1.. SPDX-License-Identifier: GPL-2.0
2.. include:: <isonum.txt>
3
4.. |struct cpuidle_state| replace:: :c:type:`struct cpuidle_state <cpuidle_state>`
5.. |cpufreq| replace:: :doc:`CPU Performance Scaling <cpufreq>`
6
7========================
8CPU Idle Time Management
9========================
10
11:Copyright: |copy| 2018 Intel Corporation
12
13:Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
14
15
16Concepts
17========
18
19Modern processors are generally able to enter states in which the execution of
20a program is suspended and instructions belonging to it are not fetched from
21memory or executed.  Those states are the *idle* states of the processor.
22
23Since part of the processor hardware is not used in idle states, entering them
24generally allows power drawn by the processor to be reduced and, in consequence,
25it is an opportunity to save energy.
26
27CPU idle time management is an energy-efficiency feature concerned about using
28the idle states of processors for this purpose.
29
30Logical CPUs
31------------
32
33CPU idle time management operates on CPUs as seen by the *CPU scheduler* (that
34is the part of the kernel responsible for the distribution of computational
35work in the system).  In its view, CPUs are *logical* units.  That is, they need
36not be separate physical entities and may just be interfaces appearing to
37software as individual single-core processors.  In other words, a CPU is an
38entity which appears to be fetching instructions that belong to one sequence
39(program) from memory and executing them, but it need not work this way
40physically.  Generally, three different cases can be consider here.
41
42First, if the whole processor can only follow one sequence of instructions (one
43program) at a time, it is a CPU.  In that case, if the hardware is asked to
44enter an idle state, that applies to the processor as a whole.
45
46Second, if the processor is multi-core, each core in it is able to follow at
47least one program at a time.  The cores need not be entirely independent of each
48other (for example, they may share caches), but still most of the time they
49work physically in parallel with each other, so if each of them executes only
50one program, those programs run mostly independently of each other at the same
51time.  The entire cores are CPUs in that case and if the hardware is asked to
52enter an idle state, that applies to the core that asked for it in the first
53place, but it also may apply to a larger unit (say a "package" or a "cluster")
54that the core belongs to (in fact, it may apply to an entire hierarchy of larger
55units containing the core).  Namely, if all of the cores in the larger unit
56except for one have been put into idle states at the "core level" and the
57remaining core asks the processor to enter an idle state, that may trigger it
58to put the whole larger unit into an idle state which also will affect the
59other cores in that unit.
60
61Finally, each core in a multi-core processor may be able to follow more than one
62program in the same time frame (that is, each core may be able to fetch
63instructions from multiple locations in memory and execute them in the same time
64frame, but not necessarily entirely in parallel with each other).  In that case
65the cores present themselves to software as "bundles" each consisting of
66multiple individual single-core "processors", referred to as *hardware threads*
67(or hyper-threads specifically on Intel hardware), that each can follow one
68sequence of instructions.  Then, the hardware threads are CPUs from the CPU idle
69time management perspective and if the processor is asked to enter an idle state
70by one of them, the hardware thread (or CPU) that asked for it is stopped, but
71nothing more happens, unless all of the other hardware threads within the same
72core also have asked the processor to enter an idle state.  In that situation,
73the core may be put into an idle state individually or a larger unit containing
74it may be put into an idle state as a whole (if the other cores within the
75larger unit are in idle states already).
76
77Idle CPUs
78---------
79
80Logical CPUs, simply referred to as "CPUs" in what follows, are regarded as
81*idle* by the Linux kernel when there are no tasks to run on them except for the
82special "idle" task.
83
84Tasks are the CPU scheduler's representation of work.  Each task consists of a
85sequence of instructions to execute, or code, data to be manipulated while
86running that code, and some context information that needs to be loaded into the
87processor every time the task's code is run by a CPU.  The CPU scheduler
88distributes work by assigning tasks to run to the CPUs present in the system.
89
90Tasks can be in various states.  In particular, they are *runnable* if there are
91no specific conditions preventing their code from being run by a CPU as long as
92there is a CPU available for that (for example, they are not waiting for any
93events to occur or similar).  When a task becomes runnable, the CPU scheduler
94assigns it to one of the available CPUs to run and if there are no more runnable
95tasks assigned to it, the CPU will load the given task's context and run its
96code (from the instruction following the last one executed so far, possibly by
97another CPU).  [If there are multiple runnable tasks assigned to one CPU
98simultaneously, they will be subject to prioritization and time sharing in order
99to allow them to make some progress over time.]
100
101The special "idle" task becomes runnable if there are no other runnable tasks
102assigned to the given CPU and the CPU is then regarded as idle.  In other words,
103in Linux idle CPUs run the code of the "idle" task called *the idle loop*.  That
104code may cause the processor to be put into one of its idle states, if they are
105supported, in order to save energy, but if the processor does not support any
106idle states, or there is not enough time to spend in an idle state before the
107next wakeup event, or there are strict latency constraints preventing any of the
108available idle states from being used, the CPU will simply execute more or less
109useless instructions in a loop until it is assigned a new task to run.
110
111
112.. _idle-loop:
113
114The Idle Loop
115=============
116
117The idle loop code takes two major steps in every iteration of it.  First, it
118calls into a code module referred to as the *governor* that belongs to the CPU
119idle time management subsystem called ``CPUIdle`` to select an idle state for
120the CPU to ask the hardware to enter.  Second, it invokes another code module
121from the ``CPUIdle`` subsystem, called the *driver*, to actually ask the
122processor hardware to enter the idle state selected by the governor.
123
124The role of the governor is to find an idle state most suitable for the
125conditions at hand.  For this purpose, idle states that the hardware can be
126asked to enter by logical CPUs are represented in an abstract way independent of
127the platform or the processor architecture and organized in a one-dimensional
128(linear) array.  That array has to be prepared and supplied by the ``CPUIdle``
129driver matching the platform the kernel is running on at the initialization
130time.  This allows ``CPUIdle`` governors to be independent of the underlying
131hardware and to work with any platforms that the Linux kernel can run on.
132
133Each idle state present in that array is characterized by two parameters to be
134taken into account by the governor, the *target residency* and the (worst-case)
135*exit latency*.  The target residency is the minimum time the hardware must
136spend in the given state, including the time needed to enter it (which may be
137substantial), in order to save more energy than it would save by entering one of
138the shallower idle states instead.  [The "depth" of an idle state roughly
139corresponds to the power drawn by the processor in that state.]  The exit
140latency, in turn, is the maximum time it will take a CPU asking the processor
141hardware to enter an idle state to start executing the first instruction after a
142wakeup from that state.  Note that in general the exit latency also must cover
143the time needed to enter the given state in case the wakeup occurs when the
144hardware is entering it and it must be entered completely to be exited in an
145ordered manner.
146
147There are two types of information that can influence the governor's decisions.
148First of all, the governor knows the time until the closest timer event.  That
149time is known exactly, because the kernel programs timers and it knows exactly
150when they will trigger, and it is the maximum time the hardware that the given
151CPU depends on can spend in an idle state, including the time necessary to enter
152and exit it.  However, the CPU may be woken up by a non-timer event at any time
153(in particular, before the closest timer triggers) and it generally is not known
154when that may happen.  The governor can only see how much time the CPU actually
155was idle after it has been woken up (that time will be referred to as the *idle
156duration* from now on) and it can use that information somehow along with the
157time until the closest timer to estimate the idle duration in future.  How the
158governor uses that information depends on what algorithm is implemented by it
159and that is the primary reason for having more than one governor in the
160``CPUIdle`` subsystem.
161
162There are four ``CPUIdle`` governors available, ``menu``, `TEO <teo-gov_>`_,
163``ladder`` and ``haltpoll``.  Which of them is used by default depends on the
164configuration of the kernel and in particular on whether or not the scheduler
165tick can be `stopped by the idle loop <idle-cpus-and-tick_>`_.  Available
166governors can be read from the :file:`available_governors`, and the governor
167can be changed at runtime.  The name of the ``CPUIdle`` governor currently
168used by the kernel can be read from the :file:`current_governor_ro` or
169:file:`current_governor` file under :file:`/sys/devices/system/cpu/cpuidle/`
170in ``sysfs``.
171
172Which ``CPUIdle`` driver is used, on the other hand, usually depends on the
173platform the kernel is running on, but there are platforms with more than one
174matching driver.  For example, there are two drivers that can work with the
175majority of Intel platforms, ``intel_idle`` and ``acpi_idle``, one with
176hardcoded idle states information and the other able to read that information
177from the system's ACPI tables, respectively.  Still, even in those cases, the
178driver chosen at the system initialization time cannot be replaced later, so the
179decision on which one of them to use has to be made early (on Intel platforms
180the ``acpi_idle`` driver will be used if ``intel_idle`` is disabled for some
181reason or if it does not recognize the processor).  The name of the ``CPUIdle``
182driver currently used by the kernel can be read from the :file:`current_driver`
183file under :file:`/sys/devices/system/cpu/cpuidle/` in ``sysfs``.
184
185
186.. _idle-cpus-and-tick:
187
188Idle CPUs and The Scheduler Tick
189================================
190
191The scheduler tick is a timer that triggers periodically in order to implement
192the time sharing strategy of the CPU scheduler.  Of course, if there are
193multiple runnable tasks assigned to one CPU at the same time, the only way to
194allow them to make reasonable progress in a given time frame is to make them
195share the available CPU time.  Namely, in rough approximation, each task is
196given a slice of the CPU time to run its code, subject to the scheduling class,
197prioritization and so on and when that time slice is used up, the CPU should be
198switched over to running (the code of) another task.  The currently running task
199may not want to give the CPU away voluntarily, however, and the scheduler tick
200is there to make the switch happen regardless.  That is not the only role of the
201tick, but it is the primary reason for using it.
202
203The scheduler tick is problematic from the CPU idle time management perspective,
204because it triggers periodically and relatively often (depending on the kernel
205configuration, the length of the tick period is between 1 ms and 10 ms).
206Thus, if the tick is allowed to trigger on idle CPUs, it will not make sense
207for them to ask the hardware to enter idle states with target residencies above
208the tick period length.  Moreover, in that case the idle duration of any CPU
209will never exceed the tick period length and the energy used for entering and
210exiting idle states due to the tick wakeups on idle CPUs will be wasted.
211
212Fortunately, it is not really necessary to allow the tick to trigger on idle
213CPUs, because (by definition) they have no tasks to run except for the special
214"idle" one.  In other words, from the CPU scheduler perspective, the only user
215of the CPU time on them is the idle loop.  Since the time of an idle CPU need
216not be shared between multiple runnable tasks, the primary reason for using the
217tick goes away if the given CPU is idle.  Consequently, it is possible to stop
218the scheduler tick entirely on idle CPUs in principle, even though that may not
219always be worth the effort.
220
221Whether or not it makes sense to stop the scheduler tick in the idle loop
222depends on what is expected by the governor.  First, if there is another
223(non-tick) timer due to trigger within the tick range, stopping the tick clearly
224would be a waste of time, even though the timer hardware may not need to be
225reprogrammed in that case.  Second, if the governor is expecting a non-timer
226wakeup within the tick range, stopping the tick is not necessary and it may even
227be harmful.  Namely, in that case the governor will select an idle state with
228the target residency within the time until the expected wakeup, so that state is
229going to be relatively shallow.  The governor really cannot select a deep idle
230state then, as that would contradict its own expectation of a wakeup in short
231order.  Now, if the wakeup really occurs shortly, stopping the tick would be a
232waste of time and in this case the timer hardware would need to be reprogrammed,
233which is expensive.  On the other hand, if the tick is stopped and the wakeup
234does not occur any time soon, the hardware may spend indefinite amount of time
235in the shallow idle state selected by the governor, which will be a waste of
236energy.  Hence, if the governor is expecting a wakeup of any kind within the
237tick range, it is better to allow the tick trigger.  Otherwise, however, the
238governor will select a relatively deep idle state, so the tick should be stopped
239so that it does not wake up the CPU too early.
240
241In any case, the governor knows what it is expecting and the decision on whether
242or not to stop the scheduler tick belongs to it.  Still, if the tick has been
243stopped already (in one of the previous iterations of the loop), it is better
244to leave it as is and the governor needs to take that into account.
245
246The kernel can be configured to disable stopping the scheduler tick in the idle
247loop altogether.  That can be done through the build-time configuration of it
248(by unsetting the ``CONFIG_NO_HZ_IDLE`` configuration option) or by passing
249``nohz=off`` to it in the command line.  In both cases, as the stopping of the
250scheduler tick is disabled, the governor's decisions regarding it are simply
251ignored by the idle loop code and the tick is never stopped.
252
253The systems that run kernels configured to allow the scheduler tick to be
254stopped on idle CPUs are referred to as *tickless* systems and they are
255generally regarded as more energy-efficient than the systems running kernels in
256which the tick cannot be stopped.  If the given system is tickless, it will use
257the ``menu`` governor by default and if it is not tickless, the default
258``CPUIdle`` governor on it will be ``ladder``.
259
260
261.. _menu-gov:
262
263The ``menu`` Governor
264=====================
265
266The ``menu`` governor is the default ``CPUIdle`` governor for tickless systems.
267It is quite complex, but the basic principle of its design is straightforward.
268Namely, when invoked to select an idle state for a CPU (i.e. an idle state that
269the CPU will ask the processor hardware to enter), it attempts to predict the
270idle duration and uses the predicted value for idle state selection.
271
272It first obtains the time until the closest timer event with the assumption
273that the scheduler tick will be stopped.  That time, referred to as the *sleep
274length* in what follows, is the upper bound on the time before the next CPU
275wakeup.  It is used to determine the sleep length range, which in turn is needed
276to get the sleep length correction factor.
277
278The ``menu`` governor maintains two arrays of sleep length correction factors.
279One of them is used when tasks previously running on the given CPU are waiting
280for some I/O operations to complete and the other one is used when that is not
281the case.  Each array contains several correction factor values that correspond
282to different sleep length ranges organized so that each range represented in the
283array is approximately 10 times wider than the previous one.
284
285The correction factor for the given sleep length range (determined before
286selecting the idle state for the CPU) is updated after the CPU has been woken
287up and the closer the sleep length is to the observed idle duration, the closer
288to 1 the correction factor becomes (it must fall between 0 and 1 inclusive).
289The sleep length is multiplied by the correction factor for the range that it
290falls into to obtain the first approximation of the predicted idle duration.
291
292Next, the governor uses a simple pattern recognition algorithm to refine its
293idle duration prediction.  Namely, it saves the last 8 observed idle duration
294values and, when predicting the idle duration next time, it computes the average
295and variance of them.  If the variance is small (smaller than 400 square
296milliseconds) or it is small relative to the average (the average is greater
297that 6 times the standard deviation), the average is regarded as the "typical
298interval" value.  Otherwise, the longest of the saved observed idle duration
299values is discarded and the computation is repeated for the remaining ones.
300Again, if the variance of them is small (in the above sense), the average is
301taken as the "typical interval" value and so on, until either the "typical
302interval" is determined or too many data points are disregarded, in which case
303the "typical interval" is assumed to equal "infinity" (the maximum unsigned
304integer value).  The "typical interval" computed this way is compared with the
305sleep length multiplied by the correction factor and the minimum of the two is
306taken as the predicted idle duration.
307
308Then, the governor computes an extra latency limit to help "interactive"
309workloads.  It uses the observation that if the exit latency of the selected
310idle state is comparable with the predicted idle duration, the total time spent
311in that state probably will be very short and the amount of energy to save by
312entering it will be relatively small, so likely it is better to avoid the
313overhead related to entering that state and exiting it.  Thus selecting a
314shallower state is likely to be a better option then.   The first approximation
315of the extra latency limit is the predicted idle duration itself which
316additionally is divided by a value depending on the number of tasks that
317previously ran on the given CPU and now they are waiting for I/O operations to
318complete.  The result of that division is compared with the latency limit coming
319from the power management quality of service, or `PM QoS <cpu-pm-qos_>`_,
320framework and the minimum of the two is taken as the limit for the idle states'
321exit latency.
322
323Now, the governor is ready to walk the list of idle states and choose one of
324them.  For this purpose, it compares the target residency of each state with
325the predicted idle duration and the exit latency of it with the computed latency
326limit.  It selects the state with the target residency closest to the predicted
327idle duration, but still below it, and exit latency that does not exceed the
328limit.
329
330In the final step the governor may still need to refine the idle state selection
331if it has not decided to `stop the scheduler tick <idle-cpus-and-tick_>`_.  That
332happens if the idle duration predicted by it is less than the tick period and
333the tick has not been stopped already (in a previous iteration of the idle
334loop).  Then, the sleep length used in the previous computations may not reflect
335the real time until the closest timer event and if it really is greater than
336that time, the governor may need to select a shallower state with a suitable
337target residency.
338
339
340.. _teo-gov:
341
342The Timer Events Oriented (TEO) Governor
343========================================
344
345The timer events oriented (TEO) governor is an alternative ``CPUIdle`` governor
346for tickless systems.  It follows the same basic strategy as the ``menu`` `one
347<menu-gov_>`_: it always tries to find the deepest idle state suitable for the
348given conditions.  However, it applies a different approach to that problem.
349
350.. kernel-doc:: drivers/cpuidle/governors/teo.c
351   :doc: teo-description
352
353.. _idle-states-representation:
354
355Representation of Idle States
356=============================
357
358For the CPU idle time management purposes all of the physical idle states
359supported by the processor have to be represented as a one-dimensional array of
360|struct cpuidle_state| objects each allowing an individual (logical) CPU to ask
361the processor hardware to enter an idle state of certain properties.  If there
362is a hierarchy of units in the processor, one |struct cpuidle_state| object can
363cover a combination of idle states supported by the units at different levels of
364the hierarchy.  In that case, the `target residency and exit latency parameters
365of it <idle-loop_>`_, must reflect the properties of the idle state at the
366deepest level (i.e. the idle state of the unit containing all of the other
367units).
368
369For example, take a processor with two cores in a larger unit referred to as
370a "module" and suppose that asking the hardware to enter a specific idle state
371(say "X") at the "core" level by one core will trigger the module to try to
372enter a specific idle state of its own (say "MX") if the other core is in idle
373state "X" already.  In other words, asking for idle state "X" at the "core"
374level gives the hardware a license to go as deep as to idle state "MX" at the
375"module" level, but there is no guarantee that this is going to happen (the core
376asking for idle state "X" may just end up in that state by itself instead).
377Then, the target residency of the |struct cpuidle_state| object representing
378idle state "X" must reflect the minimum time to spend in idle state "MX" of
379the module (including the time needed to enter it), because that is the minimum
380time the CPU needs to be idle to save any energy in case the hardware enters
381that state.  Analogously, the exit latency parameter of that object must cover
382the exit time of idle state "MX" of the module (and usually its entry time too),
383because that is the maximum delay between a wakeup signal and the time the CPU
384will start to execute the first new instruction (assuming that both cores in the
385module will always be ready to execute instructions as soon as the module
386becomes operational as a whole).
387
388There are processors without direct coordination between different levels of the
389hierarchy of units inside them, however.  In those cases asking for an idle
390state at the "core" level does not automatically affect the "module" level, for
391example, in any way and the ``CPUIdle`` driver is responsible for the entire
392handling of the hierarchy.  Then, the definition of the idle state objects is
393entirely up to the driver, but still the physical properties of the idle state
394that the processor hardware finally goes into must always follow the parameters
395used by the governor for idle state selection (for instance, the actual exit
396latency of that idle state must not exceed the exit latency parameter of the
397idle state object selected by the governor).
398
399In addition to the target residency and exit latency idle state parameters
400discussed above, the objects representing idle states each contain a few other
401parameters describing the idle state and a pointer to the function to run in
402order to ask the hardware to enter that state.  Also, for each
403|struct cpuidle_state| object, there is a corresponding
404:c:type:`struct cpuidle_state_usage <cpuidle_state_usage>` one containing usage
405statistics of the given idle state.  That information is exposed by the kernel
406via ``sysfs``.
407
408For each CPU in the system, there is a :file:`/sys/devices/system/cpu/cpu<N>/cpuidle/`
409directory in ``sysfs``, where the number ``<N>`` is assigned to the given
410CPU at the initialization time.  That directory contains a set of subdirectories
411called :file:`state0`, :file:`state1` and so on, up to the number of idle state
412objects defined for the given CPU minus one.  Each of these directories
413corresponds to one idle state object and the larger the number in its name, the
414deeper the (effective) idle state represented by it.  Each of them contains
415a number of files (attributes) representing the properties of the idle state
416object corresponding to it, as follows:
417
418``above``
419	Total number of times this idle state had been asked for, but the
420	observed idle duration was certainly too short to match its target
421	residency.
422
423``below``
424	Total number of times this idle state had been asked for, but certainly
425	a deeper idle state would have been a better match for the observed idle
426	duration.
427
428``desc``
429	Description of the idle state.
430
431``disable``
432	Whether or not this idle state is disabled.
433
434``default_status``
435	The default status of this state, "enabled" or "disabled".
436
437``latency``
438	Exit latency of the idle state in microseconds.
439
440``name``
441	Name of the idle state.
442
443``power``
444	Power drawn by hardware in this idle state in milliwatts (if specified,
445	0 otherwise).
446
447``residency``
448	Target residency of the idle state in microseconds.
449
450``time``
451	Total time spent in this idle state by the given CPU (as measured by the
452	kernel) in microseconds.
453
454``usage``
455	Total number of times the hardware has been asked by the given CPU to
456	enter this idle state.
457
458``rejected``
459	Total number of times a request to enter this idle state on the given
460	CPU was rejected.
461
462The :file:`desc` and :file:`name` files both contain strings.  The difference
463between them is that the name is expected to be more concise, while the
464description may be longer and it may contain white space or special characters.
465The other files listed above contain integer numbers.
466
467The :file:`disable` attribute is the only writeable one.  If it contains 1, the
468given idle state is disabled for this particular CPU, which means that the
469governor will never select it for this particular CPU and the ``CPUIdle``
470driver will never ask the hardware to enter it for that CPU as a result.
471However, disabling an idle state for one CPU does not prevent it from being
472asked for by the other CPUs, so it must be disabled for all of them in order to
473never be asked for by any of them.  [Note that, due to the way the ``ladder``
474governor is implemented, disabling an idle state prevents that governor from
475selecting any idle states deeper than the disabled one too.]
476
477If the :file:`disable` attribute contains 0, the given idle state is enabled for
478this particular CPU, but it still may be disabled for some or all of the other
479CPUs in the system at the same time.  Writing 1 to it causes the idle state to
480be disabled for this particular CPU and writing 0 to it allows the governor to
481take it into consideration for the given CPU and the driver to ask for it,
482unless that state was disabled globally in the driver (in which case it cannot
483be used at all).
484
485The :file:`power` attribute is not defined very well, especially for idle state
486objects representing combinations of idle states at different levels of the
487hierarchy of units in the processor, and it generally is hard to obtain idle
488state power numbers for complex hardware, so :file:`power` often contains 0 (not
489available) and if it contains a nonzero number, that number may not be very
490accurate and it should not be relied on for anything meaningful.
491
492The number in the :file:`time` file generally may be greater than the total time
493really spent by the given CPU in the given idle state, because it is measured by
494the kernel and it may not cover the cases in which the hardware refused to enter
495this idle state and entered a shallower one instead of it (or even it did not
496enter any idle state at all).  The kernel can only measure the time span between
497asking the hardware to enter an idle state and the subsequent wakeup of the CPU
498and it cannot say what really happened in the meantime at the hardware level.
499Moreover, if the idle state object in question represents a combination of idle
500states at different levels of the hierarchy of units in the processor,
501the kernel can never say how deep the hardware went down the hierarchy in any
502particular case.  For these reasons, the only reliable way to find out how
503much time has been spent by the hardware in different idle states supported by
504it is to use idle state residency counters in the hardware, if available.
505
506Generally, an interrupt received when trying to enter an idle state causes the
507idle state entry request to be rejected, in which case the ``CPUIdle`` driver
508may return an error code to indicate that this was the case. The :file:`usage`
509and :file:`rejected` files report the number of times the given idle state
510was entered successfully or rejected, respectively.
511
512.. _cpu-pm-qos:
513
514Power Management Quality of Service for CPUs
515============================================
516
517The power management quality of service (PM QoS) framework in the Linux kernel
518allows kernel code and user space processes to set constraints on various
519energy-efficiency features of the kernel to prevent performance from dropping
520below a required level.
521
522CPU idle time management can be affected by PM QoS in two ways, through the
523global CPU latency limit and through the resume latency constraints for
524individual CPUs.  Kernel code (e.g. device drivers) can set both of them with
525the help of special internal interfaces provided by the PM QoS framework.  User
526space can modify the former by opening the :file:`cpu_dma_latency` special
527device file under :file:`/dev/` and writing a binary value (interpreted as a
528signed 32-bit integer) to it.  In turn, the resume latency constraint for a CPU
529can be modified from user space by writing a string (representing a signed
53032-bit integer) to the :file:`power/pm_qos_resume_latency_us` file under
531:file:`/sys/devices/system/cpu/cpu<N>/` in ``sysfs``, where the CPU number
532``<N>`` is allocated at the system initialization time.  Negative values
533will be rejected in both cases and, also in both cases, the written integer
534number will be interpreted as a requested PM QoS constraint in microseconds.
535
536The requested value is not automatically applied as a new constraint, however,
537as it may be less restrictive (greater in this particular case) than another
538constraint previously requested by someone else.  For this reason, the PM QoS
539framework maintains a list of requests that have been made so far for the
540global CPU latency limit and for each individual CPU, aggregates them and
541applies the effective (minimum in this particular case) value as the new
542constraint.
543
544In fact, opening the :file:`cpu_dma_latency` special device file causes a new
545PM QoS request to be created and added to a global priority list of CPU latency
546limit requests and the file descriptor coming from the "open" operation
547represents that request.  If that file descriptor is then used for writing, the
548number written to it will be associated with the PM QoS request represented by
549it as a new requested limit value.  Next, the priority list mechanism will be
550used to determine the new effective value of the entire list of requests and
551that effective value will be set as a new CPU latency limit.  Thus requesting a
552new limit value will only change the real limit if the effective "list" value is
553affected by it, which is the case if it is the minimum of the requested values
554in the list.
555
556The process holding a file descriptor obtained by opening the
557:file:`cpu_dma_latency` special device file controls the PM QoS request
558associated with that file descriptor, but it controls this particular PM QoS
559request only.
560
561Closing the :file:`cpu_dma_latency` special device file or, more precisely, the
562file descriptor obtained while opening it, causes the PM QoS request associated
563with that file descriptor to be removed from the global priority list of CPU
564latency limit requests and destroyed.  If that happens, the priority list
565mechanism will be used again, to determine the new effective value for the whole
566list and that value will become the new limit.
567
568In turn, for each CPU there is one resume latency PM QoS request associated with
569the :file:`power/pm_qos_resume_latency_us` file under
570:file:`/sys/devices/system/cpu/cpu<N>/` in ``sysfs`` and writing to it causes
571this single PM QoS request to be updated regardless of which user space
572process does that.  In other words, this PM QoS request is shared by the entire
573user space, so access to the file associated with it needs to be arbitrated
574to avoid confusion.  [Arguably, the only legitimate use of this mechanism in
575practice is to pin a process to the CPU in question and let it use the
576``sysfs`` interface to control the resume latency constraint for it.]  It is
577still only a request, however.  It is an entry in a priority list used to
578determine the effective value to be set as the resume latency constraint for the
579CPU in question every time the list of requests is updated this way or another
580(there may be other requests coming from kernel code in that list).
581
582CPU idle time governors are expected to regard the minimum of the global
583(effective) CPU latency limit and the effective resume latency constraint for
584the given CPU as the upper limit for the exit latency of the idle states that
585they are allowed to select for that CPU.  They should never select any idle
586states with exit latency beyond that limit.
587
588
589Idle States Control Via Kernel Command Line
590===========================================
591
592In addition to the ``sysfs`` interface allowing individual idle states to be
593`disabled for individual CPUs <idle-states-representation_>`_, there are kernel
594command line parameters affecting CPU idle time management.
595
596The ``cpuidle.off=1`` kernel command line option can be used to disable the
597CPU idle time management entirely.  It does not prevent the idle loop from
598running on idle CPUs, but it prevents the CPU idle time governors and drivers
599from being invoked.  If it is added to the kernel command line, the idle loop
600will ask the hardware to enter idle states on idle CPUs via the CPU architecture
601support code that is expected to provide a default mechanism for this purpose.
602That default mechanism usually is the least common denominator for all of the
603processors implementing the architecture (i.e. CPU instruction set) in question,
604however, so it is rather crude and not very energy-efficient.  For this reason,
605it is not recommended for production use.
606
607The ``cpuidle.governor=`` kernel command line switch allows the ``CPUIdle``
608governor to use to be specified.  It has to be appended with a string matching
609the name of an available governor (e.g. ``cpuidle.governor=menu``) and that
610governor will be used instead of the default one.  It is possible to force
611the ``menu`` governor to be used on the systems that use the ``ladder`` governor
612by default this way, for example.
613
614The other kernel command line parameters controlling CPU idle time management
615described below are only relevant for the *x86* architecture and some of
616them affect Intel processors only.
617
618The *x86* architecture support code recognizes three kernel command line
619options related to CPU idle time management: ``idle=poll``, ``idle=halt``,
620and ``idle=nomwait``.  The first two of them disable the ``acpi_idle`` and
621``intel_idle`` drivers altogether, which effectively causes the entire
622``CPUIdle`` subsystem to be disabled and makes the idle loop invoke the
623architecture support code to deal with idle CPUs.  How it does that depends on
624which of the two parameters is added to the kernel command line.  In the
625``idle=halt`` case, the architecture support code will use the ``HLT``
626instruction of the CPUs (which, as a rule, suspends the execution of the program
627and causes the hardware to attempt to enter the shallowest available idle state)
628for this purpose, and if ``idle=poll`` is used, idle CPUs will execute a
629more or less "lightweight" sequence of instructions in a tight loop.  [Note
630that using ``idle=poll`` is somewhat drastic in many cases, as preventing idle
631CPUs from saving almost any energy at all may not be the only effect of it.
632For example, on Intel hardware it effectively prevents CPUs from using
633P-states (see |cpufreq|) that require any number of CPUs in a package to be
634idle, so it very well may hurt single-thread computations performance as well as
635energy-efficiency.  Thus using it for performance reasons may not be a good idea
636at all.]
637
638The ``idle=nomwait`` option disables the ``intel_idle`` driver and causes
639``acpi_idle`` to be used (as long as all of the information needed by it is
640there in the system's ACPI tables), but it is not allowed to use the
641``MWAIT`` instruction of the CPUs to ask the hardware to enter idle states.
642
643In addition to the architecture-level kernel command line options affecting CPU
644idle time management, there are parameters affecting individual ``CPUIdle``
645drivers that can be passed to them via the kernel command line.  Specifically,
646the ``intel_idle.max_cstate=<n>`` and ``processor.max_cstate=<n>`` parameters,
647where ``<n>`` is an idle state index also used in the name of the given
648state's directory in ``sysfs`` (see
649`Representation of Idle States <idle-states-representation_>`_), causes the
650``intel_idle`` and ``acpi_idle`` drivers, respectively, to discard all of the
651idle states deeper than idle state ``<n>``.  In that case, they will never ask
652for any of those idle states or expose them to the governor.  [The behavior of
653the two drivers is different for ``<n>`` equal to ``0``.  Adding
654``intel_idle.max_cstate=0`` to the kernel command line disables the
655``intel_idle`` driver and allows ``acpi_idle`` to be used, whereas
656``processor.max_cstate=0`` is equivalent to ``processor.max_cstate=1``.
657Also, the ``acpi_idle`` driver is part of the ``processor`` kernel module that
658can be loaded separately and ``max_cstate=<n>`` can be passed to it as a module
659parameter when it is loaded.]
660