Re: [PATCH/RFC 3/6] drivers: firmware: psci: Implement shallow suspend mode

[Geert Uytterhoeven]

Hi Sudeep,

On Tue, Feb 21, 2017 at 6:22 PM, Sudeep Holla <sudeep.holla@xxxxxxx> wrote:
> On 21/02/17 16:32, Geert Uytterhoeven wrote:
>> On Tue, Feb 21, 2017 at 12:14 PM, Sudeep Holla <sudeep.holla@xxxxxxx> wrote:
>>> On 21/02/17 11:07, Pavel Machek wrote:
>>>>> Enable support for "shallow" suspend mode, also known as "Standby" or
>>>>> "Power-On Suspend".
>>>>>
>>>>> As secondary CPU cores are taken offline, "shallow" suspend mode saves
>>>>> slightly more power than "s2idle", but less than "deep" suspend mode.
>>>>> However, unlike "deep" suspend mode, "shallow" suspend mode can be used
>>>>> regardless of the presence of support for PSCI_SYSTEM_SUSPEND, which is
>>>>> an optional API in PSCI v1.0.
>>>>
>>>> If system supports "shallow" suspend, why does not PSCI implement it?
>>>
>>> Yes it can, and IIUC it already does on this platform with CPU_SUSPEND.
>>> All it now needs is just to use existing "freeze" suspend mode in Linux.
>>
>> How can Linux know if using "deep" suspend will allow to wake-up the system
>> according to configured wake-up sources, or not?
>
> I am not sure if we have such selective configuration of wakeup source
> implemented in Linux.
>
> ACPI specification has some provisions where each device can state if it
> can specify device state in each system sleeping state that can wake the
> system.
>
> DT has no mechanism today to express this relations. I had brought up
> this discussion in plumbers(2015). Refer slide 7 in [0]
>
> And the way you are trying to do that is not correct IMO especially
> making it just PSCI specific.
>
>> Note that "it will not, ever" is an accepted answer.
>
> IIUC, it's not implemented today. I can't talk about future ;), but your

Good, so there's no need for the DT property, and drivers/firmware/psci.c
should aways call do_cpu_idle() instead of PSCI SYSTEM_SUSPEND if any
other wake-up sources are configured?

That follows the principle of least surprise: it doesn't leave the user with
a system that won't wake up the way he configured it to wake up.

> proposal is horrible hack.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds