Hi Simon,

On Mon, Jul 20, 2026 at 1:17 PM Simon Glass <[email protected]> wrote:
>
> Hi Anirudh,
>
> On 2026-07-19T18:03:31, Anirudh Srinivasan
> <[email protected]> wrote:
>
> > Patches 1-5 contain misc fixes to drivers/bios_emulator that are
> > suitable for upstream. I would like to get these reviewed and merged in
> > if possible.
> >
> > Patches 6, 7, 8 are an implementation of how I got u-boot to actually
> > talk to to the Aspeed VGA card.
>
> Thanks for looking at this. Please can you split this into two
> submissions? Patches 1-5 stand on their own; mixing them with the
> NOT-FOR-UPSTREAM material makes it awkward to pick up what is ready.
> Send the RFC portion as its own series marked [RFC] in the subject.

Thank you for your feedback on this series. I will polish it up based
on your feedback and send something over for the first 5 patches next
week.

>
> On the design: I suspect the right shape is a driver-specific hook
> rather than open-coding Aspeed knowledge inside drivers/bios_emulator.
> The emulator should stay generic and call out to the video driver (or
> a small ops table) to translate an I/O port access into whatever the
> card needs - MMIO on BAR2 here, real I/O on x86. See comment on patch
> 7.

Yes, I was thinking of something like this too, but wasn't sure what
kind of hooks/routing other VBIOSes might need. I'll see if I can try
out an AMD card with this approach.

>
> > This series contains a bunch of mischallaneous fixes to the BIOSEMU
> > codebase
>
> miscellaneous - also I noticed  'to to' and 'There some checkpatch warnings'.
>
> For upstream this will need a doc/ entry (which board, which
> defconfig, what to expect) and ideally test coverage. The QEMU RISC-V
> enablement in patch 8 is a nice way to make that reproducible - please
> consider keeping it so it can run in CI.

This setup requires you to have a PCIe VGA card and run a qemu VM with
pcie passthrough (or a board with working PCIe in upstream u-boot).
This isn't easily testable in CI, right? The doc/ entry would also
become super specific to the setup I am currently using.
What do you suggest I do here? Should I just document my specific
setup and add another defconfig with these Kconfig options? I was
thinking that we shouldn't add bloat to the existing
qemu-riscv64_smode_defconfig.

>
> Regards,
> Simon

Reply via email to