Re: [RESEND] [PATCH v2 2/2] OvmfPkg/ResetVector: Exclude SEV launch secrets page from pre-validation

["Dov Murik" <dovmurik@linux.ibm.com>]

On 23/02/2023 16:58, Dov Murik wrote:
> 
> 
> On 21/02/2023 11:38, Gerd Hoffmann wrote:
>> On Mon, Feb 20, 2023 at 08:44:23AM -0600, Tom Lendacky wrote:
>>> On 2/20/23 02:49, Dov Murik wrote:
>>>> In order to allow the VMM (such as QEMU) to add a page with hashes of
>>>> kernel/initrd/cmdline for measured direct boot on SNP, this page must
>>>> not be part of the SNP metadata list reported to the VMM.
>>>>
>>>> Check if that page is defined; if it is, skip it in the metadata list.
>>>> In such case, VMM should fill the page with the hashes content, or
>>>> explicitly update it as a zero page (if kernel hashes are not used).
>>>
>>> Would it be better to define a new section type (similar to what I did in
>>> the SVSM PoC)? This way, it remains listed in the metadata and allows the
>>> VMM to detect it and decide how to handle it.
>>
>> Explicitly describing things sounds better to me too.
>>
> 
> Thanks for the feedback Tom and Gerd.
> 
> 
> I can define a new section type OVMF_SECTION_TYPE_KERNEL_HASHES. In the AmdSev
> target it'll cover the single MEMFD page at 00F000 (after the CPUID page).  
> Now there's a question for the QEMU side -- should QEMU then fill the page
> and encrypt it (launch_update type=NORMAL)? (currently the whole hashes table
> creation and encryption is done elsewhere there)
> 
> And on regular OvmfX64 builds - should that area should be with type
> OVMF_SECTION_TYPE_SNP_SEC_MEM which is accepted as a type=ZERO page ?
> 
> 
> Playing with this idea, the metadata list will add:
> 
> 
> ; Kernel hashes section for measured direct boot
> %define OVMF_SECTION_TYPE_KERNEL_HASHES   0x5
> 
> ...
> 
> ; Kernel hashes for measured direct boot, or zero page if
> ; there are no kernel hashes / SEV secrets
> SevSnpKernelHashes:
>   DD  SEV_SNP_KERNEL_HASHES_BASE
>   DD  SEV_SNP_KERNEL_HASHES_SIZE
>   DD  SEV_SNP_KERNEL_HASHES_TYPE
> 

Or maybe this metadata section ^^^^^ should be added only if the Pcd for
secrets+hashes page is defined?

-Dov

> 
> 
> and the base/size/type of that region are defined in an
> %if statement in ResetVector.nasmb:
> 
> 
> %if (FixedPcdGet32 (PcdSevLaunchSecretBase) > 0)
>   ; There's a reserved page for SEV secrets and hashes; the VMM will fill and
>   ; validate the page
>   %define SEV_SNP_KERNEL_HASHES_TYPE    OVMF_SECTION_TYPE_KERNEL_HASHES
>   %define SEV_SNP_KERNEL_HASHES_BASE    (FixedPcdGet32 (PcdSevLaunchSecretBase))
> %else
>   ; No SEV secrets and hashes page; the VMM will validate it as another zero page
>   %define SEV_SNP_KERNEL_HASHES_TYPE    OVMF_SECTION_TYPE_SNP_SEC_MEM
>   %define SEV_SNP_KERNEL_HASHES_BASE    (CPUID_BASE + CPUID_SIZE)
> %endif
> %define SEV_SNP_KERNEL_HASHES_SIZE    (FixedPcdGet32 (PcdOvmfSecPeiTempRamBase) - SEV_SNP_KERNEL_HASHES_BASE)
> 
> 
> (I still need to figure out the point about QEMU above.)
> 
> 
> Is that what you had in mind?
> 
> -Dov