From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([2001:4830:134:3::10]:46751)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <alex.williamson@redhat.com>) id 1aO9sZ-0006lC-A6
	for qemu-devel@nongnu.org; Tue, 26 Jan 2016 15:06:22 -0500
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <alex.williamson@redhat.com>) id 1aO9sW-0001z9-0e
	for qemu-devel@nongnu.org; Tue, 26 Jan 2016 15:06:19 -0500
Received: from mx1.redhat.com ([209.132.183.28]:59596)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <alex.williamson@redhat.com>) id 1aO9sV-0001ym-Jk
	for qemu-devel@nongnu.org; Tue, 26 Jan 2016 15:06:15 -0500
Message-ID: <1453838773.15515.1.camel@redhat.com>
From: Alex Williamson <alex.williamson@redhat.com>
Date: Tue, 26 Jan 2016 13:06:13 -0700
In-Reply-To: <20160126102003.GA14400@nvidia.com>
References: <569C5071.6080004@intel.com>
	<1453092476.32741.67.camel@redhat.com> <569CA8AD.6070200@intel.com>
	<1453143919.32741.169.camel@redhat.com> <569F4C86.2070501@intel.com>
	<AADFC41AFE54684AB9EE6CBC0274A5D15F786B4B@SHSMSX101.ccr.corp.intel.com>
	<56A6083E.10703@intel.com> <1453757426.32741.614.camel@redhat.com>
	<AADFC41AFE54684AB9EE6CBC0274A5D15F78D2A3@SHSMSX101.ccr.corp.intel.com>
	<20160126102003.GA14400@nvidia.com>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Subject: Re: [Qemu-devel] VFIO based vGPU(was Re: [Announcement] 2015-Q3
 release of XenGT - a Mediated ...)
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Neo Jia <cjia@nvidia.com>, "Tian, Kevin" <kevin.tian@intel.com>
Cc: "Ruan, Shuai" <shuai.ruan@intel.com>, "Song, Jike" <jike.song@intel.com>, "kvm@vger.kernel.org" <kvm@vger.kernel.org>, "igvt-g@lists.01.org" <igvt-g@ml01.01.org>, qemu-devel <qemu-devel@nongnu.org>, Kirti Wankhede <kwankhede@nvidia.com>, Gerd Hoffmann <kraxel@redhat.com>, Paolo Bonzini <pbonzini@redhat.com>, "Lv, Zhiyuan" <zhiyuan.lv@intel.com>

On Tue, 2016-01-26 at 02:20 -0800, Neo Jia wrote:
> On Mon, Jan 25, 2016 at 09:45:14PM +0000, Tian, Kevin wrote:
> > > From: Alex Williamson [mailto:alex.williamson@redhat.com]
>=C2=A0
> Hi Alex, Kevin and Jike,
>=C2=A0
> (Seems I shouldn't use attachment, resend it again to the list, patches=
 are
> inline at the end)
>=C2=A0
> Thanks for adding me to this technical discussion, a great opportunity
> for us to design together which can bring both Intel and NVIDIA vGPU so=
lution to
> KVM platform.
>=C2=A0
> Instead of directly jumping to the proposal that we have been working o=
n
> recently for NVIDIA vGPU on KVM, I think it is better for me to put out=
 couple
> quick comments / thoughts regarding the existing discussions on this th=
read as
> fundamentally I think we are solving the same problem, DMA, interrupt a=
nd MMIO.
>=C2=A0
> Then we can look at what we have, hopefully we can reach some consensus=
 soon.
>=C2=A0
> > Yes, and since you're creating and destroying the vgpu here, this is
> > where I'd expect a struct device to be created and added to an IOMMU
> > group. =C2=A0The lifecycle management should really include links bet=
ween
> > the vGPU and physical GPU, which would be much, much easier to do wit=
h
> > struct devices create here rather than at the point where we start
> > doing vfio "stuff".
>=C2=A0
> Infact to keep vfio-vgpu to be more generic, vgpu device creation and m=
anagement
> can be centralized and done in vfio-vgpu. That also include adding to I=
OMMU
> group and VFIO group.

Is this really a good idea?=C2=A0=C2=A0The concept of a vgpu is not uniqu=
e to
vfio, we want vfio to be a driver for a vgpu, not an integral part of
the lifecycle of a vgpu.=C2=A0=C2=A0That certainly doesn't exclude adding
infrastructure to make lifecycle management of a vgpu more consistent
between drivers, but it should be done independently of vfio.=C2=A0=C2=A0=
I'll go
back to the SR-IOV model, vfio is often used with SR-IOV VFs, but vfio
does not create the VF, that's done in coordination with the PF making
use of some PCI infrastructure for consistency between drivers.

It seems like we need to take more advantage of the class and driver
core support to perhaps setup a vgpu bus and class with vfio-vgpu just
being a driver for those devices.

> Graphics driver can register with vfio-vgpu to get management and emula=
tion call
> backs to graphics driver.=C2=A0=C2=A0=C2=A0
>=C2=A0
> We already have struct vgpu_device in our proposal that keeps pointer t=
o
> physical device.=C2=A0=C2=A0
>=C2=A0
> > - vfio_pci will inject an IRQ to guest only when physical IRQ
> > generated; whereas vfio_vgpu may inject an IRQ for emulation
> > purpose. Anyway they can share the same injection interface;
>=C2=A0
> eventfd to inject the interrupt is known to vfio-vgpu, that fd should b=
e
> available to graphics driver so that graphics driver can inject interru=
pts
> directly when physical device triggers interrupt.=C2=A0
>=C2=A0
> Here is the proposal we have, please review.
>=C2=A0
> Please note the patches we have put out here is mainly for POC purpose =
to
> verify our understanding also can serve the purpose to reduce confusion=
s and speed up=C2=A0
> our design, although we are very happy to refine that to something even=
tually
> can be used for both parties and upstreamed.
>=C2=A0
> Linux vGPU kernel design
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D
>=C2=A0
> Here we are proposing a generic Linux kernel module based on VFIO frame=
work
> which allows different GPU vendors to plugin and provide their GPU virt=
ualization
> solution on KVM, the benefits of having such generic kernel module are:
>=C2=A0
> 1) Reuse QEMU VFIO driver, supporting VFIO UAPI
>=C2=A0
> 2) GPU HW agnostic management API for upper layer software such as libv=
irt
>=C2=A0
> 3) No duplicated VFIO kernel logic reimplemented by different GPU drive=
r vendor
>=C2=A0
> 0. High level overview
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D
>=C2=A0
> =C2=A0
> =C2=A0 user space:
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+-----------+=C2=A0=C2=A0VFI=
O IOMMU IOCTLs
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+---------| Q=
EMU VFIO |-------------------------+
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0VFIO IOCTLs=C2=A0=C2=A0=
=C2=A0|=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+-----------=
+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0|
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|=C2=A0
> =C2=A0---------------------|-------------------------------------------=
----|---------
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|
> =C2=A0 kernel space:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|=C2=A0=C2=
=A0+--->----------->---+=C2=A0=C2=A0(callback)=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0V
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|=C2=A0=C2=A0=
|=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0v=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+=
------V-----+
> =C2=A0 +----------+=C2=A0=C2=A0=C2=A0+----V--^--+=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+--+--+-----+=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0| VGPU=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0|
> =C2=A0 |=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|=C2=
=A0=C2=A0=C2=A0|=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0|=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+----| nvidia.ko +----->-----> TYPE1 IO=
MMU|
> =C2=A0 | VFIO Bus <=3D=3D=3D| VGPU.ko=C2=A0=C2=A0|<----|=C2=A0=C2=A0=C2=
=A0=C2=A0+-----------+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0+---++-------+=C2=A0
> =C2=A0 |=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|=C2=
=A0=C2=A0=C2=A0|=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0|=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0| (register)=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0^=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0||
> =C2=A0 +----------+=C2=A0=C2=A0=C2=A0+-------+--+=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0|=C2=A0=C2=A0=C2=A0=C2=A0+-----------+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
|=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0||
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0V=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+----| i915.ko=C2=A0=C2=
=A0=C2=A0+-----+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+---VV-------+=C2=A0
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0|=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0+-----^-----+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0| TYPE1=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0|=C2=A0=C2=A0(callback)=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0| IOMMU=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0+-->------------>---+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+------------+
> =C2=A0access flow:
>=C2=A0
> =C2=A0 Guest MMIO / PCI config access
> =C2=A0 |
> =C2=A0 -------------------------------------------------
> =C2=A0 |
> =C2=A0 +-----> KVM VM_EXITs=C2=A0=C2=A0(kernel)
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|
> =C2=A0 -------------------------------------------------
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+-----> QEM=
U VFIO driver (user)
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|=C2=A0
> =C2=A0 -------------------------------------------------
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0|
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0+---->=C2=A0=C2=A0VGPU kernel driver =
(kernel)
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0|=C2=A0=C2=A0
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0|=C2=A0
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0+----> vendor driver callback
>=C2=A0
>=C2=A0
> 1. VGPU management interface
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D
>=C2=A0
> This is the interface allows upper layer software (mostly libvirt) to q=
uery and
> configure virtual GPU device in a HW agnostic fashion. Also, this manag=
ement
> interface has provided flexibility to underlying GPU vendor to support =
virtual
> device hotplug, multiple virtual devices per VM, multiple virtual devic=
es from
> different physical devices, etc.
>=C2=A0
> 1.1 Under per-physical device sysfs:
> -----------------------------------------------------------------------=
-----------
>=C2=A0
> vgpu_supported_types - RO, list the current supported virtual GPU types=
 and its
> VGPU_ID. VGPU_ID - a vGPU type identifier returned from reads of
> "vgpu_supported_types".
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0
> vgpu_create - WO, input syntax <VM_UUID:idx:VGPU_ID>, create a virtual
> gpu device on a target physical GPU. idx: virtual device index inside a=
 VM
>=C2=A0
> vgpu_destroy - WO, input syntax <VM_UUID:idx>, destroy a virtual gpu de=
vice on a
> target physical GPU


I've noted in previous discussions that we need to separate user policy
from kernel policy here, the kernel policy should not require a "VM
UUID".=C2=A0=C2=A0A UUID simply represents a set of one or more devices a=
nd an
index picks the device within the set.=C2=A0=C2=A0Whether that UUID match=
es a VM
or is independently used is up to the user policy when creating the
device.

Personally I'd also prefer to get rid of the concept of indexes within a
UUID set of devices and instead have each device be independent.=C2=A0=C2=
=A0This
seems to be an imposition on the nvidia implementation into the kernel
interface design.


> 1.3 Under vgpu class sysfs:
> -----------------------------------------------------------------------=
-----------
>=C2=A0
> vgpu_start - WO, input syntax <VM_UUID>, this will trigger the registra=
tion
> interface to notify the GPU vendor driver to commit virtual GPU resourc=
e for
> this target VM.=C2=A0
>=C2=A0
> Also, the vgpu_start function is a synchronized call, the successful re=
turn of
> this call will indicate all the requested vGPU resource has been fully
> committed, the VMM should continue.
>=C2=A0
> vgpu_shutdown - WO, input syntax <VM_UUID>, this will trigger the regis=
tration
> interface to notify the GPU vendor driver to release virtual GPU resour=
ce of
> this target VM.
>=C2=A0
> 1.4 Virtual device Hotplug
> -----------------------------------------------------------------------=
-----------
>=C2=A0
> To support virtual device hotplug, <vgpu_create> and <vgpu_destroy> can=
 be
> accessed during VM runtime, and the corresponding registration callback=
 will be
> invoked to allow GPU vendor support hotplug.
>=C2=A0
> To support hotplug, vendor driver would take necessary action to handle=
 the
> situation when a vgpu_create is done on a VM_UUID after vgpu_start, and=
 that
> implies both create and start for that vgpu device.
>=C2=A0
> Same, vgpu_destroy implies a vgpu_shudown on a running VM only if vendo=
r driver
> supports vgpu hotplug.
>=C2=A0
> If hotplug is not supported and VM is still running, vendor driver can =
return
> error code to indicate not supported.
>=C2=A0
> Separate create from start gives flixibility to have:
>=C2=A0
> - multiple vgpu instances for single VM and
> - hotplug feature.
>=C2=A0
> 2. GPU driver vendor registration interface
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D
>=C2=A0
> 2.1 Registration interface definition (include/linux/vgpu.h)
> -----------------------------------------------------------------------=
-----------
>=C2=A0
> extern int vgpu_register_device(struct pci_dev *dev,=C2=A0
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0const struct gpu_device_ops =
*ops);
>=C2=A0
> extern void vgpu_unregister_device(struct pci_dev *dev);
>=C2=A0
> /**
> =C2=A0* struct gpu_device_ops - Structure to be registered for each phy=
sical GPU to
> =C2=A0* register the device to vgpu module.
> =C2=A0*
> =C2=A0* @owner:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
The module owner.
> =C2=A0* @vgpu_supported_config:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Call=
ed to get information about supported vgpu
> =C2=A0* types.
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@dev : pci device structure of phy=
sical GPU.=C2=A0
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@config: should return string list=
ing supported
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0config
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Returns integer: success (0) or er=
ror (< 0)
> =C2=A0* @vgpu_create:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Called to allocate basic res=
ouces in graphics
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0driver for a particular vgpu.
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@dev: physical pci device structur=
e on which
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0vgpu=C2=A0
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
should be created
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@vm_uuid: VM's uuid for which VM i=
t is intended
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0to
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@instance: vgpu instance in that V=
M
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@vgpu_id: This represents the type=
 of vgpu to be
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0created
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Returns integer: success (0) or er=
ror (< 0)
> =C2=A0* @vgpu_destroy:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Called to free resources in graphi=
cs driver for
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0a vgpu instance of that VM.
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@dev: physical pci device structur=
e to which
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0this vgpu points to.
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@vm_uuid: VM's uuid for which the =
vgpu belongs
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0to.
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@instance: vgpu instance in that V=
M
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Returns integer: success (0) or er=
ror (< 0)
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0If VM is running and vgpu_destroy =
is called that=C2=A0
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0means the vGPU is being hotunpluge=
d. Return
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0error
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0if VM is running and graphics driv=
er doesn't
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0support vgpu hotplug.
> =C2=A0* @vgpu_start:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Called to do initiate =
vGPU initialization
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0process in graphics driver when VM=
 boots before
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0qemu starts.
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@vm_uuid: VM's UUID which is booti=
ng.
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Returns integer: success (0) or er=
ror (< 0)
> =C2=A0* @vgpu_shutdown:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Called to teardown vGPU related resou=
rces for
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0the VM
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@vm_uuid: VM's UUID which is shutt=
ing down .
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Returns integer: success (0) or er=
ror (< 0)
> =C2=A0* @read:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0Read emulation callback
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@vdev: vgpu device structure
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@buf: read buffer
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@count: number bytes to read=C2=A0
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@address_space: specifies for whic=
h address
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0space
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0the request is: pci_config_space, =
IO register
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0space or MMIO space.
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Retuns number on bytes read on suc=
cess or error.
> =C2=A0* @write:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
Write emulation callback
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@vdev: vgpu device structure
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@buf: write buffer
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@count: number bytes to be written
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@address_space: specifies for whic=
h address
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0space
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0the request is: pci_config_space, =
IO register
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0space or MMIO space.
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Retuns number on bytes written on =
success or
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0error.
> =C2=A0* @vgpu_set_irqs:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0Called to send about interrupts confi=
guration
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0information that qemu set.=C2=A0
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@vdev: vgpu device structure
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0@flags, index, start, count and *d=
ata : same as
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0that of struct vfio_irq_set of
> =C2=A0*=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0VFIO_DEVICE_SET_IRQS API.=C2=A0
> =C2=A0*
> =C2=A0* Physical GPU that support vGPU should be register with vgpu mod=
ule with=C2=A0
> =C2=A0* gpu_device_ops structure.
> =C2=A0*/
>=C2=A0
> struct gpu_device_ops {
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0struct module=C2=A0=C2=A0=
=C2=A0*owner;
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0int=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0(*vgpu_supported_config)(struct pci_dev *dev, char *config);
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0int=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0(*vgpu_create)(struct pci_dev *dev, uuid_le vm_uuid,
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0uint32_t instance, uint32_t vgpu_i=
d);
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0int=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0(*vgpu_destroy)(struct pci_dev *dev, uuid_le vm_uuid,
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0uint32_t instance);
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0int=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0(*vgpu_start)(uuid_le vm_uuid);
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0int=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0(*vgpu_shutdown)(uuid_le vm_uuid);
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0ssize_t (*read) (struct=
 vgpu_device *vdev, char *buf, size_t count,
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0uint32_t address_space, loff_t pos);
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0ssize_t (*write)(struct=
 vgpu_device *vdev, char *buf, size_t count,
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0uint32_t address_space,loff_t pos);
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0int=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0(*vgpu_set_irqs)(struct vgpu_device *vdev, uint32_t flags,
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0unsigned index, unsign=
ed start, unsigned count,
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0void *data);
>=C2=A0
> };


I wonder if it shouldn't be vfio-vgpu sub-drivers (ie, Intel and Nvidia)
that register these ops with the main vfio-vgpu driver and they should
also include a probe() function which allows us to associate a given
vgpu device with a set of vendor ops.


>=C2=A0
> 2.2 Details for callbacks we haven't mentioned above.
> -----------------------------------------------------------------------=
----------
>=C2=A0
> vgpu_supported_config: allows the vendor driver to specify the supporte=
d vGPU
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0type/co=
nfiguration
>=C2=A0
> vgpu_create=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
: create a virtual GPU device, can be used for device hotplug.
>=C2=A0
> vgpu_destroy=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0: des=
troy a virtual GPU device, can be used for device hotplug.
>=C2=A0
> vgpu_start=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0: callback function to notify vendor driver vgpu device
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0come to=
 live for a given virtual machine.
>=C2=A0
> vgpu_shutdown=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0: callback=
 function to notify vendor driver=C2=A0
>=C2=A0
> read=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0: callback to vendor driver to handle vi=
rtual device config
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0space o=
r MMIO read access
>=C2=A0
> write=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0: callback to vendor driver to handle virtu=
al device config
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0space o=
r MMIO write access
>=C2=A0
> vgpu_set_irqs=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0: callback=
 to vendor driver to pass along the interrupt
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0informa=
tion for the target virtual device, then vendor
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0driver =
can inject interrupt into virtual machine for this
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0device.
>=C2=A0
> 2.3 Potential additional virtual device configuration registration inte=
rface:
> -----------------------------------------------------------------------=
----------
>=C2=A0
> callback function to describe the MMAP behavior of the virtual GPU=C2=A0
>=C2=A0
> callback function to allow GPU vendor driver to provide PCI config spac=
e backing
> memory.
>=C2=A0
> 3. VGPU TYPE1 IOMMU
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D
>=C2=A0
> Here we are providing a TYPE1 IOMMU for vGPU which will basically keep =
track the=C2=A0
> <iova, hva, size, flag> and save the QEMU mm for later reference.
>=C2=A0
> You can find the quick/ugly implementation in the attached patch file, =
which is
> actually just a simple version Alex's type1 IOMMU without actual real
> mapping when IOMMU_MAP_DMA / IOMMU_UNMAP_DMA is called.=C2=A0
>=C2=A0
> We have thought about providing another vendor driver registration inte=
rface so
> such tracking information will be sent to vendor driver and he will use=
 the QEMU
> mm to do the get_user_pages / remap_pfn_range when it is required. Afte=
r doing a
> quick implementation within our driver, I noticed following issues:
>=C2=A0
> 1) OS/VFIO logic into vendor driver which will be a maintenance issue.
>=C2=A0
> 2) Every driver vendor has to implement their own RB tree, instead of r=
eusing
> the common existing VFIO code (vfio_find/link/unlink_dma)=C2=A0
>=C2=A0
> 3) IOMMU_UNMAP_DMA is expecting to get "unmapped bytes" back to the cal=
ler/QEMU,
> better not have anything inside a vendor driver that the VFIO caller im=
mediately
> depends on.
>=C2=A0
> Based on the above consideration, we decide to implement the DMA tracki=
ng logic
> within VGPU TYPE1 IOMMU code (ideally, this should be merged into curre=
nt TYPE1
> IOMMU code) and expose two symbols to outside for MMIO mapping and page
> translation and pinning.=C2=A0
>=C2=A0
> Also, with a mmap MMIO interface between virtual and physical, this all=
ows
> para-virtualized guest driver can access his virtual MMIO without takin=
g a MMAP
> fault hit, also we can support different MMIO size between virtual and =
physical
> device.
>=C2=A0
> int vgpu_map_virtual_bar
> (
> =C2=A0=C2=A0=C2=A0=C2=A0uint64_t virt_bar_addr,
> =C2=A0=C2=A0=C2=A0=C2=A0uint64_t phys_bar_addr,
> =C2=A0=C2=A0=C2=A0=C2=A0uint32_t len,
> =C2=A0=C2=A0=C2=A0=C2=A0uint32_t flags
> )
>=C2=A0
> EXPORT_SYMBOL(vgpu_map_virtual_bar);


Per the implementation provided, this needs to be implemented in the
vfio device driver, not in the iommu interface.=C2=A0=C2=A0Finding the DM=
A mapping
of the device and replacing it is wrong.=C2=A0=C2=A0It should be remapped=
 at the
vfio device file interface using vm_ops.


> int vgpu_dma_do_translate(dma_addr_t *gfn_buffer, uint32_t count)
>=C2=A0
> EXPORT_SYMBOL(vgpu_dma_do_translate);
>=C2=A0
> Still a lot to be added and modified, such as supporting multiple VMs a=
nd=C2=A0
> multiple virtual devices, tracking the mapped / pinned region within VG=
PU IOMMU=C2=A0
> kernel driver, error handling, roll-back and locked memory size per use=
r, etc.=C2=A0

Particularly, handling of mapping changes is completely missing.=C2=A0=C2=
=A0This
cannot be a point in time translation, the user is free to remap
addresses whenever they wish and device translations need to be updated
accordingly.


> 4. Modules
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D
>=C2=A0
> Two new modules are introduced: vfio_iommu_type1_vgpu.ko and vgpu.ko
>=C2=A0
> vfio_iommu_type1_vgpu.ko - IOMMU TYPE1 driver supporting the IOMMU=C2=A0
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0TYPE1 v1 and v2 interface.=C2=A0

Depending on how intrusive it is, this can possibly by done within the
existing type1 driver.=C2=A0=C2=A0Either that or we can split out common =
code for
use by a separate module.

> vgpu.ko=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0- provide registration interfac=
e and virtual device
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0VFIO access.
>=C2=A0
> 5. QEMU note
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D
>=C2=A0
> To allow us focus on the VGPU kernel driver prototyping, we have introd=
uced a new VFIO=C2=A0
> class - vgpu inside QEMU, so we don't have to change the existing vfio/=
pci.c file and=C2=A0
> use it as a reference for our implementation. It is basically just a qu=
ick c & p
> from vfio/pci.c to quickly meet our needs.
>=C2=A0
> Once this proposal is finalized, we will move to vfio/pci.c instead of =
a new
> class, and probably the only thing required is to have a new way to dis=
cover the
> device.
>=C2=A0
> 6. Examples
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D
>=C2=A0
> On this server, we have two NVIDIA M60 GPUs.
>=C2=A0
> [root@cjia-vgx-kvm ~]# lspci -d 10de:13f2
> 86:00.0 VGA compatible controller: NVIDIA Corporation Device 13f2 (rev =
a1)
> 87:00.0 VGA compatible controller: NVIDIA Corporation Device 13f2 (rev =
a1)
>=C2=A0
> After nvidia.ko gets initialized, we can query the supported vGPU type =
by
> accessing the "vgpu_supported_types" like following:
>=C2=A0
> [root@cjia-vgx-kvm ~]# cat /sys/bus/pci/devices/0000\:86\:00.0/vgpu_sup=
ported_types=C2=A0
> 11:GRID M60-0B
> 12:GRID M60-0Q
> 13:GRID M60-1B
> 14:GRID M60-1Q
> 15:GRID M60-2B
> 16:GRID M60-2Q
> 17:GRID M60-4Q
> 18:GRID M60-8Q
>=C2=A0
> For example the VM_UUID is c0b26072-dd1b-4340-84fe-bf338c510818, and we=
 would
> like to create "GRID M60-4Q" VM on it.
>=C2=A0
> echo "c0b26072-dd1b-4340-84fe-bf338c510818:0:17" > /sys/bus/pci/devices=
/0000\:86\:00.0/vgpu_create
>=C2=A0
> Note: the number 0 here is for vGPU device index. So far the change is =
not tested
> for multiple vgpu devices yet, but we will support it.
>=C2=A0
> At this moment, if you query the "vgpu_supported_types" it will still s=
how all
> supported virtual GPU types as no virtual GPU resource is committed yet=
.
>=C2=A0
> Starting VM:
>=C2=A0
> echo "c0b26072-dd1b-4340-84fe-bf338c510818" > /sys/class/vgpu/vgpu_star=
t
>=C2=A0
> then, the supported vGPU type query will return:
>=C2=A0
> [root@cjia-vgx-kvm /home/cjia]$
> > cat /sys/bus/pci/devices/0000\:86\:00.0/vgpu_supported_types
> 17:GRID M60-4Q
>=C2=A0
> So vgpu_supported_config needs to be called whenever a new virtual devi=
ce gets
> created as the underlying HW might limit the supported types if there a=
re
> any existing VM runnings.
>=C2=A0
> Then, VM gets shutdown, writes to /sys/class/vgpu/vgpu_shutdown will in=
fo the
> GPU driver vendor to clean up resource.
>=C2=A0
> Eventually, those virtual GPUs can be removed by writing to vgpu_destro=
y under
> device sysfs.


I'd like to hear Intel's thoughts on this interface.=C2=A0=C2=A0Are there
different vgpu capacities or priority classes that would necessitate
different types of vcpus on Intel?

I think there are some gaps in translating from named vgpu types to
indexes here, along with my previous mention of the UUID/set oddity.

Does Intel have a need for start and shutdown interfaces?

Neo, wasn't there at some point information about how many of each type
could be supported through these interfaces?=C2=A0=C2=A0How does a user k=
now their
capacity limits?

Thanks,
Alex