Here's a few fixes I've been piling on during the merge window.
The first patch improves the handling of the ACTIVE registers, which
we never synchronise on the read side (the distributor state can only
be updated when the vcpu exits). Let's fix it the same way we do it on
the write side (stop-the-world, read, restart). Yes, this is
expensive.
The last two patches deal with an issue where we consider the HW state
of an interrupt when responding to a userspace access. We should never
do this, as the guest shouldn't be running at this stage and if it is,
it is absolutely fine to return random bits to userspace. It could
also be that there is no active guest context at this stage, and you
end up with an Oops, which nobody really enjoys.
Marc Zyngier (3):
KVM: arm: vgic: Synchronize the whole guest on GIC{D,R}_I{S,C}ACTIVER
read
KVM: arm: vgic: Only use the virtual state when userspace accesses
enable bits
KVM: arm: vgic-v2: Only use the virtual state when userspace accesses
pending bits
virt/kvm/arm/vgic/vgic-mmio-v2.c | 16 ++-
virt/kvm/arm/vgic/vgic-mmio-v3.c | 20 ++--
virt/kvm/arm/vgic/vgic-mmio.c | 183 +++++++++++++++++++++++++------
virt/kvm/arm/vgic/vgic-mmio.h | 19 ++++
4 files changed, 188 insertions(+), 50 deletions(-)
--
2.25.1
