Re: [edk2-devel] [edk2-staging/RiscV64QemuVirt PATCH V4 10/34] OvmfPkg/Sec: Add RISC-V support

["Chang, Abner" <abner.chang@amd.com>]

[AMD Official Use Only - General]



> -----Original Message-----
> From: Sunil V L <sunilvl@ventanamicro.com>
> Sent: Monday, October 17, 2022 12:53 PM
> To: Chang, Abner <Abner.Chang@amd.com>
> Cc: devel@edk2.groups.io; Ard Biesheuvel <ardb+tianocore@kernel.org>;
> Jiewen Yao <jiewen.yao@intel.com>; Jordan Justen
> <jordan.l.justen@intel.com>; Gerd Hoffmann <kraxel@redhat.com>; Singh,
> Brijesh <Brijesh.Singh@amd.com>; Erdem Aktas <erdemaktas@google.com>;
> James Bottomley <jejb@linux.ibm.com>; Min Xu <min.m.xu@intel.com>;
> Lendacky, Thomas <Thomas.Lendacky@amd.com>; Daniel Schaefer
> <git@danielschaefer.me>
> Subject: Re: [edk2-devel] [edk2-staging/RiscV64QemuVirt PATCH V4 10/34]
> OvmfPkg/Sec: Add RISC-V support
> 
> Caution: This message originated from an External Source. Use proper
> caution when opening attachments, clicking links, or responding.
> 
> 
> On Sat, Oct 15, 2022 at 03:50:02PM +0000, Chang, Abner wrote:
> > [AMD Official Use Only - General]
> >
> > Hi Sunil, this is my comment for both 9/34 and 10/34.
> > The original RISC-V SecCore implementation can be used by RiscVVirt and
> the RISC-V platforms under edk2-platform. So my plan back to half year ago
> was to have SecCore under UefiCpuPkg, that would be Riscv64SecCore
> because current the SecCore under UeficpuPkg is hard to be leveraged. With
> this, one RISC-V SecCore driver can be used by RiscVVirt and RISC-V
> platforms. Of course we have to move some Pcds to under PCD Riscv64 arch
> section in UEfiCpuPkg.dec. Even through having PcdRiscVDbtFvBase(Size) for
> Riscv64 in UefiCpuPkg.dec makes sense to me because Device Tree is part of
> RISC-V processor initialization.
> > Could you please take some time having this change? I think that is doable
> and also an ideal implementation for RISC-V.
> >
> Hi Abner,
> 
> Having SEC as part of SecCoreNative in UefiCpuPkg was my original proposal
> when I had sent the RFC series to discuss in the design meeting. But if you
> remember, Mike provided 2 major feedbacks in the meeting which made me
> to move to OvmfPkg.
> 
> 1) We should not add FDF related PCD variables in common packges like
> UefiCpuPkg/MdeModulePkg. OvmfPkg is the only package (ofcourse
> ArmVirtPkg also currently) which is an exception to this since it actually has
> platform implementation useful for CI coverage.
> 
> 2) Avoid new circular dependencies. We will have to use EmbeddedPkg in
> UefiCpuPkg/MdeModulePkg for FDT if we use SecCore in UefiCpuPkg.
Ok, this is convincing. Thanks.
Abner

> 
> It made sense to me and OvmfPkg already had the required PCD variables
> and modules. Also, qemu virt can support different features (boot from flash
> vs memory, number of flash devices etc) compared to real platforms.
> So, it will be hard to use the same design for both virt machine and real
> platforms. The original SecCore in edk2-platforms can be continued to be
> used by real platforms. Like Ray suggested, I plan to look at MinPlatformPkg
> design for real platforms in edk2-platforms repo which will make the design
> flexible.
> 
> Thanks
> Sunil
> 
> > Thanks
> > Abner