1.. SPDX-License-Identifier: GPL-2.0
2.. include:: <isonum.txt>
3
4==============================================
5``intel_idle`` CPU Idle Time Management Driver
6==============================================
7
8:Copyright: |copy| 2020 Intel Corporation
9
10:Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
11
12
13General Information
14===================
15
16``intel_idle`` is a part of the
17:doc:`CPU idle time management subsystem <cpuidle>` in the Linux kernel
18(``CPUIdle``).  It is the default CPU idle time management driver for the
19Nehalem and later generations of Intel processors, but the level of support for
20a particular processor model in it depends on whether or not it recognizes that
21processor model and may also depend on information coming from the platform
22firmware.  [To understand ``intel_idle`` it is necessary to know how ``CPUIdle``
23works in general, so this is the time to get familiar with :doc:`cpuidle` if you
24have not done that yet.]
25
26``intel_idle`` uses the ``MWAIT`` instruction to inform the processor that the
27logical CPU executing it is idle and so it may be possible to put some of the
28processor's functional blocks into low-power states.  That instruction takes two
29arguments (passed in the ``EAX`` and ``ECX`` registers of the target CPU), the
30first of which, referred to as a *hint*, can be used by the processor to
31determine what can be done (for details refer to Intel Software Developer’s
32Manual [1]_).  Accordingly, ``intel_idle`` refuses to work with processors in
33which the support for the ``MWAIT`` instruction has been disabled (for example,
34via the platform firmware configuration menu) or which do not support that
35instruction at all.
36
37``intel_idle`` is not modular, so it cannot be unloaded, which means that the
38only way to pass early-configuration-time parameters to it is via the kernel
39command line.
40
41
42.. _intel-idle-enumeration-of-states:
43
44Enumeration of Idle States
45==========================
46
47Each ``MWAIT`` hint value is interpreted by the processor as a license to
48reconfigure itself in a certain way in order to save energy.  The processor
49configurations (with reduced power draw) resulting from that are referred to
50as C-states (in the ACPI terminology) or idle states.  The list of meaningful
51``MWAIT`` hint values and idle states (i.e. low-power configurations of the
52processor) corresponding to them depends on the processor model and it may also
53depend on the configuration of the platform.
54
55In order to create a list of available idle states required by the ``CPUIdle``
56subsystem (see :ref:`idle-states-representation` in :doc:`cpuidle`),
57``intel_idle`` can use two sources of information: static tables of idle states
58for different processor models included in the driver itself and the ACPI tables
59of the system.  The former are always used if the processor model at hand is
60recognized by ``intel_idle`` and the latter are used if that is required for
61the given processor model (which is the case for all server processor models
62recognized by ``intel_idle``) or if the processor model is not recognized.
63[There is a module parameter that can be used to make the driver use the ACPI
64tables with any processor model recognized by it; see
65`below <intel-idle-parameters_>`_.]
66
67If the ACPI tables are going to be used for building the list of available idle
68states, ``intel_idle`` first looks for a ``_CST`` object under one of the ACPI
69objects corresponding to the CPUs in the system (refer to the ACPI specification
70[2]_ for the description of ``_CST`` and its output package).  Because the
71``CPUIdle`` subsystem expects that the list of idle states supplied by the
72driver will be suitable for all of the CPUs handled by it and ``intel_idle`` is
73registered as the ``CPUIdle`` driver for all of the CPUs in the system, the
74driver looks for the first ``_CST`` object returning at least one valid idle
75state description and such that all of the idle states included in its return
76package are of the FFH (Functional Fixed Hardware) type, which means that the
77``MWAIT`` instruction is expected to be used to tell the processor that it can
78enter one of them.  The return package of that ``_CST`` is then assumed to be
79applicable to all of the other CPUs in the system and the idle state
80descriptions extracted from it are stored in a preliminary list of idle states
81coming from the ACPI tables.  [This step is skipped if ``intel_idle`` is
82configured to ignore the ACPI tables; see `below <intel-idle-parameters_>`_.]
83
84Next, the first (index 0) entry in the list of available idle states is
85initialized to represent a "polling idle state" (a pseudo-idle state in which
86the target CPU continuously fetches and executes instructions), and the
87subsequent (real) idle state entries are populated as follows.
88
89If the processor model at hand is recognized by ``intel_idle``, there is a
90(static) table of idle state descriptions for it in the driver.  In that case,
91the "internal" table is the primary source of information on idle states and the
92information from it is copied to the final list of available idle states.  If
93using the ACPI tables for the enumeration of idle states is not required
94(depending on the processor model), all of the listed idle state are enabled by
95default (so all of them will be taken into consideration by ``CPUIdle``
96governors during CPU idle state selection).  Otherwise, some of the listed idle
97states may not be enabled by default if there are no matching entries in the
98preliminary list of idle states coming from the ACPI tables.  In that case user
99space still can enable them later (on a per-CPU basis) with the help of
100the ``disable`` idle state attribute in ``sysfs`` (see
101:ref:`idle-states-representation` in :doc:`cpuidle`).  This basically means that
102the idle states "known" to the driver may not be enabled by default if they have
103not been exposed by the platform firmware (through the ACPI tables).
104
105If the given processor model is not recognized by ``intel_idle``, but it
106supports ``MWAIT``, the preliminary list of idle states coming from the ACPI
107tables is used for building the final list that will be supplied to the
108``CPUIdle`` core during driver registration.  For each idle state in that list,
109the description, ``MWAIT`` hint and exit latency are copied to the corresponding
110entry in the final list of idle states.  The name of the idle state represented
111by it (to be returned by the ``name`` idle state attribute in ``sysfs``) is
112"CX_ACPI", where X is the index of that idle state in the final list (note that
113the minimum value of X is 1, because 0 is reserved for the "polling" state), and
114its target residency is based on the exit latency value.  Specifically, for
115C1-type idle states the exit latency value is also used as the target residency
116(for compatibility with the majority of the "internal" tables of idle states for
117various processor models recognized by ``intel_idle``) and for the other idle
118state types (C2 and C3) the target residency value is 3 times the exit latency
119(again, that is because it reflects the target residency to exit latency ratio
120in the majority of cases for the processor models recognized by ``intel_idle``).
121All of the idle states in the final list are enabled by default in this case.
122
123
124.. _intel-idle-initialization:
125
126Initialization
127==============
128
129The initialization of ``intel_idle`` starts with checking if the kernel command
130line options forbid the use of the ``MWAIT`` instruction.  If that is the case,
131an error code is returned right away.
132
133The next step is to check whether or not the processor model is known to the
134driver, which determines the idle states enumeration method (see
135`above <intel-idle-enumeration-of-states_>`_), and whether or not the processor
136supports ``MWAIT`` (the initialization fails if that is not the case).  Then,
137the ``MWAIT`` support in the processor is enumerated through ``CPUID`` and the
138driver initialization fails if the level of support is not as expected (for
139example, if the total number of ``MWAIT`` substates returned is 0).
140
141Next, if the driver is not configured to ignore the ACPI tables (see
142`below <intel-idle-parameters_>`_), the idle states information provided by the
143platform firmware is extracted from them.
144
145Then, ``CPUIdle`` device objects are allocated for all CPUs and the list of
146available idle states is created as explained
147`above <intel-idle-enumeration-of-states_>`_.
148
149Finally, ``intel_idle`` is registered with the help of cpuidle_register_driver()
150as the ``CPUIdle`` driver for all CPUs in the system and a CPU online callback
151for configuring individual CPUs is registered via cpuhp_setup_state(), which
152(among other things) causes the callback routine to be invoked for all of the
153CPUs present in the system at that time (each CPU executes its own instance of
154the callback routine).  That routine registers a ``CPUIdle`` device for the CPU
155running it (which enables the ``CPUIdle`` subsystem to operate that CPU) and
156optionally performs some CPU-specific initialization actions that may be
157required for the given processor model.
158
159
160.. _intel-idle-parameters:
161
162Kernel Command Line Options and Module Parameters
163=================================================
164
165The *x86* architecture support code recognizes three kernel command line
166options related to CPU idle time management: ``idle=poll``, ``idle=halt``,
167and ``idle=nomwait``.  If any of them is present in the kernel command line, the
168``MWAIT`` instruction is not allowed to be used, so the initialization of
169``intel_idle`` will fail.
170
171Apart from that there are four module parameters recognized by ``intel_idle``
172itself that can be set via the kernel command line (they cannot be updated via
173sysfs, so that is the only way to change their values).
174
175The ``max_cstate`` parameter value is the maximum idle state index in the list
176of idle states supplied to the ``CPUIdle`` core during the registration of the
177driver.  It is also the maximum number of regular (non-polling) idle states that
178can be used by ``intel_idle``, so the enumeration of idle states is terminated
179after finding that number of usable idle states (the other idle states that
180potentially might have been used if ``max_cstate`` had been greater are not
181taken into consideration at all).  Setting ``max_cstate`` can prevent
182``intel_idle`` from exposing idle states that are regarded as "too deep" for
183some reason to the ``CPUIdle`` core, but it does so by making them effectively
184invisible until the system is shut down and started again which may not always
185be desirable.  In practice, it is only really necessary to do that if the idle
186states in question cannot be enabled during system startup, because in the
187working state of the system the CPU power management quality of service (PM
188QoS) feature can be used to prevent ``CPUIdle`` from touching those idle states
189even if they have been enumerated (see :ref:`cpu-pm-qos` in :doc:`cpuidle`).
190Setting ``max_cstate`` to 0 causes the ``intel_idle`` initialization to fail.
191
192The ``no_acpi`` and ``use_acpi`` module parameters (recognized by ``intel_idle``
193if the kernel has been configured with ACPI support) can be set to make the
194driver ignore the system's ACPI tables entirely or use them for all of the
195recognized processor models, respectively (they both are unset by default and
196``use_acpi`` has no effect if ``no_acpi`` is set).
197
198The value of the ``states_off`` module parameter (0 by default) represents a
199list of idle states to be disabled by default in the form of a bitmask.
200
201Namely, the positions of the bits that are set in the ``states_off`` value are
202the indices of idle states to be disabled by default (as reflected by the names
203of the corresponding idle state directories in ``sysfs``, :file:`state0`,
204:file:`state1` ... :file:`state<i>` ..., where ``<i>`` is the index of the given
205idle state; see :ref:`idle-states-representation` in :doc:`cpuidle`).
206
207For example, if ``states_off`` is equal to 3, the driver will disable idle
208states 0 and 1 by default, and if it is equal to 8, idle state 3 will be
209disabled by default and so on (bit positions beyond the maximum idle state index
210are ignored).
211
212The idle states disabled this way can be enabled (on a per-CPU basis) from user
213space via ``sysfs``.
214
215
216.. _intel-idle-core-and-package-idle-states:
217
218Core and Package Levels of Idle States
219======================================
220
221Typically, in a processor supporting the ``MWAIT`` instruction there are (at
222least) two levels of idle states (or C-states).  One level, referred to as
223"core C-states", covers individual cores in the processor, whereas the other
224level, referred to as "package C-states", covers the entire processor package
225and it may also involve other components of the system (GPUs, memory
226controllers, I/O hubs etc.).
227
228Some of the ``MWAIT`` hint values allow the processor to use core C-states only
229(most importantly, that is the case for the ``MWAIT`` hint value corresponding
230to the ``C1`` idle state), but the majority of them give it a license to put
231the target core (i.e. the core containing the logical CPU executing ``MWAIT``
232with the given hint value) into a specific core C-state and then (if possible)
233to enter a specific package C-state at the deeper level.  For example, the
234``MWAIT`` hint value representing the ``C3`` idle state allows the processor to
235put the target core into the low-power state referred to as "core ``C3``" (or
236``CC3``), which happens if all of the logical CPUs (SMT siblings) in that core
237have executed ``MWAIT`` with the ``C3`` hint value (or with a hint value
238representing a deeper idle state), and in addition to that (in the majority of
239cases) it gives the processor a license to put the entire package (possibly
240including some non-CPU components such as a GPU or a memory controller) into the
241low-power state referred to as "package ``C3``" (or ``PC3``), which happens if
242all of the cores have gone into the ``CC3`` state and (possibly) some additional
243conditions are satisfied (for instance, if the GPU is covered by ``PC3``, it may
244be required to be in a certain GPU-specific low-power state for ``PC3`` to be
245reachable).
246
247As a rule, there is no simple way to make the processor use core C-states only
248if the conditions for entering the corresponding package C-states are met, so
249the logical CPU executing ``MWAIT`` with a hint value that is not core-level
250only (like for ``C1``) must always assume that this may cause the processor to
251enter a package C-state.  [That is why the exit latency and target residency
252values corresponding to the majority of ``MWAIT`` hint values in the "internal"
253tables of idle states in ``intel_idle`` reflect the properties of package
254C-states.]  If using package C-states is not desirable at all, either
255:ref:`PM QoS <cpu-pm-qos>` or the ``max_cstate`` module parameter of
256``intel_idle`` described `above <intel-idle-parameters_>`_ must be used to
257restrict the range of permissible idle states to the ones with core-level only
258``MWAIT`` hint values (like ``C1``).
259
260
261References
262==========
263
264.. [1] *Intel® 64 and IA-32 Architectures Software Developer’s Manual Volume 2B*,
265       https://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-software-developer-vol-2b-manual.html
266
267.. [2] *Advanced Configuration and Power Interface (ACPI) Specification*,
268       https://uefi.org/specifications
269