Problem with Quadro K3100M KVM Passthrough

Graphics / Linux Linux
1

Hi,

We have an HPE blade system workstation, WS480 Gen9 with the Nvidia 6x GPU side car that was working with XEN Server sharing out 6 virtual systems.

In trying to save some money, we decided to setup RHEL 7 Server with KVM and passthough the graphics cards to the systems and install RHEL 7 on the guests.

We have a proof of concept system up and running and lspci shows the graphics card and we can install the driver, but when we try to start X using RGS, we get the following error in /var/log/Xorg.0.log:

485.933] (EE) NVIDIA(0): Failed to allocate software rendering cache surface: out of
[   485.934] (EE) NVIDIA(0):     memory.
[   485.934] (EE) NVIDIA(0):  *** Aborting ***
[   485.940] (EE) 
[   485.940] (EE) NVIDIA: A GPU exception occurred during X server initialization(EE) 
[   485.940] (EE) 
[   485.940] (EE) Please also check the log file at "/var/log/Xorg.0.log" for additional information.
[   485.940] (EE) 
[   485.940] (EE) Server terminated with error (1). Closing log file.

We are also getting the following in dmesg:

8.632955] NVRM: Xid (PCI:0000:00:09): 69, Illegal Class Error: ChID 0008, Class 00000000, Offset 0000010c, Data 00000000
[    8.970383] NVRM: GPU at PCI:0000:00:09: GPU-3f856369-9e87-188e-4e26-f6f5c2997711
[    8.971100] NVRM: Xid (PCI:0000:00:09): 69, Illegal Class Error: ChID 0008, Class 00000000, Offset 0000010c, Data 00000000
[   12.872943] random: crng init done

we are installing the 390.87 driver and here is the SMI output from the system:

# nvidia-smi 
Wed Sep 26 12:04:07 2018       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 390.87                 Driver Version: 390.87                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Quadro K3100M       Off  | 00000000:00:09.0 Off |                  N/A |
| N/A   37C    P0    13W /  N/A |      0MiB /  4037MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

Any help would be great!

Let me know if you need any more info!

Thanks!
Joe Giles
nvidia-bug-report.log-ceerws0305a.gz (75 KB)

2

Did you check with another kernel (4.14+) on the guest?
Please run nvidia-bug-report.sh as root and attach the resulting .gz file to your post. Hovering the mouse over an existing post will reveal a paperclip icon.

3

Hi Generix,

I will attach to the original message. I did try another kernel on both the host and the guest.

Here is some additional information.

I get this on the terminal of the guest:

[    8.734794] NVRM: GPU at PCI:0000:00:09: GPU-3f856369-9e87-188e-4e26-f6f5c2997711
[    8.735463] NVRM: Xid (PCI:0000:00:09): 69, Illegal Class Error: ChID 0008, Class 00000000, Offset 0000010c, Data 00000000
[    9.099207] NVRM: GPU at PCI:0000:00:09: GPU-3f856369-9e87-188e-4e26-f6f5c2997711