vfio-pci-passthrough.pdf

发布时间：2022-06-11 发布人：admin 分类：说明书资料大小：1.22M 资料格式：pdf 举报版权申诉

shirleylinyuer-10898465-4744302543314662973.pdf-第1页.png

第1页 / 共33页

shirleylinyuer-10898465-4744302543314662973.pdf-第2页.png

第2页 / 共33页

shirleylinyuer-10898465-4744302543314662973.pdf-第3页.png

第3页 / 共33页

shirleylinyuer-10898465-4744302543314662973.pdf-第4页.png

第4页 / 共33页

shirleylinyuer-10898465-4744302543314662973.pdf-第5页.png

第5页 / 共33页

shirleylinyuer-10898465-4744302543314662973.pdf-第6页.png

第6页 / 共33页

shirleylinyuer-10898465-4744302543314662973.pdf-第7页.png

第7页 / 共33页

shirleylinyuer-10898465-4744302543314662973.pdf-第8页.png

第8页 / 共33页

Slide 1

Slide 2

Slide 3

Slide 4

Slide 5

Slide 6

Slide 7

Slide 8

Slide 9

Slide 10

Slide 11

Slide 12

Slide 13

Slide 14

Slide 15

Slide 16

Slide 17

Slide 18

Slide 19

Slide 20

Slide 21

Slide 22

Slide 23

Slide 24

Slide 25

Slide 26

Slide 27

Slide 28

Slide 29

Slide 30

Slide 31

Slide 32

Slide 33

vfio-pci passthrough Fei Li fli@suse.com

Summary ➢ What is VFIO/IOMMU? Why want them? ➢ VFIO – qemu part ➢ VFIO – kernel part ➢ VFIO usage: how to passthrough a pci device 2

- What is VFIO/IOMMU? Why want them?

What is VFIO/IOMMU? ➢ The VFIO (Virtual Function I/O) driver is an IOMMU/device agnostic framework for exposing direct device access to userspace, in a secure, IOMMU protected environment. ➢ For x86, it needs the I/O MMU hardware support. ➢ VFIO consists of ➢ - kernel device driver: vfio_pci_driver, vfio_iommu_driver, vfio_dma ➢ - QEMU device class: VFIODevice, VFIOPCIDevice ➢ The guest can operate the pass-throughed PCI device by: - accessing the mapped PCI config space and memory space - ioctl() on a fd of the VFIO kernel device for control operations 4

What is VFIO/IOMMU? (Continue.) ➢ In qemu, use VFIO to configure IOMMU: e.g. ioctl(VFIO_SET_IOMMU) && ioctl(VFIO_IOMMU_MAP_DMA) 5

VFIO: device, group, container ➢ A group is a set of devices which is isolatable from all other devices, specialized in IOMMU. It is the minimum granularity. ➢ Within one container, different groups can share a set of page tables to reduce the duplication. The container provides little functionality: version check and extension query. ➢ The user needs to add a group into the container. ➢ The VFIO device API includes ioctls for describing the device, the I/O regions and their read/write/mmap offsets on the device descriptor, && mechanisms for describing and registering interrupt notifications. 6

Why want them? ➢ In short, for higher I/O performance by lessening the times of VM-EXIT/VM-ENTRY when accessing PCI BAR and doing DMA. 7

Why want them? (literal) ➢ - when accessing PCI BAR. The emulated guest BIOS emulate the BAR address for the guest (On the contrary, if expose the host real BAR to the guest, there may be conflicts between host real PCI BAR and the other emulated PCI device’s BAR in the guest). For the first time when BAR is accessed and can not be visited, the VM exists and does the address mapping between GPA and HPA using EPT, and records the mapping. When later access, no VM-exit is needed. ➢ - when PCI device communicates with GPA via DMA. When initializing the vfio: vfio_realize() in qemu, a memory mapping is established via vfio_region_mmap(). Then a (ioctl(s->container, VFIO_IOMMU_MAP_DMA, &dma_map) in qemu_vfio_do_mapping will do the real mapping between VFIO iova and HPA. 8

分享到：

赞收藏

资料库

vfio-pci-passthrough.pdf

相关推荐

云计算

热门标签

最新资料