The problem seems to persist for my installation. The Ubuntu VM with passthrough Nvidia GTX 1050Ti and a USB 3.0 hub runs fine for hours, but never closes down properly. 

12 minutes ago, pederm said:

The problem seems to persist for my installation. The Ubuntu VM with passthrough Nvidia GTX 1050Ti and a USB 3.0 hub runs fine for hours, but never closes down properly. 

 

Ok, I'm going to move this to a separate topic.

17 minutes ago, pederm said:

The problem seems to persist for my installation. The Ubuntu VM with passthrough Nvidia GTX 1050Ti and a USB 3.0 hub runs fine for hours, but never closes down properly. 

 

After initiating shutdown on that VM how far does it get?  Meaning does the ubuntu OS seem to shutdown correctly, and then only qemu process never exits?

 

Helpful to post diagnostics in this state.

The Ubuntu VM closes down, all looks ok. But in htop, the qemu process is marked with a red D in the S column, and a new start of the VM gives an error stating the VM is already running. 

I have recreated the problem, but the diagnostics file does not show much as far as I can judge.

asus-diagnostics-20171230-1137.zip

Edited December 30, 20178 yr by pederm
Changed libvirtd to qemu

On 12/30/2017 at 2:58 AM, pederm said:

The Ubuntu VM closes down, all looks ok. But in htop, the qemu process is marked with a red D in the S column, and a new start of the VM gives an error stating the VM is already running. 

I have recreated the problem, but the diagnostics file does not show much as far as I can judge.

asus-diagnostics-20171230-1137.zip

 

Please retest with -rc19.

Ran an Ubuntu VM for more than 30 minutes and closed it down. No trouble, it worked fine and the qemu process exited as expected. Afterwards I was able to start a new session.

 

I will runn this session for a couple of hours and hopefully be able to close it with no trouble also.

After several hours of VM usage, no trouble closing it down. The qemu issues indeed seems to have been fixed.

44 minutes ago, pederm said:

After several hours of VM usage, no trouble closing it down. The qemu issues indeed seems to have been fixed.

 

Nice!  Thanks for the testing, much appreciated!

Hello,

Upgraded from 6.3.5 to 6.4rc19b. My virtual machine logical CPU selection went from 8 separate selections to 4. They're now labeled cpu 0/4, 1/5, 2/6, and 3/7. Seems to be similar to an older issue reported here. I've got a new Dell PowerEdge T330 with Kaby lake based Xeon cpu. 

 

cat /sys/devices/system/cpu/*/topology/thread_siblings_list
0,4
1,5
2,6
3,7
0,4
1,5
2,6
3,7

 

 

 

NVIDIA gpu passthrough is giving error 43 after installing drivers no matter what.  I am still legacy booting unraid.  As of 6.3.5 SeaBios and Q35-2.7 VM worked with my GT610 or GT710 just fine.  As of rc19b I have tried OVMF and Seabios and the latest of both machine options in addition to Q35-2.7.  I have tried with Hyper-V on or off and with a couple of Hyper-V related XML additions.  I tried loading the VBIOS.  I can't get anything to work.  The GPU somewhat works (pink lines, shaky picture) prior to installing drivers but as soon as the drivers install and I reboot the VM the GPU gives error 43.  I have attached the XML of my most recent config.

win10.xml

17 hours ago, brhersh said:

Upgraded from 6.3.5 to 6.4rc19b. My virtual machine logical CPU selection went from 8 separate selections to 4

 

Please show a screenshot of the VM page with CPU selection.

58 minutes ago, jorj4man said:

NVIDIA gpu passthrough is giving error 43 after installing drivers no matter what.  I am still legacy booting unraid.  As of 6.3.5 SeaBios and Q35-2.7 VM worked with my GT610 or GT710 just fine.  As of rc19b I have tried OVMF and Seabios and the latest of both machine options in addition to Q35-2.7.  I have tried with Hyper-V on or off and with a couple of Hyper-V related XML additions.  I tried loading the VBIOS.  I can't get anything to work.  The GPU somewhat works (pink lines, shaky picture) prior to installing drivers but as soon as the drivers install and I reboot the VM the GPU gives error 43.  I have attached the XML of my most recent config.

win10.xml

 

Are you booting your server in UEFI mode or legacy BIOS mode (CSM)?

1 hour ago, jonp said:

 

Are you booting your server in UEFI mode or legacy BIOS mode (CSM)?

 

Legacy

I just upgraded to rc20 and now one of my VM's won't start.  I see this in my log...

 

Jan 7 18:17:57 media kernel: br0: port 3(vnet1) entered disabled state
Jan 7 18:17:57 media kernel: device vnet1 left promiscuous mode
Jan 7 18:17:57 media kernel: br0: port 3(vnet1) entered disabled state
Jan 7 18:18:46 media kernel: vfio-pci 0000:02:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=io+mem:owns=io+mem
Jan 7 18:18:46 media kernel: br0: port 3(vnet1) entered blocking state
Jan 7 18:18:46 media kernel: br0: port 3(vnet1) entered disabled state
Jan 7 18:18:46 media kernel: device vnet1 entered promiscuous mode
Jan 7 18:18:46 media kernel: br0: port 3(vnet1) entered blocking state
Jan 7 18:18:46 media kernel: br0: port 3(vnet1) entered forwarding state
Jan 7 18:18:48 media kernel: vfio-pci 0000:02:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
Jan 7 18:18:51 media kernel: br0: port 3(vnet1) entered disabled state
Jan 7 18:18:51 media kernel: device vnet1 left promiscuous mode
Jan 7 18:18:51 media kernel: br0: port 3(vnet1) entered disabled state
Jan 7 18:18:52 media kernel: vfio-pci 0000:02:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=io+mem:owns=io+mem
Jan 7 18:19:02 media kernel: vfio-pci 0000:02:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=io+mem:owns=io+mem
Jan 7 18:19:02 media kernel: br0: port 3(vnet1) entered blocking state
Jan 7 18:19:02 media kernel: br0: port 3(vnet1) entered disabled state
Jan 7 18:19:02 media kernel: device vnet1 entered promiscuous mode
Jan 7 18:19:02 media kernel: br0: port 3(vnet1) entered blocking state
Jan 7 18:19:02 media kernel: br0: port 3(vnet1) entered forwarding state

 

I've changed the video card to VNC, and it starts just fine and I've checked for any windows 10 updates and device manager issues, but no updates and no failures.  it runs fine.  VNC isn't passing the mouse thru, so I have to use keyboard shortcuts with VNC.  If I connect with splashtop desktop, it and the mouse work just fine.

 

I closed it, changed the setup to use the nVidia 550Ti video card and audio, but it just doesn't start.

 

the monitor connected to it looks like it's trying to connect, as it goes black, then blue (no input is displayed by TV), then just keeps cycling.

 

My other Windows 10 VM seems to work just fine.

 

Diagnostics attached.

media-diagnostics-20180107-1826.zip

On 1/6/2018 at 3:07 PM, bonienl said:

 

Please show a screenshot of the VM page with CPU selection.

 

Looking again, maybe it's just organized differently?

 

 

2 minutes ago, brhersh said:

 

Looking again, maybe it's just organized differently?

 

 

 

You still have eight selections.  Each of the checkboxes is for a thread, it's just showing pairing.  Unless you don't have an eight thread cpu in which case I'm wrong.

32 minutes ago, brhersh said:

 

Looking again, maybe it's just organized differently?

 

 

 

This is correct display of 4 cores + 4 threads = 8 total.

A processor with hyperthreading is displayed in pairs which makes selection of a core with associated thread easier.

On 1/6/2018 at 2:56 PM, jorj4man said:

NVIDIA gpu passthrough is giving error 43 after installing drivers no matter what.  I am still legacy booting unraid.  As of 6.3.5 SeaBios and Q35-2.7 VM worked with my GT610 or GT710 just fine.  As of rc19b I have tried OVMF and Seabios and the latest of both machine options in addition to Q35-2.7.  I have tried with Hyper-V on or off and with a couple of Hyper-V related XML additions.  I tried loading the VBIOS.  I can't get anything to work.  The GPU somewhat works (pink lines, shaky picture) prior to installing drivers but as soon as the drivers install and I reboot the VM the GPU gives error 43.  I have attached the XML of my most recent config.

win10.xml

 

Well some combination of variables and it is working.  I used a OVMF and i440fx-2.10 and set everything up on the GPU instead of using VNC, and installed drivers from the NVIDIA site (390.65) instead of through device manager, which I had been doing.