[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20160407214216.GH1990@wotan.suse.de>
Date: Thu, 7 Apr 2016 23:42:16 +0200
From: "Luis R. Rodriguez" <mcgrof@...nel.org>
To: David Vrabel <david.vrabel@...rix.com>
Cc: "Luis R. Rodriguez" <mcgrof@...nel.org>, bp@...en8.de,
hpa@...or.com, tglx@...utronix.de, mingo@...hat.com,
rusty@...tcorp.com.au, xen-devel@...ts.xensource.com,
matt@...eblueprint.co.uk, andrew.cooper3@...rix.com,
x86@...nel.org, linux-kernel@...r.kernel.org, luto@...capital.net,
jlee@...e.com, lguest@...ts.ozlabs.org, glin@...e.com,
boris.ostrovsky@...cle.com, andriy.shevchenko@...ux.intel.com,
Josh Triplett <josh@...htriplett.org>, robert.moore@...el.com,
lv.zheng@...el.com, rafael.j.wysocki@...el.com, toshi.kani@...com,
linux-acpi@...r.kernel.org, tiwai@...e.de
Subject: Re: [Xen-devel] [PATCH v4 11/14] pnpbios: replace paravirt_enabled()
check with legacy device check
On Thu, Apr 07, 2016 at 10:46:11AM +0100, David Vrabel wrote:
> On 07/04/16 01:06, Luis R. Rodriguez wrote:
> > Since we are removing paravirt_enabled() replace it with a
> > logical equivalent. Even though PNPBIOS is x86 specific we
> > add an arch-specific type call, which can be implemented by
> > any architecture to show how other legacy attribute devices
> > can later be also checked for with other ACPI legacy attribute
> > flags.
> >
> > This implicates the first ACPI 5.2.9.3 IA-PC Boot Architecture
> > ACPI_FADT_LEGACY_DEVICES flag device, and shows how to add more.
> [...]
> > +struct x86_legacy_devices {
> > + int pnpbios;
> > +};
>
> It's not clear why pnpbios needs a new structure
I'm glad you asked. Dealing with placing pnpbios quirk in a more useful generic
fashion was perhaps the most difficult challenge in this series. As I reviewed
possibilities to remove paravirt_enabled() the best prospect I found was to see
if Xen could instead use ACPI 5.2.9.3 IA-PC Boot Architecture flags to annotate
some quirks. It turns out that it is possible, but there are only so many flags,
and also, we didn't want to have a solution that incurred respective upstream
Xen hypervisor change, that would be silly.
To make this quirk more useful then this folds the pnpbios quirk as a sub
quirk under the more borad ACPI_FADT_LEGACY_DEVICES ACPI flag. What this
does, as can be seen by also looking at the next patch, "x86, ACPI: parse
ACPI_FADT_LEGACY_DEVICES" is it explicitly folds pnpbios as one of the
ACPI_FADT_LEGACY_DEVICES devices, but also paves the way for further known
main legacy components to added to the list.
> and why this structure of devices does not have the bit for the rtc device.
That's because ACPI has its own dedicated flag for it, so there already
is a one-to-one mapping available. All we needed to do to replace the
RTC hack was to provide a mechanism to unify both the paravirt RTC hack
with the ACPI RTC flag.
Luis
Powered by blists - more mailing lists