This series fixes problems with initializing the VGIC. The problem is that we were initializing the state of the VGIC on-demand partially, and doing some final initializtion at the time when we were going to run a VCPU for the first time. This broke migration, because the first-vcpu-run init would overwrite restored state. We also cleanup the naming of the init functions and add checks when creating VCPUs and when injecting IRQs from userspace. This series invalidates patch 3 "KVM: arm/arm64: check vgic_initialized before VCPU creation" in Eric Auger's vgic init ioctl series. Eric's series should be applied after this one, making future ABIs (IRQFD, VFIO, GICv3, ...) require explicit userspace vgic initialization instead of this on-demand approach that we now have to maintain for legacy userspace compatiblity. I've tested this with 32-bit and 64-bit QEMU and kvmtool. Christoffer Dall (4): arm/arm64: KVM: Rename vgic_initialized to vgic_ready arm/arm64: KVM: Add (new) vgic_initialized macro arm/arm64: KVM: Don't allow creating VCPUs after vgic_initialized arm/arm64: KVM: Initialize the vgic on-demand when injecting IRQs Peter Maydell (1): arm/arm64: KVM: vgic: move reset initialization into vgic_init_maps() arch/arm/kvm/arm.c | 13 +++++--- include/kvm/arm_vgic.h | 12 +++++-- virt/kvm/arm/vgic.c | 91 +++++++++++++++++++++++--------------------------- 3 files changed, 60 insertions(+), 56 deletions(-) -- 2.1.2.330.g565301e.dirty -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
