Re: [edk2-devel] [GSoC 2022] Introducing myself & seeking for project ideas

["Pedro Falcato" <pedro.falcato@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2916 bytes --]

(CC Gerd, Isaac)

Comments inline.

On Mon, May 23, 2022 at 5:50 PM Benjamin Doron <benjamin.doron00@gmail.com>
wrote:

> Hi Theo and Nate,
> I took a brief look at this myself, because having an emulated environment
> would help me with my project. I didn't know then that QemuOpenBoardPkg was
> an accepted project this year. OvmfPkg is large, I'm unfamiliar with QEMU's
> codebase and I'm only minimally familiar with Intel's old ICH chipsets (the
> platform some emulators expose), so I looked at porting QEMU's Q35 + ICH9
> support into SimicsOpenBoardPkg. I don't know how you're preparing, but I'd
> recommend at least a look there: Q35's ICH9 and Simics' ICH10 are fairly
> similar. There are other QEMU machines, but I can't comment on those.
>
We're still trying to figure everything out and since the GSoC projects
were only announced Friday, we haven't discussed much. My idea was to try
to get something relatively smaller and simpler than current OVMF, as the
end result can be a lot more interesting than just repackaging OVMF or
straight up copying SimicsOpenBoardPkg; we're also mostly aiming for Stage
IV (Booting to an OS), so we can safely discard some of the advanced
features of OVMF for now. Again, I'd like to give you more details but it's
still too early and we're trying to introduce Theo to UEFI; hopefully we'll
get a better idea of the project for the summer this week.

>
> SimicsOpenBoardPkg can partially boot QEMU with a minimum of changes. It
> makes it into the DXE phase (where we'd eventually get a shell), but fails
> to initialise SMM, so it can't load the variable driver in there. Many
> drivers depend on the variable protocol, including critical
> UEFI-architecture ones, so the DXE core will assert and hang after printing
> a bunch of "driver GUID discovered but not loaded" messages. To fix this,
> the SMM access, maybe SMM control drivers would need to be patched; some
> register definitions differ between chipsets.
>
One issue with QEMU is that you currently have 3 machines worth supporting:
the i440fx (i440fx + PIIX), which is the default, the Q35 (Q35 + ICH9), and
the microvm (probably not going to be apart of the scope of this project,
at least for now, even though OVMF supports it). I don't know much about
Simics or SimicsOpenBoardPkg, but I imagine that taking way too much
inspiration from them would possibly negatively impact the capability of
supporting multiple platforms in one OpenBoardPkg.

>
> Anyways, I can send you my diff if you'd like, or you're welcome to
> approach this any way you'd like. SimicsOpenBoardPkg is not a true
> MinPlatform board because it implements a number of init steps itself
> rather than using some of MinPlatform's FSP-centric libraries. I'm probably
> going to skip to the step where I try Frankensteining enough MinPlatform
> code in to suffice my testing.
>
> Best regards,
> Benjamin
> 
>
>

-- 
Pedro Falcato

[-- Attachment #2: Type: text/html, Size: 3746 bytes --]