On Thu, 2014-01-30 at 17:13 +0000, Marc Zyngier wrote:

Hi Ian,

On 30/01/14 16:11, Ian Campbell wrote:

quoted

mach-virt has existed for a while but it is not written down what it actually
consists of. Although it seems a bit unusual to document a binding for an
entire platform since mach-virt is entirely virtual it is helpful to have
something to refer to in the absence of a single concrete implementation.

I've done my best to capture the requirements based on the git log and my
memory/understanding.

While here remove the xenvm dts example, the Xen tools will now build a
suitable mach-virt compatible dts when launching the guest.

Signed-off-by: Ian Campbell <redacted>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: Pawel Moll <redacted>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Kumar Gala <redacted>
Cc: Olof Johansson <redacted>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Marc Zyngier <redacted>
Cc: Will Deacon <redacted>
Cc: Stefano Stabellini <redacted>
Cc: devicetree at vger.kernel.org
Cc: linux-kernel at vger.kernel.org
Cc: linux-arm-kernel at lists.infradead.org
---
I'm not sure which tree this sort of thing should go though, sorry for the
huge Cc.
---
 .../devicetree/bindings/arm/mach-virt.txt          |   32 ++++++++
 arch/arm/boot/dts/xenvm-4.2.dts                    |   81 --------------------
 2 files changed, 32 insertions(+), 81 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/arm/mach-virt.txt
 delete mode 100644 arch/arm/boot/dts/xenvm-4.2.dts

diff --git a/Documentation/devicetree/bindings/arm/mach-virt.txt b/Documentation/devicetree/bindings/arm/mach-virt.txt
new file mode 100644
index 0000000..562bcda
--- /dev/null
+++ b/Documentation/devicetree/bindings/arm/mach-virt.txt

@@ -0,0 +1,32 @@
+* Mach-virt "Dummy Virtual Machine" platform
+
+"mach-virt" is the smallest, dumbest platform possible, to be used as
+a guest for Xen, KVM and other hypervisors. It has no
+properties/functionality of its own and is driven entirely by device
+tree.
+
+This document defines the requirements for such a platform.
+
+* Required properties:
+
+- compatible: should be one of:
+	"linux,dummy-virt"
+	"xen,xenvm"
+
+In addition to the standard nodes (chosen, cpus, memory etc) the
+platform is required to provide certain other basic functionality
+which must be described in the device tree:
+
+    The platform must provide an ARM Generic Interrupt Controller
+    (GIC), defined in Documentation/devicetree/bindings/arm/gic.txt.
+
+    The platform must provide ARM architected timer, defined in
+    Documentation/devicetree/bindings/arm/arch_timer.txt.
+
+    If the platform is SMP then it must provide the Power State
+    Coordination Interface (PSCI) described in
+    Documentation/devicetree/bindings/arm/psci.txt.

I'm afraid I disagree with most of the above. The whole point of
mach-virt is to provide a shell for DT platforms. None of this hardware
is mandated. Instead, all the necessary information should be described
in DT.

"Add support for the smallest, dumbest possible platform, to be
 used as a guest for KVM or other hypervisors.

 It only mandates a GIC and architected timers"

(your original commit message :-P)

Actually, mach-virt doesn't really stand for Virtual Machine. It stands
for virtual mach-* directory! Eventually, mach-virt should become the
default platform, the one we use when we don't match anything else in
the kernel

I can word it more like that for sure, along with the alternative
wording suggested by Christopher/Stefano to clarify the intent that
everything comes from DTB and removal of the specific requirements for
GIC/timer/PSCI I think that suit the (new) intention better?

What you've described here are requirements for a hypervisor like Xen or
KVM. mach-virt itself shouldn't have any of that.

Cheers,

	M.