https://bugzilla.kernel.org/show_bug.cgi?id=219009 --- Comment #29 from h4ck3r (michal.litwinczuk@xxxxx) --- (In reply to mlevitsk from comment #26) > But the question is - did they use nested virtualization on Linux actively > and with vls enabled? > > > The use case which causes the reboots as I understand is Hyperv enabled > Windows, in which case pretty much the whole Windows is running as a nested > VM, nested to the Hyperv hypervisor. > > Once I get my hands on a client Zen4 machine (I only have Zen2 at home), I > will also try to reproduce this but not promises when this will happen. > > Meanwhile I really hope that someone from AMD can take a look a this, and > either confirm that this is or will be fixed with a microcode patch or > confirm that we have to disable vls on the affected CPUs. > > Best regards, > Maxim Levitsky Not really - most of them are microservice type ones. That would also mean there is less chance of corruption since they occupied less host memory. And windows uses nested virt even if hyperv is not installed somehow. (installation does not, but freshly booted guest crashed my node) Im afraid it might be unresolvable issue, even with microcode. At least most things point to similar issue as memory leaks on their igpus. -- You may reply to this email to add a comment. You are receiving this mail because: You are watching the assignee of the bug.
