vglrun fails with BadValue

[Jason Edgecombe]

Hi everyone,

I'm having trouble setting up virtualGL. I'm running a server with two graphics card. I'm using the integrated graphics (Display :0) for normal X and an AMD card (Display :1) for virtualGL.

I configured the 2nd X server using the command commands:

% /opt/VirtualGL/bin/vglserver_config -config +s +f -t

% Xorg :1 -configure -isolateDevice 'PCI:3:0:0'

# xorg written as /root/xorg.conf.new

I put the command "Xorg :1 -config /root/xorg.conf.new" in a systemd service.

Display :1 is headless. I can set "DISPLAY=:1" and run xterm without any output. Taking a screenshot of display :1 confirms that the program displayed.

a simple test of glxgears with vglrun fails for me.

% vglrun +v -d :1 glxinfo

[VGL] Shared memory segment ID for vglconfig: 6193165

[VGL] VirtualGL v2.6.1 64-bit (Build 20190101)

name of display: :101.0

[VGL] Opening connection to 3D X server :1

[VGL] Using Pbuffers for rendering

failed to create drawable

X Error of failed request:  BadValue (integer parameter out of range for operation)

  Major opcode of failed request:  53 (X_CreatePixmap)

  Value in failed request:  0x1e

  Serial number of failed request:  33

  Current serial number in output stream:  34

I'm connected to the server using fastX. The goal is to use vglrun to do 3D rendering on display :1 and have the output display in my remote FastX session (display :101). I'm not using vglconnect, because the X application and graphics card are in the same machine.

I'm running on RHEL 7.6 with the RedHat-provided AMD drivers, because the proprietary drivers caused the server to crash every 5 minutes.

Any help is appreciated.

Thanks,

Jason

[DRC]

Things to try:

- Run ‘/opt/VirtualGL/bin/glreadtest -pbo’ directly on Display :1. That will validate whether Pbuffers are working properly.

- Run ‘/opt/VirtualGL/bin/glxspheres64 -i’ directly on Display :1 and take a screenshot to verify that the application is displaying properly. That will validate whether OpenGL is working properly on the 3D X server.

- Try displaying glxgears with vglrun to :0 without fastX, rather than to :101, and see if that works.

- Try displaying glxspheres64 with vglrun rather than glxgears, and see if that works (NOTE: in VirtualGL land, glxspheres/glxspheres64 is *always* the baseline test, not glxgears.)

The error message would suggest a problem with the 2D X server, since that is where the X_CreatePixmap request is being sent. It may be that glxgears (NOTE: you said you were trying to run glxgears, but your command line showed glxinfo— please clarify) doesn’t work but other OpenGL applications that don’t use Pixmaps do, so that’s why I suggested running glxspheres instead.

In general, you’re not likely to get good performance with Red Hat-supplied drivers, and I echo your experience with the AMD proprietary drivers having either poor stability or compatibility. I wish AMD would prove me wrong about that, but right now, both of my AMD GPUs are stuck on RHEL 6 for compatibility reasons. Meanwhile the nVidia GPUs I have that are of similar or older vintage continue to be supported on the latest Linux distributions. I don’t want VirtualGL to be an nVidia-only solution, but the reality is that most serious VGL users are on nVidia hardware, because there seems to be a significantly higher probability that nVidia hardware and drivers will actually work properly with VGL.

--
You received this message because you are subscribed to the Google Groups "VirtualGL User Discussion/Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to virtualgl-use...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/virtualgl-users/c9c7263c-3d16-4d14-947c-63e9fae522fc%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Jason Edgecombe]

Hello,

Thanks for the suggestion on using Nvidia. An Nvidia card worked much better. 

In case anyone needs them, here are my notes on getting VirtualGL to work:

Thanks,

Jason

==============================================================================

## These instructions are my rough notes. I still need to do a wipe and rebuild to verify and clean up the instructions.

## These instructions are tailored for installing Virtgual GL on a server running RHEL 7.6 with two graphics cards. Xorg DISPLAY :0 is an integrated Matrox graphics card that is not suitable for use with VirtualGL. The second video card is an Nvidia M4000. We set up a second Xorg instance running on DISPLAY :1 and use that for VirtualGL. Display :1 is running NOTHING, not even lightdm/gdm.

# My efforts with an AMD card failed. The 18.50 version of the AMD proprietary driver caused the kernel to crash 5 minutes after boot. Using the RedHat-supplied drivers would not work with Virtual GL.

# all commands should be run as root. for the most part, these should just be copy and paste into a command-line (bash)

# lightdm is probably not needed. The server can run the multi-user target without a GUI.

yum -y install lightdm

systemctl stop gdm

systemctl disable gdm

systemctl enable lightdm

systemctl start lightdm

systemctl isolate multi-user.target

# We need both 32bit and 64bit virtualGL if the application is 32bit.

rpm -e VirtualGL --allmatches

wget -O /tmp/VirtualGL-2.6.1.x86_64.rpm https://sourceforge.net/projects/virtualgl/files/2.6.1/VirtualGL-2.6.1.x86_64.rpm/download

yum -y localinstall /tmp/VirtualGL-2.6.1.x86_64.rpm

wget -O /tmp/VirtualGL-2.6.1.i386.rpm https://sourceforge.net/projects/virtualgl/files/2.6.1/VirtualGL-2.6.1.i386.rpm/download

yum -y localinstall /tmp/VirtualGL-2.6.1.i386.rpm

##

# install the nvidia driver from elrepo

# Remove ocl-id because it conflicts with the nvidia driver because of OpenCL

yum -y remove ocl-icd.x86_64

nvidia-detect | grep -qs kmod && yum -y install $(nvidia-detect)

# lammps will be reinstalled by puppet on the next puppet run

# reboot for the nvidia drives to detect everything.

reboot

# Grab the bus ID for the nvidia card. this should similar to "PCI:1:0:0". lspci can provide the same info, but the format is slighly different. nvidia-xconfig shuold provide a format that is compatible with "Xorg -isolateDevice"

bus=$(nvidia-xconfig  --query-gpu-info|grep PCI|awk '{print $4}')

####

### Configure a new Xorg server on Display :1. Configuration is two-stage, because nvidia-xconfig assumes DISPLAY :0 and doesn't allow you to override it on the command line.

Xorg :1 -configure -isolateDevice $bus

# xorg written as /root/xorg.conf.new

# Run "Xorg :1 -config /root/xorg.conf.new" and check for errors. Ctrl-C to cancel

cp -f /root/xorg.conf.new /etc/X11/xorg.1.conf

# Add the magic nvidia bits to xorg.conf

nvidia-xconfig -a --allow-empty-initial-configuration --busid=$bus --use-display-device=none --virtual=1920x1200 -c /etc/X11/xorg.1.conf

### Set up a systemd service to manage Xorg Display :1

### begin headlessx.service ##### Just copy and paste into bash

cat > /etc/systemd/system/headlessx.service <<EOF

[Unit]

Description=Headless X server for VirtualGL

After=network.target

[Service]

Type=simple

ExecStart=/bin/bash -c "Xorg :1 -config /etc/X11/xorg.1.conf"

Restart=on-failure

[Install]

WantedBy=multi-user.target

EOF

### end headlessx.service #####

systemctl daemon-reload

systemctl enable headlessx

systemctl restart headlessx

#configure VirtualGL

systemctl stop lightdm

# configure virtualGL to allow all users. disable XTEST

/opt/VirtualGL/bin/vglserver_config -config +s +f -t

# refresh X display permissions as changed by vglserver_config

systemctl stop headlessx

rmmod nvidia_drm nvidia_modeset nvidia

systemctl start headlessx

#rmmod nvidia_drm nvidia_modeset nvidia

#systemctl isolate graphical.target

reboot

######## example commands below for running VirtualGL as a normal user. This is intended for use with the StarNet FastX.

#Testing VirtualGL. Login to fastX. Tested by RDP to mws939, then fastX to lxs-hfss2

#vglrun -c proxy +v /opt/VirtualGL/bin/glxspheres64

VGL_DISPLAY=:1 vglrun -c proxy +v /opt/VirtualGL/bin/glxspheres64

# To accelerate an entire FastX session, login to the FastX admin web backend and add a new bookmark. For Mate, the command should be the following:

/bin/bash -c 'VGL_DISPLAY=:1 vglrun -c proxy +v mate-session'

To view this discussion on the web visit https://groups.google.com/d/msgid/virtualgl-users/EEBD0622-7594-4830-8615-77FE77C8D4F4%40virtualgl.org.