qemu for windows on docker. Run / Install / Whatever use for, it's UP On You.
./win7.qcow2 - Disk volume
./iso - ISO Folder
./iso/virtio.iso - emu virtio
Just Use docker-compose up -d
docker-compose.yml
version: '3.3'
services:
emu-windows:
image: emengweb/emu-windows:latest
privileged: false
environment:
- CPU=2 # Default 1
- MEMERY=3G # Default 1G
#- ISOFILE=virtio.iso # Default Null, Can set ios for install custem OS
#- USEKVM=true # if u run as VPS, disable it!!!
volumes:
- $PWD/win7.qcow2:/disk.qcow2 # Change Disk file name
- $PWD/iso:/iso
- /dev/bus/usb:/dev/bus/usb
- /lib/modules:/lib/modules
ports:
- 6901:5900 # VNC no password
- 9833:3389 # RDP Administrator password
devices:
#- /dev/kvm # if u run as VPS, disable it!!!
- /dev/vfio/vfio
#- /dev/vfio/1
- /dev/bus/usb
ulimits:
memlock:
soft: -1
hard: -1docker run \
--volume $PWD/win7.qcow2:/disk.qcow2 `# the persistent volume` \
--volume $PWD/iso:/iso `# the iso folder` \
--interactive --tty \
-p 6901:5900 \
-p 9833:3389 \
-e CPU='2' \
-e MEMERY='3G' \
`# -e ISOFILE='virtio.iso'` \
`# -e USEKVM='true'` \
\
`# --device /dev/kvm # use hardware acceleration` \
--device /dev/vfio/vfio ` # vfio is used for PCIe passthrough` \
`# --device /dev/vfio/1 # the vfio IOMMU group` \
--ulimit memlock=-1:-1 `# so DMA can happen for the vfio passthrough` \
--device /dev/bus/usb `# since we use usb-host device passthrough (note you can specify specific devices too)` \
--volume /dev/bus/usb:/dev/bus/usb `# to allow for hot-plugging of USB devices` \
--volume /lib/modules:/lib/modules `# needed for loading vfio` \
--privileged `# needed for allowing hot-plugging of USB devices, but should be able to replace with cgroup stuff? also needed for modprobe commands` \
emengweb/emu-windows:latestDOCKER_BUILDKIT=1 docker build -t emu-windows .qemu-img create -f qcow2 disk.qcow2 20GGiven a QEMU Windows 10 image (you can create one as you normally would, name it emugaming.qcow2), this will run it on Docker with GPU/vfio passthrough. Works great for near-native gaming.
Runs on Linux and with some changes should also be able to run on Windows or MacOS hosts (I think MacOS supports VT-d?). Note that you'll need a ROM dump of your GPU if you're passing through the primary boot video card. Also you'll need the kernel parameters iommu=pt intel_iommu=on to enable IOMMU/VT-d.
Still working on making stuff actually configurable, but make your modifications in start.sh for configuring your components. Ideally in the future you'll be able to just run docker run lg/emugaming with hardly any parameters and all will be automatic, including generating the Windows 10 image from MS.
DOCKER_BUILDKIT=1 docker build -t emugaming .- Create a disk image using
qemu-img create -f qcow2 emugaming.qcow2 100Gon the host. You might need theqemu-imgpackage. - Run the
docker runcommand below but appendingashat the end to enter the docker image. This will open ash, Alpine's bash equivalent. - Download the latest Windows 10 ISO from Microsoft.
- Download the latest stable virtio-win driver pack from Fedora or their github.
- Run a
qemucommand similar tostart.sh, except:- I'd recommend not using the GPU passthrough just yet. Remove the
vfio-pci,usb-host,vgaandnographiclines, and instead pass through-vnc 0.0.0.0:0 -vga qxl -device usb-tabletand using VNC to install Windows and getting the image stable. - Note you need to mount the two ISOs you downloaded
-cdrom win10.iso -cdrom virtio.iso - When you install Windows it won't see your hard drive, click Have Disk and select the
viostordriver in the VirtIO driver ISO - You can use the
quitorsystem_powerdowncommands in qemu to do a hard shutdown or a soft one
- I'd recommend not using the GPU passthrough just yet. Remove the
- Once Windows is installed, I'd recommend some things to change:
- Disable hibernation and standby (its ok to keep monitor standby on though)
- Set
bcdedit /set '{current}' recoveryenabled Nosince it can get confusing when debugging video card passthrough problems - Install the remainder of the Virtio drivers, plus other missing drivers too (like video card)
docker run \
--volume $PWD/emugaming.qcow2:/emugaming.qcow2 `# the persistent volume` \
--interactive --tty \
\
--device /dev/kvm `# use hardware acceleration` \
--device /dev/vfio/vfio ` # vfio is used for PCIe passthrough` \
--device /dev/vfio/1 `# the vfio IOMMU group` \
--ulimit memlock=-1:-1 `# so DMA can happen for the vfio passthrough` \
--device /dev/bus/usb `# since we use usb-host device passthrough (note you can specify specific devices too)` \
--volume /dev/bus/usb:/dev/bus/usb `# to allow for hot-plugging of USB devices` \
--volume /lib/modules:/lib/modules `# needed for loading vfio` \
--privileged `# needed for allowing hot-plugging of USB devices, but should be able to replace with cgroup stuff? also needed for modprobe commands` \
emugamingTo use the QEMU Monitor command-line:
docker exec -it <CONTAINER_ID> socat - UNIX-CONNECT:/var/run/qemu_monitorTo start this docker image at boot:
docker run \
...
--detach \
--restart unless-stopped \
...- Make GPU params configurable (like address of devices and ROM file)
- Make USB devices passed-through configurable
- Make CPU cores configurable (or autodetect max?)
- Detect and complain when IOMMU isnt on or devices not passthrough-able
- Generate Windows 10 image automatically based on official Microsoft images
- Remove the need for
--privilegedand broad permissions. Should be possible... - Remove the sleep statement in
start.sh - Fix windows launching recover mode
- Get this built for k3s or another Kubernetes distro
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent