Re: [PATCH v5] ACPI: PM: Introduce CONFIG_ACPI_S2IDLE for platform-independent S2Idle support

[Riwen Lu]

Hi Rafael,

Gentle ping on this thread.

Having reworked the patch based on your feedback, I believe the main
problem with v5 was the new user-visible Kconfig option together with
the conditional select: an s2idle toggle is not something users or
distros should ever need to set.  Before posting the next revision,
I'd like to align with you on one of two directions:

(a) Keep the code split into drivers/acpi/s2idle.c, but turn
    ACPI_S2IDLE into a hidden symbol:

        config ACPI_S2IDLE
                bool
                depends on SUSPEND
                default y

    following the existing ACPI_LPIT pattern.  No prompt, no select
    and no arch changes: arm64 gets s2idle automatically.

(b) Don't split sleep.c at all.  Have arm64 select
    ACPI_SYSTEM_POWER_STATES_SUPPORT and add two tiny stubs in
    arch/arm64 for the S3-only references in sleep.c.
    acpi_sleep_state_supported() already rejects all firmware
    S-states on HW-reduced platforms, so only s2idle becomes
    functional.  This is 16 lines in total with no code motion, but
    the symbol name becomes loose and /proc/acpi/* would appear on
    arm64.

I slightly prefer (a).  Unless you prefer (b) or have a different
direction in mind, I'll post (a) as v6 next week.

Thanks!

Riwen