Comments on: The Actual Performance Impact of Spectre/Meltdown Hyper-V Updates

By: Eric Siron

Eric Siron — Fri, 12 Mar 2021 18:18:00 +0000

In reply to Nessie. Portions of the fixes in question only exist in the operating system patches. When you pass up a thread from a guest to the hypervisor, it bypasses a lot of the controls in the management OS. Microcode (firmware) fixes always apply, of course. This isn't a blanket statement because the hypervisor portion also gets some updates, but you should always expect the hypervisor to minimize its involvement in performance-sensitive operations.

By: Nessie

Nessie — Mon, 08 Feb 2021 15:51:00 +0000

What I have not understood well so far is the relationship we have with virtual CPUs. As explained in this article
https://www.altaro.com/hyper-v/hyper-v-virtual-cpus-explained/,
and how I have understood it, the vCPU send a thread to the Thread Scheduler and the thread gets then passed over to be executed on one of the physical CPUs. Should it therefore not enough that only the host, on which the hypervisor is running, is getting patched? Since from my understanding it is only the ‘physical’ thread that could make use of the branch prediction. Are there some further information available that describes in more detail how instructions are passed from a vCPU to a pCPUs to better understand this.

Cheers

By: James

James — Fri, 19 Jan 2018 15:04:00 +0000

Good write up but no mention of x86 (32bit meltdown) update not being available.

We have 25 windows 7 PC’s on 32bit windows 7, and even though meldown KB has been installed there is no meltdown protection.

Another useful thing to mention is a free test tool (much easier than the microsoft way of testing for protection) is available from grc.com called InSpectre and works really well.