September 29, 20178 yr Noticed the Windows 10 VM could not reboot properly and was constantly going to the diagnostics screen in a boot loop. Using rather customised config that could be the issue, although I am focusing on low latency DPC for real-time audio/video/3d workloads, hopefully, with internal Windows timer below 0.5ms. It would be great if we can estimate the best VM performance configuration for Windows 10 on the Ryzen platform and have it in future iterations of unRAID. unRAID kernel options: append isolcpus=8-15 nohz_full=8-15 rcu_nocbs=8-15 initrd=/bzroot Attaching VM XML and diagnostics. win10kvm.xml Edited September 27, 20187 yr by realies
September 29, 20178 yr Author Further testing and it seems the XML config has nothing to do with the issue. Any other option other than safe mode from the automatic repair window results in a restart, excluding the "no automatic reboot on error" type of option, which makes the machine hang. Windows works absolutely fine when booting from the same drive barebones. What would be an appropriate way of detecting where it fails? Selecting "Boot log" within msconfig->Boot does not produce any logs. Edited September 29, 20178 yr by realies
September 30, 20178 yr Hi realies, If you remove the extra parameters (isolcpus is ok) from your syslinux config, does that change the behavior with the VM at all? - Jon
October 1, 20178 yr Did you install windows from the getgo with those kernel options enabled? has windows ever booted? or are the bootloops after a windows update? anything meaningful in the system event logs in the windows VM when you boot into safe mode?
October 1, 20178 yr Author Assumed virtualisation could be inherited, although it turned out that enabling the Hyper-V components within Windows 10 and rebooting causes this particular issue with boot looping. Unfortunately, I have found that creating a system restore point before enabling the components does not help with returning to a working state and boot logging does not work at all. Nothing meaningful in the windows VM event logs.
October 2, 20178 yr 23 hours ago, realies said: Assumed virtualisation could be inherited, although it turned out that enabling the Hyper-V components within Windows 10 and rebooting causes this particular issue with boot looping. Unfortunately, I have found that creating a system restore point before enabling the components does not help with returning to a working state and boot logging does not work at all. Nothing meaningful in the windows VM event logs. Question regarding the bolded statement. Are you stating that in the VM configuration (XML) that if you enable Hyper-V, this causes boot looping, or are you talking about enabling something from within the Windows 10 VM itself? I have not been able to recreate this issue on our own test systems.
October 2, 20178 yr Author I mean exactly what is stated, enabling Hyper-V within Windows (from add/remove features) and rebooting causes the machine to go into a boot loop that does not go past the TianoCore screen. After three cycles, the Windows 10 recovery options menu is loaded up from which only the safe mode options are able to boot the VM. Have tried switching between versions of i440fx and Q35 without any difference. Bare metal booting from the same problematic setup has no issues at all. Perhaps this is a KVM/QEMU issue? Edited October 2, 20178 yr by realies
October 2, 20178 yr Why are you enabling Hyper-V inside Windows 10? Are trying to utilize nested virtualization?
October 2, 20178 yr Author Indeed I have assumed that nested virtualisation can be utilised. Would it be possible to transfer 6.4.0-rc9f configuration to 6.3.5 without anything breaking, including the current trial period?
October 2, 20178 yr see here: https://ladipro.wordpress.com/2017/02/24/running-hyperv-in-kvm-guest/ Nested virtualization inside a KVM VM requires an additional kernel parameter, which by default isnt enabled in unraid. the following command would return 'Y' if it was enabled. cat /sys/module/kvm_intel/parameters/nested It also requires a fairly recent linux kernel\qemu\kvm version (which rc9 has), but im not sure 6.3.5 ticks all the boxes for it to work? i've not tried nested virtualization so i cant vouch for if this will even work if you enabled the parameter, or what performance would be like in the nested VMs. I'm not sure what the use case is though when you can spin up VMs inside unraid anyway? Edited October 2, 20178 yr by billington.mark words
October 3, 20178 yr Author It seems nested visualisation is enabled in rc9f. # cat /sys/module/kvm_amd/parameters/nested 1
October 3, 20178 yr may have found the issue. remember you mentioning you were on ryzen.... Hyper-V on KVM does not work on AMD CPUs Fixed in kernel 4.12 for 1 vCPU405a353a0e and in kernel 4.13 for >1 vCPU4aebd0e9caab2f4d73eb9b61174793a12713c25b1a5e185294 RHBZ 1440025 rc9f is using kernel 4.12.14. You should be able to boot if you drop down to 1 vCPU on the VM. Assuming the next RC has a bump to 4.13, your issue should hopefully be resolved. Mark
October 7, 20178 yr may have found the issue. remember you mentioning you were on ryzen.... Hyper-V on KVM does not work on AMD CPUs Fixed in kernel 4.12 for 1 vCPU405a353a0e and in kernel 4.13 for >1 vCPU4aebd0e9caab2f4d73eb9b61174793a12713c25b1a5e185294 RHBZ 1440025 rc9f is using kernel 4.12.14. You should be able to boot if you drop down to 1 vCPU on the VM. Assuming the next RC has a bump to 4.13, your issue should hopefully be resolved. MarkThanks Mark!!!Sent from my SM-G930P using Tapatalk
Archived
This topic is now archived and is closed to further replies.