Re: [PATCH v6 4/5] arm64: add support for ACPI Low Power Idle(LPI)

[Sudeep Holla <sudeep.holla@arm.com>]

On 27/06/16 17:29, Daniel Lezcano wrote:
> On 06/22/2016 04:17 PM, Lorenzo Pieralisi wrote:
>> Hi Sudeep,
>>
>> On Tue, Jun 14, 2016 at 03:48:38PM +0100, Sudeep Holla wrote:
>>> This patch adds appropriate callbacks to support ACPI Low Power Idle
>>> (LPI) on ARM64.
>>>
>>> Now that arm_enter_idle_state is exactly same in both generic ARM{32,64}
>>> CPUIdle driver and ARM64 backend for ACPI processor idle driver, we can
>>> unify it and move to cpuidle-arm.h header.
>>>
>>> Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
>>> Cc: Mark Rutland <mark.rutland@arm.com>
>>> Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
>>> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
>>> Cc: linux-arm-kernel@lists.infradead.org
>>> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
>>> ---
>>>   arch/arm64/kernel/cpuidle.c   | 17 +++++++++++++
>>>   drivers/cpuidle/cpuidle-arm.c | 23 ++----------------
>>>   drivers/firmware/psci.c       | 56
>>> +++++++++++++++++++++++++++++++++++++++++++
>>>   include/linux/cpuidle-arm.h   | 30 +++++++++++++++++++++++
>>>   4 files changed, 105 insertions(+), 21 deletions(-)
>>>   create mode 100644 include/linux/cpuidle-arm.h
>>
>> This patch seems fine by me, it would be good if Daniel can have
>> a look too.
>>
>> Some minor comments below.
>>
>> [...]
>>
>>> diff --git a/drivers/firmware/psci.c b/drivers/firmware/psci.c
>>> index 03e04582791c..c6caa863d156 100644
>>> --- a/drivers/firmware/psci.c
>>> +++ b/drivers/firmware/psci.c
>>> @@ -13,6 +13,7 @@
>>>
>>>   #define pr_fmt(fmt) "psci: " fmt
>>>
>>> +#include <linux/acpi.h>
>>>   #include <linux/arm-smccc.h>
>>>   #include <linux/cpuidle.h>
>>>   #include <linux/errno.h>
>>> @@ -310,11 +311,66 @@ static int psci_dt_cpu_init_idle(struct
>>> device_node *cpu_node, int cpu)
>>>       return ret;
>>>   }
>>>
>>> +#ifdef CONFIG_ACPI
>>> +#include <acpi/processor.h>
>>> +
>>> +static int __maybe_unused psci_acpi_cpu_init_idle(unsigned int cpu)
>>> +{
>>> +    int i, count;
>>> +    u32 *psci_states;
>>> +    struct acpi_processor *pr;
>>> +    struct acpi_lpi_state *lpi;
>>> +
>>> +    pr = per_cpu(processors, cpu);
>>> +    if (unlikely(!pr || !pr->flags.has_lpi))
>>> +        return -EINVAL;
>>> +
>>> +    /*
>>> +     * If the PSCI cpu_suspend function hook has not been initialized
>>> +     * idle states must not be enabled, so bail out
>>> +     */
>>> +    if (!psci_ops.cpu_suspend)
>>> +        return -EOPNOTSUPP;
>>> +
>>> +    count = pr->power.count - 1;
>>> +    if (count <= 0)
>>> +        return -ENODEV;
>>> +
>>> +    psci_states = kcalloc(count, sizeof(*psci_states), GFP_KERNEL);
>>> +    if (!psci_states)
>>> +        return -ENOMEM;
>>> +
>>> +    for (i = 0; i < count; i++) {
>>> +        u32 state;
>>> +
>>> +        lpi = &pr->power.lpi_states[i + 1];
>>> +        state = lpi->address & 0xFFFFFFFF;
>
> Why mask 'address' if 'state' is u32 ?
>

Agreed, I overlooked it.

>>> +        if (!psci_power_state_is_valid(state)) {
>>> +            pr_warn("Invalid PSCI power state %#x\n", state);
>>> +            kfree(psci_states);
>>> +            return -EINVAL;
>>> +        }
>>> +        psci_states[i] = state;
>>> +    }
>>> +    /* Idle states parsed correctly, initialize per-cpu pointer */
>>> +    per_cpu(psci_power_state, cpu) = psci_states;
>>> +    return 0;
>
> The ACPI and the PSCI code are not self contained here.
>
> It would be nice to move this function to the ACPI code.
>
>>> +}
>>> +#else
>>> +static int __maybe_unused psci_acpi_cpu_init_idle(unsigned int cpu)
>>> +{
>>> +    return -EINVAL;
>>> +}
>>> +#endif
>>> +
>>>   int psci_cpu_init_idle(unsigned int cpu)
>>>   {
>>>       struct device_node *cpu_node;
>>>       int ret;
>>>
>>> +    if (!acpi_disabled)
>>> +        return psci_acpi_cpu_init_idle(cpu);
>>> +
>
> acpi_disabled - acpi_disabled - acpi_disabled everywhere :/
>
> The enable-method approach is not straightforward and now it is polluted
> by acpi-disabled.
>
> So IIUC,
>
> smp_init_cpus (contains acpi_disabled)
>    smp_cpu_setup
>      cpu_read_ops
>        cpu_read_enable_method (contains acpi_disabled)
>          acpi_get_enable_method (returns 'psci' after checking
> psci_is_present)
>
> Then psci_cpu_init_idle is called... and check again acpi_disabled.
>
> IMO, the circumlocution with the psci vs acpi vs acpi_disabled is
> getting unnecessary too complex, is prone to error and will lead to
> unmaintainable code very soon.
>
> I suggest to sort out encapsulation and self-contained code before
> adding more feature in this area.
>

I understand your concern but I have no idea on how to clean up. Lorenzo
asked to factor our common code between psci_{dt,acpi}_cpu_init_idle and
I think you might not like the refactoring[1]. I didn't want to change
cpuidle_ops and hence psci_dt_cpu_init_idle parameters. I will see if
changing that simplifies things.

>>>       cpu_node = of_get_cpu_node(cpu, NULL);
>>>       if (!cpu_node)
>>>           return -ENODEV;
>>> diff --git a/include/linux/cpuidle-arm.h b/include/linux/cpuidle-arm.h
>>> new file mode 100644
>>> index 000000000000..b99bcb3f43dd
>>> --- /dev/null
>>> +++ b/include/linux/cpuidle-arm.h
>>
>> arm-cpuidle.h for consistency with other (ARM) include/linux files ?
>>
>>> @@ -0,0 +1,30 @@
>>> +#include <linux/cpu_pm.h>
>>> +
>>> +#include <asm/cpuidle.h>
>>> +
>>> +/*
>>> + * arm_enter_idle_state - Programs CPU to enter the specified state
>>> + */
>>> +static int arm_generic_enter_idle_state(int idx)
>>> +{
>>> +    int ret;
>>> +
>>> +    if (!idx) {
>>> +        cpu_do_idle();
>>> +        return idx;
>>> +    }
>>> +
>>> +    ret = cpu_pm_enter();
>>> +    if (!ret) {
>>> +        /*
>>> +         * Pass idle state index to cpu_suspend which in turn will
>>> +         * call the CPU ops suspend protocol with idle index as a
>>> +         * parameter.
>>> +         */
>>> +        ret = arm_cpuidle_suspend(idx);
>>> +
>>> +        cpu_pm_exit();
>>> +    }
>>> +
>>> +    return ret ? -1 : idx;
>>> +}
>>
>> Either you do this, or we have to add it somehow somewhere in
>> drivers/cpuidle to avoid duplicating it.
>>
>> @Daniel: do you have an opinion on this please ?
>
> I don't like the idea to add an ARM arch specific header in
> include/linux. I thought this directory was supposed to contain as much
> as possible arch agnostic headers.
>

While I agree, but what we have is ARM specific code here and calling it
generic might not make it any usable on other architecture unless they 
have the same semantics. I am fine if you and Rafael are OK with that.

> May be the name can be changed to something more generic:
>
> eg.
>
> int cpuidle_generic_enter(int idx);
>
> and then add an option:
>
> HAVE_CPUIDLE_GENERIC_ENTER
>
> , then in the generic header:
>
> #ifdef HAVE_CPUIDLE_GENERIC_ENTER
> int cpuidle_generic_enter(int idx);
> #endif
>
> , change the function name in cpuidle-arm .c
>
> and finally add in the ARM and ARM64 Kconfig's option
> HAVE_CPUIDLE_GENERIC_ENTER.
>
>
>    -- Daniel
>

[1] 
http://git.kernel.org/cgit/linux/kernel/git/sudeep.holla/linux.git/commit/?h=for_review/arm64_lpi&id=9b516ad4442b4168e962ba4ca87bd568d605053b
-- 
Regards,
Sudeep