Re: [PATCH v4 0/5] Xen/ACPI: support sleep state entering on hardware reduced systems

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Thursday, June 27, 2013 11:01:58 AM Ben Guthro wrote:
> In version 3.4 acpi_os_prepare_sleep() got introduced in parallel with
> reduced hardware sleep support, and the two changes didn't get
> synchronized: The new code doesn't call the hook function (if so
> requested). Fix this, requiring a boolean parameter to be added to the
> hook function to distinguish "extended" from "legacy" sleep.
> 
> This requires adjusting TXT, but the adjustments only go as far as
> failing the extended mode call (since, looking at the TXT interface,
> there doesn't even appear to be precautions to deal with that
> alternative interface).
> 
> The hypervisor change underlying this is commit 62d1a69 ("ACPI: support
> v5 (reduced HW) sleep interface") on the master branch of
> git://xenbits.xen.org/xen.git.
> 
> Signed-off-by: Ben Guthro <benjamin.guthro@citrix.com>
> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> Cc: Richard L Maliszewski <richard.l.maliszewski@intel.com>
> Cc: Gang Wei <gang.wei@intel.com>
> Cc: Shane Wang <shane.wang@intel.com>
> Cc: Bob Moore <robert.moore@intel.com>
> Cc: Rafaell J. Wysocki <rjw@sisk.pl>
> Cc: linux-acpi@vger.kernel.org
> Cc: tboot-devel@lists.sourceforge.net 
> 
> v2: Extend description to include reference to hypervisor side change
> v3: Split into multiple patches, separating subsystems
>     Remove bool parameters, in favor of u8
> v4: Remove linux/acpi.h dependencies
>     Further patch split to break out acpica from OSL
>     More bool vs u8 fixes
> 
> Ben Guthro (5):
>   acpi: Remove need to include linux/acpi.h in common acpica code
>   acpi: Call acpi_os_prepare_sleep hook in reduced hardware sleep path
>   acpi: Adjust linux acpi OS functions to new extended parameter
>   x86/tboot: Fail extended mode reduced hardware sleep
>   xen/acpi: notify xen when reduced hardware sleep is available

The ongoing discussion means to me that the ACPICA maintainers don't want
acpi_os_prepare_sleep() and quite frankly I understand them, because ACPICA
is about implementing the spec and not about things beyond it.

This means that patch [1/5] goes away.

That said, at the same time we need to address the problem at hand, which
is to make Xen work with the reduced HW sleep.

For that, I don't honestly think that modifying acpi_os_prepare_sleep() the
way the patchset is doing it is appropriate and the change of the meaning of
the arguments is simply disgusting.

To me, it would be much cleaner to add acpi_os_prepare_extended_sleep()
specifically to be called by acpi_hw_extended_sleep() and make tboot and Xen
use that.

This way or another, we'll need to live with one more divergence between the
upstream ACPICA and the Linux ACPICA code because of that, but that'd be just
a few added lines in acpi_hw_extended_sleep(), so I suppose it wouldn't be
such a big deal.

Thanks,
Rafael


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.