Kernel panic basics on Linux

What you will achieve

Recognise kernel panics, capture evidence from journald and dmesg, and apply common fixes — bad modules, full disks, failing RAM.

1) What it looks like

Frozen console, stack trace on screen, caps lock blinking — system completely dead except magic sysrq.

2) Capture logs after reboot

journalctl -k -b -1 --no-pager | tail -100
sudo dmesg -T | grep -i panic

3) Common causes

• Recent kernel upgrade — boot previous entry from GRUB Advanced options.
• Bad DKMS module (NVIDIA, VirtualBox) — remove module or use LTS kernel.
• Filesystem corruption — boot live USB, run fsck.
• RAM — run memtest86+ from GRUB.

4) Blacklist module temporarily

echo "blacklist nouveau" | sudo tee /etc/modprobe.d/blacklist-nouveau.conf
sudo update-initramfs -u

Verify

Stable uptime after fix: uptime and no new panic lines in journalctl -k.

5) kdump for post-mortem

sudo apt install linux-crashdump
# captures vmcore on next panic

6) Kernel parameters to test

intel_idle.max_cstate=1
nouveau.modeset=0
iommu=soft

Add one at a time at GRUB — binary search which option stabilises boot.

7) Hardware checklist

• Reseat RAM and power cables.
• Test with minimal USB peripherals.
• Check CPU thermals — sensors after successful boot.

When to escalate

Repeating panics on stock kernel with memtest clean point to failing SSD or motherboard — software fixes hit diminishing returns.

8) mce (machine check) errors

grep mce /var/log/kern.log

Hardware CPU/RAM errors precede panics — replace hardware when MCE logs accumulate.

Prerequisites

Console or serial log capture. Previous kernel in GRUB. Live USB for fsck. memtest86+ on USB optional. Recent hardware/software change timeline (driver, RAM, kernel update).

Preserve panic log

journalctl -k -b -1 --no-pager > ~/last-boot-kernel.log

sysctl panic settings

kernel.panic=10
kernel.panic_on_oops=1

Auto-reboot after panic on headless servers — paired with kdump for analysis.

Silent boot hide panic

Remove quiet splash from GRUB_CMDLINE_LINUX_DEFAULT to see full panic oops on screen — photograph or log for vendor bug report with exact call trace symbol names.

kexec rapid reboot

kexec loads new kernel without full firmware init — faster recovery loop when testing kernel params but does not fix hardware panics.

vendor support bundle

Before RMA RAM or SSD collect sosreport Ubuntu or redhat-support-tool on RHEL — vendor wants full hardware inventory attached to panic ticket.

netconsole remote panic log

Configure netconsole to send panic oops to syslog server on another host — captures oops when local disk too corrupt to write journal after reboot.

Related guides

• Recover root password (single-user mode)