vhost-device-gpu - Man Page

vhost-device-gpu --socket-path <path> --gpu-mode <GPU_MODE>

A virtio-gpu device using the vhost-user protocol.

• -s, --socket-path <SOCKET>

vhost-user Unix domain socket.

• -g, --gpu-mode <GPU_MODE>

The mode specifies which backend implementation to use.

• -c, --capset <CAPSET>

Comma-separated list of enabled capsets.

• --use-egl <USE_EGL>

Enable backend to use EGL.

Default: true

Possible values: true, false

• --use-glx <USE_GLX>

Enable backend to use GLX.

Default: false

Possible values: true, false

• --use-gles <USE_GLES>

Enable backend to use GLES.

Default: true

Possible values: true, false

• --use-surfaceless <USE_SURFACELESS>

Enable surfaceless backend option.

Default: true

Possible values: true, false

• -h, --help

Print help and usage information.

• -V, --version

Print version information.

• This device links native libraries (due to the usage of Rutabaga) compiled with GNU libc. The CI is set up to not build this device for musl targets.
• It might be possible to build those libraries using musl and then build the GPU device, but this is untested.
• Currently, only sharing the display output to QEMU through a socket using the transfer_read operation (VIRTIO_GPU_CMD_TRANSFER_FROM_HOST_3D) is supported.
• Directly sharing display output resources using dmabuf is not yet supported.

• The following features are not yet supported:

  • VIRTIO_GPU_CMD_RESOURCE_CREATE_BLOB
  • VIRTIO_GPU_CMD_SET_SCANOUT_BLOB
  • VIRTIO_GPU_CMD_RESOURCE_ASSIGN_UUID These require rust-vmm/vhost#251\c , which in turn requires QEMU API stabilization.

• Because blob resources are not yet supported, some capsets are limited:

  • Venus (Vulkan implementation in virglrenderer) support is unavailable.
  • gfxstream-vulkan and gfxstream-gles are exposed but can only be used for display output (no hardware acceleration yet).

The device leverages the rutabaga_gfx\c  crate to provide rendering with virglrenderer and gfxstream.

• Gfxstream support is compiled by default but can be disabled by building without the gfxstream feature flag:

  cargo build --no-default-features

• With Virglrenderer, Rutabaga translates OpenGL and Vulkan calls to an intermediate representation, allowing OpenGL acceleration on the host.
• With the gfxstream rendering mode, GLES and Vulkan calls are forwarded to the host with minimal modification.

Start the daemon on the host machine using either GPU mode:

1. Virglrenderer

2. Gfxstream (if the crate has been compiled with the gfxstream feature)

   host# vhost-device-gpu --socket-path /tmp/gpu.socket --gpu-mode virglrenderer

With QEMU, two device frontends can be used:

1. vhost-user-gpu-pci
2. vhost-user-vga (also implements VGA, allowing visibility of boot messages before GPU initialization)

By default, QEMU adds another VGA output. Disable it using -vga none.

qemu-system-x86_64 \
  -chardev socket,id=vgpu,path=/tmp/gpu.socket \
  -device vhost-user-gpu-pci,chardev=vgpu,id=vgpu \
  -object memory-backend-memfd,share=on,id=mem0,size=4G \
  -machine q35,memory-backend=mem0,accel=kvm \
  -display gtk,gl=on,show-cursor=on \
  -vga none

qemu-system-x86_64 \
  -chardev socket,id=vgpu,path=/tmp/gpu.socket \
  -device vhost-user-vga,chardev=vgpu,id=vgpu \
  -object memory-backend-memfd,share=on,id=mem0,size=4G \
  -machine q35,memory-backend=mem0,accel=kvm \
  -display gtk,gl=on,show-cursor=on \
  -vga none

RUST_LOG

Logging level. Set to debug for maximum output.

See GitHub Issues: https://github.com/rust-vmm/vhost-device/issues\c

Dorinda Bassey dbassey@redhat.com\c

Matej Hrica mhrica@redhat.com\c

qemu(1)