Re: [RFC PATCH v3 0/5] x86/kvm: Virtual suspend time injection support
From: Marc Zyngier <maz@kernel.org>
Date: 2021-10-20 13:52:36
Also in:
kvm, kvmarm, linux-arm-kernel, lkml
From: Marc Zyngier <maz@kernel.org>
Date: 2021-10-20 13:52:36
Also in:
kvm, kvmarm, linux-arm-kernel, lkml
Hi Hikaru, On Wed, 20 Oct 2021 13:04:25 +0100, Hikaru Nishida [off-list ref] wrote:
Hi, This patch series adds virtual suspend time injection support to KVM. It is an updated version of the following series: v2: https://lore.kernel.org/kvm/20210806100710.2425336-1-hikalium@chromium.org/ (local) v1: https://lore.kernel.org/kvm/20210426090644.2218834-1-hikalium@chromium.org/ (local) Please take a look again. To kvm/arm64 folks: I'm going to implement this mechanism to ARM64 as well but not sure which function should be used to make an IRQ (like kvm_apic_set_irq in x86) and if it is okay to use kvm_gfn_to_hva_cache / kvm_write_guest_cached for sharing the suspend duration.
Before we discuss interrupt injection, I want to understand what this is doing, and how this is doing it. And more precisely, I want to find out how you solve the various problems described by Thomas here [1]. Assuming you solve these, you should model the guest memory access similarly to what we do for stolen time. As for injecting an interrupt, why can't this be a userspace thing? Thanks, M. [1] https://lore.kernel.org/all/871r557jls.ffs@tglx (local) -- Without deviation from the norm, progress is not possible.