1 .. SPDX-License-Identifier: GPL-2.0 1 .. SPDX-License-Identifier: GPL-2.0
2 2
3 ============================================== 3 ===============================================
4 How to Implement a new CPUFreq Processor Drive 4 How to Implement a new CPUFreq Processor Driver
5 ============================================== 5 ===============================================
6 6
7 Authors: 7 Authors:
8 8
9 9
10 - Dominik Brodowski <linux@brodo.de> 10 - Dominik Brodowski <linux@brodo.de>
11 - Rafael J. Wysocki <rafael.j.wysocki@i 11 - Rafael J. Wysocki <rafael.j.wysocki@intel.com>
12 - Viresh Kumar <viresh.kumar@linaro.org 12 - Viresh Kumar <viresh.kumar@linaro.org>
13 13
14 .. Contents 14 .. Contents
15 15
16 1. What To Do? 16 1. What To Do?
17 1.1 Initialization 17 1.1 Initialization
18 1.2 Per-CPU Initialization 18 1.2 Per-CPU Initialization
19 1.3 verify 19 1.3 verify
20 1.4 target/target_index or setpolicy? 20 1.4 target/target_index or setpolicy?
21 1.5 target/target_index 21 1.5 target/target_index
22 1.6 setpolicy 22 1.6 setpolicy
23 1.7 get_intermediate and target_intermedia 23 1.7 get_intermediate and target_intermediate
24 2. Frequency Table Helpers 24 2. Frequency Table Helpers
25 25
26 26
27 27
28 1. What To Do? 28 1. What To Do?
29 ============== 29 ==============
30 30
31 So, you just got a brand-new CPU / chipset wit 31 So, you just got a brand-new CPU / chipset with datasheets and want to
32 add cpufreq support for this CPU / chipset? Gr 32 add cpufreq support for this CPU / chipset? Great. Here are some hints
33 on what is necessary: 33 on what is necessary:
34 34
35 35
36 1.1 Initialization 36 1.1 Initialization
37 ------------------ 37 ------------------
38 38
39 First of all, in an __initcall level 7 (module 39 First of all, in an __initcall level 7 (module_init()) or later
40 function check whether this kernel runs on the 40 function check whether this kernel runs on the right CPU and the right
41 chipset. If so, register a struct cpufreq_driv 41 chipset. If so, register a struct cpufreq_driver with the CPUfreq core
42 using cpufreq_register_driver() 42 using cpufreq_register_driver()
43 43
44 What shall this struct cpufreq_driver contain? 44 What shall this struct cpufreq_driver contain?
45 45
46 .name - The name of this driver. 46 .name - The name of this driver.
47 47
48 .init - A pointer to the per-policy initializ 48 .init - A pointer to the per-policy initialization function.
49 49
50 .verify - A pointer to a "verification" funct 50 .verify - A pointer to a "verification" function.
51 51
52 .setpolicy _or_ .fast_switch _or_ .target _or 52 .setpolicy _or_ .fast_switch _or_ .target _or_ .target_index - See
53 below on the differences. 53 below on the differences.
54 54
55 And optionally 55 And optionally
56 56
57 .flags - Hints for the cpufreq core. 57 .flags - Hints for the cpufreq core.
58 58
59 .driver_data - cpufreq driver specific data. 59 .driver_data - cpufreq driver specific data.
60 60
>> 61 .resolve_freq - Returns the most appropriate frequency for a target
>> 62 frequency. Doesn't change the frequency though.
>> 63
61 .get_intermediate and target_intermediate - U 64 .get_intermediate and target_intermediate - Used to switch to stable
62 frequency while changing CPU frequency. 65 frequency while changing CPU frequency.
63 66
64 .get - Returns current frequency of the CPU. 67 .get - Returns current frequency of the CPU.
65 68
66 .bios_limit - Returns HW/BIOS max frequency l 69 .bios_limit - Returns HW/BIOS max frequency limitations for the CPU.
67 70
68 .exit - A pointer to a per-policy cleanup fun 71 .exit - A pointer to a per-policy cleanup function called during
69 CPU_POST_DEAD phase of cpu hotplug process. 72 CPU_POST_DEAD phase of cpu hotplug process.
>> 73
>> 74 .stop_cpu - A pointer to a per-policy stop function called during
>> 75 CPU_DOWN_PREPARE phase of cpu hotplug process.
70 76
71 .suspend - A pointer to a per-policy suspend 77 .suspend - A pointer to a per-policy suspend function which is called
72 with interrupts disabled and _after_ the gove 78 with interrupts disabled and _after_ the governor is stopped for the
73 policy. 79 policy.
74 80
75 .resume - A pointer to a per-policy resume fu 81 .resume - A pointer to a per-policy resume function which is called
76 with interrupts disabled and _before_ the gov 82 with interrupts disabled and _before_ the governor is started again.
77 83
78 .ready - A pointer to a per-policy ready func 84 .ready - A pointer to a per-policy ready function which is called after
79 the policy is fully initialized. 85 the policy is fully initialized.
80 86
81 .attr - A pointer to a NULL-terminated list o 87 .attr - A pointer to a NULL-terminated list of "struct freq_attr" which
82 allow to export values to sysfs. 88 allow to export values to sysfs.
83 89
84 .boost_enabled - If set, boost frequencies ar 90 .boost_enabled - If set, boost frequencies are enabled.
85 91
86 .set_boost - A pointer to a per-policy functi 92 .set_boost - A pointer to a per-policy function to enable/disable boost
87 frequencies. 93 frequencies.
88 94
89 95
90 1.2 Per-CPU Initialization 96 1.2 Per-CPU Initialization
91 -------------------------- 97 --------------------------
92 98
93 Whenever a new CPU is registered with the devi 99 Whenever a new CPU is registered with the device model, or after the
94 cpufreq driver registers itself, the per-polic 100 cpufreq driver registers itself, the per-policy initialization function
95 cpufreq_driver.init is called if no cpufreq po 101 cpufreq_driver.init is called if no cpufreq policy existed for the CPU.
96 Note that the .init() and .exit() routines are 102 Note that the .init() and .exit() routines are called only once for the
97 policy and not for each CPU managed by the pol 103 policy and not for each CPU managed by the policy. It takes a ``struct
98 cpufreq_policy *policy`` as argument. What to 104 cpufreq_policy *policy`` as argument. What to do now?
99 105
100 If necessary, activate the CPUfreq support on 106 If necessary, activate the CPUfreq support on your CPU.
101 107
102 Then, the driver must fill in the following va 108 Then, the driver must fill in the following values:
103 109
104 +-----------------------------------+--------- 110 +-----------------------------------+--------------------------------------+
105 |policy->cpuinfo.min_freq _and_ | 111 |policy->cpuinfo.min_freq _and_ | |
106 |policy->cpuinfo.max_freq | the mini 112 |policy->cpuinfo.max_freq | the minimum and maximum frequency |
107 | | (in kHz) 113 | | (in kHz) which is supported by |
108 | | this CPU 114 | | this CPU |
109 +-----------------------------------+--------- 115 +-----------------------------------+--------------------------------------+
110 |policy->cpuinfo.transition_latency | the time 116 |policy->cpuinfo.transition_latency | the time it takes on this CPU to |
111 | | switch b 117 | | switch between two frequencies in |
112 | | nanoseco 118 | | nanoseconds (if appropriate, else |
113 | | specify 119 | | specify CPUFREQ_ETERNAL) |
114 +-----------------------------------+--------- 120 +-----------------------------------+--------------------------------------+
115 |policy->cur | The curr 121 |policy->cur | The current operating frequency of |
116 | | this CPU 122 | | this CPU (if appropriate) |
117 +-----------------------------------+--------- 123 +-----------------------------------+--------------------------------------+
118 |policy->min, | 124 |policy->min, | |
119 |policy->max, | 125 |policy->max, | |
120 |policy->policy and, if necessary, | 126 |policy->policy and, if necessary, | |
121 |policy->governor | must con 127 |policy->governor | must contain the "default policy" for|
122 | | this CPU 128 | | this CPU. A few moments later, |
123 | | cpufreq_ 129 | | cpufreq_driver.verify and either |
124 | | cpufreq_ 130 | | cpufreq_driver.setpolicy or |
125 | | cpufreq_ 131 | | cpufreq_driver.target/target_index is|
126 | | called w 132 | | called with these values. |
127 +-----------------------------------+--------- 133 +-----------------------------------+--------------------------------------+
128 |policy->cpus | Update t 134 |policy->cpus | Update this with the masks of the |
129 | | (online 135 | | (online + offline) CPUs that do DVFS |
130 | | along wi 136 | | along with this CPU (i.e. that share|
131 | | clock/vo 137 | | clock/voltage rails with it). |
132 +-----------------------------------+--------- 138 +-----------------------------------+--------------------------------------+
133 139
134 For setting some of these values (cpuinfo.min[ 140 For setting some of these values (cpuinfo.min[max]_freq, policy->min[max]), the
135 frequency table helpers might be helpful. See 141 frequency table helpers might be helpful. See the section 2 for more information
136 on them. 142 on them.
137 143
138 144
139 1.3 verify 145 1.3 verify
140 ---------- 146 ----------
141 147
142 When the user decides a new policy (consisting 148 When the user decides a new policy (consisting of
143 "policy,governor,min,max") shall be set, this 149 "policy,governor,min,max") shall be set, this policy must be validated
144 so that incompatible values can be corrected. 150 so that incompatible values can be corrected. For verifying these
145 values cpufreq_verify_within_limits(``struct c 151 values cpufreq_verify_within_limits(``struct cpufreq_policy *policy``,
146 ``unsigned int min_freq``, ``unsigned int max_ 152 ``unsigned int min_freq``, ``unsigned int max_freq``) function might be helpful.
147 See section 2 for details on frequency table h 153 See section 2 for details on frequency table helpers.
148 154
149 You need to make sure that at least one valid 155 You need to make sure that at least one valid frequency (or operating
150 range) is within policy->min and policy->max. 156 range) is within policy->min and policy->max. If necessary, increase
151 policy->max first, and only if this is no solu 157 policy->max first, and only if this is no solution, decrease policy->min.
152 158
153 159
154 1.4 target or target_index or setpolicy or fas 160 1.4 target or target_index or setpolicy or fast_switch?
155 ---------------------------------------------- 161 -------------------------------------------------------
156 162
157 Most cpufreq drivers or even most cpu frequenc 163 Most cpufreq drivers or even most cpu frequency scaling algorithms
158 only allow the CPU frequency to be set to pred 164 only allow the CPU frequency to be set to predefined fixed values. For
159 these, you use the ->target(), ->target_index( 165 these, you use the ->target(), ->target_index() or ->fast_switch()
160 callbacks. 166 callbacks.
161 167
162 Some cpufreq capable processors switch the fre 168 Some cpufreq capable processors switch the frequency between certain
163 limits on their own. These shall use the ->set 169 limits on their own. These shall use the ->setpolicy() callback.
164 170
165 171
166 1.5. target/target_index 172 1.5. target/target_index
167 ------------------------ 173 ------------------------
168 174
169 The target_index call has two arguments: ``str 175 The target_index call has two arguments: ``struct cpufreq_policy *policy``,
170 and ``unsigned int`` index (into the exposed f 176 and ``unsigned int`` index (into the exposed frequency table).
171 177
172 The CPUfreq driver must set the new frequency 178 The CPUfreq driver must set the new frequency when called here. The
173 actual frequency must be determined by freq_ta 179 actual frequency must be determined by freq_table[index].frequency.
174 180
175 It should always restore to earlier frequency 181 It should always restore to earlier frequency (i.e. policy->restore_freq) in
176 case of errors, even if we switched to interme 182 case of errors, even if we switched to intermediate frequency earlier.
177 183
178 Deprecated 184 Deprecated
179 ---------- 185 ----------
180 The target call has three arguments: ``struct 186 The target call has three arguments: ``struct cpufreq_policy *policy``,
181 unsigned int target_frequency, unsigned int re 187 unsigned int target_frequency, unsigned int relation.
182 188
183 The CPUfreq driver must set the new frequency 189 The CPUfreq driver must set the new frequency when called here. The
184 actual frequency must be determined using the 190 actual frequency must be determined using the following rules:
185 191
186 - keep close to "target_freq" 192 - keep close to "target_freq"
187 - policy->min <= new_freq <= policy->max (THIS 193 - policy->min <= new_freq <= policy->max (THIS MUST BE VALID!!!)
188 - if relation==CPUFREQ_REL_L, try to select a 194 - if relation==CPUFREQ_REL_L, try to select a new_freq higher than or equal
189 target_freq. ("L for lowest, but no lower th 195 target_freq. ("L for lowest, but no lower than")
190 - if relation==CPUFREQ_REL_H, try to select a 196 - if relation==CPUFREQ_REL_H, try to select a new_freq lower than or equal
191 target_freq. ("H for highest, but no higher 197 target_freq. ("H for highest, but no higher than")
192 198
193 Here again the frequency table helper might as 199 Here again the frequency table helper might assist you - see section 2
194 for details. 200 for details.
195 201
196 1.6. fast_switch 202 1.6. fast_switch
197 ---------------- 203 ----------------
198 204
199 This function is used for frequency switching 205 This function is used for frequency switching from scheduler's context.
200 Not all drivers are expected to implement it, 206 Not all drivers are expected to implement it, as sleeping from within
201 this callback isn't allowed. This callback mus 207 this callback isn't allowed. This callback must be highly optimized to
202 do switching as fast as possible. 208 do switching as fast as possible.
203 209
204 This function has two arguments: ``struct cpuf 210 This function has two arguments: ``struct cpufreq_policy *policy`` and
205 ``unsigned int target_frequency``. 211 ``unsigned int target_frequency``.
206 212
207 213
208 1.7 setpolicy 214 1.7 setpolicy
209 ------------- 215 -------------
210 216
211 The setpolicy call only takes a ``struct cpufr 217 The setpolicy call only takes a ``struct cpufreq_policy *policy`` as
212 argument. You need to set the lower limit of t 218 argument. You need to set the lower limit of the in-processor or
213 in-chipset dynamic frequency switching to poli 219 in-chipset dynamic frequency switching to policy->min, the upper limit
214 to policy->max, and -if supported- select a pe 220 to policy->max, and -if supported- select a performance-oriented
215 setting when policy->policy is CPUFREQ_POLICY_ 221 setting when policy->policy is CPUFREQ_POLICY_PERFORMANCE, and a
216 powersaving-oriented setting when CPUFREQ_POLI 222 powersaving-oriented setting when CPUFREQ_POLICY_POWERSAVE. Also check
217 the reference implementation in drivers/cpufre 223 the reference implementation in drivers/cpufreq/longrun.c
218 224
219 1.8 get_intermediate and target_intermediate 225 1.8 get_intermediate and target_intermediate
220 -------------------------------------------- 226 --------------------------------------------
221 227
222 Only for drivers with target_index() and CPUFR 228 Only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset.
223 229
224 get_intermediate should return a stable interm 230 get_intermediate should return a stable intermediate frequency platform wants to
225 switch to, and target_intermediate() should se 231 switch to, and target_intermediate() should set CPU to that frequency, before
226 jumping to the frequency corresponding to 'ind 232 jumping to the frequency corresponding to 'index'. Core will take care of
227 sending notifications and driver doesn't have 233 sending notifications and driver doesn't have to handle them in
228 target_intermediate() or target_index(). 234 target_intermediate() or target_index().
229 235
230 Drivers can return '0' from get_intermediate() 236 Drivers can return '0' from get_intermediate() in case they don't wish to switch
231 to intermediate frequency for some target freq 237 to intermediate frequency for some target frequency. In that case core will
232 directly call ->target_index(). 238 directly call ->target_index().
233 239
234 NOTE: ->target_index() should restore to polic 240 NOTE: ->target_index() should restore to policy->restore_freq in case of
235 failures as core would send notifications for 241 failures as core would send notifications for that.
236 242
237 243
238 2. Frequency Table Helpers 244 2. Frequency Table Helpers
239 ========================== 245 ==========================
240 246
241 As most cpufreq processors only allow for bein 247 As most cpufreq processors only allow for being set to a few specific
242 frequencies, a "frequency table" with some fun 248 frequencies, a "frequency table" with some functions might assist in
243 some work of the processor driver. Such a "fre 249 some work of the processor driver. Such a "frequency table" consists of
244 an array of struct cpufreq_frequency_table ent 250 an array of struct cpufreq_frequency_table entries, with driver specific
245 values in "driver_data", the corresponding fre 251 values in "driver_data", the corresponding frequency in "frequency" and
246 flags set. At the end of the table, you need t 252 flags set. At the end of the table, you need to add a
247 cpufreq_frequency_table entry with frequency s 253 cpufreq_frequency_table entry with frequency set to CPUFREQ_TABLE_END.
248 And if you want to skip one entry in the table 254 And if you want to skip one entry in the table, set the frequency to
249 CPUFREQ_ENTRY_INVALID. The entries don't need 255 CPUFREQ_ENTRY_INVALID. The entries don't need to be in sorted in any
250 particular order, but if they are cpufreq core 256 particular order, but if they are cpufreq core will do DVFS a bit
251 quickly for them as search for best match is f 257 quickly for them as search for best match is faster.
252 258
253 The cpufreq table is verified automatically by 259 The cpufreq table is verified automatically by the core if the policy contains a
254 valid pointer in its policy->freq_table field. 260 valid pointer in its policy->freq_table field.
255 261
256 cpufreq_frequency_table_verify() assures that 262 cpufreq_frequency_table_verify() assures that at least one valid
257 frequency is within policy->min and policy->ma 263 frequency is within policy->min and policy->max, and all other criteria
258 are met. This is helpful for the ->verify call 264 are met. This is helpful for the ->verify call.
259 265
260 cpufreq_frequency_table_target() is the corres 266 cpufreq_frequency_table_target() is the corresponding frequency table
261 helper for the ->target stage. Just pass the v 267 helper for the ->target stage. Just pass the values to this function,
262 and this function returns the of the frequency 268 and this function returns the of the frequency table entry which
263 contains the frequency the CPU shall be set to 269 contains the frequency the CPU shall be set to.
264 270
265 The following macros can be used as iterators 271 The following macros can be used as iterators over cpufreq_frequency_table:
266 272
267 cpufreq_for_each_entry(pos, table) - iterates 273 cpufreq_for_each_entry(pos, table) - iterates over all entries of frequency
268 table. 274 table.
269 275
270 cpufreq_for_each_valid_entry(pos, table) - ite 276 cpufreq_for_each_valid_entry(pos, table) - iterates over all entries,
271 excluding CPUFREQ_ENTRY_INVALID frequencies. 277 excluding CPUFREQ_ENTRY_INVALID frequencies.
272 Use arguments "pos" - a ``cpufreq_frequency_ta 278 Use arguments "pos" - a ``cpufreq_frequency_table *`` as a loop cursor and
273 "table" - the ``cpufreq_frequency_table *`` yo 279 "table" - the ``cpufreq_frequency_table *`` you want to iterate over.
274 280
275 For example:: 281 For example::
276 282
277 struct cpufreq_frequency_table *pos, * 283 struct cpufreq_frequency_table *pos, *driver_freq_table;
278 284
279 cpufreq_for_each_entry(pos, driver_fre 285 cpufreq_for_each_entry(pos, driver_freq_table) {
280 /* Do something with pos */ 286 /* Do something with pos */
281 pos->frequency = ... 287 pos->frequency = ...
282 } 288 }
283 289
284 If you need to work with the position of pos w 290 If you need to work with the position of pos within driver_freq_table,
285 do not subtract the pointers, as it is quite c 291 do not subtract the pointers, as it is quite costly. Instead, use the
286 macros cpufreq_for_each_entry_idx() and cpufre 292 macros cpufreq_for_each_entry_idx() and cpufreq_for_each_valid_entry_idx().
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.