It is also possible to convert an MBR disk to GPT in-place and install the GRUB EFI bootloader after-the-fact. Be forewarned, however, that this is very tricky.
20201210/Google convert mbr to gpt
20201210/Google linux convert mbr to gpt in-place
20201210/https://serverfault.com/questions/963178/how-do-i-convert-my-linux-disk-from-mbr-to-gpt-with-uefi
20201210/Google grub install gpt partition
20201210/https://superuser.com/questions/903112/grub2-install-this-gpt-partition-lavel-contains-no-bios-boot-partition
20201210/Google grub install efi boot
20201210/https://askubuntu.com/questions/831216/how-can-i-reinstall-grub-to-the-efi-partition
Run all these commands as root.
# First, we assume the existing disk layout looks something like this:
# Part# Size Mount FS
# 1 1074MB /boot ext4
# 2 8.5GB / ext4
# 3 8GB /home ext4
# 5 4GB swap
# Note that LVM is not in use, this will make your life so much
# easier. If you are using LVM, that makes things more complicated.
# Convert from MBR to GPT.
sgdisk -g /dev/sda
# PLEASE NOTE: Until you reinstall the bootloader, your system is
# unbootable!
# Shrink /boot filesystem to 940 MiB, as measured in 4KiB blocks.
resize2fs /dev/sda1 $((940 * 1024 / 4))
# Shrink /boot partition to 1GB.
parted /dev/sda resizepart 1 1000MB
# Create the 1MB BIOS boot and 16MB EFI boot partitions.
parted /dev/sda mkpart bios-boot fat32 1002MB 1003MB
parted /dev/sda toggle 4 bios_grub
parted /dev/sda mkpart efiboot fat32 1004MB 1020MB
parted /dev/sda toggle 6 boot
partprobe
# Cleanup any files erroneously in `/boot/efi`, format, update
# `/etc/fstab`, mount the efiboot partition.
rm -rf /boot/efi
mkdir -p /boot/efi
mkfs -t vfat /dev/sda6
echo /dev/disk/by-partlabel/efiboot /boot/efi vfat defaults 0 2 >>/etc/fstab
mount /boot/efi
# Important! Make sure you regenerate your `initramfs` after all this
# partition manipulation, especially if you're using LVM, or else you
# could easily end up with a partially unbootable system!
dracut --regenerate-all --force
# Install required packages and update the bootloader.
yum -y install grub2-pc-modules grub2-efi-x64-modules efibootmgr
# Note that we must specify `--target` explicitly because it defaults
# to `i386-pc`. Also, we must explicitly specify both targets since
# the `x86_64-efi` target does not install the legacy BIOS boot code.
grub2-install --target=i386-pc /dev/sda
# grub2-install --target=x86_64-efi /dev/sda # Does not support Secure Boot.
# grub2-mkconfig -o /boot/grub2/grub.cfg # Will do later.
# PLEASE NOTE: You will get warnings and not be able to update EFI
# vars at this point because we did not boot with EFI yet.
# Copy to default EFI boot filename so that we can boot with EFI for
# the first time.
# mkdir -p /boot/efi/EFI/BOOT
# cp -p /boot/efi/EFI/centos/grubx64.efi /boot/efi/EFI/BOOT/BOOTX64.EFI
# In lieu of my old recommendation, simply use the Secure Boot shim
# installation of GRUB, this makes UEFI boot much easier, plus it also
# works with Secure Boot.
yum -y install grub2-efi-x64 shim-x64
rm /boot/grub2/grubenv # Ensure we don't symlink into the ESP.
grub2-mkconfig -o /boot/grub2/grub.cfg
cp -p /boot/grub2/{grub.cfg,grubenv} /boot/efi/EFI/centos/
# Note that if we want to do anything fancy, you still need the
# `grub2-efi-x64-modules` package, but you also have to copy these
# modules to the EFI System Partition (ESP).
echo Double check your configuration and reboot.
# echo Within EFI, you must run the bootloader install command again
# echo so that EFI vars can be set to update the EFI boot menu.
# echo
# echo grub2-install --target=x86_64-efi /dev/sda
Regarding LVM. If you almost brick…
If you get stuck at the recovery console before switch_root, do something like this:
mount /dev/dm-3 /sysroot
^D
Then in CentOS semi-booted recovery console, first run:
vgscan --mknodes
Now, immediately regenerate your initramfs to prevent this from
happening again.
dracut --regenerate-all --force
Other random web notes say…
vgchange -a y vgname...
One last helpful tip, if you are cloning ext2/3/4 partitions but need them to have unique UUIDs for the sake of finding the right ones at boot, that is an easy fix:
tune2fs /dev/sda1 -U `cat /proc/sys/kernel/random/uuid`
To create more loop devices if you need them on CentOS/Fedora:
# To create /dev/loop4:
mknod -m 0660 /dev/loop4 b 7 4
20201210/Google losetup create new loop device as needed
20201210/https://unix.stackexchange.com/questions/98742/how-to-add-more-dev-loop-devices-on-fedora-19