虚拟化容器,大数据,DBA,中间件,监控。

Intel iGVT-g vGPU in KVM – KVMGT howto

14 03月
作者:admin|分类:容器虚拟化

We are pleased to announce the first release of KVMGT project. KVMGT is the implementation of Intel GVT-g technology, a full GPU virtualization solution. Under Intel GVT-g, a virtual GPU instance is maintained for each VM, with part of performance critical resources directly assigned. The capability of running a native graphics driver inside a VM, without hypervisor intervention in performance critical paths, achieves a good balance of performance, feature, and sharing capability.

 

KVMGT is still in the early stages:

  • Basic functions of full GPU virtualization works; guests can see a full-featured vGPU.
    We ran several 3D workloads such as lightsmark, nexuiz, urbanterror and warsow.
  • Only Linux guests are supported so far, and PPGTT must be disabled in guest through a kernel parameter(see README.kvmgt in QEMU).
  • This drop also includes some Xen specific changes, which will be cleaned up later.
  • Our end goal is to upstream both XenGT and KVMGT, which shares ~90% logic for vGPU device model (will be part of i915 driver), with only difference in hypervisor specific services.
  • There is currently insufficient test coverage, so please bear with some stability issues :)

 

There are things that need to be improved, esp. the KVM interfacing:

  1. A domid was added to each KVMGT guest

    An ID is needed for foreground OS switching, e.g.

     # echo <domid>    >    /sys/kernel/vgt/control/foreground_vm

    domid 0 is reserved for host OS.
     

  2. SRCU workarounds.

    Some KVM functions, such as:

  •  kvm_io_bus_register_dev
  •  install_new_memslots

must be called *without* &kvm->srcu read-locked. Otherwise it hangs.

In KVMGT, we need to register an iodev only *after* BAR registers are written by guest. That means, we already have &kvm->srcu hold -trapping/emulating PIO(BAR registers) makes us in such a condition.

That will make kvm_io_bus_register_dev hangs.

Currently we have to disable rcu_assign_pointer() in such functions.

These were dirty workarounds, so your suggestions are highly welcome!

  1. syscalls were called to access "/dev/mem" from kernel

An in-kernel memslot was added for aperture, but using syscalls like Open and mmap to open and access the character device "/dev/mem",for pass-through.

The source codes(kernel, qemu as well as seabios) are available at github:

In the KVMGT-qemu repository, there is a "README.kvmgt" for reference.

More information about Intel GVT-g and KVMGT can be found at:

 

We appreciate your comments, BUG reports, and contributions!

Project: 

浏览4526 评论0
返回
目录
返回
首页
在 ESXi 上安装和配置 NVIDIA VIB 实现3D显卡 GPU虚拟化 Example using SPICE and QXL for improved Graphics experience in the guest