You are not logged in.
From htop, I see [kworker/0:4+events] is always using 40%+ CPU and in either D or R state. The system is still usable but with significant lag. The issue happens since boot, and rebooting does not help.
I first observed this problem when upgrading linux 5.2.14 -> 5.3.1. At that time I installed linux-hardend 5.2.18 which "fixed" the problem. But when upgrading linux-hardend 5.2.18 -> 5.3.7 this issue appers again. So now I think this might be a kernel 5.3.x problem. I've checked journalctl and dmseg, but they are almost the same on 5.2.x and 5.3.x. What should i do for next step?
Here's the diff from dmesg:
https://gist.github.com/wangqr/83d2d34e … 2e49f178fc
Last edited by wangqr (2019-11-02 04:26:17)
Offline
This issue may be related, or may be even the same as this issue: "one CPU core at 100% usage after kernel upgrade".
Offline
And I just tested linux-5.3.arch1-1 and linux-mainline-5.4rc4-1, both version has the issue.
Offline
I am observing same behaviour.
It is interesting, I can NOT see kworker is taking cca 42% of cpu in HTOP. But in TOP I can see it.
Also I did check like this:
ps -eo pcpu,pid,user,args | sort -k1 -r -n | head -10
42.9 144 root [kworker/3:1+events]
...
other processes are bellow 1.5
I am observing that on my home server. I am running monitorix there so I did check monthly chart and and load is same for last month.
At the moment I have 5.3.8-arch1-1 kernel version and I assume that that problem was with previous kernels too.
Offline
See the link above. This is a intel i915 driver bug, and reported at https://bugs.freedesktop.org/show_bug.cgi?id=112125 . It is fixed in drm-tip, hopefully it will be in kernel soon. In the meantime you can manually apply the patch and build own kernel, or just plug in a monitor if your server is running headless.
Offline