Would it be helpful to test this on hardware?  I have a UEFI aarch64 machine I 
could try booting it on, though I believe it still needs patches not yet in 
mainline Linux.

  — Ian

On December 22, 2025 11:18:12 AM PST, Rutherther <[email protected]> wrote:
>
>Hey Guix,
>
>I have been getting Guix ready to be able to generate AArch64 installer
>iso image and qcow2 vm image. During this I reworked some parts of how
>images are generated and related things, namely:
>
>- Images currently remove /boot/efi from file-systems of
>  operating-system, I have added it back so it gets mounted
>
>The rationale for this has been that the generated images might not
>boot, when you specify the /boot/efi device wrongly, ie. the case in
>desktop.tmpl. I added it back, because you cannot reconfigure EFI system
>without esp mounted. But I have specified the correct label, according
>to the esp. So boot is fine now.
>
>In case you wanted to test this for qcow2 images, use newly available
>qcow2-gpt that is part of the PR. Or you could use efi-raw.
>
>- For iso9660, bootloader was unconditionally replaced with mkrescue
>  bootloader, I have conditioned this with grub-hybrid being
>  supported-package, it isn't for AArch64. (specifically grub-pc is not)
>
>The iso is bootable in QEMU and real hardware even with this change.
>What makes it use mkrescue mode of grub is make-iso9660-image in (gnu
>build image), not the bootloader being grub-mkrescue. That just helps us
>to get both EFI and Legacy BIOS support.
>
>- I have changed %default-xorg-modules to depend on the system
>
>Similarly to how %desktop-service or %base-initrd-modules do. This is
>because a package in them is not supported on AArch64.
>%default-xorg-modules typically go to services that is thunked, so I
>don't expect any problems.
>
>- I have added virtio_mmio and virtio_scsi to base-initrd-modules
>
>This is to support virtio-blk and virtio-scsi-pci on VMs by default,
>Other virtio modules are already added, so it felt right adding them
>here.
>
>On x86_64, these can be used, while on AArch64 they are pretty much
>necessary for any disks (either one or the other).
>
>--
>
>I would be interested if someone has any notes about setups this could
>potentially break. I cannot think of any, but if you are experienced
>with some non-regular setups, it would help to know. So I hope this will
>be just welcome improvements and not otherwise.
>
>I am not going to merge this to master for couple more days, but then I
>was planning to, so it would be great to know if there is something this
>could break beforehand, to fix it beforehand.
>
>Here is the PR: <https://codeberg.org/guix/guix/pulls/4591>
>
>Rutherther
>

Reply via email to