KernelCI rootfs
#1426
Replies: 1 comment 1 reply
-
I wonder if we just comment package_management, build by hand and compare rootfs size difference. Another approach might be to check kselftest.sh, why it is calling apt-get, probably it detects that some software/tool missing. If it is small package, required by kselftests - we might just add it in build recipe. |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
From experiments it is observed that when using rootfs.ext4.xz on FVP, we are having a stripped image (with apt and dpkg missing).
Taking a closer look over kernelci-core and the history of commits it seems that:
• The intention was to have a full debian system in rootfs.ext4.xz when the functionality was introduced ("646a8199 Add three new images with the full debian system")
• That all changed with "f68edce0 config/rootfs/debos: fix ext4 corrupt images". It seems that the move and compression actions on the image happen before the image is unmounted, which happens after rootfs.yaml is completed. The solution was to run those with postprocess: true which will run those after the filesystem is unmounted, but
• At that point scripts/crush.sh would have run and stripped the rootfs.
We need rootfs with apt and dpkg because, the newly built bullseye-kselftest rootfs when deployed in lava job the below blocker is observed.
Any recommendation to prevent removing the packages before unmounting is much appreciated.
Or should we build uncrushed images as well?
Beta Was this translation helpful? Give feedback.
All reactions