Noticeable typing lag in Linux VM terminals since v16.2 upgrade on Linux host

[ajgringo619]

Upgraded my EndeavourOS host to version 16.2 yesterday. Now I'm noticing a lag when typing in any VM terminal window - Arch, Ubuntu, openSUSE are experiencing the same thing. I'm using open-vm-tools, which has never been a problem before. I have not changed any settings on the VMs since their creations: 8GB RAM, 4GB vGPU, 8 vCPUs.

Anyone else seeing this? I haven't tried with my Windows 10 host yet (also on v16.2).

UPDATE: working great with Windows 10 (Home) host.

[goldeneye_007]

I've had a suspicion for awhile that it is a video driver problem. The fact that when glx gears runs it goes away dovetails with my observation that you can reduce lag by constantly moving the mouse which tells me it is an interrupt issue. I wonder if there is a video mode that can dumb it down enough to where the lag goes away? Like maybe running -nomodeset at boot time? Or perhaps reducing/increasing or matching to the host the video resolution of the guest? Just guessing at this point since its been awhile since I have run into the issue and don't have time at the moment to run tests.

[vmwareuser11321]

Got the same issue. Looking at the log file, I see thousands of entries like these:

---

2023-09-29T13:09:15.244Z In(05) vcpu-0 VIDE: (0x170) Unexpected IN for cmd 0xa3 len 2
2023-09-29T13:09:15.244Z In(05) vcpu-0 VIDE: (0x170) Unexpected IN for cmd 0xa3 len 2
2023-09-29T13:09:15.244Z In(05) vcpu-0 VIDE: (0x170) Unexpected IN for cmd 0xa3 len 2
2023-09-29T13:09:15.244Z In(05) vcpu-0 VIDE: (0x170) Unexpected IN for cmd 0xa3 len 2

---

Does anyone else also have these entries?

[tbn-mm]

That's what I'm doing now:

Start glxgears,
set Always on Top (no effect if not visible),
minimize to about 1 pixel.

Lag is gone. Costs a little bit of energy.

Here is a script, that can be placed at the end of ~/.profile

sudo apt-get install mesa-utils wmctrl

#!/bin/bash
pgrep glxgears > /dev/null
PGREP_GLXGEARS=$?
if [ $PGREP_GLXGEARS == 1 ]; then
  DISPLAY=:0 sleep 5 && pgrep glxgears || nohup glxgears -geometry 1x1+0+0 > /dev/null &
  nohup sleep 10 && wmctrl -r "glxgears" -b add,above &
fi

 Tested on Ubuntu 22.04

[goldeneye_007]

Fairly certain at this point that it is a virtual video driver issue. Maybe it is polling for a power saving feature. Anyone played around with disabling power management features? The fact you can ssh into a machine indicates that there must be an advanced function in the desktop environments that is triggering it.

[JMA100]

Using the nvidia control panel, I switched to performance mode and validated I was in a higher clock state for both the gpu and its memory.  It had no effect inside the virtual machine.

Goldeneye_007 - I'm not sure if you caught some of the earlier discussion - if you turn off GPU acceleration, then there's no lag in the GUI.  It appears to be specific to running with the "Accelerate 3d graphics" option on.  I do agree with you that it's related to the 3d accelerated code path in the vmware svga driver. 

I'm guessing it's interrupt related.  (I think someone else suggested that earlier in the discussion.)

[omglolwtf1337]

Disabling VBS (Memory Integrity) in Windows completely fixed the issue for me. Guide how to make it - https://www.digitaltrends.com/computing/how-to-disable-vbs-windows-11/

VMWare 17.5

Host: Windows 11

Tested on 2 guests - ubuntu and debian based.