Re: [PATCH] Add QemuIGDHelperDxe

[Laszlo Ersek <lersek@redhat.com>]

Hi Gerd,

On 04/25/18 14:13, Gerd Hoffmann wrote:
> Little driver to setup the OpRegion for Intel vgpu devices.
> 
> Contributed-under: TianoCore Contribution Agreement 1.1
> Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
> ---
>  OvmfPkg/QemuIGDHelperDxe/QemuIGDHelper.c      | 262 ++++++++++++++++++++++++++
>  OvmfPkg/OvmfPkgIa32.dsc                       |   1 +
>  OvmfPkg/OvmfPkgIa32.fdf                       |   1 +
>  OvmfPkg/OvmfPkgIa32X64.dsc                    |   1 +
>  OvmfPkg/OvmfPkgIa32X64.fdf                    |   1 +
>  OvmfPkg/OvmfPkgX64.dsc                        |   1 +
>  OvmfPkg/OvmfPkgX64.fdf                        |   1 +
>  OvmfPkg/QemuIGDHelperDxe/QemuIGDHelperDxe.inf |  32 ++++
>  8 files changed, 300 insertions(+)
>  create mode 100644 OvmfPkg/QemuIGDHelperDxe/QemuIGDHelper.c
>  create mode 100644 OvmfPkg/QemuIGDHelperDxe/QemuIGDHelperDxe.inf

I wrote the exact same driver last week, functionally speaking, and
indeed it "works" (reportedly).

But, it's not the right thing to do. Please see:

  https://bugzilla.tianocore.org/show_bug.cgi?id=935

As I wrote there, the right thing to do is one of the following two options:

(a) OS-level drivers for the GVT-g VGPU should be liberated of ACPI and
other firmware-level dependencies, and become pure PCI(e) drivers.

(b) alternatively, Intel should provide a standalone, complete UEFI
driver for the GVT-g VGPU, that sets up all the artifacts for the
OS-level driver as well, without relying on fw_cfg or other platform
hacks. This driver should exist as a file on the EFI system partition,
or (better) in the ROM BAR of the assigned device.

Changbin Du's comment at
<https://bugzilla.tianocore.org/show_bug.cgi?id=935#c8> seems to imply
that (b) could happen in practice.

Thanks!
Laszlo