sure gcc understands that this is a 32x32->64 multiply,
* followed by a 64/32->64 divide.
*/
scaled = div_u64((u64) (u32) stime * (u64) (u32) rtime, (u32)total);
sched: Lower chances of cputime scaling overflow Some users have reported that after running a process with hundreds of threads on intensive CPU-bound loads, the cputime of the group started to freeze after a few days. This is due to how we scale the tick-based cputime against the scheduler precise execution time value. We add the values of all threads in the group and we multiply that against the sum of the scheduler exec runtime of the whole group. This easily overflows after a few days/weeks of execution. A proposed solution to solve this was to compute that multiplication on stime instead of utime: 62188451f0d63add7ad0cd2a1ae269d600c1663d ("cputime: Avoid multiplication overflow on utime scaling") The rationale behind that was that it's easy for a thread to spend most of its time in userspace under intensive CPU-bound workload but it's much harder to do CPU-bound intensive long run in the kernel. This postulate got defeated when a user recently reported he was still seeing cputime freezes after the above patch. The workload that triggers this issue relates to intensive networking workloads where most of the cputime is consumed in the kernel. To reduce much more the opportunities for multiplication overflow, lets reduce the multiplication factors to the remainders of the division between sched exec runtime and cputime. Assuming the difference between these shouldn't ever be that large, it could work on many situations. This gets the same results as in the upstream scaling code except for a small difference: the upstream code always rounds the results to the nearest integer not greater to what would be the precise result. The new code rounds to the nearest integer either greater or not greater. In practice this difference probably shouldn't matter but it's worth mentioning. If this solution appears not to be enough in the end, we'll need to partly revert back to the behaviour prior to commit 0cf55e1ec08bb5a22e068309e2d8ba1180ab4239 ("sched, cputime: Introduce thread_group_times()") Back then, the scaling was done on exit() time before adding the cputime of an exiting thread to the signal struct. And then we'll need to scale one-by-one the live threads cputime in thread_group_cputime(). The drawback may be a slightly slower code on exit time. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Stanislaw Gruszka <sgruszka@redhat.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org>
2013-02-20 21:54:55 +04:00
return (__force cputime_t) scaled;
}
/*
* Adjust tick based cputime random precision against scheduler
* runtime accounting.
*/
static void cputime_adjust(struct task_cputime *curr,
struct cputime *prev,
cputime_t *ut, cputime_t *st)
{
cputime_t rtime, stime, utime;
/*
* Tick based cputime accounting depend on random scheduling
* timeslices of a task to be interrupted or not by the timer.
* Depending on these circumstances, the number of these interrupts
* may be over or under-optimistic, matching the real user and system
* cputime with a variable precision.
*
* Fix this by scaling these tick based values against the total
* runtime accounted by the CFS scheduler.
*/
rtime = nsecs_to_cputime(curr->sum_exec_runtime);
/*
* Update userspace visible utime/stime values only if actual execution
* time is bigger than already exported. Note that can happen, that we
* provided bigger values due to scaling inaccuracy on big numbers.
*/
if (prev->stime + prev->utime >= rtime)
goto out;
stime = curr->stime;
utime = curr->utime;
if (utime == 0) {
stime = rtime;
} else if (stime == 0) {
utime = rtime;
} else {
cputime_t total = stime + utime;
sched: Lower chances of cputime scaling overflow Some users have reported that after running a process with hundreds of threads on intensive CPU-bound loads, the cputime of the group started to freeze after a few days. This is due to how we scale the tick-based cputime against the scheduler precise execution time value. We add the values of all threads in the group and we multiply that against the sum of the scheduler exec runtime of the whole group. This easily overflows after a few days/weeks of execution. A proposed solution to solve this was to compute that multiplication on stime instead of utime: 62188451f0d63add7ad0cd2a1ae269d600c1663d ("cputime: Avoid multiplication overflow on utime scaling") The rationale behind that was that it's easy for a thread to spend most of its time in userspace under intensive CPU-bound workload but it's much harder to do CPU-bound intensive long run in the kernel. This postulate got defeated when a user recently reported he was still seeing cputime freezes after the above patch. The workload that triggers this issue relates to intensive networking workloads where most of the cputime is consumed in the kernel. To reduce much more the opportunities for multiplication overflow, lets reduce the multiplication factors to the remainders of the division between sched exec runtime and cputime. Assuming the difference between these shouldn't ever be that large, it could work on many situations. This gets the same results as in the upstream scaling code except for a small difference: the upstream code always rounds the results to the nearest integer not greater to what would be the precise result. The new code rounds to the nearest integer either greater or not greater. In practice this difference probably shouldn't matter but it's worth mentioning. If this solution appears not to be enough in the end, we'll need to partly revert back to the behaviour prior to commit 0cf55e1ec08bb5a22e068309e2d8ba1180ab4239 ("sched, cputime: Introduce thread_group_times()") Back then, the scaling was done on exit() time before adding the cputime of an exiting thread to the signal struct. And then we'll need to scale one-by-one the live threads cputime in thread_group_cputime(). The drawback may be a slightly slower code on exit time. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Stanislaw Gruszka <sgruszka@redhat.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org>
2013-02-20 21:54:55 +04:00
stime = scale_stime((__force u64)stime,
(__force u64)rtime, (__force u64)total);
utime = rtime - stime;
sched: Lower chances of cputime scaling overflow Some users have reported that after running a process with hundreds of threads on intensive CPU-bound loads, the cputime of the group started to freeze after a few days. This is due to how we scale the tick-based cputime against the scheduler precise execution time value. We add the values of all threads in the group and we multiply that against the sum of the scheduler exec runtime of the whole group. This easily overflows after a few days/weeks of execution. A proposed solution to solve this was to compute that multiplication on stime instead of utime: 62188451f0d63add7ad0cd2a1ae269d600c1663d ("cputime: Avoid multiplication overflow on utime scaling") The rationale behind that was that it's easy for a thread to spend most of its time in userspace under intensive CPU-bound workload but it's much harder to do CPU-bound intensive long run in the kernel. This postulate got defeated when a user recently reported he was still seeing cputime freezes after the above patch. The workload that triggers this issue relates to intensive networking workloads where most of the cputime is consumed in the kernel. To reduce much more the opportunities for multiplication overflow, lets reduce the multiplication factors to the remainders of the division between sched exec runtime and cputime. Assuming the difference between these shouldn't ever be that large, it could work on many situations. This gets the same results as in the upstream scaling code except for a small difference: the upstream code always rounds the results to the nearest integer not greater to what would be the precise result. The new code rounds to the nearest integer either greater or not greater. In practice this difference probably shouldn't matter but it's worth mentioning. If this solution appears not to be enough in the end, we'll need to partly revert back to the behaviour prior to commit 0cf55e1ec08bb5a22e068309e2d8ba1180ab4239 ("sched, cputime: Introduce thread_group_times()") Back then, the scaling was done on exit() time before adding the cputime of an exiting thread to the signal struct. And then we'll need to scale one-by-one the live threads cputime in thread_group_cputime(). The drawback may be a slightly slower code on exit time. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Stanislaw Gruszka <sgruszka@redhat.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org>
2013-02-20 21:54:55 +04:00
}
/*
* If the tick based count grows faster than the scheduler one,
* the result of the scaling may go backward.
* Let's enforce monotonicity.
*/
cputime: Avoid multiplication overflow on utime scaling We scale stime, utime values based on rtime (sum_exec_runtime converted to jiffies). During scaling we multiple rtime * utime, which seems to be fine, since both values are converted to u64, but it's not. Let assume HZ is 1000 - 1ms tick. Process consist of 64 threads, run for 1 day, threads utilize 100% cpu on user space. Machine has 64 cpus. Process rtime = utime will be 64 * 24 * 60 * 60 * 1000 jiffies, which is 0x149970000. Multiplication rtime * utime result is 0x1a855771100000000, which can not be covered in 64 bits. Result of overflow is stall of utime values visible in user space (prev_utime in kernel), even if application still consume lot of CPU time. A solution to solve this is to perform the multiplication on stime instead of utime. It's easy to grow the utime value fast with a CPU bound thread in userspace for example. Now we assume that doing so with stime is much harder. In most cases a task shouldn't ever spend much time in kernel space as it tends to sleep waiting for jobs completion when they take long to achieve. IO is the typical example of that. Hence scaling the cputime by performing the multiplication on stime instead of utime should considerably reduce the chances of an overflow on most workloads. This is largely inspired by a patch from Stanislaw Gruszka: http://lkml.kernel.org/r/20130107113144.GA7544@redhat.com Inspired-by: Stanislaw Gruszka <sgruszka@redhat.com> Reported-by: Stanislaw Gruszka <sgruszka@redhat.com> Acked-by: Stanislaw Gruszka <sgruszka@redhat.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Andrew Morton <akpm@linux-foundation.org> Link: http://lkml.kernel.org/r/1359217182-25184-1-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2013-01-26 20:19:42 +04:00
prev->stime = max(prev->stime, stime);
prev->utime = max(prev->utime, utime);
out:
*ut = prev->utime;
*st = prev->stime;
}
void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st)
{
struct task_cputime cputime = {
.sum_exec_runtime = p->se.sum_exec_runtime,
};
task_cputime(p, &cputime.utime, &cputime.stime);
cputime_adjust(&cputime, &p->prev_cputime, ut, st);
}
/*
* Must be called with siglock held.
*/
void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st)
{
struct task_cputime cputime;
thread_group_cputime(p, &cputime);
cputime_adjust(&cputime, &p->signal->prev_cputime, ut, st);
}
#endif /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
static unsigned long long vtime_delta(struct task_struct *tsk)
{
unsigned long long clock;
cputime: Use local_clock() for full dynticks cputime accounting Running the full dynticks cputime accounting with preemptible kernel debugging trigger the following warning: [ 4.488303] BUG: using smp_processor_id() in preemptible [00000000] code: init/1 [ 4.490971] caller is native_sched_clock+0x22/0x80 [ 4.493663] Pid: 1, comm: init Not tainted 3.8.0+ #13 [ 4.496376] Call Trace: [ 4.498996] [<ffffffff813410eb>] debug_smp_processor_id+0xdb/0xf0 [ 4.501716] [<ffffffff8101e642>] native_sched_clock+0x22/0x80 [ 4.504434] [<ffffffff8101db99>] sched_clock+0x9/0x10 [ 4.507185] [<ffffffff81096ccd>] fetch_task_cputime+0xad/0x120 [ 4.509916] [<ffffffff81096dd5>] task_cputime+0x35/0x60 [ 4.512622] [<ffffffff810f146e>] acct_update_integrals+0x1e/0x40 [ 4.515372] [<ffffffff8117d2cf>] do_execve_common+0x4ff/0x5c0 [ 4.518117] [<ffffffff8117cf14>] ? do_execve_common+0x144/0x5c0 [ 4.520844] [<ffffffff81867a10>] ? rest_init+0x160/0x160 [ 4.523554] [<ffffffff8117d457>] do_execve+0x37/0x40 [ 4.526276] [<ffffffff810021a3>] run_init_process+0x23/0x30 [ 4.528953] [<ffffffff81867aac>] kernel_init+0x9c/0xf0 [ 4.531608] [<ffffffff8188356c>] ret_from_fork+0x7c/0xb0 We use sched_clock() to perform and fixup the cputime accounting. However we are calling it with preemption enabled from the read side, which trigger the bug above. To fix this up, use local_clock() instead. It takes care of preemption and also provide a more reliable clock source. This is welcome for this kind of statistic that is widely relied on in userspace. Reported-by: Thomas Gleixner <tglx@linutronix.de> Reported-by: Ingo Molnar <mingo@kernel.org> Suggested-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Kevin Hilman <khilman@linaro.org> Link: http://lkml.kernel.org/r/1361636925-22288-3-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2013-02-23 20:28:45 +04:00
clock = local_clock();
if (clock < tsk->vtime_snap)
return 0;
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
return clock - tsk->vtime_snap;
}
static cputime_t get_vtime_delta(struct task_struct *tsk)
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
{
unsigned long long delta = vtime_delta(tsk);
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
WARN_ON_ONCE(tsk->vtime_snap_whence == VTIME_SLEEPING);
tsk->vtime_snap += delta;
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
/* CHECKME: always safe to convert nsecs to cputime? */
return nsecs_to_cputime(delta);
}
static void __vtime_account_system(struct task_struct *tsk)
{
cputime_t delta_cpu = get_vtime_delta(tsk);
account_system_time(tsk, irq_count(), delta_cpu, cputime_to_scaled(delta_cpu));
}
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
void vtime_account_system(struct task_struct *tsk)
{
write_seqlock(&tsk->vtime_seqlock);
__vtime_account_system(tsk);
write_sequnlock(&tsk->vtime_seqlock);
}
void vtime_gen_account_irq_exit(struct task_struct *tsk)
{
write_seqlock(&tsk->vtime_seqlock);
__vtime_account_system(tsk);
if (context_tracking_in_user())
tsk->vtime_snap_whence = VTIME_USER;
write_sequnlock(&tsk->vtime_seqlock);
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
}
void vtime_account_user(struct task_struct *tsk)
{
cputime_t delta_cpu;
write_seqlock(&tsk->vtime_seqlock);
delta_cpu = get_vtime_delta(tsk);
tsk->vtime_snap_whence = VTIME_SYS;
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
account_user_time(tsk, delta_cpu, cputime_to_scaled(delta_cpu));
write_sequnlock(&tsk->vtime_seqlock);
}
void vtime_user_enter(struct task_struct *tsk)
{
write_seqlock(&tsk->vtime_seqlock);
__vtime_account_system(tsk);
tsk->vtime_snap_whence = VTIME_USER;
write_sequnlock(&tsk->vtime_seqlock);
}
void vtime_guest_enter(struct task_struct *tsk)
{
/*
* The flags must be updated under the lock with
* the vtime_snap flush and update.
* That enforces a right ordering and update sequence
* synchronization against the reader (task_gtime())
* that can thus safely catch up with a tickless delta.
*/
write_seqlock(&tsk->vtime_seqlock);
__vtime_account_system(tsk);
current->flags |= PF_VCPU;
write_sequnlock(&tsk->vtime_seqlock);
}
EXPORT_SYMBOL_GPL(vtime_guest_enter);
void vtime_guest_exit(struct task_struct *tsk)
{
write_seqlock(&tsk->vtime_seqlock);
__vtime_account_system(tsk);
current->flags &= ~PF_VCPU;
write_sequnlock(&tsk->vtime_seqlock);
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
}
EXPORT_SYMBOL_GPL(vtime_guest_exit);
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
void vtime_account_idle(struct task_struct *tsk)
{
cputime_t delta_cpu = get_vtime_delta(tsk);
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
account_idle_time(delta_cpu);
}
void arch_vtime_task_switch(struct task_struct *prev)
{
write_seqlock(&prev->vtime_seqlock);
prev->vtime_snap_whence = VTIME_SLEEPING;
write_sequnlock(&prev->vtime_seqlock);
write_seqlock(&current->vtime_seqlock);
current->vtime_snap_whence = VTIME_SYS;
current->vtime_snap = sched_clock_cpu(smp_processor_id());
write_sequnlock(&current->vtime_seqlock);
}
void vtime_init_idle(struct task_struct *t, int cpu)
{
unsigned long flags;
write_seqlock_irqsave(&t->vtime_seqlock, flags);
t->vtime_snap_whence = VTIME_SYS;
t->vtime_snap = sched_clock_cpu(cpu);
write_sequnlock_irqrestore(&t->vtime_seqlock, flags);
}
cputime_t task_gtime(struct task_struct *t)
{
unsigned int seq;
cputime_t gtime;
do {
seq = read_seqbegin(&t->vtime_seqlock);
gtime = t->gtime;
if (t->flags & PF_VCPU)
gtime += vtime_delta(t);
} while (read_seqretry(&t->vtime_seqlock, seq));
return gtime;
}
/*
* Fetch cputime raw values from fields of task_struct and
* add up the pending nohz execution time since the last
* cputime snapshot.
*/
static void
fetch_task_cputime(struct task_struct *t,
cputime_t *u_dst, cputime_t *s_dst,
cputime_t *u_src, cputime_t *s_src,
cputime_t *udelta, cputime_t *sdelta)
{
unsigned int seq;
unsigned long long delta;
do {
*udelta = 0;
*sdelta = 0;
seq = read_seqbegin(&t->vtime_seqlock);
if (u_dst)
*u_dst = *u_src;
if (s_dst)
*s_dst = *s_src;
/* Task is sleeping, nothing to add */
if (t->vtime_snap_whence == VTIME_SLEEPING ||
is_idle_task(t))
continue;
delta = vtime_delta(t);
/*
* Task runs either in user or kernel space, add pending nohz time to
* the right place.
*/
if (t->vtime_snap_whence == VTIME_USER || t->flags & PF_VCPU) {
*udelta = delta;
} else {
if (t->vtime_snap_whence == VTIME_SYS)
*sdelta = delta;
}
} while (read_seqretry(&t->vtime_seqlock, seq));
}
void task_cputime(struct task_struct *t, cputime_t *utime, cputime_t *stime)
{
cputime_t udelta, sdelta;
fetch_task_cputime(t, utime, stime, &t->utime,
&t->stime, &udelta, &sdelta);
if (utime)
*utime += udelta;
if (stime)
*stime += sdelta;
}
void task_cputime_scaled(struct task_struct *t,
cputime_t *utimescaled, cputime_t *stimescaled)
{
cputime_t udelta, sdelta;
fetch_task_cputime(t, utimescaled, stimescaled,
&t->utimescaled, &t->stimescaled, &udelta, &sdelta);
if (utimescaled)
*utimescaled += cputime_to_scaled(udelta);
if (stimescaled)
*stimescaled += cputime_to_scaled(sdelta);
}
cputime: Generic on-demand virtual cputime accounting If we want to stop the tick further idle, we need to be able to account the cputime without using the tick. Virtual based cputime accounting solves that problem by hooking into kernel/user boundaries. However implementing CONFIG_VIRT_CPU_ACCOUNTING require low level hooks and involves more overhead. But we already have a generic context tracking subsystem that is required for RCU needs by archs which plan to shut down the tick outside idle. This patch implements a generic virtual based cputime accounting that relies on these generic kernel/user hooks. There are some upsides of doing this: - This requires no arch code to implement CONFIG_VIRT_CPU_ACCOUNTING if context tracking is already built (already necessary for RCU in full tickless mode). - We can rely on the generic context tracking subsystem to dynamically (de)activate the hooks, so that we can switch anytime between virtual and tick based accounting. This way we don't have the overhead of the virtual accounting when the tick is running periodically. And one downside: - There is probably more overhead than a native virtual based cputime accounting. But this relies on hooks that are already set anyway. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2012-07-25 09:56:04 +04:00
#endif /* CONFIG_VIRT_CPU_ACCOUNTING_GEN */