Re: GRUB 2.12 release, etc.

[Glenn Washburn]

On Fri, 1 Jul 2022 18:26:05 +0800
Xiaotian Wu <wuxiaotian@loongson.cn> wrote:

> 
> 在 2022/6/30 下午9:12, Daniel Kiper 写道:
> > On Tue, Jun 14, 2022 at 11:30:47AM -0500, Glenn Washburn wrote:
> >> On Tue, 14 Jun 2022 17:48:39 +0800
> >> Xiaotian Wu <wuxiaotian@loongson.cn> wrote:
> >>
> >>> Hi Daniel.
> >>>
> >>> Thank you for your release plan.
> >>>
> >>> I would like to ask, can GRUB-2.12 add a new LoongArch architecture 
> >>> support?
> >>>
> >>> LoongArch is a RISC instruction set from China. Currently, Linux-5.19
> >>> adds LoongArch CPU support and will support complete EFIStub in version
> >>> 5.20.
> >>>
> >>> On GRUB, including 2 patch sets:
> >>>
> >>> 1. Add the basic support of LoongArch so that GRUB can run on the
> >>> LoongArch machine. See
> >>> https://github.com/loongarch64/grub/tree/dev/patchwork/base
> >>>
> >>> 2. Add the loader for kernel, but it depends on a series of patch of
> >>> Loadfile2. See https://github.com/loongarch64/grub/tree/dev/patchwork/efi
> >>>
> >>> These patchs are already available, and tested on CLFS and Archlinux for
> >>> LoongArch.
> >>>
> >>> If we plan to support LoongArch in GRUB-2.12, I will update these patchs
> >>> and release the 4th version of the patch for review.
> >> I think it would be great to have LoongArch in GRUB. It should be a hard
> >> requirement that support for the architecture be added for the make
> >> check tests. This will require that QEMU have support for LoongArch. It
> >> looks like the current release of QEMU does not have support for
> >> LoongArch, but there is a patch series[1] which adds it. I presume that
> > What is the status of this patch series right now? If it is in or close
> > by I would ask for at least minimal tests for the LoongArch support in
> > the GRUB.
> 
> These patchs can be worked, and I just sent the V4 version of the patch 
> to the email list.
> 
> About the QEMU, we have maintained a branch[1] to generate 
> qemu-system-loongarch64, but there are still some jobs that are not 
> completed. The expected result is to add it to QEMU-7.1.
> 
> I have added related tests in the patch of the second part[2], but 

Great, this is what I was looking for. Especially important is adding
support for LoongArch via QEMU in grub-shell, which that commit appears
to do.

> because the QEMU work is not completed, it cannot pass the all testcase, 
> although it works well on the physical machine.

I would like something like that commit mentioned to be included with
the series adding support for LoongArch. The GRUB tests that you don't
plan on having support in QEMU to pass anytime soon should be skipped
(as it looks like has been done from that commit). For tests that need
support that is currently lacking in QEMU but is planned to be added
soon, I'm fine with them failing. It would be nice if this situation
could be detected (perhaps QEMU version check?) and a hard error (exit
99) returned, rather than a normal failure. But if its a big pain, is
not super important for me.

Glenn

> [1] https://github.com/loongson/qemu/tree/tcg-full-system
> 
> [2] 
> https://github.com/loongarch64/grub/commit/fe92204a681d9d550b724bd9fdc31971eebc77e0
> 
> 
> >> will get accepted at some point. I'd suggest the tests be made assuming
> >> that (or some other patch LoongArch patch series) will get accepted.
> >> The INSTALL file can be updated to indicate that QEMU > 7 is required,
> >> or something like that. Also, if there's any special build requirements
> >> those should also be noted in the INSTALL file and docs/grub*.texi
> >> should be updated for the new arch. For example, what's the minimum GCC
> >> version needed to compile?
> > Yeah, that would be nice if INSTALL file is updated too.
> >
> > Daniel
> >
> > _______________________________________________
> > Grub-devel mailing list
> > Grub-devel@gnu.org
> > https://lists.gnu.org/mailman/listinfo/grub-devel