linux-freeze-triagelisted

# /linux-freeze-triage Find the real cause of a Linux freeze / black-screen by evidence, not by guessing from the symptom. ## Master TOC - [Core discipline](#core-discipline) - [Step 0: capture before you guess](#step-0-capture-before-you-guess) - [Step 1: rule out sleep/suspend](#step-1-rule-out-sleepsuspend) - [Step 2: read the freeze moment](#step-2-read-the-freeze-moment) - [The usual suspects (ranked)](#the-usual-suspects-ranked) - [Recovery without losing work](#recovery-without-losing-work) - [Bundled scripts](#bundled-scripts) - [War story](#war-story) ## Core discipline - Do NOT commit to a hypothesis from the symptom alone. A blinking power LED is NOT proof of sleep (it is usually the monitor's no-signal standby). "Happens after idle" is NOT proof of a DPMS/suspend bug. - Instrument and CAPTURE the next occurrence; let evidence pick the cause. - A log-churning service (e.g. a unit crash-looping every ~2s) rotates journald so fast that the real freeze logs roll off — find and fix it first, or you will never see the evidence. - Same failure on BOTH Xorg and Wayland => the cause is BELOW the display server (driver/kernel/hardware); do not blame the compositor. ## Step 0: capture before you guess Run `diagnose.sh` (read-only battery). Arm `freeze-watch.sh` (near-zero-cost: nice 19 / ionice idle / 60s) so the NEXT freeze is recorded — its last log line is the state at death: - `NVSMI_FAIL` => GPU fell off the bus / driver broken. - `DRIVER_MISMATCH ... REBOOT-NEEDED`