Re: [PATCH v5 2/5] ACPI / processor_idle: Add support for Low Power Idle(LPI) states

["Prakash, Prashanth" <pprakash@codeaurora.org>]

On 5/18/2016 11:37 AM, Sudeep Holla wrote:
>
>
> On 17/05/16 18:46, Prakash, Prashanth wrote:
>> Hi Sudeep,
>>
>> On 5/11/2016 9:37 AM, Sudeep Holla wrote:
>>> +
>>> +static int acpi_processor_get_lpi_info(struct acpi_processor *pr)
>>> +{
>>> +    int ret, i;
>>> +    struct acpi_lpi_states_array *info;
>>> +    struct acpi_device *d = NULL;
>>> +    acpi_handle handle = pr->handle, pr_ahandle;
>>> +    acpi_status status;
>>> +
>>> +    if (!osc_pc_lpi_support_confirmed)
>>> +        return -EOPNOTSUPP;
>>> +
>>> +    max_leaf_depth = 0;
>>> +    if (!acpi_has_method(handle, "_LPI"))
>>> +        return -EINVAL;
>>> +    flat_state_cnt = 0;
>>> +
>>> +    while (ACPI_SUCCESS(status = acpi_get_parent(handle, &pr_ahandle))) {
>>> +        if (!acpi_has_method(handle, "_LPI"))
>>> +            continue;
>>> +
>>> +        acpi_bus_get_device(handle, &d);
>>> +        if (!strcmp(acpi_device_hid(d), ACPI_PROCESSOR_CONTAINER_HID))
>>> +            break;
>>> +
>>> +        max_leaf_depth++;
>>> +        handle = pr_ahandle;
>>> +    }
>>> +
>> In the above loop, we break when we find a device with HID ==
>> ACPI_PROCESSOR_CONTAINER_HID. Shouldn't we continue to parse as long as the
>> parent HID == ACPI_PROCESSOR_CONTAINER_HID? This is required to make sure we
>> parse states in levels higher than cluster level in processor hierarchy.
>>
>
> Yes, thanks for pointing that out. With just clusters in _LPI on my dev
> board, I missed it.
>
Same reason, I failed to notice it all this time :)
>> Also, I think it might be safe to break out of the loop if we didn't find
>> _LPI package, instead of continuing. Given  the presence of LPI entry:
>> "Enabled Parent State", I can't think of a non-ambiguous scenario where we
>> might find LPI packages in state N and N+2, but not in N+1, as we will not
>> be able to figure out which state in N enables which states in N+2.
>> Thoughts?
>
> Though I admit I haven't thought in detail on how to deal with the
> asymmetric topology, but that was the reason why I continue instead of
> breaking.
>
> Excerpts from the spec: "... This example is symmetric but that is not a
> requirement. For example, a system may contain a different number of
> processors in different containers or an asymmetric hierarchy where one
> side of the topology tree is deeper than another...."
>
If it addresses asymmetric topology, sure we can keep as it doesn't impact other
scenarios. Also, we need to set handle=pr_ahandle prior to the continue statement.


Thanks,
Prashanth