Factory reset is not working as described on raspberrypi

[c0ffee]

Kairos version:
NAME="kairos-opensuse-arm-rpi"
VERSION="v1.4.0-k3sv1.24.9+k3s1"
ID="kairos"
ID_LIKE="kairos-opensuse-arm-rpi"
VERSION_ID="v1.4.0-k3sv1.24.9+k3s1"
PRETTY_NAME="kairos-opensuse-arm-rpi v1.4.0-k3sv1.24.9+k3s1"
ANSI_COLOR="0;32"
BUG_REPORT_URL="https://github.com/kairos-io/kairos/issues/new/choose"
HOME_URL="https://github.com/kairos-io/provider-kairos"
IMAGE_REPO="quay.io/kairos/kairos-opensuse-arm-rpi"
IMAGE_LABEL="latest"
GITHUB_REPO="kairos-io/provider-kairos"
VARIANT="core"
FLAVOR="opensuse"

CPU architecture, OS, and Version:
Linux k8s-rpi3-1 5.14.21-150400.24.38-default #1 SMP PREEMPT_DYNAMIC Fri Dec 9 09:29:22 UTC 2022 (e9c5676) aarch64 aarch64 aarch64 GNU/Linux

Describe the bug
Factory reset from userspace is not working as described in the documentation. /oem is actually RO and is missing the grubenv file. Actually even the grub entry is missing at all too.

To Reproduce
grub2-editenv /oem/grubenv set next_entry=statereset

Expected behavior
Device does a factory reset :)

Logs

Additional context

[mudler]

Thanks @c0ffee for filing an issue!

That's a good point - by default ARM images don't come with an OEM partition so docs aren't correct in such case. We should move to LVM, or either document how to set an OEM partition externally.

[c0ffee]

The reset option is missing in the grub menu too

What would be then the way to do a reset on Wich a device?

In general I think LVM would be the best approach to have the same lock and feel

The limitation is probably the partition table at the moment right?

[mudler]

The reset option is missing in the grub menu too

What would be then the way to do a reset on Wich a device?

Once a partition labeled "COS_OEM" is found it is automatically loaded, however that may need further tweaking. Indeed the best approach is to move over something more flexible, like LVM.

The limitation is probably the partition table at the moment right?

Correct

[jimmykarily]

Waiting for this: #1271

[mauromorales]

#1271 has been merged. I tested this and it is possible to create file /oem/grubenv and it persist over reboots, however the factory reset was not initiated on next reboot

[mudler]

#1271 has been merged. I tested this and it is possible to create file /oem/grubenv and it persist over reboots, however the factory reset was not initiated on next reboot

I think that has to do to the fact that there is no installation phase while flashing, so we miss the default grub configuration:

kairos/overlay/files/system/oem/21_grub.yaml

Line 13 in b42d350

cp -rfv /etc/kairos/branding/grubmenu.cfg /tmp/mnt/STATE/grubmenu

[jimmykarily]

Isn't this a problem too?

kairos/overlay/files/system/oem/08_boot_assessment.yaml

Line 41 in a5a8a5c

after-install:

[jimmykarily]

I copied the grub config in place and I now get this at boot:

[jimmykarily]

Regarding the missing grubenv file (see the description of the issue at the top), I think there is some detection happening here: https://github.com/kairos-io/packages/blob/1cfec31b9ca1097e1fbd96d9bac990c42d04e678/packages/static/grub-config/files/grub.cfg#L5

so the grub2-editenv /oem/grubenv command actually creates that file, and the detection in the grub.cfg finds that and uses that. So as long as the grub config is in place, it should work. I need to try this out.

[jimmykarily]

Indeed, with the fix in the draft PR in osbuilder (see linked PR above), when I run grub2-editenv /oem/grubenv and then reboot, the reset menu is selected when it boots.

Then is fails with this:

[jimmykarily]

Actually it doesn't fail with the error above. After a while it continues. The problem seems to be that it can't even boot into recovery (not just "reset"). I flashed this image from quay.io: kairos-opensuse-leap-arm-rpi-v2.2.0-rc1.img.xz and this can't boot into recovery either. So, it doesn't have to do with any of my changes.

[mudler]

looks like a regression in using LVM as partitioning. I can boot into recovery just fine on Nvidia AGX Orin

[jimmykarily]

Maybe we need an insmod lvm line below this one: https://github.com/kairos-io/packages/blob/df06817de4b5bab8c1c0b572f9318880e4908659/packages/static/grub-config/files/grub.cfg#L56 ? I will try things

• https://wiki.archlinux.org/title/GRUB#LVM

[jimmykarily]

insmod lvm took the recovery boot a little bit further. It started printing things and then at some point it rebooted again. I will try with the serial cable to see if there was some error printed which I missed.

[jimmykarily]

This should fix booting to recovery: #1471 (comment)

I'm not sure if insmod lvm is also needed. I will test with and without.