Just spent 25 minutes in front of a frozen screen waiting for the kernel to kill something. Had to restart with holding a power button, losing some work. Not even TTY shortcuts worked.
System should preserve at least some megabytes of memory for itself so if Reddit suddenly decides it needs 600 MiB for one Firefox tab (true story) you could at least close the browser window without putting down the entire system.
Or at least record what apps were open during the last boot.
When a system hangs like that it’s usually caused by a graphics (or other) kernel driver with direct hardware access crashing. You would have to identify the driver causing the crash and file a bug report upstream with the appropriate development team.
600MB for a browser tab isn’t that unheard of nowadays.
It went over 98% RAM usage right before freezing. So I believe memory is a problem.
It is normal for a linux system to maximise its use of memory.
For example file system data is cached after being read.
You would need to look at how the memory is being used to know if you do not have enough memory for the work load you are running.
But as you are see OOM killer run that suggests either not enough memory or an app with high memory usage or a leak.
The kernel has algorithms to handle memory pressure efficiently.
Using the OOM killer is the last resort when every other strategy has failed.
The journal will tell you which process the OOM killer killed.
What app was killed?
The system froze around 18:14 and remained frozen for 25 minutes until I initiated a crash reboot in 18:38. Here’s what logs (listed from newer to older) have to say.
дек 06 18:39:06 fedora kernel: BIOS-e820: [mem 0x0000000000000000-0x000000000009efff] usable
дек 06 18:39:06 fedora kernel: BIOS-provided physical RAM map:
дек 06 18:39:06 fedora kernel: Command line: BOOT_IMAGE=(hd0,gpt4)/vmlinuz-6.11.7-300.fc41.x86_64 root=UUID=6df17643-5906-4857-b7cb-6b7183f3617e ro rootflags>
дек 06 18:39:06 fedora kernel: Linux version 6.11.7-300.fc41.x86_64 (mockbuild@6c49ea81e72a4908b02fd96c22f92edf) (gcc (GCC) 14.2.1 20240912 (Red Hat 14.2.1-3>
-- Boot 0fc0a63559a1425a8ed7f378168f53f9 --
дек 06 18:13:30 fedora pipewire[2854]: spa.alsa: hw:sofhdadspp: (0 suppressed) snd_pcm_avail after recover: Broken pipe
дек 06 18:13:08 fedora geoclue[3052]: Failed to query location: Query location SOUP error: Not Found
дек 06 18:09:17 fedora PackageKit[349977]: resolve transaction /17384_edbeaccb from uid 1000 finished with success after 2714ms
дек 06 18:09:16 fedora systemd[1]: fwupd.service: Consumed 1.670s CPU time, 36.1M memory peak.
Look at the logs before the hang for clues.