Draft: main/postmarketos-initramfs: do not re-mount boot partition into sysroot

Since quite a while we use UUIDs and not just labels in /etc/fstab, making this unnecessary. Instead of re-implementing the /etc/fstab feature, let the rootfs do its thing.

This has actually hidden a serious bug that /boot from /etc/fstab could not be mounted: https://gitlab.com/postmarketOS/pmbootstrap/-/merge_requests/2304 I guess the main question is how to transition current installations with fat32 or fat16 in /boot where /etc/fstab cannot mount it.

I guess the main question is how to transition current installations

All "current installations" or specific ones e.g. that don't have an fstab or ?

By clayton craft on 2024-04-30T21:36:57

Updated the details. The problem is that fstab with fat32 or fat16 under boot cannot mount them due to https://gitlab.com/postmarketOS/pmbootstrap/-/merge_requests/2304 So if we stop mounting in the initramfs, and fstab is buggy, then could be that /boot would not be able to be mounted.

By Pablo Correa Gomez on 2024-04-30T21:36:57

this is kinda... eww.. but since the rootfs is mounted at this point then maybe init.sh could detect if /etc/fstab is sane and fall back to mounting /boot if it's not?

By clayton craft on 2024-04-30T21:56:47

That might be a bit slow? I'm thinking to in the post-upgrade of postmarketos-base to do sed -i s%/boot fat32 defaults%/boot vfat codepage-437,iocharset=iso8859-1%g and also with 16?

By Pablo Correa Gomez on 2024-04-30T22:37:11

Though even with that we cannot remove it. After discussion in chat with @jenneron and @Arnavion seems like devices in #1898 (old installs if they've been ported, new installs if they have not yet) do not even have an /etc/fstab (https://gitlab.postmarketos.org/postmarketos/pmaports/-/blob/master/main/postmarketos-base/rootfs-etc-fstab?ref_type=heads). Though we might be able to detect which those are by searching for pmos_boot_uuid in the commandline, which is not available for those

By Pablo Correa Gomez on 2024-05-01T06:14:51

changed the description

By Pablo Correa Gomez on 2024-04-30T21:35:23

mentioned in merge request !5085 (merged)

By Pablo Correa Gomez on 2024-05-01T13:28:58

Closing, since I completely changed approach

By Pablo Correa Gomez on 2024-05-01T16:53:24

closed

mentioned in merge request !5089 (merged)

By Pablo Correa Gomez on 2024-05-01T17:50:40