Re: [PATCH][RFC,v4] ACPI / PM: Introduce efi poweroff for HW-full platforms without _S5

[Matt Fleming]

On Wed, 16 Mar, at 05:59:29AM, Chen, Yu C wrote:
> Hi Matt,
> 
> > -----Original Message-----
> > From: Matt Fleming [mailto:matt@xxxxxxxxxxxxxxxxxxx]
> > Sent: Tuesday, March 15, 2016 4:01 AM
> > To: Chen, Yu C
> > Cc: linux-acpi@xxxxxxxxxxxxxxx; linux-pm@xxxxxxxxxxxxxxx; Rafael J. Wysocki;
> > Len Brown; Thomas Gleixner; Ingo Molnar; H. Peter Anvin; Zhang, Rui; linux-
> > efi@xxxxxxxxxxxxxxx; x86@xxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Ard
> > Biesheuvel; Mark Salter
> > Subject: Re: [PATCH][RFC,v4] ACPI / PM: Introduce efi poweroff for HW-full
> > platforms without _S5
> > 
> > On Fri, 11 Mar, at 04:33:46PM, Chen, Yu C wrote:
> > >
> > > There is  a future Base-IA platform, we are planning to skip
> > > implementing the SLP_TYP register and the S5 object.  (already there
> > > will be no S3 and no S4)
> > 
> > Cool. This is really valuable information that should go into the commit
> > message.
> > 
> > Because if this is the rationale for the change, I don't see why we'd need to
> > provide the default stuff. Instead we should just enforce EFI reboot, and
> > only add the pm_poweroff_default hook if there is an explicit user in the
> > future, IMO.
> 
> Do you mean the patch v3 make sense
> https://patchwork.kernel.org/patch/8514751/
> and we should use efi power off as our first choice, if there is no _S5  available(no acpi_power_off),
> even there is a customized  poweroff(driver provided, eg)? 

Unless someone can point to a platform driver that is in the upstream
kernel where this is actually a problem, the answer is: yes.
 
For that matter, unless someone can do the same for pm_power_off
overriding efi_reboot() (which on x86 would only happen for ACPI
HW-reduced platforms), I would be much prefer the original patch,
where you had,

bool efi_poweroff_reqired(voi)
{
	return acpi_gbl_reduced_hardware || acpi_no_s5;
}

since you've already explained that this change won't break legacy
platforms that are missing _S5 (if any even exist in the wild).