Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux

dt-bindings: virtio: Add virtio-pci-iommu node

Some systems implement virtio-iommu as a PCI endpoint. The operating
system needs to discover the relationship between IOMMU and masters long
before the PCI endpoint gets probed. Add a PCI child node to describe the
virtio-iommu device.

The virtio-pci-iommu is conceptually split between a PCI programming
interface and a translation component on the parent bus. The latter
doesn't have a node in the device tree. The virtio-pci-iommu node
describes both, by linking the PCI endpoint to "iommus" property of DMA
master nodes and to "iommu-map" properties of bus nodes.

Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Jean-Philippe Brucker <jean-philippe.brucker@arm.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

authored by

Jean-Philippe Brucker and committed by
Michael S. Tsirkin
6c9e92ef 8447d84e

+66
+66
Documentation/devicetree/bindings/virtio/iommu.txt
··· 1 + * virtio IOMMU PCI device 2 + 3 + When virtio-iommu uses the PCI transport, its programming interface is 4 + discovered dynamically by the PCI probing infrastructure. However the 5 + device tree statically describes the relation between IOMMU and DMA 6 + masters. Therefore, the PCI root complex that hosts the virtio-iommu 7 + contains a child node representing the IOMMU device explicitly. 8 + 9 + Required properties: 10 + 11 + - compatible: Should be "virtio,pci-iommu" 12 + - reg: PCI address of the IOMMU. As defined in the PCI Bus 13 + Binding reference [1], the reg property is a five-cell 14 + address encoded as (phys.hi phys.mid phys.lo size.hi 15 + size.lo). phys.hi should contain the device's BDF as 16 + 0b00000000 bbbbbbbb dddddfff 00000000. The other cells 17 + should be zero. 18 + - #iommu-cells: Each platform DMA master managed by the IOMMU is assigned 19 + an endpoint ID, described by the "iommus" property [2]. 20 + For virtio-iommu, #iommu-cells must be 1. 21 + 22 + Notes: 23 + 24 + - DMA from the IOMMU device isn't managed by another IOMMU. Therefore the 25 + virtio-iommu node doesn't have an "iommus" property, and is omitted from 26 + the iommu-map property of the root complex. 27 + 28 + Example: 29 + 30 + pcie@10000000 { 31 + compatible = "pci-host-ecam-generic"; 32 + ... 33 + 34 + /* The IOMMU programming interface uses slot 00:01.0 */ 35 + iommu0: iommu@0008 { 36 + compatible = "virtio,pci-iommu"; 37 + reg = <0x00000800 0 0 0 0>; 38 + #iommu-cells = <1>; 39 + }; 40 + 41 + /* 42 + * The IOMMU manages all functions in this PCI domain except 43 + * itself. Omit BDF 00:01.0. 44 + */ 45 + iommu-map = <0x0 &iommu0 0x0 0x8> 46 + <0x9 &iommu0 0x9 0xfff7>; 47 + }; 48 + 49 + pcie@20000000 { 50 + compatible = "pci-host-ecam-generic"; 51 + ... 52 + /* 53 + * The IOMMU also manages all functions from this domain, 54 + * with endpoint IDs 0x10000 - 0x1ffff 55 + */ 56 + iommu-map = <0x0 &iommu0 0x10000 0x10000>; 57 + }; 58 + 59 + ethernet@fe001000 { 60 + ... 61 + /* The IOMMU manages this platform device with endpoint ID 0x20000 */ 62 + iommus = <&iommu0 0x20000>; 63 + }; 64 + 65 + [1] Documentation/devicetree/bindings/pci/pci.txt 66 + [2] Documentation/devicetree/bindings/iommu/iommu.txt