Re: [PATCH v3 0/3] OvmfPkg: mark flash memory range as MMIO

[Laszlo Ersek <lersek@redhat.com>]

On 07/05/18 21:12, Brijesh Singh wrote:
> The Qemu flash range is marked as 'system ram' in EFI runtime memmap
> but it is actually an IO address. The patch series updates the
> EFI runtime memmap to add this range as Memory Mapped IO address.
> 
> Changes since v2:
>  - added cover letter and extend CC list to add all OvmfPkg
>    maintainers/reviewers.
> 
> Changes since v1:
>  - split the OvmfPkg single patch into three patches based on Laszlo's feedback.
>   a) mark memory as MMIO for non SEV case
>   b) do not adding EFI runtime mapping for SMM build
>   c) clear C-bit when SEV is active for non SMM builds only
> 
> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
> Cc: Anthony Perard <anthony.perard@citrix.com>
> Cc: Julien Grall <julien.grall@linaro.org>
> Cc: Justen Jordan L <jordan.l.justen@intel.com>
> Cc: Laszlo Ersek <lersek@redhat.com>
> Contributed-under: TianoCore Contribution Agreement 1.1
> Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
> 
> Complete tree is available:
> url: https://github.com/codomania/edk2.git
> branch: qemu-flash-mmio
> 
> Brijesh Singh (3):
>   OvmfPkg/QemuFlashFvbServicesRuntimeDxe: mark Flash memory range as
>     MMIO
>   OvmfPkg/QemuFlashFvbServicesRuntimeDxe: Do not expose MMIO in SMM
>     build
>   OvmfPkg/QemuFlashFvbServicesRuntimeDxe: Restore C-bit when SEV is
>     active
> 
>  .../FvbServicesRuntimeDxe.inf                      |  1 +
>  .../FwBlockService.c                               | 38 +-----------
>  .../FwBlockService.h                               |  7 +++
>  .../FwBlockServiceDxe.c                            | 68 ++++++++++++++++++++++
>  .../FwBlockServiceSmm.c                            | 13 +++++
>  5 files changed, 90 insertions(+), 37 deletions(-)
> 

Regression tested -- only without SEV -- on top of commit 9090c8b53301,
using the configs below.

The "rdmsr" and "efibootmgr" checks are explained at
<https://github.com/tianocore/tianocore.github.io/wiki/Testing-SMM-with-QEMU,-KVM-and-libvirt>.


* i440fx, X64, without SMM:

(01) Fedora 28 guest, efibootmgr + rdmsr checks after boot and S3: PASS

(02) RHEL 6 guest, efibootmgr check after boot: PASS

(03) RHEL 7 guest, efibootmgr check after boot and S3: PASS

(04) Windows 7 guest, boot and S3: PASS


* q35, IA32, with SMM:

(05) Fedora 26 guest, efibootmgr check after boot and S3: PASS


* q35, IA32X64, with SMM:

(06) Fedora 28 guest, efibootmgr + rdmsr checks after boot and S3: PASS

(07) RHEL 7 guest, efibootmgr check after boot and S3: PASS

(08) Windows 10 guest, boot and S3: PASS

(09) Windows Server 2016 guest, boot: PASS
     (can't test S3 due to signing requirements for the QXL driver)

(10) Windows 8.1 guest, boot and S3: PASS

(11) Windows Server 2012 R2 guest, boot and S3: PASS

(12) Windows 7 guest, boot and S3: PASS

(13) Windows Server 2008 R2 guest, boot and S3: PASS


Once you submit v4 with the superficial updates requested, I will add:

Regression-tested-by: Laszlo Ersek <lersek@redhat.com>

I'll leave the SEV testing (with and without SMM) up to you :)

Thanks!
Laszlo