Diagnose and recover non-booting Windows machines from a Linux recovery environment (e.g., SystemRescue USB). Use this skill whenever a user or agent mentions: a Windows machine that won't boot, BSOD loops, NTFS filesystem errors, hibernation/Fast Startup issues, GPU driver crashes preventing boot, or needs to back up data from a dead Windows drive. Also triggers on: "my laptop won't boot", "Windows is stuck", "ntfsfix", "dirty filesystem", "rsync from Windows", "recover data from Windows drive", or any scenario involving Linux-based diagnosis of a Windows disk. When in doubt, use this skill — it is designed to run safely and will ask the operator before any destructive action.
Diagnose and recover non-booting Windows systems using a Linux recovery environment. Covers: NTFS filesystem repair, hibernation state clearing, GPU driver crash recovery, data backup via rsync, and BCD/boot configuration repair.
A Windows machine will not boot -- stuck in BSOD loops, hung on startup, or reporting NTFS filesystem errors -- and the only available recovery environment is a Linux live USB (e.g., SystemRescue). The user needs to diagnose the root cause, back up critical data before it is lost, and repair the system to a bootable state, all without access to Windows itself.
This skill supports three operators:
Always do this first. Before running any commands:
ntfsfix — clears hibernation state, empties NTFS journal (irreversible)rsync with --delete — destructive sync (only if used)dd, mkfs, fdisk, gdisk writeschntpwInterrupt modes:
[A] Prompt before every action
[B] Prompt before dangerous actions only ← recommended default
[C] Run fully — log everything, interrupt only on unexpected errors
Do not proceed until the operator selects a mode. Default to [B] if no response (agent context).
lsblk # Confirm NVMe/SATA drive is visible
fdisk -l /dev/nvme0n1 # Verify GPT partition table intact
efibootmgr -v # Read-only: confirm UEFI boot entries
Expected: NVMe visible, GPT with EFI + NTFS + Recovery partitions.
If drive not visible: Check BIOS NVMe mode (RAID vs AHCI). Run nvme list. If still missing — possible dead drive. Stop and report.
mount -t ntfs-3g -o ro /dev/nvme0n1p4 /mnt/win
mount | grep nvme
Outcomes:
fuseblk ro) → hibernation metadata present. Proceed to Step 3.⚠️ FUSE mount blocks raw device access.
ntfsinfo,ntfsfix, and similar tools require the volume to be unmounted first.
Run all diagnostics before any repairs.
xxd -l 4 /mnt/win/hiberfil.sys
RSTR → Fast Startup hibernation (most common boot blocker)hibr/HIBR → Full hibernationwake/WAKE → Already resumed, not the issue⚠️ Copy hives to writable location first — chntpw cannot read from read-only FUSE mount.
mkdir -p /mnt/tmp/diagnostic
cp /mnt/win/Windows/System32/config/SOFTWARE /mnt/tmp/diagnostic/
cp /mnt/win/Windows/System32/config/SYSTEM /mnt/tmp/diagnostic/
chntpw -e /mnt/tmp/diagnostic/SOFTWARE # Check Windows version/build
chntpw -e /mnt/tmp/diagnostic/SYSTEM # Check boot state, driver config
In SYSTEM hive: cd Select → check Current (active ControlSet) and Failed value.
For driver state: cd ControlSet001\Services\<drivername> → Start value: 0=boot, 1=system, 2=auto, 3=demand, 4=disabled.
ls -la /mnt/win/Windows/Minidump/
xxd -s 0x38 -l 4 -e <file>.dmp # Read bugcheck code
GPU crash codes: 0x116 = VIDEO_TDR_FAILURE, 0x119 = VIDEO_SCHEDULER_INTERNAL_ERROR
ls -la /mnt/win/Windows/System32/ntoskrnl.exe
ls -la /mnt/win/Windows/System32/winload.efi
Do this before any repair that modifies the volume.
⚠️ DANGEROUS if
--deleteis used. Confirm exclusions with operator before running. Note: Gather ALL exclusions before starting. Restarting rsync re-scans the full tree.
# Identify large directories worth excluding first
du -sh /mnt/win/Users/<username>/*/
# Mount external drive
lsblk # identify external drive (e.g., /dev/sdb1)
mkdir -p /mnt/backup
mount /dev/sdb1 /mnt/backup
# Run backup
rsync -av --progress \
--exclude='AppData' \
--exclude='NTUSER*' \
--exclude='ntuser*' \
/mnt/win/Users/<username>/ /mnt/backup/<backup-name>/
Known non-fatal error: rsync exit code 23 with symlink errors on exFAT target = Windows junction points. Not real user data. Safe to ignore or add --no-links.
I/O errors on specific files = possible bad NVMe sectors. Note affected files. rsync will skip and continue. Run chkdsk C: /r after Windows boots.
⚠️ DANGEROUS — irreversible. Clears hibernation state and empties NTFS journal. Only proceed if hibernation signature is
RSTRorhibrand state is known corrupt.
umount /mnt/win
ntfsfix /dev/nvme0n1p4
mount -t ntfs-3g /dev/nvme0n1p4 /mnt/win
mount | grep nvme # Verify now RW
If still read-only after first run:
ntfsfix -d /dev/nvme0n1p4
If crash dumps show 0x116/0x119 and Windows boots after ntfsfix:
# From Windows elevated CMD after booting Safe Mode:
bcdedit /set {current} safeboot minimal
# Reboot → Safe Mode → uninstall NVIDIA from Device Manager
bcdedit /deletevalue {current} safeboot
# Reboot normally → install fresh driver from nvidia.com
Also: always disable Fast Startup after recovery:
powercfg /h off
Boot to recovery (F12 → recovery partition, or Shift+Restart from Windows):
Troubleshoot → Advanced Options → Command Prompt
bcdedit /set {current} safeboot minimal
powercfg /h off
chkdsk C: /f
WinRE → Troubleshoot → Advanced Options → System Restore. Choose restore point from before crash cascade. If no restore points or restore fails → Windows Reset or clean install. Ensure backup from Step 4 is complete first.
Drive visible? → NO → Check BIOS NVMe mode. Stop if still missing.
↓ YES
Partition table intact? → NO → Run gdisk, testdisk. Stop if unrecoverable.
↓ YES
BitLocker? → YES → Need recovery key. Stop.
↓ NO
hiberfil.sys signature?
RSTR/hibr → Step 5 (ntfsfix) [after Step 4 backup]
wake/missing → Step 3c crash dumps
↓
Crash dumps show 0x116/0x119? → YES → Step 6 (GPU driver)
↓ NO
BCD/boot files missing? → YES → Step 7 (WinRE)
↓ NO
ntfsfix + WinRE both fail → Step 8 (System Restore / reinstall)
[ ] powercfg /h off # Disable Fast Startup permanently
[ ] Uninstall NVIDIA driver (Safe Mode) # If GPU BSODs were occurring
[ ] Install latest NVIDIA driver # nvidia.com
[ ] sfc /scannow # System file check
[ ] DISM /Online /Cleanup-Image /RestoreHealth
[ ] chkdsk C: /f (add /r if I/O errors seen)
[ ] Windows Update
[ ] Razer firmware update (support.razer.com, search RZ09-0409)
[ ] Install HWiNFO64, monitor GPU temps under load (>90C sustained = thermal issue)
[ ] Verify backup integrity on external drive
mount -t ntfs-3g /dev/nvme0n1p4 /mnt/win && mount | grep nvme shows rw (not ro).chkdsk C: /f from within Windows and verify it reports a clean filesystem with no errors.powercfg /a should not list "Fast Startup" as available, or powercfg /h confirms hibernation is off.| Pitfall | Symptom | Fix |
|---|---|---|
| chntpw on FUSE mount | "Access denied" | Copy hive to writable path first |
| ntfsinfo on mounted volume | "Exclusively opened" | Unmount first |
| ntfsfix on mounted volume | Fails/no-op | Unmount first |
| exFAT rsync symlink errors | Exit code 23, symlink failed | Non-fatal; add --no-links |
| I/O errors on specific files | Input/output error (5) | Note files; run chkdsk /r after boot |
| rsync restart rescans tree | Slow restart | Gather all exclusions before first run |
| Razer splash hides POST | Can't see boot errors | Use F12 to interrupt splash |
Runbook derived from live incident 2026-02-28 (Razer Blade 15 RZ09-0409). Generalized for Windows NTFS boot failures diagnosed from Linux.