1#@!#!123s

: / proc / thread-self / root / proc / self / root / usr / lib / python3.6 / site-packages / tuned / plugins /
Filename : plugin_cpu.py
back
from . import hotplug
from .decorators import *
import tuned.logs
from tuned.utils.commands import commands
import tuned.consts as consts

import os
import errno
import struct
import errno
import platform
import procfs

log = tuned.logs.get()

cpuidle_states_path = "/sys/devices/system/cpu/cpu0/cpuidle"

class CPULatencyPlugin(hotplug.Plugin):
	"""
	`cpu`::
	
	Sets the CPU governor to the value specified by the [option]`governor`
	option and dynamically changes the Power Management Quality of
	Service (PM QoS) CPU Direct Memory Access (DMA) latency according
	to the CPU load.
	
	`governor`:::
	The [option]`governor` option of the 'cpu' plug-in supports specifying
	CPU governors. Multiple governors are separated using '|'. The '|'
	character is meant to represent a logical 'or' operator. Note that the
	same syntax is used for the [option]`energy_perf_bias` option. *TuneD*
	will set the first governor that is available on the system.
	+    
	For example, with the following profile, *TuneD* will set the 'ondemand'
	governor, if it is available. If it is not available, but the 'powersave'
	governor is available, 'powersave' will be set. If neither of them are
	available, the governor will not be changed.
	+
	.Specifying a CPU governor
	====
	----
	[cpu]
	governor=ondemand|powersave
	----
	====
	
	`sampling_down_factor`:::
	The sampling rate determines how frequently the governor checks
	to tune the CPU. The [option]`sampling_down_factor` is a tunable
	that multiplies the sampling rate when the CPU is at its highest
	clock frequency thereby delaying load evaluation and improving
	performance. Allowed values for sampling_down_factor are 1 to 100000.
	+
	.The recommended setting for jitter reduction
	====
	----
	[cpu]
	sampling_down_factor = 100
	----
	====
	
	`energy_perf_bias`:::
	[option]`energy_perf_bias` supports managing energy
	vs. performance policy via x86 Model Specific Registers using the
	`x86_energy_perf_policy` tool. Multiple alternative Energy Performance
	Bias (EPB) values are supported. The alternative values are separated
	using the '|' character. The following EPB values are supported
	starting with kernel 4.13: "performance", "balance-performance",
	"normal", "balance-power" and "power". On newer processors is value
	writen straight to file (see rhbz#2095829)
	+
	.Specifying alternative Energy Performance Bias values
	====
	----
	[cpu]
	energy_perf_bias=powersave|power
	----
	
	*TuneD* will try to set EPB to 'powersave'. If that fails, it will
	try to set it to 'power'.
	====
	
	`energy_performance_preference`:::
	[option]`energy_performance_preference` supports managing energy
	vs. performance hints on newer Intel and AMD processors with active P-State
	CPU scaling drivers (intel_pstate or amd-pstate). Multiple alternative
	Energy Performance Preferences (EPP) values are supported. The alternative
	values are separated using the '|' character. Available values can be found
	in `energy_performance_available_preferences` file in `CPUFreq` policy
	directory in `sysfs`.
	in
	+
	.Specifying alternative Energy Performance Hints values
	====
	----
	[cpu]
	energy_performance_preference=balance_power|power
	----
	
	*TuneD* will try to set EPP to 'balance_power'. If that fails, it will
	try to set it to 'power'.
	====
	
	`latency_low, latency_high, load_threshold`:::
	+
	If the CPU load is lower than the value specified by
	the[option]`load_threshold` option, the latency is set to the value
	specified either by the [option]`latency_high` option or by the
	[option]`latency_low` option.
	+
	`force_latency`:::
	You can also force the latency to a specific value and prevent it from
	dynamically changing further. To do so, set the [option]`force_latency`
	option to the required latency value.
	+
	The maximum latency value can be specified in several ways:
	+
	 * by a numerical value in microseconds (for example, `force_latency=10`)
	 * as the kernel CPU idle level ID of the maximum C-state allowed
	   (for example, force_latency = cstate.id:1)
	 * as a case sensitive name of the maximum C-state allowed
	   (for example, force_latency = cstate.name:C1)
	 * by using 'None' as a fallback value to prevent errors when alternative
	   C-state IDs/names do not exist. When 'None' is used in the alternatives
	   pipeline, all the alternatives that follow 'None' are ignored.
	+
	It is also possible to specify multiple fallback values separated by '|' as
	the C-state names and/or IDs may not be available on some systems.
	+
	.Specifying fallback C-state values
	====
	----
	[cpu]
	force_latency=cstate.name:C6|cstate.id:4|10
	----
	This configuration tries to obtain and set the latency of C-state named C6.
	If the C-state C6 does not exist, kernel CPU idle level ID 4 (state4) latency
	is searched for in sysfs. Finally, if the state4 directory in sysfs is not found,
	the last latency fallback value is `10` us. The value is encoded and written into
	the kernel's PM QoS file `/dev/cpu_dma_latency`.
	====
	+
	.Specifying fallback C-state values using 'None'.
	====
	----
	[cpu]
	force_latency=cstate.name:XYZ|None
	----
	In this case, if C-state with the name `XYZ` does not exist
	[option]`force_latency`, no latency value will be written into the
	kernel's PM QoS file, and no errors will be reported due to the
	presence of 'None'.
	====
	
	`min_perf_pct, max_perf_pct, no_turbo`:::
	These options set the internals of the Intel P-State driver exposed via the kernel's
	`sysfs` interface.
	+
	.Adjusting the configuration of the Intel P-State driver
	====
	----
	[cpu]
	min_perf_pct=100
	----
	Limit the minimum P-State that will be requested by the driver. It states
	it as a percentage of the max (non-turbo) performance level.
	====
	+
	`pm_qos_resume_latency_us`:::
	This option allow to set specific latency for all cpus or specific ones.
	====
	----
	[cpu]
	pm_qos_resume_latency_us=n/a
	----
	Special value that disables C-states completely.
	====
	----
	[cpu]
	pm_qos_resume_latency_us=0
	----
	Allows all C-states.
	====
	----
	[cpu]
	pm_qos_resume_latency_us=100
	----
	Allows any C-state with a resume latency less than value.
	"""

	def __init__(self, *args, **kwargs):
		super(CPULatencyPlugin, self).__init__(*args, **kwargs)

		self._has_pm_qos = True
		self._arch = "x86_64"
		self._is_x86 = False
		self._is_intel = False
		self._is_amd = False
		self._has_energy_perf_bias = False
		self._has_intel_pstate = False
		self._has_amd_pstate = False
		self._has_pm_qos_resume_latency_us = None

		self._min_perf_pct_save = None
		self._max_perf_pct_save = None
		self._no_turbo_save = None
		self._governors_map = {}
		self._cmd = commands()

		self._flags = None

	def _init_devices(self):
		self._devices_supported = True
		self._free_devices = set()
		# current list of devices
		for device in self._hardware_inventory.get_devices("cpu"):
			self._free_devices.add(device.sys_name)

		self._assigned_devices = set()

	def _get_device_objects(self, devices):
		return [self._hardware_inventory.get_device("cpu", x) for x in devices]

	@classmethod
	def _get_config_options(self):
		return {
			"load_threshold"       : 0.2,
			"latency_low"          : 100,
			"latency_high"         : 1000,
			"force_latency"        : None,
			"governor"             : None,
			"sampling_down_factor" : None,
			"energy_perf_bias"     : None,
			"min_perf_pct"         : None,
			"max_perf_pct"         : None,
			"no_turbo"             : None,
			"pm_qos_resume_latency_us": None,
			"energy_performance_preference" : None,
		}

	def _check_arch(self):
		intel_archs = [ "x86_64", "i686", "i585", "i486", "i386" ]
		self._arch = platform.machine()

		if self._arch in intel_archs:
			# Possible other x86 vendors (from arch/x86/kernel/cpu/*):
			# "CentaurHauls", "CyrixInstead", "Geode by NSC", "HygonGenuine", "GenuineTMx86",
			# "TransmetaCPU", "UMC UMC UMC"
			cpu = procfs.cpuinfo()
			vendor = cpu.tags.get("vendor_id")
			if vendor == "GenuineIntel":
				self._is_intel = True
			elif vendor == "AuthenticAMD" or vendor == "HygonGenuine":
				self._is_amd = True
			else:
				# We always assign Intel, unless we know better
				self._is_intel = True
			log.info("We are running on an x86 %s platform" % vendor)
		else:
			log.info("We are running on %s (non x86)" % self._arch)

		if self._is_intel:
			# Check for x86_energy_perf_policy, ignore if not available / supported
			self._check_energy_perf_bias()
			# Check for intel_pstate
			self._check_intel_pstate()

		if self._is_amd:
			# Check for amd-pstate
			self._check_amd_pstate()

	def _check_energy_perf_bias(self):
		self._has_energy_perf_bias = False
		retcode_unsupported = 1
		retcode, out = self._cmd.execute(["x86_energy_perf_policy", "-r"], no_errors = [errno.ENOENT, retcode_unsupported])
		# With recent versions of the tool, a zero exit code is
		# returned even if EPB is not supported. The output is empty
		# in that case, however.
		if retcode == 0 and out != "":
			self._has_energy_perf_bias = True
		elif retcode < 0:
			log.warning("unable to run x86_energy_perf_policy tool, ignoring CPU energy performance bias, is the tool installed?")
		else:
			log.warning("your CPU doesn't support MSR_IA32_ENERGY_PERF_BIAS, ignoring CPU energy performance bias")

	def _check_intel_pstate(self):
		self._has_intel_pstate = os.path.exists("/sys/devices/system/cpu/intel_pstate")
		if self._has_intel_pstate:
			log.info("intel_pstate detected")

	def _check_amd_pstate(self):
		self._has_amd_pstate = os.path.exists("/sys/devices/system/cpu/amd_pstate")
		if self._has_amd_pstate:
			log.info("amd-pstate detected")

	def _get_cpuinfo_flags(self):
		if self._flags is None:
			self._flags = procfs.cpuinfo().tags.get("flags", [])
		return self._flags

	def _is_cpu_online(self, device):
		sd = str(device)
		return self._cmd.is_cpu_online(str(device).replace("cpu", ""))

	def _cpu_has_scaling_governor(self, device):
		return os.path.exists("/sys/devices/system/cpu/%s/cpufreq/scaling_governor" % device)

	def _check_cpu_can_change_governor(self, device):
		if not self._is_cpu_online(device):
			log.debug("'%s' is not online, skipping" % device)
			return False
		if not self._cpu_has_scaling_governor(device):
			log.debug("there is no scaling governor fo '%s', skipping" % device)
			return False
		return True

	def _instance_init(self, instance):
		instance._has_static_tuning = True
		instance._has_dynamic_tuning = False

		# only the first instance of the plugin can control the latency
		if list(self._instances.values())[0] == instance:
			instance._first_instance = True
			try:
				self._cpu_latency_fd = os.open(consts.PATH_CPU_DMA_LATENCY, os.O_WRONLY)
			except OSError:
				log.error("Unable to open '%s', disabling PM_QoS control" % consts.PATH_CPU_DMA_LATENCY)
				self._has_pm_qos = False
			self._latency = None

			if instance.options["force_latency"] is None and instance.options["pm_qos_resume_latency_us"] is None:
				instance._load_monitor = self._monitors_repository.create("load", None)
				instance._has_dynamic_tuning = True
			else:
				instance._load_monitor = None

			self._check_arch()
		else:
			instance._first_instance = False
			log.info("Latency settings from non-first CPU plugin instance '%s' will be ignored." % instance.name)

		try:
			instance._first_device = list(instance.assigned_devices)[0]
		except IndexError:
			instance._first_device = None

	def _instance_cleanup(self, instance):
		if instance._first_instance:
			if self._has_pm_qos:
				os.close(self._cpu_latency_fd)
			if instance._load_monitor is not None:
				self._monitors_repository.delete(instance._load_monitor)

	def _get_intel_pstate_attr(self, attr):
		return self._cmd.read_file("/sys/devices/system/cpu/intel_pstate/%s" % attr, None).strip()

	def _set_intel_pstate_attr(self, attr, val):
		if val is not None:
			self._cmd.write_to_file("/sys/devices/system/cpu/intel_pstate/%s" % attr, val)

	def _getset_intel_pstate_attr(self, attr, value):
		if value is None:
			return None
		v = self._get_intel_pstate_attr(attr)
		self._set_intel_pstate_attr(attr, value)
		return v

	def _instance_apply_static(self, instance):
		super(CPULatencyPlugin, self)._instance_apply_static(instance)

		if not instance._first_instance:
			return

		force_latency_value = self._variables.expand(
			instance.options["force_latency"])
		if force_latency_value is not None:
			self._set_latency(force_latency_value)
		if self._has_intel_pstate:
			new_value = self._variables.expand(
				instance.options["min_perf_pct"])
			self._min_perf_pct_save = self._getset_intel_pstate_attr(
				"min_perf_pct", new_value)
			new_value = self._variables.expand(
				instance.options["max_perf_pct"])
			self._max_perf_pct_save = self._getset_intel_pstate_attr(
				"max_perf_pct", new_value)
			new_value = self._variables.expand(
				instance.options["no_turbo"])
			self._no_turbo_save = self._getset_intel_pstate_attr(
				"no_turbo", new_value)

	def _instance_unapply_static(self, instance, rollback = consts.ROLLBACK_SOFT):
		super(CPULatencyPlugin, self)._instance_unapply_static(instance, rollback)

		if instance._first_instance and self._has_intel_pstate:
			self._set_intel_pstate_attr("min_perf_pct", self._min_perf_pct_save)
			self._set_intel_pstate_attr("max_perf_pct", self._max_perf_pct_save)
			self._set_intel_pstate_attr("no_turbo", self._no_turbo_save)

	def _instance_apply_dynamic(self, instance, device):
		self._instance_update_dynamic(instance, device)

	def _instance_update_dynamic(self, instance, device):
		assert(instance._first_instance)
		if device != instance._first_device:
			return

		load = instance._load_monitor.get_load()["system"]
		if load < instance.options["load_threshold"]:
			self._set_latency(instance.options["latency_high"])
		else:
			self._set_latency(instance.options["latency_low"])

	def _instance_unapply_dynamic(self, instance, device):
		pass

	def _str2int(self, s):
		try:
			return int(s)
		except (ValueError, TypeError):
			return None

	def _read_cstates_latency(self):
		self.cstates_latency = {}
		for d in os.listdir(cpuidle_states_path):
			cstate_path = cpuidle_states_path + "/%s/" % d
			name = self._cmd.read_file(cstate_path + "name", err_ret = None, no_error = True)
			latency = self._cmd.read_file(cstate_path + "latency", err_ret = None, no_error = True)
			if name is not None and latency is not None:
				latency = self._str2int(latency)
				if latency is not None:
					self.cstates_latency[name.strip()] = latency

	def _get_latency_by_cstate_name(self, name, no_zero=False):
		log.debug("getting latency for cstate with name '%s'" % name)
		if self.cstates_latency is None:
			log.debug("reading cstates latency table")
			self._read_cstates_latency()
		latency = self.cstates_latency.get(name, None)
		if no_zero and latency == 0:
			log.debug("skipping latency 0 as set by param")
			return None
		log.debug("cstate name mapped to latency: %s" % str(latency))
		return latency

	def _get_latency_by_cstate_id(self, lid, no_zero=False):
		log.debug("getting latency for cstate with ID '%s'" % str(lid))
		lid = self._str2int(lid)
		if lid is None:
			log.debug("cstate ID is invalid")
			return None
		latency_path = cpuidle_states_path + "/%s/latency" % ("state%d" % lid)
		latency = self._str2int(self._cmd.read_file(latency_path, err_ret = None, no_error = True))
		if no_zero and latency == 0:
			log.debug("skipping latency 0 as set by param")
			return None
		log.debug("cstate ID mapped to latency: %s" % str(latency))
		return latency

	# returns (latency, skip), skip means we want to skip latency settings
	def _parse_latency(self, latency, allow_na=False):
		self.cstates_latency = None
		latencies = str(latency).split("|")
		log.debug("parsing latency '%s', allow_na '%s'" % (latency, allow_na))
		for latency in latencies:
			try:
				latency = int(latency)
				log.debug("parsed directly specified latency value: %d" % latency)
			except ValueError:
				if latency[0:18] == "cstate.id_no_zero:":
					latency = self._get_latency_by_cstate_id(latency[18:], no_zero=True)
				elif latency[0:10] == "cstate.id:":
					latency = self._get_latency_by_cstate_id(latency[10:])
				elif latency[0:20] == "cstate.name_no_zero:":
					latency = self._get_latency_by_cstate_name(latency[20:], no_zero=True)
				elif latency[0:12] == "cstate.name:":
					latency = self._get_latency_by_cstate_name(latency[12:])
				elif latency in ["none", "None"]:
					log.debug("latency 'none' specified")
					return None, True
				elif allow_na and latency == "n/a":
					log.debug("latency 'n/a' specified")
					pass
				else:
					log.debug("invalid latency specified: '%s'" % str(latency))
					latency = None
			if latency is not None:
				break
		return latency, False

	def _set_latency(self, latency):
		latency, skip = self._parse_latency(latency)
		if not skip and self._has_pm_qos:
			if latency is None:
				log.error("unable to evaluate latency value (probably wrong settings in the 'cpu' section of current profile), disabling PM QoS")
				self._has_pm_qos = False
			elif self._latency != latency:
				log.info("setting new cpu latency %d" % latency)
				latency_bin = struct.pack("i", latency)
				os.write(self._cpu_latency_fd, latency_bin)
				self._latency = latency

	def _get_available_governors(self, device):
		return self._cmd.read_file("/sys/devices/system/cpu/%s/cpufreq/scaling_available_governors" % device).strip().split()

	@command_set("governor", per_device=True)
	def _set_governor(self, governors, device, sim, remove):
		if not self._check_cpu_can_change_governor(device):
			return None
		governors = str(governors)
		governors = governors.split("|")
		governors = [governor.strip() for governor in governors]
		for governor in governors:
			if len(governor) == 0:
				log.error("The 'governor' option contains an empty value.")
				return None
		available_governors = self._get_available_governors(device)
		for governor in governors:
			if governor in available_governors:
				if not sim:
					log.info("setting governor '%s' on cpu '%s'"
							% (governor, device))
					self._cmd.write_to_file("/sys/devices/system/cpu/%s/cpufreq/scaling_governor"
							% device, governor, no_error = [errno.ENOENT] if remove else False)
				break
			elif not sim:
				log.debug("Ignoring governor '%s' on cpu '%s', it is not supported"
						% (governor, device))
		else:
			log.warn("None of the scaling governors is supported: %s"
					% ", ".join(governors))
			governor = None
		return governor

	@command_get("governor")
	def _get_governor(self, device, ignore_missing=False):
		governor = None
		if not self._check_cpu_can_change_governor(device):
			return None
		data = self._cmd.read_file("/sys/devices/system/cpu/%s/cpufreq/scaling_governor" % device, no_error=ignore_missing).strip()
		if len(data) > 0:
			governor = data

		if governor is None:
			log.error("could not get current governor on cpu '%s'" % device)

		return governor

	def _sampling_down_factor_path(self, governor = "ondemand"):
		return "/sys/devices/system/cpu/cpufreq/%s/sampling_down_factor" % governor

	@command_set("sampling_down_factor", per_device = True, priority = 10)
	def _set_sampling_down_factor(self, sampling_down_factor, device, sim, remove):
		val = None

		# hack to clear governors map when the profile starts unloading
		# TODO: this should be handled better way, by e.g. currently non-implemented
		# Plugin.profile_load_finished() method
		if device in self._governors_map:
			self._governors_map.clear()

		self._governors_map[device] = None
		governor = self._get_governor(device)
		if governor is None:
			log.debug("ignoring sampling_down_factor setting for CPU '%s', cannot match governor" % device)
			return None
		if governor not in list(self._governors_map.values()):
			self._governors_map[device] = governor
			path = self._sampling_down_factor_path(governor)
			if not os.path.exists(path):
				log.debug("ignoring sampling_down_factor setting for CPU '%s', governor '%s' doesn't support it" % (device, governor))
				return None
			val = str(sampling_down_factor)
			if not sim:
				log.info("setting sampling_down_factor to '%s' for governor '%s'" % (val, governor))
				self._cmd.write_to_file(path, val, no_error = [errno.ENOENT] if remove else False)
		return val

	@command_get("sampling_down_factor")
	def _get_sampling_down_factor(self, device, ignore_missing=False):
		governor = self._get_governor(device, ignore_missing=ignore_missing)
		if governor is None:
			return None
		path = self._sampling_down_factor_path(governor)
		if not os.path.exists(path):
			return None
		return self._cmd.read_file(path).strip()

	def _try_set_energy_perf_bias(self, cpu_id, value):
		(retcode, out, err_msg) = self._cmd.execute(
				["x86_energy_perf_policy",
				"-c", cpu_id,
				str(value)
				],
				return_err = True)
		return (retcode, err_msg)

	def _pstate_preference_path(self, cpu_id, available = False):
		return "/sys/devices/system/cpu/cpufreq/policy%s/energy_performance_%s" % (cpu_id, "available_preferences" if available else "preference")

	def _energy_perf_bias_path(self, cpu_id):
		return "/sys/devices/system/cpu/cpu%s/power/energy_perf_bias" % cpu_id

	@command_set("energy_perf_bias", per_device=True)
	def _set_energy_perf_bias(self, energy_perf_bias, device, sim, remove):
		if not self._is_cpu_online(device):
			log.debug("%s is not online, skipping" % device)
			return None
		cpu_id = device.lstrip("cpu")
		vals = energy_perf_bias.split('|')

		# It should be writen straight to sysfs energy_perf_bias file if requested on newer processors
		# see rhbz#2095829
		if consts.CFG_CPU_EPP_FLAG in self._get_cpuinfo_flags():
			energy_perf_bias_path = self._energy_perf_bias_path(cpu_id)
			if os.path.exists(energy_perf_bias_path):
				if not sim:
					for val in vals:
						val = val.strip()
						if self._cmd.write_to_file(energy_perf_bias_path, val, \
							no_error = [errno.ENOENT] if remove else False):
								log.info("energy_perf_bias successfully set to '%s' on cpu '%s'"
										 % (val, device))
								break
					else:
						log.error("Failed to set energy_perf_bias on cpu '%s'. Is the value in the profile correct?"
								  % device)
					
				return str(energy_perf_bias)
			else:
				log.error("Failed to set energy_perf_bias on cpu '%s' because energy_perf_bias file does not exist."
						  % device)
				return None
		elif self._has_energy_perf_bias:
			if not sim:
				for val in vals:
					val = val.strip()
					log.debug("Trying to set energy_perf_bias to '%s' on cpu '%s'"
							% (val, device))
					(retcode, err_msg) = self._try_set_energy_perf_bias(
							cpu_id, val)
					if retcode == 0:
						log.info("energy_perf_bias successfully set to '%s' on cpu '%s'"
								% (val, device))
						break
					elif retcode < 0:
						log.error("Failed to set energy_perf_bias: %s"
								% err_msg)
						break
					else:
						log.debug("Could not set energy_perf_bias to '%s' on cpu '%s', trying another value"
								% (val, device))
				else:
					log.error("Failed to set energy_perf_bias on cpu '%s'. Is the value in the profile correct?"
							% device)
			return str(energy_perf_bias)
		else:
			return None

	def _try_parse_num(self, s):
		try:
			v = int(s)
		except ValueError as e:
			try:
				v = int(s, 16)
			except ValueError as e:
				v = s
		return v

	# Before Linux 4.13
	def _energy_perf_policy_to_human(self, s):
		return {0:"performance", 6:"normal", 15:"powersave"}.get(self._try_parse_num(s), s)

	# Since Linux 4.13
	def _energy_perf_policy_to_human_v2(self, s):
		return {0:"performance",
				4:"balance-performance",
				6:"normal",
				8:"balance-power",
				15:"power",
				}.get(self._try_parse_num(s), s)

	@command_get("energy_perf_bias")
	def _get_energy_perf_bias(self, device, ignore_missing=False):
		energy_perf_bias = None
		if not self._is_cpu_online(device):
			log.debug("%s is not online, skipping" % device)
			return None
		cpu_id = device.lstrip("cpu")
		if consts.CFG_CPU_EPP_FLAG in self._get_cpuinfo_flags():
			energy_perf_bias_path = self._energy_perf_bias_path(cpu_id)
			if os.path.exists(energy_perf_bias_path):
				energy_perf_bias = self._energy_perf_policy_to_human_v2(self._cmd.read_file(energy_perf_bias_path))
		elif self._has_energy_perf_bias:
			retcode, lines = self._cmd.execute(["x86_energy_perf_policy", "-c", cpu_id, "-r"])
			if retcode == 0:
				for line in lines.splitlines():
					l = line.split()
					if len(l) == 2:
						energy_perf_bias = self._energy_perf_policy_to_human(l[1])
						break
					elif len(l) == 3:
						energy_perf_bias = self._energy_perf_policy_to_human_v2(l[2])
						break

		return energy_perf_bias

	def _pm_qos_resume_latency_us_path(self, device):
		return "/sys/devices/system/cpu/%s/power/pm_qos_resume_latency_us" % device

	def _check_pm_qos_resume_latency_us(self, device):
		if self._has_pm_qos_resume_latency_us is None:
			self._has_pm_qos_resume_latency_us = os.path.exists(self._pm_qos_resume_latency_us_path(device))
			if not self._has_pm_qos_resume_latency_us:
				log.info("Option 'pm_qos_resume_latency_us' is not supported on current hardware.")
		return self._has_pm_qos_resume_latency_us

	@command_set("pm_qos_resume_latency_us", per_device=True)
	def _set_pm_qos_resume_latency_us(self, pm_qos_resume_latency_us, device, sim, remove):
		if not self._is_cpu_online(device):
			log.debug("%s is not online, skipping" % device)
			return None
		latency, skip = self._parse_latency(pm_qos_resume_latency_us, allow_na=True)
		if skip or not self._check_pm_qos_resume_latency_us(device):
			return None
		if latency is None or (latency != "n/a" and latency < 0):
			log.warning("Invalid pm_qos_resume_latency_us specified: '%s', cpu: '%s'." % (pm_qos_resume_latency_us, device))
			return None
		if not sim:
			self._cmd.write_to_file(self._pm_qos_resume_latency_us_path(device), latency, \
				no_error = [errno.ENOENT] if remove else False)
		return latency

	@command_get("pm_qos_resume_latency_us")
	def _get_pm_qos_resume_latency_us(self, device, ignore_missing=False):
		if not self._is_cpu_online(device):
			log.debug("%s is not online, skipping" % device)
			return None
		if not self._check_pm_qos_resume_latency_us(device):
			return None
		return self._cmd.read_file(self._pm_qos_resume_latency_us_path(device), no_error=ignore_missing).strip()

	@command_set("energy_performance_preference", per_device=True)
	def _set_energy_performance_preference(self, energy_performance_preference, device, sim, remove):
		if not self._is_cpu_online(device):
			log.debug("%s is not online, skipping" % device)
			return None
		cpu_id = device.lstrip("cpu")
		if os.path.exists(self._pstate_preference_path(cpu_id, True)):
			vals = energy_performance_preference.split('|')
			if not sim:
				avail_vals = set(self._cmd.read_file(self._pstate_preference_path(cpu_id, True)).split())
				for val in vals:
					if val in avail_vals:
						self._cmd.write_to_file(self._pstate_preference_path(cpu_id), val, \
							no_error = [errno.ENOENT] if remove else False)
						log.info("Setting energy_performance_preference value '%s' for cpu '%s'" % (val, device))
						break
					else:
						log.warn("energy_performance_preference value '%s' unavailable for cpu '%s'" % (val, device))
				else:
					log.error("Failed to set energy_performance_preference on cpu '%s'. Is the value in the profile correct?"
							  % device)
			return str(energy_performance_preference)
		else:
			log.debug("energy_performance_available_preferences file missing, which can happen if the system is booted without a P-state driver.")
		return None

	@command_get("energy_performance_preference")
	def _get_energy_performance_preference(self, device, ignore_missing=False):
		if not self._is_cpu_online(device):
			log.debug("%s is not online, skipping" % device)
			return None
		cpu_id = device.lstrip("cpu")
		# read the EPP hint used by the intel_pstate and amd-pstate CPU scaling drivers
		if os.path.exists(self._pstate_preference_path(cpu_id, True)):
			return self._cmd.read_file(self._pstate_preference_path(cpu_id)).strip()
		else:
			log.debug("energy_performance_available_preferences file missing, which can happen if the system is booted without a P-state driver.")
		return None