RE: [PATCH v4 0/6] ACPI: Add _OST support for ACPI hotplug

[Toshi Kani <toshi.kani@hp.com>]

On Wed, 2012-05-30 at 02:56 +0000, Moore, Robert wrote:
> It does not make any difference. Essentially, a get_handle is performed by evaluate_object anyway.
> 

Thanks Robert for the confirmation!
-Toshi


> >-----Original Message-----
> >From: Toshi Kani [mailto:toshi.kani@hp.com]
> >Sent: Tuesday, May 29, 2012 4:43 PM
> >To: shuahkhan@gmail.com
> >Cc: Moore, Robert; lenb@kernel.org; linux-acpi@vger.kernel.org;
> >bhelgaas@google.com; liuj97@gmail.com; andi@firstfloor.org; linux-
> >kernel@vger.kernel.org
> >Subject: RE: [PATCH v4 0/6] ACPI: Add _OST support for ACPI hotplug
> >
> >On Tue, 2012-05-29 at 16:44 -0600, Shuah Khan wrote:
> >> On Tue, 2012-05-29 at 22:27 +0000, Moore, Robert wrote:
> >> > > > 2. Calling acpi_get_handle() on _OST prior to executing the method.
> >> > > > This will ensure that this method only gets run if it is present
> >> > > under
> >> > > > the device in question. Coupled with what is already outlined in #1
> >> > > > above, now _OST gets executed only when it is defined under the
> >> > > device object.
> >> > > > Example case in the existing code, please see
> >> > > acpi_processor_ppc_ost()
> >> > > > implementation.
> >> > >
> >> > > Yes, I did look at acpi_processor_ppc_ost() when I implemented the
> >> > > function.  I believe calling acpi_get_handle() is redundant since
> >> > > acpi_ns_get_node() is called within acpi_evaluate_object() as well.
> >> > > acpi_evaluate_object() simply returns with AE_NOT_FOUND when _OST
> >> > > method does not exist.
> >> > >
> >> >
> >> > This is correct. If _OST does not exist, AE_NOT_FOUND will be returned
> >from evaluate_object.
> >>
> >> Yes that is correct from the ACPI Spec and implementation point of view.
> >> My thinking is that a call to acpi_get_handle() might not penalize the
> >> OS as much as acpi_evaluate_object() would on systems that don't
> >> actually implement _OST. In other words, acpi_get_handle() might not go
> >> as deep as acpi_evaluate_object() would go into the ACPI layer, hence
> >> might be a safer measure on platforms that don't actually implement this
> >> optional method under all devices included in this patch set.
> >>
> >
> >I do not think we need to worry about it.  The code difference is not
> >that much, and this _OST path is limited to ACPI hotplug operations,
> >which are infrequent events.  Automatic workload balancing can make
> >frequent use of the operations, but is not frequent enough to make any
> >difference here.  I think simpler code works fine.
> >
> >Thanks,
> >-Toshi
>