On Wed, Aug 31, 2016 at 07:25:11PM +0300, David Kiarie wrote: > Hello all, > > Changes since V2 > -formating fixes. > -fixed an issue where the right IOAPIC id was not being correctly set when > using kernel_irqchip=off > > The following patchset implements AMD-Vi interrupt remapping logic and hooks > it onto existing IR infrastucture. > > I have bundled this patchset together with the "Explicit SID for > IOAPIC"."Explicit SID for IOAPIC" functions to > affiliate MSI routes with a requester ID and a PCI device if present which > enables platform devices like IOAPIC to > make interrupt requests using an explicit SID as required by both VT-d and > AMD-Vi.
Could you please Cc maintainers of all affected subsystems? E.g. I see changes in VFIO and kvm and neither Paolo nor Alex were Cc'd. > David Kiarie (6): > hw/msi: Allow platform devices to use explicit SID > hw/i386: enforce SID verification > hw/iommu: Prepare for AMD IOMMU interrupt remapping > hw/iommu: AMD IOMMU interrupt remapping > hw/acpi: report IOAPIC on IVRS > hw/iommu: share common code between IOMMUs > > hw/i386/acpi-build.c | 2 + > hw/i386/amd_iommu.c | 241 > +++++++++++++++++++++++++++++++++++++- > hw/i386/amd_iommu.h | 82 +++++++++++++ > hw/i386/intel_iommu.c | 89 +++++++------- > hw/i386/kvm/pci-assign.c | 12 +- > hw/i386/trace-events | 7 ++ > hw/i386/x86-iommu.c | 8 ++ > hw/intc/ioapic.c | 30 ++++- > hw/misc/ivshmem.c | 6 +- > hw/vfio/pci.c | 6 +- > hw/virtio/virtio-pci.c | 7 +- > include/hw/i386/ioapic_internal.h | 1 + > include/hw/i386/x86-iommu.h | 1 + > include/sysemu/kvm.h | 25 ++-- > kvm-all.c | 10 +- > kvm-stub.c | 5 +- > qemu-version.h | 1 + > target-i386/kvm.c | 15 ++- > 18 files changed, 462 insertions(+), 86 deletions(-) > create mode 100644 qemu-version.h > > -- > 2.1.4