On 26.09.2023 16:21 +0300, Yu Kuai wrote: > This means cpu is busy with something, in this case you must use top or > perf to figure out what all your cpus are doing, probably issue io and > handle io interrupt. OK, here is the last output of 'perf top -d 1' before system froze: 4.00% [kernel] ledtrig_disk_activity 3.86% [kernel] led_trigger_blink_oneshot 1.93% [kernel] read_tsc 1.68% perf hist_entry__sort 1.62% [kernel] menu_select 1.57% [kernel] psi_group_change 1.19% [kernel] native_sched_clock 0.96% [kernel] scsi_complete 0.94% [kernel] __raw_spin_lock_irqsave 0.80% perf perf_hpp__is_dynamic_entry ... And here is the last output of 'top': load average: 1.48, 0.90, 0.38 %Cpu(s): 0.1 us, 0.1 sy, 0.0 ni, 93.5 id, 6.3 wa, 0.0 hi, 0.1 si, 0.0 st MiB Mem: 32005.3 total, 29138.3 free, 1327.0 used, 1540.0 buff/cache MiB Swap: 16382.0 total, 16382.0 free, 0.0 used. 30244.5 avail Mem S %CPU COMMAND S 0.7 [mdX_raid1] S 0.7 cinnamon --replace I 0.3 [kworker/u64:2-events_unbound] I 0.3 [kworker/u64:6-events_freezable_power_] S 0.3 [gfx_low] S 0.3 /usr/bin/containerd S 0.3 /usr/lib/xorg/Xorg -core :0 ... S 0.3 /usr/libexec/gnome-terminal-server R 0.3 top D 0.3 fstrim /home ... I think, there are only two possibilities: either CPU is not that busy, or it gets busy very quickly, so we can not see it like that. I have no experience in kernel debugging. Maybe someone knows, how I can get more accurate data when system freezes? On 26.09.2023 16:21 +0300, Yu Kuai wrote: > I'm not sure what 'disk activity indicator goes off' means It means, the LED on my computer case, indicating disk activity, goes off. According to 'perf' output, LED control is the largest contributor to CPU load. :)
