On 2015/3/17 13:59, Jon Masters wrote:

On 03/17/2015 12:10 AM, Hanjun Guo wrote:

quoted

On 2015/3/17 11:23, Rafael J. Wysocki wrote:

quoted

On Tuesday, March 17, 2015 10:36:47 AM Hanjun Guo wrote:
Well, almost.  There is one problem with that, becuase sleep.c contains code
outside of the ACPI_SLEEP-dependent blocks.  That code is used for powering
off ACPI platforms.

I guess you don't want that code on ARM too, right?

Yes, you are right.

quoted

Perhaps we can use ACPI_REDUCED_HARDWARE_ONLY for that?  ARM64 will be the

Sorry, I can't fully understand your intention here, could you please
explain it more?

Let me guess a little bit. Do you mean use ACPI_REDUCED_HARDWARE_ONLY for
powering off ACPI platforms? if so, I guess it's not a good idea, ACPI spec
only says that S4BIOS is not supported on HW-reduced ACPI platforms, S5
has no such limitation, if I miss something here, please let me know.

If helpful to the discussion, current SBBR (Server Base Boot
Requirements[0]) design guidance is that for power off itself, we will
prefer calling an EFI Runtime Service (that will preferentially call an
PSCI - ARM Power State Coordination Interface - Secure Monitor Call (SMC
- think SMI-like) internally to perform the shutdown/reboot) for the
action of powering off or resetting 64-bit ARM SBBR platforms.

Agreed, PSCI is the prefer method for power off on ARM64 I think.

Therefore if the alternative of an ACPI-based power off solution were
not initially supported, I don't think it would have much practical
impact, and it could be addressed after the initial support merged.

I agree. Actually we already removed ACPI power off code for ARM64
in v9 and v10 regardless the ACPI sepc statement about S5, I just want to
confirm with Rafael that how to use ACPI_REDUCED_HARDWARE_ONLY properly
to do the same thing as we do in v10 (patch - ACPI / sleep: Introduce
CONFIG_ACPI_GENERIC_SLEEP), or if we have some other way to do that :)

Thanks
Hanjun