@coconutit I made a mistake with the command, it should be:

yum update --enablerepo=xcp-ng-ndinh1

(updated up top)

@coconutit Did you run thie command?

wget https://koji.xcp-ng.org/repos/user/8/8.3/xcpng-users.repo -O /etc/yum.repos.d/xcpng-users.repo

@mickwilli You can extract the ISO, run the following command:

bcdedit /store <extracted ISO path>\Boot\BCD /set debug on
bcdedit /store <extracted ISO path>\Boot\BCD /dbgsettings serial debugport:1 baudrate:115200

Then repack the ISO.

You can also create a new ISO using this guide: https://learn.microsoft.com/en-us/windows-hardware/manufacture/desktop/winpe-create-usb-bootable-drive then run the two commands above on media\Boot\BCD before you run MakeWinPEMedia.

Hello all, the new test build qemu-4.2.1-5.2.12.2~fixvwc2.1.xcpng8.3 is now available.

If you have tested the fix before, you can update with this command:

yum update --enablerepo=xcp-ng-ndinh1

@umbradark Hello, there are new fixed drivers in the latest bnxt_en driver disk. Could you specify modprobe.blacklist=bnxt_en then apply the linked driver disk to see if it works?

@francesco-maria I haven't determined what exact conditions would trigger the issue. It seems to only happen on new installations for now, but it could very well change in the future. Though as long as you keep the Xen drivers installed, things should be OK.

Hi all, here's a pre-build of the fix for anyone who wants to test (8.3 only):

wget https://koji.xcp-ng.org/repos/user/8/8.3/xcpng-users.repo -O /etc/yum.repos.d/xcpng-users.repo
yum update --enablerepo=xcp-ng-ndinh1

@Henrik I see, that's unfortunately a known issue when upgrading from older versions. I think the problem should have been fixed when upgrading from 9.4.1 to a newer version. So I'd appreciate feedback on that too.

@mickwilli Do you know how to create custom Windows PE ISOs? It might be possible to put the boot kernel in the ISO into debugging mode, then using the Connecting to guests using serial console guide to collect a crash analysis.

@KPS said in Updating XenTools on Windows 2022 - duplicate NIC:

When I update XenTools (Citrix-ones) from e.6. 9.3.2 to 9.4.2, there are two NICs visible in Windows 2022 after the first reboot.

After the second reboot, there is only one (as in XOA).

This is normal for Citrix tools due to the way they are upgraded.

@Henrik said in Updating XenTools on Windows 2022 - duplicate NIC:

@KPS i have seen the same on WS 2022. when upgraded to version 9.4.1 and 9.4.2 create a new nic and then the old disappears. and if static IP is set it also loses the gateway for the ip setting in some cases.

Does the gateway setting loss still happen when upgrading from 9.4.1 to 9.4.2?

The cause has been found, and a fix is under way. Stay tuned.

I've narrowed down the problem to a set of patches. Will continue to diagnose.

@Andrew Still, having an unsupported CPU makes it somewhat difficult to troubleshoot the issue, especially when it's a recent ISO. SYSTEM_THREAD_EXCEPTION_NOT_HANDLED is sometimes associated with execution of unsupported instructions (though I don't know if that's the case here). For instance, there's a poke-of-death on some newer Intel CPUs related to the Architectural LBR feature (https://docs.xcp-ng.org/troubleshooting/windows-pv-tools/#windows-bug-check-0x3b-system_service_exception-on-systems-with-newer-intel-cpus).

@flakpyro It might be related to the boot-initialization logic of the NVMe driver. If that hypothesis is correct, then uninstalling Xen tools will give you the same problem once the emulated NVMe reappears.

@mickwilli Forget that, I just realized E5-2660 v2 is way too old for Windows 10. See the supported processors.

Could you provide a crash dump?

@flakpyro Confirmed the issue. It looks like an incompatibility between Windows's NVMe driver and QEMU's NVMe emulation, but I'm not sure what exactly. The NVMe controller and namespace are detected just fine, but Windows's NVMe driver refuses to start them.

@mickwilli What host hardware do you have and which ISO did you use for the Windows installation? I assume you used the default Windows 10 template?

@mickwilli Could you provide the host hardware configuration and ISO being used?

@okynnor IIRC, it would take a reboot before the problem showed up. I'd check after a reboot if XenSvc wasn't stuck in the starting state.