linux/drivers/cpuidle/cpuidle.c
Mel Gorman 7a25759eaa cpuidle: Select polling interval based on a c-state with a longer target residency
It was noted that a few workloads that idle rapidly regressed when commit
36fcb42924 ("cpuidle: use first valid target residency as poll time")
was merged. The workloads in question were heavy communicators that idle
rapidly and were impacted by the c-state exit latency as the active CPUs
were not polling at the time of wakeup. As they were not particularly
realistic workloads, it was not considered to be a major problem.

Unfortunately, a bug was reported for a real workload in a production
environment that relied on large numbers of threads operating in a worker
pool pattern. These threads would idle for periods of time longer than the
C1 target residency and so incurred the c-state exit latency penalty. The
application is very sensitive to wakeup latency and indirectly relying
on behaviour prior to commit on a37b969a61 ("cpuidle: poll_state: Add
time limit to poll_idle()") to poll for long enough to avoid the exit
latency cost.

The target residency of C1 is typically very short. On some x86 machines,
it can be as low as 2 microseconds. In poll_idle(), the clock is checked
every POLL_IDLE_RELAX_COUNT interations of cpu_relax() and even one
iteration of that loop can be over 1 microsecond so the polling interval is
very close to the granularity of what poll_idle() can detect. Furthermore,
a basic ping pong workload like perf bench pipe has a longer round-trip
time than the 2 microseconds meaning that the CPU will almost certainly
not be polling when the ping-pong completes.

This patch selects a polling interval based on an enabled c-state that
has an target residency longer than 10usec. If there is no enabled-cstate
then polling will be up to a TICK_NSEC/16 similar to what it was up until
kernel 4.20. Polling for a full tick is unlikely (rescheduling event)
and is much longer than the existing target residencies for a deep c-state.

As an example, consider a CPU with the following c-state information from
an Intel CPU;

	residency	exit_latency
C1	2		2
C1E	20		10
C3	100		33
C6	400		133

The polling interval selected is 20usec. If booted with
intel_idle.max_cstate=1 then the polling interval is 250usec as the deeper
c-states were not available.

On an AMD EPYC machine, the c-state information is more limited and
looks like

	residency	exit_latency
C1	2		1
C2	800		400

The polling interval selected is 250usec. While C2 was considered, the
polling interval was clamped by CPUIDLE_POLL_MAX.

Note that it is not expected that polling will be a universal win. As
well as potentially trading power for performance, the performance is not
guaranteed if the extra polling prevented a turbo state being reached.
Making it a tunable was considered but it's driver-specific, may be
overridden by a governor and is not a guaranteed polling interval making
it difficult to describe without knowledge of the implementation.

tbench4
			     vanilla		    polling
Hmean     1        497.89 (   0.00%)      543.15 *   9.09%*
Hmean     2        975.88 (   0.00%)     1059.73 *   8.59%*
Hmean     4       1953.97 (   0.00%)     2081.37 *   6.52%*
Hmean     8       3645.76 (   0.00%)     4052.95 *  11.17%*
Hmean     16      6882.21 (   0.00%)     6995.93 *   1.65%*
Hmean     32     10752.20 (   0.00%)    10731.53 *  -0.19%*
Hmean     64     12875.08 (   0.00%)    12478.13 *  -3.08%*
Hmean     128    21500.54 (   0.00%)    21098.60 *  -1.87%*
Hmean     256    21253.70 (   0.00%)    21027.18 *  -1.07%*
Hmean     320    20813.50 (   0.00%)    20580.64 *  -1.12%*

Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2020-12-01 17:59:18 +01:00

780 lines
18 KiB
C

/*
* cpuidle.c - core cpuidle infrastructure
*
* (C) 2006-2007 Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
* Shaohua Li <shaohua.li@intel.com>
* Adam Belay <abelay@novell.com>
*
* This code is licenced under the GPL.
*/
#include <linux/clockchips.h>
#include <linux/kernel.h>
#include <linux/mutex.h>
#include <linux/sched.h>
#include <linux/sched/clock.h>
#include <linux/notifier.h>
#include <linux/pm_qos.h>
#include <linux/cpu.h>
#include <linux/cpuidle.h>
#include <linux/ktime.h>
#include <linux/hrtimer.h>
#include <linux/module.h>
#include <linux/suspend.h>
#include <linux/tick.h>
#include <linux/mmu_context.h>
#include <trace/events/power.h>
#include "cpuidle.h"
DEFINE_PER_CPU(struct cpuidle_device *, cpuidle_devices);
DEFINE_PER_CPU(struct cpuidle_device, cpuidle_dev);
DEFINE_MUTEX(cpuidle_lock);
LIST_HEAD(cpuidle_detected_devices);
static int enabled_devices;
static int off __read_mostly;
static int initialized __read_mostly;
int cpuidle_disabled(void)
{
return off;
}
void disable_cpuidle(void)
{
off = 1;
}
bool cpuidle_not_available(struct cpuidle_driver *drv,
struct cpuidle_device *dev)
{
return off || !initialized || !drv || !dev || !dev->enabled;
}
/**
* cpuidle_play_dead - cpu off-lining
*
* Returns in case of an error or no driver
*/
int cpuidle_play_dead(void)
{
struct cpuidle_device *dev = __this_cpu_read(cpuidle_devices);
struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
int i;
if (!drv)
return -ENODEV;
/* Find lowest-power state that supports long-term idle */
for (i = drv->state_count - 1; i >= 0; i--)
if (drv->states[i].enter_dead)
return drv->states[i].enter_dead(dev, i);
return -ENODEV;
}
static int find_deepest_state(struct cpuidle_driver *drv,
struct cpuidle_device *dev,
u64 max_latency_ns,
unsigned int forbidden_flags,
bool s2idle)
{
u64 latency_req = 0;
int i, ret = 0;
for (i = 1; i < drv->state_count; i++) {
struct cpuidle_state *s = &drv->states[i];
if (dev->states_usage[i].disable ||
s->exit_latency_ns <= latency_req ||
s->exit_latency_ns > max_latency_ns ||
(s->flags & forbidden_flags) ||
(s2idle && !s->enter_s2idle))
continue;
latency_req = s->exit_latency_ns;
ret = i;
}
return ret;
}
/**
* cpuidle_use_deepest_state - Set/unset governor override mode.
* @latency_limit_ns: Idle state exit latency limit (or no override if 0).
*
* If @latency_limit_ns is nonzero, set the current CPU to use the deepest idle
* state with exit latency within @latency_limit_ns (override governors going
* forward), or do not override governors if it is zero.
*/
void cpuidle_use_deepest_state(u64 latency_limit_ns)
{
struct cpuidle_device *dev;
preempt_disable();
dev = cpuidle_get_device();
if (dev)
dev->forced_idle_latency_limit_ns = latency_limit_ns;
preempt_enable();
}
/**
* cpuidle_find_deepest_state - Find the deepest available idle state.
* @drv: cpuidle driver for the given CPU.
* @dev: cpuidle device for the given CPU.
* @latency_limit_ns: Idle state exit latency limit
*
* Return: the index of the deepest available idle state.
*/
int cpuidle_find_deepest_state(struct cpuidle_driver *drv,
struct cpuidle_device *dev,
u64 latency_limit_ns)
{
return find_deepest_state(drv, dev, latency_limit_ns, 0, false);
}
#ifdef CONFIG_SUSPEND
static void enter_s2idle_proper(struct cpuidle_driver *drv,
struct cpuidle_device *dev, int index)
{
ktime_t time_start, time_end;
struct cpuidle_state *target_state = &drv->states[index];
time_start = ns_to_ktime(local_clock());
tick_freeze();
/*
* The state used here cannot be a "coupled" one, because the "coupled"
* cpuidle mechanism enables interrupts and doing that with timekeeping
* suspended is generally unsafe.
*/
stop_critical_timings();
if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
rcu_idle_enter();
target_state->enter_s2idle(dev, drv, index);
if (WARN_ON_ONCE(!irqs_disabled()))
local_irq_disable();
if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
rcu_idle_exit();
tick_unfreeze();
start_critical_timings();
time_end = ns_to_ktime(local_clock());
dev->states_usage[index].s2idle_time += ktime_us_delta(time_end, time_start);
dev->states_usage[index].s2idle_usage++;
}
/**
* cpuidle_enter_s2idle - Enter an idle state suitable for suspend-to-idle.
* @drv: cpuidle driver for the given CPU.
* @dev: cpuidle device for the given CPU.
*
* If there are states with the ->enter_s2idle callback, find the deepest of
* them and enter it with frozen tick.
*/
int cpuidle_enter_s2idle(struct cpuidle_driver *drv, struct cpuidle_device *dev)
{
int index;
/*
* Find the deepest state with ->enter_s2idle present, which guarantees
* that interrupts won't be enabled when it exits and allows the tick to
* be frozen safely.
*/
index = find_deepest_state(drv, dev, U64_MAX, 0, true);
if (index > 0) {
enter_s2idle_proper(drv, dev, index);
local_irq_enable();
}
return index;
}
#endif /* CONFIG_SUSPEND */
/**
* cpuidle_enter_state - enter the state and update stats
* @dev: cpuidle device for this cpu
* @drv: cpuidle driver for this cpu
* @index: index into the states table in @drv of the state to enter
*/
int cpuidle_enter_state(struct cpuidle_device *dev, struct cpuidle_driver *drv,
int index)
{
int entered_state;
struct cpuidle_state *target_state = &drv->states[index];
bool broadcast = !!(target_state->flags & CPUIDLE_FLAG_TIMER_STOP);
ktime_t time_start, time_end;
/*
* Tell the time framework to switch to a broadcast timer because our
* local timer will be shut down. If a local timer is used from another
* CPU as a broadcast timer, this call may fail if it is not available.
*/
if (broadcast && tick_broadcast_enter()) {
index = find_deepest_state(drv, dev, target_state->exit_latency_ns,
CPUIDLE_FLAG_TIMER_STOP, false);
if (index < 0) {
default_idle_call();
return -EBUSY;
}
target_state = &drv->states[index];
broadcast = false;
}
if (target_state->flags & CPUIDLE_FLAG_TLB_FLUSHED)
leave_mm(dev->cpu);
/* Take note of the planned idle state. */
sched_idle_set_state(target_state);
trace_cpu_idle(index, dev->cpu);
time_start = ns_to_ktime(local_clock());
stop_critical_timings();
if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
rcu_idle_enter();
entered_state = target_state->enter(dev, drv, index);
if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
rcu_idle_exit();
start_critical_timings();
sched_clock_idle_wakeup_event();
time_end = ns_to_ktime(local_clock());
trace_cpu_idle(PWR_EVENT_EXIT, dev->cpu);
/* The cpu is no longer idle or about to enter idle. */
sched_idle_set_state(NULL);
if (broadcast) {
if (WARN_ON_ONCE(!irqs_disabled()))
local_irq_disable();
tick_broadcast_exit();
}
if (!cpuidle_state_is_coupled(drv, index))
local_irq_enable();
if (entered_state >= 0) {
s64 diff, delay = drv->states[entered_state].exit_latency_ns;
int i;
/*
* Update cpuidle counters
* This can be moved to within driver enter routine,
* but that results in multiple copies of same code.
*/
diff = ktime_sub(time_end, time_start);
dev->last_residency_ns = diff;
dev->states_usage[entered_state].time_ns += diff;
dev->states_usage[entered_state].usage++;
if (diff < drv->states[entered_state].target_residency_ns) {
for (i = entered_state - 1; i >= 0; i--) {
if (dev->states_usage[i].disable)
continue;
/* Shallower states are enabled, so update. */
dev->states_usage[entered_state].above++;
break;
}
} else if (diff > delay) {
for (i = entered_state + 1; i < drv->state_count; i++) {
if (dev->states_usage[i].disable)
continue;
/*
* Update if a deeper state would have been a
* better match for the observed idle duration.
*/
if (diff - delay >= drv->states[i].target_residency_ns)
dev->states_usage[entered_state].below++;
break;
}
}
} else {
dev->last_residency_ns = 0;
dev->states_usage[index].rejected++;
}
return entered_state;
}
/**
* cpuidle_select - ask the cpuidle framework to choose an idle state
*
* @drv: the cpuidle driver
* @dev: the cpuidle device
* @stop_tick: indication on whether or not to stop the tick
*
* Returns the index of the idle state. The return value must not be negative.
*
* The memory location pointed to by @stop_tick is expected to be written the
* 'false' boolean value if the scheduler tick should not be stopped before
* entering the returned state.
*/
int cpuidle_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
bool *stop_tick)
{
return cpuidle_curr_governor->select(drv, dev, stop_tick);
}
/**
* cpuidle_enter - enter into the specified idle state
*
* @drv: the cpuidle driver tied with the cpu
* @dev: the cpuidle device
* @index: the index in the idle state table
*
* Returns the index in the idle state, < 0 in case of error.
* The error code depends on the backend driver
*/
int cpuidle_enter(struct cpuidle_driver *drv, struct cpuidle_device *dev,
int index)
{
int ret = 0;
/*
* Store the next hrtimer, which becomes either next tick or the next
* timer event, whatever expires first. Additionally, to make this data
* useful for consumers outside cpuidle, we rely on that the governor's
* ->select() callback have decided, whether to stop the tick or not.
*/
WRITE_ONCE(dev->next_hrtimer, tick_nohz_get_next_hrtimer());
if (cpuidle_state_is_coupled(drv, index))
ret = cpuidle_enter_state_coupled(dev, drv, index);
else
ret = cpuidle_enter_state(dev, drv, index);
WRITE_ONCE(dev->next_hrtimer, 0);
return ret;
}
/**
* cpuidle_reflect - tell the underlying governor what was the state
* we were in
*
* @dev : the cpuidle device
* @index: the index in the idle state table
*
*/
void cpuidle_reflect(struct cpuidle_device *dev, int index)
{
if (cpuidle_curr_governor->reflect && index >= 0)
cpuidle_curr_governor->reflect(dev, index);
}
/*
* Min polling interval of 10usec is a guess. It is assuming that
* for most users, the time for a single ping-pong workload like
* perf bench pipe would generally complete within 10usec but
* this is hardware dependant. Actual time can be estimated with
*
* perf bench sched pipe -l 10000
*
* Run multiple times to avoid cpufreq effects.
*/
#define CPUIDLE_POLL_MIN 10000
#define CPUIDLE_POLL_MAX (TICK_NSEC / 16)
/**
* cpuidle_poll_time - return amount of time to poll for,
* governors can override dev->poll_limit_ns if necessary
*
* @drv: the cpuidle driver tied with the cpu
* @dev: the cpuidle device
*
*/
u64 cpuidle_poll_time(struct cpuidle_driver *drv,
struct cpuidle_device *dev)
{
int i;
u64 limit_ns;
BUILD_BUG_ON(CPUIDLE_POLL_MIN > CPUIDLE_POLL_MAX);
if (dev->poll_limit_ns)
return dev->poll_limit_ns;
limit_ns = CPUIDLE_POLL_MAX;
for (i = 1; i < drv->state_count; i++) {
u64 state_limit;
if (dev->states_usage[i].disable)
continue;
state_limit = drv->states[i].target_residency_ns;
if (state_limit < CPUIDLE_POLL_MIN)
continue;
limit_ns = min_t(u64, state_limit, CPUIDLE_POLL_MAX);
break;
}
dev->poll_limit_ns = limit_ns;
return dev->poll_limit_ns;
}
/**
* cpuidle_install_idle_handler - installs the cpuidle idle loop handler
*/
void cpuidle_install_idle_handler(void)
{
if (enabled_devices) {
/* Make sure all changes finished before we switch to new idle */
smp_wmb();
initialized = 1;
}
}
/**
* cpuidle_uninstall_idle_handler - uninstalls the cpuidle idle loop handler
*/
void cpuidle_uninstall_idle_handler(void)
{
if (enabled_devices) {
initialized = 0;
wake_up_all_idle_cpus();
}
/*
* Make sure external observers (such as the scheduler)
* are done looking at pointed idle states.
*/
synchronize_rcu();
}
/**
* cpuidle_pause_and_lock - temporarily disables CPUIDLE
*/
void cpuidle_pause_and_lock(void)
{
mutex_lock(&cpuidle_lock);
cpuidle_uninstall_idle_handler();
}
EXPORT_SYMBOL_GPL(cpuidle_pause_and_lock);
/**
* cpuidle_resume_and_unlock - resumes CPUIDLE operation
*/
void cpuidle_resume_and_unlock(void)
{
cpuidle_install_idle_handler();
mutex_unlock(&cpuidle_lock);
}
EXPORT_SYMBOL_GPL(cpuidle_resume_and_unlock);
/* Currently used in suspend/resume path to suspend cpuidle */
void cpuidle_pause(void)
{
mutex_lock(&cpuidle_lock);
cpuidle_uninstall_idle_handler();
mutex_unlock(&cpuidle_lock);
}
/* Currently used in suspend/resume path to resume cpuidle */
void cpuidle_resume(void)
{
mutex_lock(&cpuidle_lock);
cpuidle_install_idle_handler();
mutex_unlock(&cpuidle_lock);
}
/**
* cpuidle_enable_device - enables idle PM for a CPU
* @dev: the CPU
*
* This function must be called between cpuidle_pause_and_lock and
* cpuidle_resume_and_unlock when used externally.
*/
int cpuidle_enable_device(struct cpuidle_device *dev)
{
int ret;
struct cpuidle_driver *drv;
if (!dev)
return -EINVAL;
if (dev->enabled)
return 0;
if (!cpuidle_curr_governor)
return -EIO;
drv = cpuidle_get_cpu_driver(dev);
if (!drv)
return -EIO;
if (!dev->registered)
return -EINVAL;
ret = cpuidle_add_device_sysfs(dev);
if (ret)
return ret;
if (cpuidle_curr_governor->enable) {
ret = cpuidle_curr_governor->enable(drv, dev);
if (ret)
goto fail_sysfs;
}
smp_wmb();
dev->enabled = 1;
enabled_devices++;
return 0;
fail_sysfs:
cpuidle_remove_device_sysfs(dev);
return ret;
}
EXPORT_SYMBOL_GPL(cpuidle_enable_device);
/**
* cpuidle_disable_device - disables idle PM for a CPU
* @dev: the CPU
*
* This function must be called between cpuidle_pause_and_lock and
* cpuidle_resume_and_unlock when used externally.
*/
void cpuidle_disable_device(struct cpuidle_device *dev)
{
struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
if (!dev || !dev->enabled)
return;
if (!drv || !cpuidle_curr_governor)
return;
dev->enabled = 0;
if (cpuidle_curr_governor->disable)
cpuidle_curr_governor->disable(drv, dev);
cpuidle_remove_device_sysfs(dev);
enabled_devices--;
}
EXPORT_SYMBOL_GPL(cpuidle_disable_device);
static void __cpuidle_unregister_device(struct cpuidle_device *dev)
{
struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
list_del(&dev->device_list);
per_cpu(cpuidle_devices, dev->cpu) = NULL;
module_put(drv->owner);
dev->registered = 0;
}
static void __cpuidle_device_init(struct cpuidle_device *dev)
{
memset(dev->states_usage, 0, sizeof(dev->states_usage));
dev->last_residency_ns = 0;
dev->next_hrtimer = 0;
}
/**
* __cpuidle_register_device - internal register function called before register
* and enable routines
* @dev: the cpu
*
* cpuidle_lock mutex must be held before this is called
*/
static int __cpuidle_register_device(struct cpuidle_device *dev)
{
struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
int i, ret;
if (!try_module_get(drv->owner))
return -EINVAL;
for (i = 0; i < drv->state_count; i++) {
if (drv->states[i].flags & CPUIDLE_FLAG_UNUSABLE)
dev->states_usage[i].disable |= CPUIDLE_STATE_DISABLED_BY_DRIVER;
if (drv->states[i].flags & CPUIDLE_FLAG_OFF)
dev->states_usage[i].disable |= CPUIDLE_STATE_DISABLED_BY_USER;
}
per_cpu(cpuidle_devices, dev->cpu) = dev;
list_add(&dev->device_list, &cpuidle_detected_devices);
ret = cpuidle_coupled_register_device(dev);
if (ret)
__cpuidle_unregister_device(dev);
else
dev->registered = 1;
return ret;
}
/**
* cpuidle_register_device - registers a CPU's idle PM feature
* @dev: the cpu
*/
int cpuidle_register_device(struct cpuidle_device *dev)
{
int ret = -EBUSY;
if (!dev)
return -EINVAL;
mutex_lock(&cpuidle_lock);
if (dev->registered)
goto out_unlock;
__cpuidle_device_init(dev);
ret = __cpuidle_register_device(dev);
if (ret)
goto out_unlock;
ret = cpuidle_add_sysfs(dev);
if (ret)
goto out_unregister;
ret = cpuidle_enable_device(dev);
if (ret)
goto out_sysfs;
cpuidle_install_idle_handler();
out_unlock:
mutex_unlock(&cpuidle_lock);
return ret;
out_sysfs:
cpuidle_remove_sysfs(dev);
out_unregister:
__cpuidle_unregister_device(dev);
goto out_unlock;
}
EXPORT_SYMBOL_GPL(cpuidle_register_device);
/**
* cpuidle_unregister_device - unregisters a CPU's idle PM feature
* @dev: the cpu
*/
void cpuidle_unregister_device(struct cpuidle_device *dev)
{
if (!dev || dev->registered == 0)
return;
cpuidle_pause_and_lock();
cpuidle_disable_device(dev);
cpuidle_remove_sysfs(dev);
__cpuidle_unregister_device(dev);
cpuidle_coupled_unregister_device(dev);
cpuidle_resume_and_unlock();
}
EXPORT_SYMBOL_GPL(cpuidle_unregister_device);
/**
* cpuidle_unregister: unregister a driver and the devices. This function
* can be used only if the driver has been previously registered through
* the cpuidle_register function.
*
* @drv: a valid pointer to a struct cpuidle_driver
*/
void cpuidle_unregister(struct cpuidle_driver *drv)
{
int cpu;
struct cpuidle_device *device;
for_each_cpu(cpu, drv->cpumask) {
device = &per_cpu(cpuidle_dev, cpu);
cpuidle_unregister_device(device);
}
cpuidle_unregister_driver(drv);
}
EXPORT_SYMBOL_GPL(cpuidle_unregister);
/**
* cpuidle_register: registers the driver and the cpu devices with the
* coupled_cpus passed as parameter. This function is used for all common
* initialization pattern there are in the arch specific drivers. The
* devices is globally defined in this file.
*
* @drv : a valid pointer to a struct cpuidle_driver
* @coupled_cpus: a cpumask for the coupled states
*
* Returns 0 on success, < 0 otherwise
*/
int cpuidle_register(struct cpuidle_driver *drv,
const struct cpumask *const coupled_cpus)
{
int ret, cpu;
struct cpuidle_device *device;
ret = cpuidle_register_driver(drv);
if (ret) {
pr_err("failed to register cpuidle driver\n");
return ret;
}
for_each_cpu(cpu, drv->cpumask) {
device = &per_cpu(cpuidle_dev, cpu);
device->cpu = cpu;
#ifdef CONFIG_ARCH_NEEDS_CPU_IDLE_COUPLED
/*
* On multiplatform for ARM, the coupled idle states could be
* enabled in the kernel even if the cpuidle driver does not
* use it. Note, coupled_cpus is a struct copy.
*/
if (coupled_cpus)
device->coupled_cpus = *coupled_cpus;
#endif
ret = cpuidle_register_device(device);
if (!ret)
continue;
pr_err("Failed to register cpuidle device for cpu%d\n", cpu);
cpuidle_unregister(drv);
break;
}
return ret;
}
EXPORT_SYMBOL_GPL(cpuidle_register);
/**
* cpuidle_init - core initializer
*/
static int __init cpuidle_init(void)
{
if (cpuidle_disabled())
return -ENODEV;
return cpuidle_add_interface(cpu_subsys.dev_root);
}
module_param(off, int, 0444);
module_param_string(governor, param_governor, CPUIDLE_NAME_LEN, 0444);
core_initcall(cpuidle_init);