Hi!
I have iommu enabled in the system, so my plan is to pass my GPU to VM and test it there, first for reproduction and then solution. I saw this issue before I updated the kernel (now running latest) also... yes the ps output is strange, but I can see CPU usage in top and also as system load.
Also running the original script once again shows
Code:
1000 0.0 0.0 00:00:10 \_ sshd
1000 0.0 0.0 00:00:04 \_ virt-ssh-helper
1000 0.0 0.0 00:00:00 systemd
1000 0.0 0.0 00:00:00 | \_ sort
1000 0.0 0.0 00:00:00 \_ sh
1000 0.0 0.0 00:00:00 \_ (sd-pam)
1000 0.0 0.0 00:00:00 | \_ ps
1000 0.0 0.0 00:00:00 | \_ head
1000 0.0 0.0 00:00:00 | \_ capture_bug.sh
1000 0.0 0.0 00:00:00 \_ bash
1000 0.0 0.0 00:00:00 \_ bash
1000 0.0 0.0 00:00:00 | \_ awk
0 2.2 0.0 00:00:04 \_ kworker/3:1-events
0 1.9 0.0 00:00:04 \_ kworker/u8:2-events_power_efficient
0 1.1 0.0 00:07:36 \_ kworker/3:3-events
0 0.5 0.0 00:07:21 \_ vhost-39281
0 0.3 0.2 00:04:48 ovs-vswitchd
0 0.2 0.0 00:04:06 libvirtd
0 0.2 0.0 00:03:05 \_ ksmd
0 0.1 0.0 00:02:34 \_ l2arc_feed
0 0.1 0.0 00:02:14 \_ z_wr_iss
0 0.1 0.0 00:01:34 \_ z_wr_int
0 0.1 0.0 00:00:00 \_ kworker/2:1-events
0 0.1 0.0 00:00:00 \_ kworker/0:2-events
Previous logs reported ~12% of usage for those processes, so that's something 
I'll try to prepare a VM with PCI passthrough and will see what will happen.
