1.. SPDX-License-Identifier: GPL-2.0
2
3===============================================
4How to Implement a new CPUFreq Processor Driver
5===============================================
6
7Authors:
8
9
10	- Dominik Brodowski  <linux@brodo.de>
11	- Rafael J. Wysocki <rafael.j.wysocki@intel.com>
12	- Viresh Kumar <viresh.kumar@linaro.org>
13
14.. Contents
15
16   1.   What To Do?
17   1.1  Initialization
18   1.2  Per-CPU Initialization
19   1.3  verify
20   1.4  target/target_index or setpolicy?
21   1.5  target/target_index
22   1.6  setpolicy
23   1.7  get_intermediate and target_intermediate
24   2.   Frequency Table Helpers
25
26
27
281. What To Do?
29==============
30
31So, you just got a brand-new CPU / chipset with datasheets and want to
32add cpufreq support for this CPU / chipset? Great. Here are some hints
33on what is necessary:
34
35
361.1 Initialization
37------------------
38
39First of all, in an __initcall level 7 (module_init()) or later
40function check whether this kernel runs on the right CPU and the right
41chipset. If so, register a struct cpufreq_driver with the CPUfreq core
42using cpufreq_register_driver()
43
44What shall this struct cpufreq_driver contain?
45
46 .name - The name of this driver.
47
48 .init - A pointer to the per-policy initialization function.
49
50 .verify - A pointer to a "verification" function.
51
52 .setpolicy _or_ .fast_switch _or_ .target _or_ .target_index - See
53 below on the differences.
54
55And optionally
56
57 .flags - Hints for the cpufreq core.
58
59 .driver_data - cpufreq driver specific data.
60
61 .get_intermediate and target_intermediate - Used to switch to stable
62 frequency while changing CPU frequency.
63
64 .get - Returns current frequency of the CPU.
65
66 .bios_limit - Returns HW/BIOS max frequency limitations for the CPU.
67
68 .exit - A pointer to a per-policy cleanup function called during
69 CPU_POST_DEAD phase of cpu hotplug process.
70
71 .suspend - A pointer to a per-policy suspend function which is called
72 with interrupts disabled and _after_ the governor is stopped for the
73 policy.
74
75 .resume - A pointer to a per-policy resume function which is called
76 with interrupts disabled and _before_ the governor is started again.
77
78 .attr - A pointer to a NULL-terminated list of "struct freq_attr" which
79 allow to export values to sysfs.
80
81 .boost_enabled - If set, boost frequencies are enabled.
82
83 .set_boost - A pointer to a per-policy function to enable/disable boost
84 frequencies.
85
86
871.2 Per-CPU Initialization
88--------------------------
89
90Whenever a new CPU is registered with the device model, or after the
91cpufreq driver registers itself, the per-policy initialization function
92cpufreq_driver.init is called if no cpufreq policy existed for the CPU.
93Note that the .init() and .exit() routines are called only once for the
94policy and not for each CPU managed by the policy. It takes a ``struct
95cpufreq_policy *policy`` as argument. What to do now?
96
97If necessary, activate the CPUfreq support on your CPU.
98
99Then, the driver must fill in the following values:
100
101+-----------------------------------+--------------------------------------+
102|policy->cpuinfo.min_freq _and_	    |					   |
103|policy->cpuinfo.max_freq	    | the minimum and maximum frequency	   |
104|				    | (in kHz) which is supported by	   |
105|				    | this CPU				   |
106+-----------------------------------+--------------------------------------+
107|policy->cpuinfo.transition_latency | the time it takes on this CPU to	   |
108|				    | switch between two frequencies in	   |
109|				    | nanoseconds (if appropriate, else	   |
110|				    | specify CPUFREQ_ETERNAL)		   |
111+-----------------------------------+--------------------------------------+
112|policy->cur			    | The current operating frequency of   |
113|				    | this CPU (if appropriate)		   |
114+-----------------------------------+--------------------------------------+
115|policy->min,			    |					   |
116|policy->max,			    |					   |
117|policy->policy and, if necessary,  |					   |
118|policy->governor		    | must contain the "default policy" for|
119|				    | this CPU. A few moments later,       |
120|				    | cpufreq_driver.verify and either     |
121|				    | cpufreq_driver.setpolicy or          |
122|				    | cpufreq_driver.target/target_index is|
123|				    | called with these values.		   |
124+-----------------------------------+--------------------------------------+
125|policy->cpus			    | Update this with the masks of the	   |
126|				    | (online + offline) CPUs that do DVFS |
127|				    | along with this CPU (i.e.  that share|
128|				    | clock/voltage rails with it).	   |
129+-----------------------------------+--------------------------------------+
130
131For setting some of these values (cpuinfo.min[max]_freq, policy->min[max]), the
132frequency table helpers might be helpful. See the section 2 for more information
133on them.
134
135
1361.3 verify
137----------
138
139When the user decides a new policy (consisting of
140"policy,governor,min,max") shall be set, this policy must be validated
141so that incompatible values can be corrected. For verifying these
142values cpufreq_verify_within_limits(``struct cpufreq_policy *policy``,
143``unsigned int min_freq``, ``unsigned int max_freq``) function might be helpful.
144See section 2 for details on frequency table helpers.
145
146You need to make sure that at least one valid frequency (or operating
147range) is within policy->min and policy->max. If necessary, increase
148policy->max first, and only if this is no solution, decrease policy->min.
149
150
1511.4 target or target_index or setpolicy or fast_switch?
152-------------------------------------------------------
153
154Most cpufreq drivers or even most cpu frequency scaling algorithms
155only allow the CPU frequency to be set to predefined fixed values. For
156these, you use the ->target(), ->target_index() or ->fast_switch()
157callbacks.
158
159Some cpufreq capable processors switch the frequency between certain
160limits on their own. These shall use the ->setpolicy() callback.
161
162
1631.5. target/target_index
164------------------------
165
166The target_index call has two arguments: ``struct cpufreq_policy *policy``,
167and ``unsigned int`` index (into the exposed frequency table).
168
169The CPUfreq driver must set the new frequency when called here. The
170actual frequency must be determined by freq_table[index].frequency.
171
172It should always restore to earlier frequency (i.e. policy->restore_freq) in
173case of errors, even if we switched to intermediate frequency earlier.
174
175Deprecated
176----------
177The target call has three arguments: ``struct cpufreq_policy *policy``,
178unsigned int target_frequency, unsigned int relation.
179
180The CPUfreq driver must set the new frequency when called here. The
181actual frequency must be determined using the following rules:
182
183- keep close to "target_freq"
184- policy->min <= new_freq <= policy->max (THIS MUST BE VALID!!!)
185- if relation==CPUFREQ_REL_L, try to select a new_freq higher than or equal
186  target_freq. ("L for lowest, but no lower than")
187- if relation==CPUFREQ_REL_H, try to select a new_freq lower than or equal
188  target_freq. ("H for highest, but no higher than")
189
190Here again the frequency table helper might assist you - see section 2
191for details.
192
1931.6. fast_switch
194----------------
195
196This function is used for frequency switching from scheduler's context.
197Not all drivers are expected to implement it, as sleeping from within
198this callback isn't allowed. This callback must be highly optimized to
199do switching as fast as possible.
200
201This function has two arguments: ``struct cpufreq_policy *policy`` and
202``unsigned int target_frequency``.
203
204
2051.7 setpolicy
206-------------
207
208The setpolicy call only takes a ``struct cpufreq_policy *policy`` as
209argument. You need to set the lower limit of the in-processor or
210in-chipset dynamic frequency switching to policy->min, the upper limit
211to policy->max, and -if supported- select a performance-oriented
212setting when policy->policy is CPUFREQ_POLICY_PERFORMANCE, and a
213powersaving-oriented setting when CPUFREQ_POLICY_POWERSAVE. Also check
214the reference implementation in drivers/cpufreq/longrun.c
215
2161.8 get_intermediate and target_intermediate
217--------------------------------------------
218
219Only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset.
220
221get_intermediate should return a stable intermediate frequency platform wants to
222switch to, and target_intermediate() should set CPU to that frequency, before
223jumping to the frequency corresponding to 'index'. Core will take care of
224sending notifications and driver doesn't have to handle them in
225target_intermediate() or target_index().
226
227Drivers can return '0' from get_intermediate() in case they don't wish to switch
228to intermediate frequency for some target frequency. In that case core will
229directly call ->target_index().
230
231NOTE: ->target_index() should restore to policy->restore_freq in case of
232failures as core would send notifications for that.
233
234
2352. Frequency Table Helpers
236==========================
237
238As most cpufreq processors only allow for being set to a few specific
239frequencies, a "frequency table" with some functions might assist in
240some work of the processor driver. Such a "frequency table" consists of
241an array of struct cpufreq_frequency_table entries, with driver specific
242values in "driver_data", the corresponding frequency in "frequency" and
243flags set. At the end of the table, you need to add a
244cpufreq_frequency_table entry with frequency set to CPUFREQ_TABLE_END.
245And if you want to skip one entry in the table, set the frequency to
246CPUFREQ_ENTRY_INVALID. The entries don't need to be in sorted in any
247particular order, but if they are cpufreq core will do DVFS a bit
248quickly for them as search for best match is faster.
249
250The cpufreq table is verified automatically by the core if the policy contains a
251valid pointer in its policy->freq_table field.
252
253cpufreq_frequency_table_verify() assures that at least one valid
254frequency is within policy->min and policy->max, and all other criteria
255are met. This is helpful for the ->verify call.
256
257cpufreq_frequency_table_target() is the corresponding frequency table
258helper for the ->target stage. Just pass the values to this function,
259and this function returns the of the frequency table entry which
260contains the frequency the CPU shall be set to.
261
262The following macros can be used as iterators over cpufreq_frequency_table:
263
264cpufreq_for_each_entry(pos, table) - iterates over all entries of frequency
265table.
266
267cpufreq_for_each_valid_entry(pos, table) - iterates over all entries,
268excluding CPUFREQ_ENTRY_INVALID frequencies.
269Use arguments "pos" - a ``cpufreq_frequency_table *`` as a loop cursor and
270"table" - the ``cpufreq_frequency_table *`` you want to iterate over.
271
272For example::
273
274	struct cpufreq_frequency_table *pos, *driver_freq_table;
275
276	cpufreq_for_each_entry(pos, driver_freq_table) {
277		/* Do something with pos */
278		pos->frequency = ...
279	}
280
281If you need to work with the position of pos within driver_freq_table,
282do not subtract the pointers, as it is quite costly. Instead, use the
283macros cpufreq_for_each_entry_idx() and cpufreq_for_each_valid_entry_idx().
284