Linux - Power Loss Meltdown - Rico贾若童的博客

🚑 What broke

Sudden power-loss left the system in an inconsistent state. (while /roota was being modified)
1. First boot stopped in BusyBox with ALERT! UUID=255a5309-… does not exist
2. initramfs could not see the NVMe root partition.
  - initramfs is “initial RAM-file-system”. It runs as the very first userspace code that the kernel starts—before your real root disk is mounted.
  - It’s a tiny, compressed cpio archive (usually /boot/initrd.img-…) unpacked directly into RAM. It contains a minimal Linux userland (BusyBox), essential kernel modules, and scripts in /init.
  - Functionalities:
    - Loads drivers for storage, USB, encryption, LVM, etc.
    - Locates and mounts the real root filesystem by UUID/label.
    - Switches the kernel’s root from the RAM disk to your SSD/NVMe.
    - Hands control to /sbin/init (systemd).
  - How initramfs is built:
    - update-initramfs gathers drivers listed in /etc/initramfs-tools/modules, plus what mkinitramfs can auto-detect, then packs them into one file per installed kernel.
Solution:
1. Booted Live-USB, ran blkid, confirmed root = /dev/nvme0n1p2 with correct UUID.
2. Ran fsck.ext4 -yf /dev/nvme0n1p2.
  - fsck = File System Check.
  - ext4 partitions the helper binary is fsck.ext4
3. Chrooted into the install and forced modules into initramfs:
```
1
 nvme nvme_core usbhid hid_generic xhci_pci xhci_hcd. 
```
4. Added rootdelay=10 temporarily to grub?
  - grub is a bootloader. A bootloader takes care of getting the operating system started up. It is also responsible for allowing the user to select between multiple operating systems at boot
5. Normally, we need to use the keyboard in BusyBox to type recovery commands. However, it was dead. It turned out that system76 computers need to follow this official guide. When running as root through chroot
  
``` apt install –reinstall linux-image-generic linux-headers-generic update-initramfs -c -k all update-grub exit
After we fixed storage detection, Ubuntu reached the purple GNOME “Oh no!” screen

gnome-shell or GDM crashed because the NVIDIA kernel module and user libraries were mismatched.
- Re-installing the Nvidia driver gave a blue background with a blinking cursor—GDM started but never showed the greeter.
  1. Purged every old NVIDIA package & module (apt purge '^nvidia-.*', modprobe -r …)
  2. Re-installed one clean version (sudo ubuntu-drivers autoinstall → driver 570).
  3. Enrolled MOK for Secure Boot.
    - MOK (Machine Owner Key) is a personal signing key you enroll so Secure Boot will allow third-party kernel modules like the NVIDIA driver.
  4. One can use nvidia-smi to verify that the nvidia driver is successfully installed
With LightDM in place the greeter appeared, but logging in looped straight back because of stale caches / permissions in $HOME.
1. Booted with systemd.unit=multi-user.target (at Grub modification page. Get into it by pressing <F8>) to guarantee a console.
2. Installed and activated LightDM: apt install lightdm; systemctl disable gdm3; systemctl enable lightdm.
  - LightDM and GDM (system default) are Display manager
  - LightDM is a quick life-raft when GDM / Wayland misbehave—you can keep it installed as a fallback.
Fix login loop
1. Freed / checked space, ensured / and $HOME were read-write.
2. Re-owned files: sudo chown -R $USER:$USER $HOME.
3. Reset GNOME caches: mv ~/.cache ~/.cache.bak, mv ~/.config/dconf ~/.config/dconf.bak, etc.
4. Completed pending upgrades: apt full-upgrade.
Confirm desktop
1. Logged in via LightDM on Xorg → desktop works.
2. Optional future step: reinstall/enable GDM once everything is stable.

Linux - Power Loss Meltdown

initramfs

🚑 What broke

CATALOG

FEATURED TAGS

FRIENDS