How to Implement a new CPUFreq Processor Driver

Authors:

1. What To Do?

So, you just got a brand-new CPU / chipset with datasheets and want toadd cpufreq support for this CPU / chipset? Great. Here are some hintson what is necessary:

1.1 Initialization

First of all, in an __initcall level 7 (module_init()) or laterfunction check whether this kernel runs on the right CPU and the rightchipset. If so, register a struct cpufreq_driver with the CPUfreq coreusing cpufreq_register_driver()

What shall this struct cpufreq_driver contain?

.name - The name of this driver.

.init - A pointer to the per-policy initialization function.

.verify - A pointer to a “verification” function.

.setpolicy _or_ .fast_switch _or_ .target _or_ .target_index - Seebelow on the differences.

And optionally

.flags - Hints for the cpufreq core.

.driver_data - cpufreq driver specific data.

.resolve_freq - Returns the most appropriate frequency for a targetfrequency. Doesn’t change the frequency though.

.get_intermediate and target_intermediate - Used to switch to stablefrequency while changing CPU frequency.

.get - Returns current frequency of the CPU.

.bios_limit - Returns HW/BIOS max frequency limitations for the CPU.

.exit - A pointer to a per-policy cleanup function called duringCPU_POST_DEAD phase of cpu hotplug process.

.stop_cpu - A pointer to a per-policy stop function called duringCPU_DOWN_PREPARE phase of cpu hotplug process.

.suspend - A pointer to a per-policy suspend function which is calledwith interrupts disabled and _after_ the governor is stopped for thepolicy.

.resume - A pointer to a per-policy resume function which is calledwith interrupts disabled and _before_ the governor is started again.

.ready - A pointer to a per-policy ready function which is called afterthe policy is fully initialized.

.attr - A pointer to a NULL-terminated list of “struct freq_attr” whichallow to export values to sysfs.

.boost_enabled - If set, boost frequencies are enabled.

.set_boost - A pointer to a per-policy function to enable/disable boostfrequencies.

1.2 Per-CPU Initialization

Whenever a new CPU is registered with the device model, or after thecpufreq driver registers itself, the per-policy initialization functioncpufreq_driver.init is called if no cpufreq policy existed for the CPU.Note that the .init() and .exit() routines are called only once for thepolicy and not for each CPU managed by the policy. It takes astructcpufreq_policy*policy as argument. What to do now?

If necessary, activate the CPUfreq support on your CPU.

Then, the driver must fill in the following values:

policy->cpuinfo.min_freq _and_policy->cpuinfo.max_freqthe minimum and maximum frequency(in kHz) which is supported bythis CPU
policy->cpuinfo.transition_latencythe time it takes on this CPU toswitch between two frequencies innanoseconds (if appropriate, elsespecify CPUFREQ_ETERNAL)
policy->curThe current operating frequency ofthis CPU (if appropriate)
policy->min,policy->max,policy->policy and, if necessary,policy->governormust contain the “default policy” forthis CPU. A few moments later,cpufreq_driver.verify and eithercpufreq_driver.setpolicy orcpufreq_driver.target/target_index iscalled with these values.
policy->cpusUpdate this with the masks of the(online + offline) CPUs that do DVFSalong with this CPU (i.e. that shareclock/voltage rails with it).

For setting some of these values (cpuinfo.min[max]_freq, policy->min[max]), thefrequency table helpers might be helpful. See the section 2 for more informationon them.

1.3 verify

When the user decides a new policy (consisting of“policy,governor,min,max”) shall be set, this policy must be validatedso that incompatible values can be corrected. For verifying thesevalues cpufreq_verify_within_limits(structcpufreq_policy*policy,unsignedintmin_freq,unsignedintmax_freq) function might be helpful.See section 2 for details on frequency table helpers.

You need to make sure that at least one valid frequency (or operatingrange) is within policy->min and policy->max. If necessary, increasepolicy->max first, and only if this is no solution, decrease policy->min.

1.4 target or target_index or setpolicy or fast_switch?

Most cpufreq drivers or even most cpu frequency scaling algorithmsonly allow the CPU frequency to be set to predefined fixed values. Forthese, you use the ->target(), ->target_index() or ->fast_switch()callbacks.

Some cpufreq capable processors switch the frequency between certainlimits on their own. These shall use the ->setpolicy() callback.

1.5. target/target_index

The target_index call has two arguments:structcpufreq_policy*policy,andunsignedint index (into the exposed frequency table).

The CPUfreq driver must set the new frequency when called here. Theactual frequency must be determined by freq_table[index].frequency.

It should always restore to earlier frequency (i.e. policy->restore_freq) incase of errors, even if we switched to intermediate frequency earlier.

Deprecated

The target call has three arguments:structcpufreq_policy*policy,unsigned int target_frequency, unsigned int relation.

The CPUfreq driver must set the new frequency when called here. Theactual frequency must be determined using the following rules:

  • keep close to “target_freq”
  • policy->min <= new_freq <= policy->max (THIS MUST BE VALID!!!)
  • if relation==CPUFREQ_REL_L, try to select a new_freq higher than or equaltarget_freq. (“L for lowest, but no lower than”)
  • if relation==CPUFREQ_REL_H, try to select a new_freq lower than or equaltarget_freq. (“H for highest, but no higher than”)

Here again the frequency table helper might assist you - see section 2for details.

1.6. fast_switch

This function is used for frequency switching from scheduler’s context.Not all drivers are expected to implement it, as sleeping from withinthis callback isn’t allowed. This callback must be highly optimized todo switching as fast as possible.

This function has two arguments:structcpufreq_policy*policy andunsignedinttarget_frequency.

1.7 setpolicy

The setpolicy call only takes astructcpufreq_policy*policy asargument. You need to set the lower limit of the in-processor orin-chipset dynamic frequency switching to policy->min, the upper limitto policy->max, and -if supported- select a performance-orientedsetting when policy->policy is CPUFREQ_POLICY_PERFORMANCE, and apowersaving-oriented setting when CPUFREQ_POLICY_POWERSAVE. Also checkthe reference implementation in drivers/cpufreq/longrun.c

1.8 get_intermediate and target_intermediate

Only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset.

get_intermediate should return a stable intermediate frequency platform wants toswitch to, and target_intermediate() should set CPU to that frequency, beforejumping to the frequency corresponding to ‘index’. Core will take care ofsending notifications and driver doesn’t have to handle them intarget_intermediate() or target_index().

Drivers can return ‘0’ from get_intermediate() in case they don’t wish to switchto intermediate frequency for some target frequency. In that case core willdirectly call ->target_index().

NOTE: ->target_index() should restore to policy->restore_freq in case offailures as core would send notifications for that.

2. Frequency Table Helpers

As most cpufreq processors only allow for being set to a few specificfrequencies, a “frequency table” with some functions might assist insome work of the processor driver. Such a “frequency table” consists ofan array of struct cpufreq_frequency_table entries, with driver specificvalues in “driver_data”, the corresponding frequency in “frequency” andflags set. At the end of the table, you need to add acpufreq_frequency_table entry with frequency set to CPUFREQ_TABLE_END.And if you want to skip one entry in the table, set the frequency toCPUFREQ_ENTRY_INVALID. The entries don’t need to be in sorted in anyparticular order, but if they are cpufreq core will do DVFS a bitquickly for them as search for best match is faster.

The cpufreq table is verified automatically by the core if the policy contains avalid pointer in its policy->freq_table field.

cpufreq_frequency_table_verify() assures that at least one validfrequency is within policy->min and policy->max, and all other criteriaare met. This is helpful for the ->verify call.

cpufreq_frequency_table_target() is the corresponding frequency tablehelper for the ->target stage. Just pass the values to this function,and this function returns the of the frequency table entry whichcontains the frequency the CPU shall be set to.

The following macros can be used as iterators over cpufreq_frequency_table:

cpufreq_for_each_entry(pos, table) - iterates over all entries of frequencytable.

cpufreq_for_each_valid_entry(pos, table) - iterates over all entries,excluding CPUFREQ_ENTRY_INVALID frequencies.Use arguments “pos” - acpufreq_frequency_table* as a loop cursor and“table” - thecpufreq_frequency_table* you want to iterate over.

For example:

struct cpufreq_frequency_table *pos, *driver_freq_table;cpufreq_for_each_entry(pos, driver_freq_table) {        /* Do something with pos */        pos->frequency = ...}

If you need to work with the position of pos within driver_freq_table,do not subtract the pointers, as it is quite costly. Instead, use themacros cpufreq_for_each_entry_idx() and cpufreq_for_each_valid_entry_idx().