January 21, 20242 yr Looks like the issue is fixed in latest kernel 6.5.x OP here https://forum.proxmox.com/threads/nested-esxi-8-0-fails-with-amd-v-implementation-is-incompatible.119874/ Might have to wait a while before we get 6.5 😞 Edited January 21, 20242 yr by mikeyosm
January 23, 20242 yr Author That is great to hear that there is a solution in site. Now we just need to get there.
January 24, 20242 yr Author It looks like getting Kernel 6.5 or greater is actually is being discussed on reddit here. The first response links to a walkthrough that discusses how to apply the kernel change with a link to compiled kernels. Looks like the last release is actually on kernel 6.7. Think I will try the 6.6.9 kernel version and see how it goes. It would be nice if we could get a moderator or someone to look at this and see about getting this into the regular build before to long if this works.
January 24, 20242 yr 56 minutes ago, mavrrick said: It looks like getting Kernel 6.5 or greater is actually is being discussed on reddit here. The first response links to a walkthrough that discusses how to apply the kernel change with a link to compiled kernels. Looks like the last release is actually on kernel 6.7. Think I will try the 6.6.9 kernel version and see how it goes. It would be nice if we could get a moderator or someone to look at this and see about getting this into the regular build before to long if this works. I managed to get nested esxi 8 working on my Intel 11th gen with 6.12.6. Waiting for UNRAID 6.13 with newer kernel before trying on my X570 board. One major issue I have with the Intel is that when I passthrough my Telsa P4 to ESXI and power on a VM, the ESXi host PSODs. I am using the following qemu arg in my ESXi template. <qemu:commandline> <qemu:arg value='-machine'/> <qemu:arg value='kernel-irqchip=split'/> <qemu:arg value='-device'/> <qemu:arg value='intel-iommu,intremap=on,caching-mode=on'/> </qemu:commandline> If I dont use this, ESXi does not detect vt-d IOMMU correctly and none of the passed through devices are selectable.
January 25, 20242 yr Author I tried the precompiled kernals from the links in that reddit post and it didn't work and nothing changed. I tried both the 6.6.9 and 6.7.0 kernals. Hopefully when the official release is done whatever work is completed will fix it, but the lack of results with that gave me a l little pause. Hopefully that is all that is needed and nothing else.
January 25, 20242 yr 5 hours ago, mavrrick said: I tried the precompiled kernals from the links in that reddit post and it didn't work and nothing changed. I tried both the 6.6.9 and 6.7.0 kernals. Hopefully when the official release is done whatever work is completed will fix it, but the lack of results with that gave me a l little pause. Hopefully that is all that is needed and nothing else. That's good to know and saved me alot of time testing against my x570 3950x unraid host. Looks like Intel is the safest way for the moment but I hit a snag on my z490d4u 11900k build. I can passthrough my tesla p4 gpu to a nested vm in esxi 8 but vgpu psods the esxi host. No matter what I try, the same result. Seems vgpu passthrough is not working but direct gpu passthrough does.
February 13, 20242 yr On 1/25/2024 at 1:51 PM, mavrrick said: I tried the precompiled kernals from the links in that reddit post and it didn't work and nothing changed. I tried both the 6.6.9 and 6.7.0 kernals. Hopefully when the official release is done whatever work is completed will fix it, but the lack of results with that gave me a l little pause. Hopefully that is all that is needed and nothing else. The 6.8rc3 and rc4 kernels enable nested virtualisation, tested and it works.
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.