sys/power.h: `enum power_sstate_transition`

[obiwac]

Summary

Turn POWER_SLEEP_STATE_* defines into enum power_sstate_transition.

Maybe I missed some places where just an int is being used for the sleep state, gonna look at this deeper later.

[obiwac]

sys/sys/power.h

55–57

Is there a particular reason we'd want to explicitly set these values? Is it preferred style in FreeBSD to explicitly set these when an enum is used in an interface or something?

[markj]

sys/dev/acpi_support/acpi_ibm.c

1454

Does this perhaps suggest that it isn't very useful in general to distinguish between SSTATE_TRANSITION_* and STYPE_*?

sys/sys/power.h

55–57

That's not necessary for internal kernel interfaces.

[obiwac]

sys/dev/acpi_support/acpi_ibm.c

1454

can you elaborate on in what way this suggests the distinction is not useful? Other places need to know if the suspend type is s2idle or s2mem, I just added this comment to note that the use of the name "IBM_EVENT_SUSPEND_TO_RAM" here doesn't necessarily mean s2mem, but could also be s2idle

[markj]

Sorry for the delay. I think this is ok.

sys/dev/acpi_support/acpi_ibm.c

1454

I think I was wondering what things would look like if we had a suspend-to-disk implementation. Presumably you'd end up with both POWER_SSTATE_TRANSITION_SUSPEND_TO_DISK and POWER_SSTATE_TRANSITION_SUSPEND?

sys/sys/power.h

54

This should be removed before committing.

[obiwac]

sys/dev/acpi_support/acpi_ibm.c

1454

you'd use POWER_SSTATE_TRANSITION_HIBERNATE for suspend-to-disk. I agree this is not the clearest thing ever considering "suspend" could also be understood to mean "suspend to disk", but these are the existing names for POWER_SSTATE_* anyway and I'm not really sure of a better nomenclature here.