xref: /linux/Documentation/cpu-freq/core.rst (revision da1d9caf95def6f0320819cf941c9fd1069ba9e1)
1.. SPDX-License-Identifier: GPL-2.0
2
3=============================================================
4General description of the CPUFreq core and CPUFreq notifiers
5=============================================================
6
7Authors:
8	- Dominik Brodowski  <linux@brodo.de>
9	- David Kimdon <dwhedon@debian.org>
10	- Rafael J. Wysocki <rafael.j.wysocki@intel.com>
11	- Viresh Kumar <viresh.kumar@linaro.org>
12
13.. Contents:
14
15   1.  CPUFreq core and interfaces
16   2.  CPUFreq notifiers
17   3.  CPUFreq Table Generation with Operating Performance Point (OPP)
18
191. General Information
20======================
21
22The CPUFreq core code is located in drivers/cpufreq/cpufreq.c. This
23cpufreq code offers a standardized interface for the CPUFreq
24architecture drivers (those pieces of code that do actual
25frequency transitions), as well as to "notifiers". These are device
26drivers or other part of the kernel that need to be informed of
27policy changes (ex. thermal modules like ACPI) or of all
28frequency changes (ex. timing code) or even need to force certain
29speed limits (like LCD drivers on ARM architecture). Additionally, the
30kernel "constant" loops_per_jiffy is updated on frequency changes
31here.
32
33Reference counting of the cpufreq policies is done by cpufreq_cpu_get
34and cpufreq_cpu_put, which make sure that the cpufreq driver is
35correctly registered with the core, and will not be unloaded until
36cpufreq_put_cpu is called. That also ensures that the respective cpufreq
37policy doesn't get freed while being used.
38
392. CPUFreq notifiers
40====================
41
42CPUFreq notifiers conform to the standard kernel notifier interface.
43See linux/include/linux/notifier.h for details on notifiers.
44
45There are two different CPUFreq notifiers - policy notifiers and
46transition notifiers.
47
48
492.1 CPUFreq policy notifiers
50----------------------------
51
52These are notified when a new policy is created or removed.
53
54The phase is specified in the second argument to the notifier.  The phase is
55CPUFREQ_CREATE_POLICY when the policy is first created and it is
56CPUFREQ_REMOVE_POLICY when the policy is removed.
57
58The third argument, a ``void *pointer``, points to a struct cpufreq_policy
59consisting of several values, including min, max (the lower and upper
60frequencies (in kHz) of the new policy).
61
62
632.2 CPUFreq transition notifiers
64--------------------------------
65
66These are notified twice for each online CPU in the policy, when the
67CPUfreq driver switches the CPU core frequency and this change has no
68any external implications.
69
70The second argument specifies the phase - CPUFREQ_PRECHANGE or
71CPUFREQ_POSTCHANGE.
72
73The third argument is a struct cpufreq_freqs with the following
74values:
75
76======	======================================
77policy	a pointer to the struct cpufreq_policy
78old	old frequency
79new	new frequency
80flags	flags of the cpufreq driver
81======	======================================
82
833. CPUFreq Table Generation with Operating Performance Point (OPP)
84==================================================================
85For details about OPP, see Documentation/power/opp.rst
86
87dev_pm_opp_init_cpufreq_table -
88	This function provides a ready to use conversion routine to translate
89	the OPP layer's internal information about the available frequencies
90	into a format readily providable to cpufreq.
91
92	.. Warning::
93
94	   Do not use this function in interrupt context.
95
96	Example::
97
98	 soc_pm_init()
99	 {
100		/* Do things */
101		r = dev_pm_opp_init_cpufreq_table(dev, &freq_table);
102		if (!r)
103			policy->freq_table = freq_table;
104		/* Do other things */
105	 }
106
107	.. note::
108
109	   This function is available only if CONFIG_CPU_FREQ is enabled in
110	   addition to CONFIG_PM_OPP.
111
112dev_pm_opp_free_cpufreq_table
113	Free up the table allocated by dev_pm_opp_init_cpufreq_table
114