On 10/16/25 00:29, Aithal, Srikanth wrote:
>
> On 10/15/2025 5:46 PM, Gerd Hoffmann wrote:
>> On Wed, Oct 15, 2025 at 04:16:10PM +0530, Aithal, Srikanth wrote:
>>> Hello Gred,
>>>
>>> AMD SEV-ES guest boot with only OVMF_CODE file on current master crashes.
>> Booting with OVMF_CODE only is not a valid configuration.
>
>
> I knew that it was not supported, but only for AMD SEV-SNP. But now, if
> you're saying it's not a supported configuration for any configs, I'll
> modify my tests accordingly.
Btw, the patch that caused this behavior is actually a51255072e72
("OvmfPkg/EmuVariableFvbRuntimeDxe: initialize emu variable fvb from rom")
Theissue Srikanth encountered that pointed at 38370cf492 was actually
fixed with 0f0b472ae54b ("OvmfPkg/PlatformInitLib: add sanity checks to
igvm code"), which was applied after a51255072e72.
What is occurring is ValidateFvHeader() is supplied the address from
PcdOvmfFlashNvStorageVariableBase, which is an MMIO address. A #VC is
generated when the code accesses this memory. However, the memory is
accessed with a CMP instruction, which is an unsupported MMIO
instruction, resulting in the assert.
Thanks,
Tom
>
> Thank you
>
> Srikanth
>
>
>> take care,
>> Gerd
>>
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#121667): https://edk2.groups.io/g/devel/message/121667
Mute This Topic: https://groups.io/mt/115616009/21656
Group Owner: [email protected]
Unsubscribe: https://edk2.groups.io/g/devel/unsub [[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
