[SOLVED] 7to8 upgrade issue with DKMS modules: update-initramfs trying to generate image for non-installed kernel?

[myxal]

Cheers. I tried to upgrade yesterday, and had the `apt dist-upgrade` process fail on the new kernel packages, because the amdgpu module (installed from AMD repo) failed to build via DKMS.

# dpkg.log
2023-07-14 00:48:35 configure pve-kernel-6.2.16-4-pve:amd64 6.2.16-4 <none>
2023-07-14 00:48:35 status half-configured pve-kernel-6.2.16-4-pve:amd64 6.2.16-4

# term.log
...
checking for linux/pgtable.h... yes
checking for drm/drm_simple_kms_helper.h... yes
checking for supported chips... done
checking for module configuration... done
configure: creating ./config.status
config.status: creating config/config.h
Building module:
Cleaning build area...(bad exit status: 2)
make -j4 KERNELRELEASE=6.2.16-4-pve -j4 TTM_NAME=amdttm SCHED_NAME=amd-sched -C /lib/modules/6.2.16-4-pve/build M=/var/lib/dkms/amdgpu/5.13.11.21.50-1384495/build...(bad exit status: 2)
Error! Bad return status for module build on kernel: 6.2.16-4-pve (x86_64)
Consult /var/lib/dkms/amdgpu/5.13.11.21.50-1384495/build/make.log for more information.
Error! One or more modules failed to install during autoinstall.
Refer to previous errors for more information.
dkms: autoinstall for kernel: 6.2.16-4-pve ESC[31mfailed!ESC[0m
run-parts: /etc/kernel/postinst.d/dkms exited with return code 11
Failed to process /etc/kernel/postinst.d at /var/lib/dpkg/info/pve-kernel-6.2.16-4-pve.postinst line 20.
ESC[1mdpkg:ESC[0m error processing package pve-kernel-6.2.16-4-pve (--configure):
installed pve-kernel-6.2.16-4-pve package post-installation script subprocess returned error exit status 2
ESC[1mdpkg:ESC[0m dependency problems prevent configuration of pve-kernel-6.2:
pve-kernel-6.2 depends on pve-kernel-6.2.16-4-pve; however:
Package pve-kernel-6.2.16-4-pve is not configured yet.
ESC[1mdpkg:ESC[0m error processing package pve-kernel-6.2 (--configure):
dependency problems - leaving unconfigured

I'm currently stuck in a state where manipulating kernel packages fails, because when update-initramfs is triggered from the package scripts, it tries to build initramfs image for a kernel version which has long since been uninstalled, 5.13.19-5. Notably, this does not happen when I simply issue `update-initramfs -u -k all`.

update-initramfs: Generating /boot/initrd.img-5.13.19-5-pve
grep: /boot/config-5.13.19-5-pve: No such file or directory
W: zstd compression (CONFIG_RD_ZSTD) not supported by kernel, using gzip
grep: /boot/config-5.13.19-5-pve: No such file or directory
E: gzip compression (CONFIG_RD_GZIP) not supported by kernel
update-initramfs: failed for /boot/initrd.img-5.13.19-5-pve with 1

I figure I could fix the issue by removing whatever is causing update-initramfs to try, but I'm not sure where to look. The only filename with this version number I found are cached package for the kernel and its headers:

# find /var/ /etc /boot -name "*5.13.19-5*"
/var/cache/apt/archives/pve-headers-5.13.19-5-pve_5.13.19-13_amd64.deb
/var/cache/apt/archives/pve-kernel-5.13.19-5-pve_5.13.19-13_amd64.deb

In terms of file contents, the only reference I could find was some temp file for the amdgpu dkms package:

# grep -r "5.13.19-5" /var /etc /boot/
... (a few log files)
/var/tmp/amdgpu-dkms-kernels:5.13.19-5-pve

Any hints on fixing this? (Even if I have to uninstall the amdgpu module)

 

[myxal]

So this was pretty simple in the end - the call to `update-initramfs` is part of `postrm` script of the `amdgpu-dkms` package, where it simply reads the `amdgpu-dkms-kernels` file line by line. The old kernel version was the only line in that file.

I replaced that line with "all" and the package uninstallation succeeded. Following that, `apt dist-upgrade` and finally a check for any incomplete dpkg operations with `dpkg -l | grep -v ^ii`, and now `pve7to8 --full` reports that everything's OK