Need help with my terrible small /boot/efi-partition (100 MB)

English Install/Boot/Login
21

I don’t see why I should create and have yet another partition (and type) when Btrfs. At least I know SBC’s e.g. with Rockchip that have SPI-flash programmable with own bootloader, can have support for Btrfs, so you can then have just 1 Btrfs partition , no boot partition needed. That is just a bit background how I experience things.

But I have not read the whole topic thread, but if Btrfs for rootfs, you can fix this problem live without reboot. A properly implemented UEFI firmware can have the ESP anywhere, does not need to be number 1 nor physically located first in the block storage (NVME or SD-card or HDD or etc). So at least if using LABEL= or UUID= (not PARTUUID=) in fstab:

  • note down label and uuid of the ESP
  • copy its content to some backup folder on rootfs and umount ESP
  • shrink rootfs and partition by 512M or so
  • wipefs the ESP and set part type to empty
  • create new FAT partition type ESP with label and uuid of the former/older
  • mount and copy the files backed up to it
  • then use efibootmgr or bootctl to recreate a bootentry as the URI in the firmware still is refering to the old. Else only option is EFI/boot/bootx64.efi (in case of Intel)
  • I mostly just reinstall grub (in Debian) or also 'update-bootloader ’ should work (Tumbleweed)
  • for a more nice parttion setup, one can temprary replace the rootfs partition to somewhere else and then cut that 512M the beginning.
  • for secure boot, I have not used/done this
  • for Ext4, one need unmounted rootfs to shrink, so need boot with liveCD or so.
22

This description lacks all safety precautions.

  • always have a backup of everything before shrinking the rootfs
  • never do this from a running system
  • …

In the best case it will work without safety precautions…but don’t start crying when it fails and you don’t have any backup.

23

So what has your answer to do with my question ?

24

I nowadays do not write ‘make sure you have a backup’. That should be obvious IMO in a Tumbleweed subforum. I use snapper as core for some scripts for more than a decade and can generate a complete OS installation from NAS/backup/clone stores to the storage of PC/computer/SBC.
In practice, I actually would just use my backup/cloning scripts to clone back, with the fix/change that the partition for ESP is larger. Is numbers in my sfdisk --dump text essentially.
But in case of complex multi-boot (I have 3 distros on an ARM SBC RK3588 including Tumbleweed), then easier to change things while keeping things running. like HomeAssitent, PiHole, and so on.

25

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.