VMware Cloud Community
marietto2008
Enthusiast
Enthusiast

The error 43 that I get when I try to pass my graphic card in a ESXI vm can be caused by the audio sub device reset interrupts ?

Hello.

I'm trying to fix the error 43 that comes out when i try to passthrough my graphic card RTX 2080 ti as well as the GTX 1060 with esxi,doesn't matter which version,I tried a lot of versions,6.5,6.7,7 to Windows 10. I've investigated more and I tried to repeat the same error in different circumstances to understand what are the repetitive patterns,and I think that I found something interesting. I think that I have found a lot of similarities between a XEN vm where I passed the GTX 1060 and a ESXI vm where I passed the GTX 1060 as well as the RTX 2080 ti. But but let's go slow. As first thing I created a XEN vm (using the EFI boot loader) and as I said I have passed the GTX 1060 adding these parameters in the XEN cfg file :

nano win10-gtx-headless.cfg

builder='hvm'

bios='ovmf'

bios_override = '/usr/lib/xen-4.11/boot/ovmf.bin'

memory = 4096

this is what it says when it starts :

root@ziomario-z390aoruspro:/etc/xen# ./create-win10-gtx-headless.sh

Parsing config from win10-gtx-headless.cfg

got a tsc mode string: "default"

libxl: error: libxl_pci.c:1162:libxl__device_pci_reset: Domain 0:write to /sys/bus/pci/devices/0000:02:00.0/reset returned -1: Inappropriate ioctl for device

libxl: error: libxl_pci.c:1167:libxl__device_pci_reset: Domain 0:The kernel doesn't support reset from sysfs for PCI device 0000:02:00.1

lspci :

02:00.0 VGA compatible controller: NVIDIA Corporation GP106 [GeForce GTX 1060 3GB] (rev a1)

02:00.1 Audio device: NVIDIA Corporation GP106 High Definition Audio Controller

In addition to this,I have got an nvidia driver and I have applied the kvm patcher to it ; I have signed the driver again and I have disabled the driver signature enforcement. The first time that the vm starts,I don't see the error 43,BUT,the external monitor does not turn on.

11-2020-09-24_11-21-03.png

the driver is properly signed also with XEN and it does not show the error 43. But when I reboot the VM,it will give the error 43. Check below :

2-2020-09-24_11-15-34.png


ok. this is the same behaviour that I see when I turn on a VM with esxi ! Someone suggested that I should use EFI instead of the old legacy MBR. I did it,but the problem is still there. Good. Now,I did the same with my Geforce 2080 ti and,as u can see below,there isn't any error 43 :

3-2020-09-24_11-15-59.png


this is what happens when I boot the vm with the working 2080 ti :

root@ziomario-z390aoruspro:/etc/xen# ./create-win10-rtx-body.sh

Parsing config from win10-rtx-body.cfg

got a tsc mode string: "default"

libxl: error: libxl_pci.c:1167:libxl__device_pci_reset: Domain 0:The kernel doesn't support reset from sysfs for PCI device 0000:01:00.1

libxl: error: libxl_pci.c:1167:libxl__device_pci_reset: Domain 0:The kernel doesn't support reset from sysfs for PCI device 0000:01:00.2

libxl: error: libxl_pci.c:1167:libxl__device_pci_reset: Domain 0:The kernel doesn't support reset from sysfs for PCI device 0000:01:00.3

lspci :

01:00.0 VGA compatible controller: NVIDIA Corporation TU102 [GeForce RTX 2080 Ti] (rev a1)

01:00.1 Audio device: NVIDIA Corporation TU102 High Definition Audio Controller (rev a1)

01:00.2 USB controller: NVIDIA Corporation TU102 USB 3.1 Host Controller (rev a1)

01:00.3 Serial Bus controller [0c80]: NVIDIA Corporation TU102 USB Type-C UCSI Controller (rev a1)

It works for the 2080 ti (it means that I don't have any error 43 after the reboot and I can use an externals screen); it does not work for the 1060 and for esxi.

Hypothesis : the reasons why I can't pass my 1060 as well as my 2080 ti with esxi are the same of the reasons why the gtx 1060 does not work inside a XEN vm because I see an identical behavioral pattern.

But it's not ended here. I see another repetitive pattern.

Check the upper and right corner of the screenshot where it runs the VM with the 1060 and the right corner of the screenshot where it runs the VM with the 2080 ti. In the first one you will see a red exclamation mark. On the second one there isn't. Well,I see the red exclamation mark also on the upper and right corner of a windows 10 / esxi VM. And this is another identical behavior.

Final conclusion : since I see some relevant identical patterns between the two situations,I suspect that the technical reason why I get the error 43 on a Windows 10 + esxi VM could be included inside these error messages :

libxl: error: libxl_pci.c:1162:libxl__device_pci_reset: Domain 0:write to /sys/bus/pci/devices/0000:02:00.0/reset returned -1: Inappropriate ioctl for device

libxl: error: libxl_pci.c:1167:libxl__device_pci_reset: Domain 0:The kernel doesn't support reset from sysfs for PCI device 0000:02:00.1

I feel that the file rouge between the two situations is the audio device integrated with the other sub components of the graphic card. For some reason ESXI does not support  some feature of the audio sub-device included on the 1060 and on the 2080 ti and I think in a lot of different nvidia consumer graphic cards.

Take in consideration that I always pass all the sub-components of the graphic cards. I tried to don't pass the audio device,but I didn't fix anything.

Why I see the red exclamation mark on the VM with the 1060 and with the VM with ESXI and I don't see it on the VM with the 2080 ti ? Below u can see some screenshots that shows that the behavior of a windows 10 vm with esxi is the same :

19-Screenshot_1.png

20-Screenshot_1.png

21-Screenshot_1.png

Tags (1)
0 Kudos
0 Replies