Rootfs build script returned, corrupt rootfs

General Discussions
1

Hi Everyone,

Having a very strange mender issue that I can’t seem to find a solution to. I’m using an intel nuc running ubuntu that I converted to using menders partitioning scheme using this guide (INTEL NUC x86-64 Ubuntu 18.04) with the main difference being that I increased the MENDER_STORAGE_TOTAL_SIZE_MB and IMAGE_ROOTFS_SIZE to accommodate my image. I deployed this and it works fine and boots well. The next step is that I’m then trying to use mender to update the OS to a yocto image built with mender. When I install this new image and reboot I get an error saying that it returned from rootfs boot script and that this could mean that the currently selected rootfs is corrupt. It then rolls back.

Some things I’ve tried

  • mounting the inactive partition and confirming that it isn’t corrupt and mounts fine
  • created a live USB with the yocto image and booted it on the NUC to confirm it does boot fine
  • tried installing the existing ubuntu image as a mender update, which worked successfully without the rootfs error

Is there any way to get logs from that rootfs boot script? it seems fairly opaque, even any info on what its checking would be useful to see what could be causing the issue.

2

Hi @pauladam316,

Thanks for reaching out. For your question:

  • logs of problems at these early stages of booting are hard to obtain, as the bootloader usually does not have any persistent storage to put them. However you might be able to see what’s going on via console or screen.
  • In this specific case, my guess is that an older GRUB can’t mount an ext4 filesystem created with current features. This troubleshoot hint might help you: Yocto Project runtime | Mender documentation

Greetz,
Josef

3

Hi @TheYoctoJester

Thanks for the reply! I took a look and yes I am in fact using different grub versions. mender-convert is installing grub 2.04 and my kirkstone-based yocto build is on grub 2.06.

I followed the instructions and added EXTRA_IMAGECMD:ext4:append = " -O ^metadata_csum_seed" to my local.conf, but unfortunately that did not fix the problem. I’m still seeing the same error

4

Hi @pauladam316,

Can you try to capture the output by grub during startup?

Greetz,
Josef