kernel-install 90-uki-copy.install fails with ... is missing .efi suffix.

[05storm26]

systemd version the issue has been seen with

core/systemd 253-1 [installed]

Used distribution

Arch

Linux kernel version used

6.1.12-arch1-1

CPU architectures issue was seen on

x86_64

Component

kernel-install

Expected behaviour you didn't see

90-uki-copy.install shouldn't fail when the mkinitcpio hook generates the uki image to the proper place.

Unexpected behaviour you saw

$ sudo kernel-install -v add 6.1.11-hardened1-1-hardened /usr/lib/modules/6.1.11-hardened1-1-hardened/vmlinuz /boot/intel-ucode.img
[...]
+ mkinitcpio -k '6.1.11-hardened1-1-hardened' --kernelimage /usr/lib/modules/6.1.11-hardened1-1-hardened/vmlinuz --microcode /boot/intel-ucode.img -U '/efi/EFI/Linux/46e5b7b51f69426fade60fe0cc165d7a-6.1.11-hardened1-1-hardened.efi'
==> Starting build: 6.1.11-hardened1-1-hardened
  -> Running build hook: [base]
  -> Running build hook: [systemd]
  -> Running build hook: [sd-vconsole]
  -> Running build hook: [kms]
==> WARNING: Possibly missing firmware for module: ast
  -> Running build hook: [sd-plymouth]
  -> Running build hook: [btrfs]
  -> Running build hook: [autodetect]
  -> Running build hook: [keyboard]
==> WARNING: Possibly missing firmware for module: xhci_pci
  -> Running build hook: [modconf]
  -> Running build hook: [block]
  -> Running build hook: [lvm2]
  -> Running build hook: [sd-encrypt]
==> WARNING: Possibly missing firmware for module: qat_4xxx
  -> Running build hook: [filesystems]
  -> Running build hook: [fsck]
  -> Running build hook: [sd-shutdown]
==> Generating module dependencies
==> Creating zstd-compressed initcpio image: /tmp/mkinitcpio.yhhdte
==> Image generation successful
==> Creating unified kernel image: /efi/EFI/Linux/46e5b7b51f69426fade60fe0cc165d7a-6.1.11-hardened1-1-hardened.efi
  -> Using UEFI stub: /usr/lib/systemd/boot/efi/linuxx64.efi.stub
  -> Using cmdline file: /etc/kernel/cmdline
  -> Using os-release file: /etc/os-release
  -> Using microcode image: /boot/intel-ucode.img
==> Unified kernel image generation successful
+/usr/lib/kernel/install.d/90-loaderentry.install add 6.1.11-hardened1-1-hardened /efi/46e5b7b51f69426fade60fe0cc165d7a/6.1.11-hardened1-1-hardened /usr/lib/modules/6.1.11-hardened1-1-hardened/vmlinuz /boot/intel-ucode.img
+/usr/lib/kernel/install.d/90-uki-copy.install add 6.1.11-hardened1-1-hardened /efi/46e5b7b51f69426fade60fe0cc165d7a/6.1.11-hardened1-1-hardened /usr/lib/modules/6.1.11-hardened1-1-hardened/vmlinuz /boot/intel-ucode.img
Error: /usr/lib/modules/6.1.11-hardened1-1-hardened/vmlinuz is missing .efi suffix.

Steps to reproduce the problem

When the layout is set to uki the 90-uki-copy.install fails with Error: /usr/lib/modules/6.1.11-hardened1-1-hardened/vmlinuz is missing .efi suffix., when the initramfs generation generates the uki image directly to the destination <esp>/EFI/Linux/<target>. 90-uki-copy shouldn't fail when the mkinitcpio hook generates the uki image to the proper place, and there is nothing to be copied by 90-uki-copy.

Additional program output to the terminal or log subsystem illustrating the issue

No response

[05storm26]

cc @behrmann

[YHNdnzj]

IIUC isn't this a mkinitcpio issue? The resulting files should be stored in $KERNEL_INSTALL_STAGING_AREA rather than being installed to esp/EFI/Linux directly.

[05storm26]

I was under the impression that that is only optional and that the initrd generator can do whatever they want including installing the image to where they wish.

[behrmann]

This is something to be fixed in mkinitcpio and is mostly an artifact that we reused the name uki for the layout before a systemd version with 90-uki-copy.install was released, which maybe wasn't the greatest idea, so putting the file into $KERNEL_INSTALL_STAGING_AREA didn't make sense, because from there it wouldn't get anywhere. As a workaround until this is fixed I recommend masking 90-uki-copy.intall in /etc/kernel/install.d (make a symlink of the same name pointing to /dev/null).

[05storm26]

Okay that works, not sure if there is a mkinitcpio bug about this yet or not. I have added this to the wiki.

[behrmann]

Cool, thanks! I'll submit a MR to address this to mkinitcpio on Monday.

[nougad]

I just created a issue for mkinicpio: https://gitlab.archlinux.org/archlinux/mkinitcpio/mkinitcpio/-/issues/163

[behrmann]

I think this bug can be closed now, since this has a pending fix downstream

[poettering]

Ok, let's close this then.