kernel panic on lenevo server

[manojdev-ai]

Will reported - Running this on lenovo server resulted in kernel panic on reboot
@claude

[rsingh-alt]

Kernel Panic on Reboot After Running the Tool

Running the script to install NVIDIA drivers + CUDA drivers on a Lenovo machine causes a kernel panic on reboot.

Few reasons :

1. Secure Boot Is Enabled

NVIDIA kernel modules installed via apt are not signed. If Secure Boot is on, the kernel rejects
the module on reboot and panics. The script has no check for this.

Fix: Exit if the Secure Boot is on.

---

2. Runfile Installed Driver Not Detected

The script checks dpkg -l to detect an existing driver. This misses drivers installed via
a .run file since dpkg has no record of them. The check passes, a second driver is installed
on top, and both conflict on reboot causing a panic.

The runfile driver is a perfectly valid installation — the script should use it if it meets the
minimum version requirement for the detected GPU, not install over it.

Fix: Read /proc/driver/nvidia/version to detect the runfile driver and extract its
version. If the version meets the version required for the detected GPU architecture, skip driver
installation. Only warn if the existing version is too old — do not uninstall or replace it.

---

3. DKMS Build Failure Not Checked Before Reboot

The NVIDIA driver compiles a kernel module on the machine during install via DKMS. This can
fail silently if kernel headers are missing or the driver version is incompatible with the
running kernel. The script reboots regardless, the kernel tries to load a module that was
never built, and panics.

Fix: After driver install, run three checks in before starting the reboot , exit if any of these fail :

1. Check dkms status for a successful installed entry for the running kernel. If the build failed, exit the script.
2. Run modprobe nvidia and verify via lsmod that the module is actually loaded in the current session. A successful DKMS build does not guarantee the module loads — it can still fail due to conflicting kernel flags or a mismatched ABI.
3. Run nvidia-smi to confirm the driver is functional end-to-end. The module loading but nvidia-smi failing indicates a partially broken driver install.

All three must pass before the reboot countdown starts. If any fail, the script exits with a message.

---

Summary

#	Issue	Impact	Fix
1	Secure Boot blocks unsigned NVIDIA module	Kernel panic on reboot	Detect → exit with message to disable in BIOS
2	Runfile driver invisible to dpkg	Kernel panic on reboot	Detect version → use if sufficient
3	DKMS build not verified before reboot	Kernel panic on reboot	Check dkms status + modprobe + lsmod + nvidia-smi → exit if any of these fail