On 18/02/2022 09:10, Edwin Chiu 邱垂峰 wrote:
>
>
>
>> -----Original Message-----
>> From: Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxxxxx>
>> Sent: Monday, February 14, 2022 2:45 PM
>> To: Edwin Chiu 邱垂峰 <edwin.chiu@xxxxxxxxxxx>; Edwin Chiu <edwinchiu0505tw@xxxxxxxxx>;
>> robh+dt@xxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; rafael@xxxxxxxxxx;
>> daniel.lezcano@xxxxxxxxxx; linux-pm@xxxxxxxxxxxxxxx
>> Subject: Re: [PATCH v4] cpuidle: sunplus: Create cpuidle driver for sunplus sp7021
>>
>> On 14/02/2022 03:55, Edwin Chiu 邱垂峰 wrote:
>>> Hi Krzysztof:
>>>
>>> Please see below answer.
>>>
>>>>> +static struct cpuidle_driver sp7021_idle_driver = {
>>>>> + .name = "sp7021_idle",
>>>>> + .owner = THIS_MODULE,
>>>>> + /*
>>>>> + * State at index 0 is standby wfi and considered standard
>>>>> + * on all ARM platforms. If in some platforms simple wfi
>>>>> + * can't be used as "state 0", DT bindings must be implemented
>>>>> + * to work around this issue and allow installing a special
>>>>> + * handler for idle state index 0.
>>>>> + */
>>>>> + .states[0] = {
>>>>> + .enter = sp7021_enter_idle_state,
>>>>> + .exit_latency = 1,
>>>>> + .target_residency = 1,
>>>>> + .power_usage = UINT_MAX,
>>>>> + .name = "WFI",
>>>>> + .desc = "ARM WFI",
>>>>
>>>> I have impression that there is no point in having custom driver with WFI...
>>>>
>>>> Still the main question from Daniel and Sudeep stays: why do you need
>>>> this? You copied exactly cpuildle-arm driver, there is nothing
>>>> different here. At least I could not spot differences. Maybe except that you use cpu_v7_do_idle
>> explicitly.
>>>>
>>>> Unfortunately I cannot understand the explanation here:
>>>> https://lore.kernel.org/all/0812c44f777d4026b79df2e3698294be@sphcmbx0
>>>> 2.sunplus.com.tw/ Why exactly cpuidle-arm does not work in your case?
>>>>
>>> Edwin=> I mean cpuidle-arm driver can't directly use with no modified.
>>> If someone want to use cpuidle-arm driver, below modification seems necessary.
>>>
>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>> Static int sp7021_cpuidle_suspend_enter(unsigned long index) {~}
>>> Static int __init sp7021_cpuidle_init(struct device_node *cpu_node, int cpu) {~}
>>> Static const struct cpuidle_ops sc_smp_ops __initconst = {
>>> .suspend = sp7021_cpuidle_suspend_enter,
>>> .init = sp7021_cpuidle_init,
>>> };
>>> CPUIDLE_METHOD_OF_DECLARE(sc_smp, "sunplus,sc-smp", &sc_smp_ops); //declare
>> enable method
>>>
>>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>> ~~~~~~~~~~
>>>
>>> But change cpuilde-arm.c for sunplus driver seems not reasonable.
>>> That is why I want to submit cpuidle-sunplus.c
>>> Althought sunplus cpuidle only come in WFI, but it can complete the cpuidle framework.
>>
>> I don't think it is correct. You can use cpuidle-arm, because it is being always initialized with
>> device_initcall(). You either use appropriate compatible in DT or add your compatible to cpuidle-arm.
>>
>> Even if this did not work, then the solution is to use common parts, not to duplicate entire driver.
>> Duplicating is not acceptable.
>>
>> Best regards,
>> Krzysztof
>
>
> I do used compatible = "arm,idle-state" in DT and enable generic arm cpuidle driver in menuconfig.
> But there have mount driver fail message due to no cpuidle_ops functions.
> That is why I need added patch code to complete cpuidle driver.
> According your comment, I will try to use common parts and hook some custom code, later.
I think I understood the motivation behind your driver. The cpuidle-arm
requires enable-method property which usually uses for ARMv7
CPU_METHOD_OF_DECLARE(). You don't have CPU_METHOD_OF_DECLARE().
Now the question: why can't you define CPU_METHOD_OF_DECLARE() just like
many other ARMv7 platforms do?
Best regards,
Krzysztof
