Re: [edk2-devel] [PATCH 0/3] OvmfPkg/Tcg2ConfigPei: fix ARM/AARCH64 build failure

["Laszlo Ersek" <lersek@redhat.com>]

On 05/21/20 10:26, Laszlo Ersek wrote:
> On 05/21/20 00:58, Laszlo Ersek wrote:
>> Ref:    https://bugzilla.tianocore.org/show_bug.cgi?id=2728
>> Repo:   https://pagure.io/lersek/edk2.git
>> Branch: restrict_tpm12_to_x86_bz_2728
>>
>> Another regression fix for edk2-stable202005.
>>
>> End of February 2020, Ard and Marc-André worked on two TPM-related
>> features in parallel. Respectively:
>>
>> - [edk2-devel] [PATCH v4 00/11]
>>   ArmVirtPkg: implement measured boot for ArmVirtQemu
>>
>>   http://mid.mail-archive.com/20200227144056.56988-1-ard.biesheuvel@linaro.org
>>   https://edk2.groups.io/g/devel/message/55004
>>
>> - [edk2-devel] [PATCH v4 0/5]
>>   Ovmf: enable TPM 1.2
>>
>>   http://mid.mail-archive.com/20200226152433.1295789-1-marcandre.lureau@redhat.com
>>   https://edk2.groups.io/g/devel/message/54894
>>
>> Both series were merged tightly one after the other. There was no merge
>> conflict, and standing alone (without rebasing one on the other), each
>> series was self-contained and correct. Their combination however led to
>> an ArmVirtQemu build regression. There never was an intent to support
>> TPM-1.2 in ArmVirtQemu, but the TPM-1.2 series for OVMF kind of made
>> that "mandatory".
>>
>> Worse, the build regression has remained hidden for 2+ months because
>> (a) I didn't expect Marc-André's series to affect any ArmVirtPkg
>> platform, (b) my ArmVirtQemu build script did not set TPM2_ENABLE.
>>
>> This series fixes the build regression, and intends no functional
>> changes at all.
>>
>> Functional regression-testing would be appreciated:
>>
>> - from Simon regarding their TPM-1.2 passthrough use case,
>>
>> - from Marc-André regarding vTPM-2.0 on X64,
>>
>> - from Eric regarding vTPM-2.0 on AARCH64.
>>
>> This is a regression fix, therefore it is eligible for merging during
>> the edk2-stable202005 Hard Feature Freeze too
>> <https://github.com/tianocore/tianocore.github.io/wiki/EDK-II-Release-Planning>.
>>
>> If you plan to regression-test this series, then please say so soon,
>> otherwise I wouldn't like to wait for long -- assuming an R-b from Ard
>> or Jordan -- even without Tested-by's.
>>
>> In the future we should likely set some "-D" flags somewhere under
>> "ArmVirtPkg/PlatformCI/" (so that our CI coverage grow). The best I can
>> personally do about that is maybe file a BZ?...
>>
>> Cc: Ard Biesheuvel <ard.biesheuvel@arm.com>
>> Cc: Eric Auger <eric.auger@redhat.com>
>> Cc: Jordan Justen <jordan.l.justen@intel.com>
>> Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
>> Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
>> Cc: Simon Hardy <simon.hardy@itdev.co.uk>
>> Cc: Stefan Berger <stefanb@linux.ibm.com>
>>
>> Thanks,
>> Laszlo
>>
>> Laszlo Ersek (3):
>>   OvmfPkg/Tcg2ConfigPei: clean up some lib class dependencies
>>   OvmfPkg/Tcg2ConfigPei: factor out InternalTpm12Detect()
>>   OvmfPkg/Tcg2ConfigPei: skip TPM-1.2 detection when building for
>>     ARM/AARCH64
>>
>>  OvmfPkg/Tcg/Tcg2Config/Tcg2ConfigPei.inf  | 13 +++-
>>  OvmfPkg/Tcg/Tcg2Config/Tcg2ConfigPeim.c   | 46 +-----------
>>  OvmfPkg/Tcg/Tcg2Config/Tpm12Support.c     | 79 ++++++++++++++++++++
>>  OvmfPkg/Tcg/Tcg2Config/Tpm12Support.h     | 34 +++++++++
>>  OvmfPkg/Tcg/Tcg2Config/Tpm12SupportNull.c | 25 +++++++
>>  5 files changed, 153 insertions(+), 44 deletions(-)
>>  create mode 100644 OvmfPkg/Tcg/Tcg2Config/Tpm12Support.c
>>  create mode 100644 OvmfPkg/Tcg/Tcg2Config/Tpm12Support.h
>>  create mode 100644 OvmfPkg/Tcg/Tcg2Config/Tpm12SupportNull.c
>>
> 
> It seems like noone has started reviewing / testing yet. That's good:
> 
> Nacked-by: Laszlo Ersek <lersek@redhat.com>
> 
> I'll send a v2 with the following changes:
> 
> - the first patch will also clean up some comments that are now stale
> (after the TPM-1.2 addition)
> 
> - the last patch will restrict the BaseLib dependency as well to IA32/X64

I'm rescinding my self-NAK in light of Ard's Tested-by; the latter is
something I won't waste.

The above-listed v2 updates can perfectly well wait until after the
stable tag.

Thanks!
Laszlo