single-gpu-passthrough: A repository from Mageas

This tutorial allows to create a KVM for gaming.

The VM is close to native performance.

Table Of Contents
Thanks to
Enable & Verify IOMMU
Configuring Libvirt
Setup Guest OS
Install Windows
Attaching devices
Libvirt Hook Helper
Config Libvirt Hooks
Start/Stop Libvirt Hooks
Video card driver virtualisation detection
vBIOS Patching
CPU Pinning
Hyper-V Enlightenments
Disk Tuning
Hugepages
CPU Governor
Update drivers on Windows
Enable Hyper-V on Windows

Thanks to

Arch wiki
The best way to learn how GPU passthrough is working.

bryansteiner
The best tutorial on GPU passthrough!

QaidVoid
The best tutorial to use VFIO!

joeknock90
Really good tutorial on the NVIDIA GPU patch.

SomeOrdinaryGamers
Bring me in the VFIO community.

Zeptic
How to get good performances in nested virtualization.

Quentin Franchi
The scripts for AMD GPUs.

Enable & Verify IOMMU

Ensure that AMD-Vi or Intel VT-d is supported by the CPU and enabled in the BIOS settings.

Enable IOMMU support by setting the kernel parameter depending on your CPU.

/etc/default/grub
`GRUB_CMDLINE_LINUX_DEFAULT="... amd_iommu=on iommu=pt ..."`
OR
`GRUB_CMDLINE_LINUX_DEFAULT="... intel_iommu=on iommu=pt ..."`

Generate grub.cfg

grub-mkconfig -o /boot/grub/grub.cfg

After rebooting, check that the groups are valid.

for g in `find /sys/kernel/iommu_groups/* -maxdepth 0 -type d | sort -V`; do
    echo "IOMMU Group ${g##*/}:"
    for d in $g/devices/*; do
        echo -e "\t$(lspci -nns ${d##*/})"
    done;
done;

Example output:

IOMMU Group 2:
    00:03.0 Host bridge [0600]: Advanced Micro Devices, Inc. [AMD] Starship/Matisse PCIe Dummy Host Bridge [1022:1482]
    00:03.1 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] Starship/Matisse GPP Bridge [1022:1483]
    09:00.0 VGA compatible controller [0300]: NVIDIA Corporation TU104 [GeForce RTX 2070 SUPER] [10de:1e84] (rev a1)
    09:00.1 Audio device [0403]: NVIDIA Corporation TU104 HD Audio Controller [10de:10f8] (rev a1)
    09:00.2 USB controller [0c03]: NVIDIA Corporation TU104 USB 3.1 Host Controller [10de:1ad8] (rev a1)
    09:00.3 Serial bus controller [0c80]: NVIDIA Corporation TU104 USB Type-C UCSI Controller [10de:1ad9] (rev a1)

If your card is not in an isolated group, you need to perform ACS override patch.

Configuring Libvirt

Install the packages.

pacman -S --needed qemu libvirt edk2-ovmf virt-manager dnsmasq ebtables

For windows 11, install the swtpm package.

pacman -S --needed swtpm

Update the permissions of libvirt to run the virtual machine as non root.

/etc/libvirt/libvirtd.conf
`unix_sock_group = "libvirt"`
`unix_sock_rw_perms = "0770"`

Add your user to the libvirt group.

sudo usermod -a -G libvirt $(whoami)

sudo systemctl enable --now libvirtd

Auto start the virsh internal network

sudo virsh net-autostart default

If you prefer, you can manually start the virsh internal network

sudo virsh net-start default

Setup Guest OS

Download virtio driver.

Create your storage volume with the raw format. Select Customize before install on Final Step.

In Overview
set Chipset to Q35
for Windows 10 set Firmware to UEFI
for Windows 11 set Firmware to UEFI secboot

In CPUs
set CPU model to host-passthrough
set CPU Topology match your cpu topology -1 core

In Sata
set Disk Bus to virtio

In NIC
set Device Model to virtio

In Add Hardware
select CDROM and point to `/path/to/virtio-driver.iso`
for Windows 11 add TPM and set Version to 2.0

Install Windows

Windows can't detect the virtio disk, so you need to Load Driver and select virtio-iso/amd64/win10 when prompted.

Windows can't connect to the internet, we will activate internet later in this tutorial.

Attaching devices

The devices you want to passthrough.

In Add PCI Host Device
PCI Host devices for your GPU
PCI Host device for your soundcard

In Add USB Host Device
Add your keyboard & mouse

Remove
`Display spice`
`Channel spice`
`Video QXL`
`Sound ich*`

Libvirt Hook Helper

Libvirt hooks automate the process of running specific tasks during VM state change.

More documentation on The Passthrough Post website.

Create Libvirt Hook Helper

sudo mkdir /etc/libvirt/hooks
sudo vim /etc/libvirt/hooks/qemu
sudo chmod +x /etc/libvirt/hooks/qemu

/etc/libvirt/hooks/qemu
#!/bin/bash # # Author: Sebastiaan Meijer (sebastiaan@passthroughpo.st) # GUEST_NAME="$1" HOOK_NAME="$2" STATE_NAME="$3" MISC="${@:4}" BASEDIR="$(dirname $0)" HOOKPATH="$BASEDIR/qemu.d/$GUEST_NAME/$HOOK_NAME/$STATE_NAME" set -e # If a script exits with an error, we should as well. # check if it's a non-empty executable file if [ -f "$HOOKPATH" ] && [ -s "$HOOKPATH"] && [ -x "$HOOKPATH" ]; then eval \"$HOOKPATH\" "$@" elif [ -d "$HOOKPATH" ]; then while read file; do # check for null string if [ ! -z "$file" ]; then eval \"$file\" "$@" fi done <<< "$(find -L "$HOOKPATH" -maxdepth 1 -type f -executable -print;)" fi

/etc/libvirt/hooks/qemu

#!/bin/bash
#
# Author: Sebastiaan Meijer (sebastiaan@passthroughpo.st)
#

GUEST_NAME="$1"
HOOK_NAME="$2"
STATE_NAME="$3"
MISC="${@:4}"

BASEDIR="$(dirname $0)"

HOOKPATH="$BASEDIR/qemu.d/$GUEST_NAME/$HOOK_NAME/$STATE_NAME"

set -e # If a script exits with an error, we should as well.

# check if it's a non-empty executable file
if [ -f "$HOOKPATH" ] && [ -s "$HOOKPATH"] && [ -x "$HOOKPATH" ]; then
    eval \"$HOOKPATH\" "$@"
elif [ -d "$HOOKPATH" ]; then
    while read file; do
        # check for null string
        if [ ! -z "$file" ]; then
          eval \"$file\" "$@"
        fi
    done <<< "$(find -L "$HOOKPATH" -maxdepth 1 -type f -executable -print;)"
fi

Config Libvirt Hooks

This configuration file allows you to create variables that can be read by the scripts below.

/etc/libvirt/hooks/kvm.conf
`# VIRSH VIRSH_GPU_VIDEO=pci_0000_09_00_0 VIRSH_GPU_AUDIO=pci_0000_09_00_1 VIRSH_USB=pci_0000_09_00_2 VIRSH_SERIAL_BUS=pci_0000_09_00_3`

Make sure to substitute the correct bus addresses for the devices you'd like to passthrough to your VM. Just in case it's still unclear, you get the virsh PCI device IDs from the Enable & Verify IOMMU script. Translate the address for each device as follows: IOMMU Group 1 01:00.0 ... --> VIRSH_...=pci_0000_01_00_0.

Start/Stop Libvirt Hooks

This command will set the variable KVM_NAME so you can execute the rest of the commands without changing the name of the VM.

KVM_NAME="YOUR_VM_NAME"

If the scripts are not working, use the scripts as template and write your own.

Choose the Start/Stop scripts that most closely match your hardware.

My hardware:

AMD Ryzen 7 3700X
NVIDIA GeForce RTX 2070 SUPER

Create Start Script

mkdir -p /etc/libvirt/hooks/qemu.d/${KVM_NAME}/prepare/begin
vim /etc/libvirt/hooks/qemu.d/${KVM_NAME}/prepare/begin/start.sh
chmod +x /etc/libvirt/hooks/qemu.d/${KVM_NAME}/prepare/begin/start.sh

/etc/libvirt/hooks/qemu.d/VM_NAME/prepare/begin/start.sh
#!/bin/bash # Helpful to read output when debugging set -x # Load variables source "/etc/libvirt/hooks/kvm.conf" # Stop display manager systemctl stop lightdm.service # Unbind VTconsoles echo 0 > /sys/class/vtconsole/vtcon0/bind echo 0 > /sys/class/vtconsole/vtcon1/bind # Unbind EFI-Framebuffer echo efi-framebuffer.0 > /sys/bus/platform/drivers/efi-framebuffer/unbind # Avoid a Race condition sleep 5 # Unload all Nvidia drivers modprobe -r nvidia_drm modprobe -r nvidia_modeset modprobe -r nvidia_uvm modprobe -r nvidia # Unbind the GPU from display driver virsh nodedev-detach $VIRSH_GPU_VIDEO virsh nodedev-detach $VIRSH_GPU_AUDIO virsh nodedev-detach $VIRSH_USB virsh nodedev-detach $VIRSH_SERIAL_BUS # Load VFIO Kernel Module modprobe vfio-pci

/etc/libvirt/hooks/qemu.d/VM_NAME/prepare/begin/start.sh

#!/bin/bash
# Helpful to read output when debugging
set -x

# Load variables
source "/etc/libvirt/hooks/kvm.conf"

# Stop display manager
systemctl stop lightdm.service

# Unbind VTconsoles
echo 0 > /sys/class/vtconsole/vtcon0/bind
echo 0 > /sys/class/vtconsole/vtcon1/bind

# Unbind EFI-Framebuffer
echo efi-framebuffer.0 > /sys/bus/platform/drivers/efi-framebuffer/unbind

# Avoid a Race condition
sleep 5

# Unload all Nvidia drivers
modprobe -r nvidia_drm
modprobe -r nvidia_modeset
modprobe -r nvidia_uvm
modprobe -r nvidia

# Unbind the GPU from display driver
virsh nodedev-detach $VIRSH_GPU_VIDEO
virsh nodedev-detach $VIRSH_GPU_AUDIO
virsh nodedev-detach $VIRSH_USB
virsh nodedev-detach $VIRSH_SERIAL_BUS

# Load VFIO Kernel Module  
modprobe vfio-pci

Create Stop Script

mkdir -p /etc/libvirt/hooks/qemu.d/${KVM_NAME}/release/end
vim /etc/libvirt/hooks/qemu.d/${KVM_NAME}/release/end/stop.sh
chmod +x /etc/libvirt/hooks/qemu.d/${KVM_NAME}/release/end/stop.sh

/etc/libvirt/hooks/qemu.d/VM_NAME/release/end/stop.sh
#!/bin/bash set -x # Load variables source "/etc/libvirt/hooks/kvm.conf" # Unload VFIO-PCI Kernel Driver modprobe -r vfio-pci modprobe -r vfio_iommu_type1 modprobe -r vfio # Re-Bind GPU to Nvidia Driver virsh nodedev-reattach $VIRSH_GPU_VIDEO virsh nodedev-reattach $VIRSH_GPU_AUDIO virsh nodedev-reattach $VIRSH_USB virsh nodedev-reattach $VIRSH_SERIAL_BUS # Rebind VT consoles echo 1 > /sys/class/vtconsole/vtcon0/bind echo 1 > /sys/class/vtconsole/vtcon1/bind # Bind EFI-Framebuffer nvidia-xconfig --query-gpu-info > /dev/null 2>&1 echo "efi-framebuffer.0" > /sys/bus/platform/drivers/efi-framebuffer/bind # Load all Nvidia drivers modprobe nvidia_drm modprobe nvidia_modeset modprobe drm_kms_helper modprobe drm modprobe nvidia_uvm modprobe nvidia # Restart Display Manager systemctl start lightdm.service

/etc/libvirt/hooks/qemu.d/VM_NAME/release/end/stop.sh

#!/bin/bash
set -x

# Load variables
source "/etc/libvirt/hooks/kvm.conf"

# Unload VFIO-PCI Kernel Driver
modprobe -r vfio-pci
modprobe -r vfio_iommu_type1
modprobe -r vfio

# Re-Bind GPU to Nvidia Driver
virsh nodedev-reattach $VIRSH_GPU_VIDEO
virsh nodedev-reattach $VIRSH_GPU_AUDIO
virsh nodedev-reattach $VIRSH_USB
virsh nodedev-reattach $VIRSH_SERIAL_BUS

# Rebind VT consoles
echo 1 > /sys/class/vtconsole/vtcon0/bind
echo 1 > /sys/class/vtconsole/vtcon1/bind

# Bind EFI-Framebuffer
nvidia-xconfig --query-gpu-info > /dev/null 2>&1
echo "efi-framebuffer.0" > /sys/bus/platform/drivers/efi-framebuffer/bind

# Load all Nvidia drivers
modprobe nvidia_drm
modprobe nvidia_modeset
modprobe drm_kms_helper
modprobe drm
modprobe nvidia_uvm
modprobe nvidia

# Restart Display Manager
systemctl start lightdm.service

Quentin hardware:

AMD Ryzen 5 2600
Radeon RX 590 Series

Create Start Script

mkdir -p /etc/libvirt/hooks/qemu.d/${KVM_NAME}/prepare/begin
vim /etc/libvirt/hooks/qemu.d/${KVM_NAME}/prepare/begin/start.sh
chmod +x /etc/libvirt/hooks/qemu.d/${KVM_NAME}/prepare/begin/start.sh

/etc/libvirt/hooks/qemu.d/VM_NAME/prepare/begin/start.sh
#!/bin/bash set -x # load variables source "/etc/libvirt/hooks/kvm.conf" # Stop display manager systemctl stop lightdm.service # Stop pipewire pulse_pid=$(pgrep -u quentin pulseaudio) pipewire_pid=$(pgrep -u quentin pipewire-media) kill $pulse_pid kill $pipewire_pid # Unbind VTconsoles echo 0 > /sys/class/vtconsole/vtcon0/bind echo 0 > /sys/class/vtconsole/vtcon1/bind # Unbind EFI Framebuffer # echo efi-framebuffer.0 > /sys/bus/platform/drivers/efi-framebuffer/unbind # Avoid a race condition by waiting a couple of seconds. This can be calibrated to be shorter or longer if required for your system sleep 5 # Unload AMD kernel module modprobe -r amdgpu # Detach GPU devices from host virsh nodedev-detach $VIRSH_GPU_VIDEO virsh nodedev-detach $VIRSH_GPU_AUDIO # Load vfio module modprobe vfio modprobe vfio_pci modprobe vfio_iommu_type1

/etc/libvirt/hooks/qemu.d/VM_NAME/prepare/begin/start.sh

#!/bin/bash
set -x

# load variables
source "/etc/libvirt/hooks/kvm.conf"

# Stop display manager
systemctl stop lightdm.service

# Stop pipewire
pulse_pid=$(pgrep -u quentin pulseaudio)
pipewire_pid=$(pgrep -u quentin pipewire-media)
kill $pulse_pid
kill $pipewire_pid

# Unbind VTconsoles
echo 0 > /sys/class/vtconsole/vtcon0/bind
echo 0 > /sys/class/vtconsole/vtcon1/bind

# Unbind EFI Framebuffer
# echo efi-framebuffer.0 > /sys/bus/platform/drivers/efi-framebuffer/unbind

# Avoid a race condition by waiting a couple of seconds. This can be calibrated to be shorter or longer if required for your system
sleep 5

# Unload AMD kernel module
modprobe -r amdgpu

# Detach GPU devices from host
virsh nodedev-detach $VIRSH_GPU_VIDEO
virsh nodedev-detach $VIRSH_GPU_AUDIO

# Load vfio module
modprobe vfio
modprobe vfio_pci
modprobe vfio_iommu_type1

Create Stop Script

mkdir -p /etc/libvirt/hooks/qemu.d/${KVM_NAME}/release/end
vim /etc/libvirt/hooks/qemu.d/${KVM_NAME}/release/end/stop.sh
chmod +x /etc/libvirt/hooks/qemu.d/${KVM_NAME}/release/end/stop.sh

/etc/libvirt/hooks/qemu.d/VM_NAME/release/end/stop.sh
#!/bin/bash set -x # load variables source "/etc/libvirt/hooks/kvm.conf" # Unload all the vfio modules modprobe -r vfio_pci modprobe -r vfio_iommu_type1 modprobe -r vfio # Attach GPU devices to host # Use your GPU and HDMI Audio PCI host device virsh nodedev-reattach $VIRSH_GPU_VIDEO virsh nodedev-reattach $VIRSH_GPU_AUDIO # Rebind VTconsoles echo 1 > /sys/class/vtconsole/vtcon0/bind echo 1 > /sys/class/vtconsole/vtcon1/bind # Rebind framebuffer to host # echo "efi-framebuffer.0" > /sys/bus/platform/drivers/efi-framebuffer/bind # Load AMD kernel module modprobe amdgpu modprobe gpu_sched modprobe ttm modprobe drm_kms_helper modprobe i2c_algo_bit modprobe drm modprobe snd_hda_intel # Restart Display Manager systemctl start lightdm.service

/etc/libvirt/hooks/qemu.d/VM_NAME/release/end/stop.sh

#!/bin/bash
set -x

# load variables
source "/etc/libvirt/hooks/kvm.conf"

# Unload all the vfio modules
modprobe -r vfio_pci
modprobe -r vfio_iommu_type1
modprobe -r vfio

# Attach GPU devices to host
# Use your GPU and HDMI Audio PCI host device
virsh nodedev-reattach $VIRSH_GPU_VIDEO
virsh nodedev-reattach $VIRSH_GPU_AUDIO

# Rebind VTconsoles
echo 1 > /sys/class/vtconsole/vtcon0/bind
echo 1 > /sys/class/vtconsole/vtcon1/bind

# Rebind framebuffer to host
# echo "efi-framebuffer.0" > /sys/bus/platform/drivers/efi-framebuffer/bind

# Load AMD kernel module
modprobe  amdgpu
modprobe  gpu_sched
modprobe  ttm
modprobe  drm_kms_helper
modprobe  i2c_algo_bit
modprobe  drm
modprobe  snd_hda_intel

# Restart Display Manager
systemctl start lightdm.service

Video card driver virtualisation detection

Video Card drivers refuse to run in Virtual Machine, so you need to spoof Hyper-V Vendor ID.

XML
... <features> ... <hyperv> ... <vendor_id state='on' value='buttplug'/> ... </hyperv> ... </features> ...

NVIDIA guest drivers also require hiding the KVM CPU leaf:

XML
... <features> ... <kvm> <hidden state='on'/> </kvm> <ioapic driver="kvm"/> ... </features> ...

vBIOS Patching

How to patch NVIDIA vBIOS

Only NVIDIA GPU's need to be patched

To get a rom for your GPU you can either download one from here or use nvflash to dump the bios currently on your GPU.

Use the dumped/downloaded vbios and open it in a hex editor.

Search for the strings "VIDEO".

Then you have to search for the first U. that is in front of VIDEO.

Delete all of the code above the U. then save your patched vbios.

To add the patched rom, in hostdev add rom, only for the VGA pci:

XML
... <hostdev mode='subsystem' type='pci' managed='yes'> <source> ... </source> <rom file="/path/to/patched-vbios.rom"/> ... </hostdev> ...

CPU Pinning

My setup is an AMD Ryzen 7 3700X which has 8 physical cores and 16 threads (2 threads per core).

How to bind the threads to the core

It's very important that when we passthrough a core, we include its sibling. To get a sense of your cpu topology, use the command lscpu -e. A matching core id (i.e. "CORE" column) means that the associated threads (i.e. "CPU" column) run on the same physical core.

CPU NODE SOCKET CORE L1d:L1i:L2:L3 ONLINE    MAXMHZ    MINMHZ
  0    0      0    0 0:0:0:0          yes 4823.4370 2200.0000
  1    0      0    1 1:1:1:0          yes 4559.7651 2200.0000
  2    0      0    2 2:2:2:0          yes 4689.8428 2200.0000
  3    0      0    3 3:3:3:0          yes 4426.1709 2200.0000
  4    0      0    4 4:4:4:1          yes 5224.2178 2200.0000
  5    0      0    5 5:5:5:1          yes 5090.6250 2200.0000
  6    0      0    6 6:6:6:1          yes 5224.2178 2200.0000
  7    0      0    7 7:7:7:1          yes 4957.0308 2200.0000
  8    0      0    0 0:0:0:0          yes 4823.4370 2200.0000
  9    0      0    1 1:1:1:0          yes 4559.7651 2200.0000
 10    0      0    2 2:2:2:0          yes 4689.8428 2200.0000
 11    0      0    3 3:3:3:0          yes 4426.1709 2200.0000
 12    0      0    4 4:4:4:1          yes 5224.2178 2200.0000
 13    0      0    5 5:5:5:1          yes 5090.6250 2200.0000
 14    0      0    6 6:6:6:1          yes 5224.2178 2200.0000
 15    0      0    7 7:7:7:1          yes 4957.0308 2200.0000

According to the logic seen above, here are my core and their threads binding.

Core 1: 0, 8
Core 2: 1, 9
Core 3: 2, 10
Core 4: 3, 11
Core 5: 4, 12
Core 6: 5, 13
Core 7: 6, 14
Core 8: 7, 15

In this example, I want to get 1 core for the host and 7 cores for the guest. I will let the core 1 for my host, so 0 and 8 are the logical threads.

I show you the final result, everything will be explained below.

XML
... <vcpu placement="static">14</vcpu> <iothreads>1</iothreads> <cputune> <vcpupin vcpu="0" cpuset="1"/> <vcpupin vcpu="1" cpuset="9"/> <vcpupin vcpu="2" cpuset="2"/> <vcpupin vcpu="3" cpuset="10"/> <vcpupin vcpu="4" cpuset="3"/> <vcpupin vcpu="5" cpuset="11"/> <vcpupin vcpu="6" cpuset="4"/> <vcpupin vcpu="7" cpuset="12"/> <vcpupin vcpu="8" cpuset="5"/> <vcpupin vcpu="9" cpuset="13"/> <vcpupin vcpu="10" cpuset="6"/> <vcpupin vcpu="11" cpuset="14"/> <vcpupin vcpu="12" cpuset="7"/> <vcpupin vcpu="13" cpuset="15"/> <emulatorpin cpuset="0,8"/> <iothreadpin iothread="1" cpuset="0,8"/> </cputune> ... </domain>

XML

...
  <vcpu placement="static">14</vcpu>
  <iothreads>1</iothreads>
  <cputune>
    <vcpupin vcpu="0" cpuset="1"/>
    <vcpupin vcpu="1" cpuset="9"/>
    <vcpupin vcpu="2" cpuset="2"/>
    <vcpupin vcpu="3" cpuset="10"/>
    <vcpupin vcpu="4" cpuset="3"/>
    <vcpupin vcpu="5" cpuset="11"/>
    <vcpupin vcpu="6" cpuset="4"/>
    <vcpupin vcpu="7" cpuset="12"/>
    <vcpupin vcpu="8" cpuset="5"/>
    <vcpupin vcpu="9" cpuset="13"/>
    <vcpupin vcpu="10" cpuset="6"/>
    <vcpupin vcpu="11" cpuset="14"/>
    <vcpupin vcpu="12" cpuset="7"/>
    <vcpupin vcpu="13" cpuset="15"/>
    <emulatorpin cpuset="0,8"/>
    <iothreadpin iothread="1" cpuset="0,8"/>
  </cputune>
  ...
</domain>

Explanations of cpu pinning

Number of threads to passthrough

<vcpu placement="static">14</vcpu>

Same number as the iothreadpin below

<iothreads>1</iothreads>

cpuset corresponds to the bindings of your host core

<cputune>
  ...
  <emulatorpin cpuset="0,8"/>
  <iothreadpin iothread="1" cpuset="0,8"/>
</cputune>

vcpu corresponds to the guest cores, increment by 1 starting with 0.

cpuset correspond to your threads you want to passthrough. It is necessary that your core and their threads binding follow each other.

<cputune>
  <vcpupin vcpu="0" cpuset="1"/>
  <vcpupin vcpu="1" cpuset="9"/>
  <vcpupin vcpu="2" cpuset="2"/>
  <vcpupin vcpu="3" cpuset="10"/>
  <vcpupin vcpu="4" cpuset="3"/>
  <vcpupin vcpu="5" cpuset="11"/>
  <vcpupin vcpu="6" cpuset="4"/>
  <vcpupin vcpu="7" cpuset="12"/>
  <vcpupin vcpu="8" cpuset="5"/>
  <vcpupin vcpu="9" cpuset="13"/>
  <vcpupin vcpu="10" cpuset="6"/>
  <vcpupin vcpu="11" cpuset="14"/>
  <vcpupin vcpu="12" cpuset="7"/>
  <vcpupin vcpu="13" cpuset="15"/>
  ...
</cputune>

You need to match your CPU pathrough (Update cores and threads if needed).

XML
... <cpu mode="host-passthrough" check="none" migratable="on"> <topology sockets="1" dies="1" cores="7" threads="2"/> <cache mode="passthrough"/> <feature policy="require" name="topoext"/> </cpu> ... </domain>

XML

...
  <cpu mode="host-passthrough" check="none" migratable="on">
    <topology sockets="1" dies="1" cores="7" threads="2"/>
    <cache mode="passthrough"/>
    <feature policy="require" name="topoext"/>
  </cpu>
  ...
</domain>

Hyper-V Enlightenments

Hyper-V enlightenments help the guest VM handle virtualization tasks.

More documentation on fossies.org for qemu enlightenments.

XML
... <qemu:commandline> ... <qemu:arg value="-rtc"/> <qemu:arg value="base=localtime"/> <qemu:arg value="-cpu"/> <qemu:arg value="host,host-cache-info=on,kvm=off,l3-cache=on,kvm-hint-dedicated=on,migratable=no,hv_relaxed,hv_spinlocks=0x1fff,hv_vapic,hv_time,hv_vendor_id=buttplug,+invtsc,+topoext"/> </qemu:commandline> </devices>

XML

...
  <qemu:commandline>
    ...
    <qemu:arg value="-rtc"/>
    <qemu:arg value="base=localtime"/>
    <qemu:arg value="-cpu"/>
    <qemu:arg value="host,host-cache-info=on,kvm=off,l3-cache=on,kvm-hint-dedicated=on,migratable=no,hv_relaxed,hv_spinlocks=0x1fff,hv_vapic,hv_time,hv_vendor_id=buttplug,+invtsc,+topoext"/>
  </qemu:commandline>
</devices>

You can alternatively use this config

I do not use this configuration because I experienced mouse latency in games.

XML
<features> ... <hyperv> ... <vpindex state='on'/> <synic state='on'/> <stimer state='on'/> <reset state='on'/> <frequencies state='on'/> </hyperv> ... </features>

Disk Tuning

For more explanations on virtio scsi, check bryansteiner tutorial.

Make sure you have iothreads in your xml.

XML
<devices> ... <disk type='file' device='disk'> <driver name='qemu' type='raw' cache='none' io='threads' discard='unmap' queues='8'/> <source dev='/path/to/your/vm.img'/> <target dev='sdc' bus='scsi'/> <address type='drive' controller='0' bus='0' target='0' unit='2'/> </disk> ... <controller type='scsi' index='0' model='virtio-scsi'> <driver iothread='1' queues='8'/> <address type='pci' domain='0x0000' bus='0x03' slot='0x00' function='0x0'/> </controller> ... </devices>

XML

<devices>
  ...
  <disk type='file' device='disk'>
      <driver name='qemu' type='raw' cache='none' io='threads' discard='unmap' queues='8'/>
      <source dev='/path/to/your/vm.img'/>
      <target dev='sdc' bus='scsi'/>
      <address type='drive' controller='0' bus='0' target='0' unit='2'/>
  </disk>
  ...
  <controller type='scsi' index='0' model='virtio-scsi'>
      <driver iothread='1' queues='8'/>
      <address type='pci' domain='0x0000' bus='0x03' slot='0x00' function='0x0'/>
  </controller>       
  ...
</devices>

Hugepages

Memory (RAM) is divided up into basic segments called pages. By default, the x86 architecture has a page size of 4KB. CPUs utilize pages within the built in memory management unit (MMU). Although the standard page size is suitable for many tasks, hugepages are a mechanism that allow the Linux kernel to take advantage of large amounts of memory with reduced overhead. Hugepages can vary in size anywhere from 2MB to 1GB.

Many tutorials will have you reserve hugepages for your guest VM at host boot-time. There's a significant downside to this approach: a portion of RAM will be unavailable to your host even when the VM is inactive. In bryansteiner setup, he chose to allocate hugepages before the VM starts and deallocate those pages on VM shutdown.

Update your kvm config.

/etc/libvirt/hooks/kvm.conf
`# CONFIG VM_MEMORY=13312 # VIRSH ...`

VM_MEMORY in MiB is the memory allocated tho the guest.

Create Alloc Hugepages Script

vim /etc/libvirt/hooks/qemu.d/${KVM_NAME}/prepare/begin/alloc_hugepages.sh
chmod +x /etc/libvirt/hooks/qemu.d/${KVM_NAME}/prepare/begin/alloc_hugepages.sh

/etc/libvirt/hooks/qemu.d/VM_NAME/prepare/begin/alloc_hugepages.sh
#!/bin/bash ## Load the config file source "/etc/libvirt/hooks/kvm.conf" ## Calculate number of hugepages to allocate from memory (in MB) HUGEPAGES="$(($VM_MEMORY/$(($(grep Hugepagesize /proc/meminfo \| awk '{print $2}')/1024))))" echo "Allocating hugepages..." echo $HUGEPAGES > /proc/sys/vm/nr_hugepages ALLOC_PAGES=$(cat /proc/sys/vm/nr_hugepages) TRIES=0 while (( $ALLOC_PAGES != $HUGEPAGES && $TRIES < 1000 )) do echo 1 > /proc/sys/vm/compact_memory ## defrag ram echo $HUGEPAGES > /proc/sys/vm/nr_hugepages ALLOC_PAGES=$(cat /proc/sys/vm/nr_hugepages) echo "Succesfully allocated $ALLOC_PAGES / $HUGEPAGES" let TRIES+=1 done if [ "$ALLOC_PAGES" -ne "$HUGEPAGES" ] then echo "Not able to allocate all hugepages. Reverting..." echo 0 > /proc/sys/vm/nr_hugepages exit 1 fi

/etc/libvirt/hooks/qemu.d/VM_NAME/prepare/begin/alloc_hugepages.sh

#!/bin/bash

## Load the config file
source "/etc/libvirt/hooks/kvm.conf"

## Calculate number of hugepages to allocate from memory (in MB)
HUGEPAGES="$(($VM_MEMORY/$(($(grep Hugepagesize /proc/meminfo | awk '{print $2}')/1024))))"

echo "Allocating hugepages..."
echo $HUGEPAGES > /proc/sys/vm/nr_hugepages
ALLOC_PAGES=$(cat /proc/sys/vm/nr_hugepages)

TRIES=0
while (( $ALLOC_PAGES != $HUGEPAGES && $TRIES < 1000 ))
do
  echo 1 > /proc/sys/vm/compact_memory            ## defrag ram
  echo $HUGEPAGES > /proc/sys/vm/nr_hugepages
  ALLOC_PAGES=$(cat /proc/sys/vm/nr_hugepages)
  echo "Succesfully allocated $ALLOC_PAGES / $HUGEPAGES"
  let TRIES+=1
done

if [ "$ALLOC_PAGES" -ne "$HUGEPAGES" ]
then
  echo "Not able to allocate all hugepages. Reverting..."
  echo 0 > /proc/sys/vm/nr_hugepages
  exit 1
fi

Create Dealloc Hugepages Script

vim /etc/libvirt/hooks/qemu.d/${KVM_NAME}/release/end/dealloc_hugepages.sh
chmod +x /etc/libvirt/hooks/qemu.d/${KVM_NAME}/release/end/dealloc_hugepages.sh

/etc/libvirt/hooks/qemu.d/VM_NAME/release/end/dealloc_hugepages.sh
#!/bin/bash echo 0 > /proc/sys/vm/nr_hugepages

XML
... <memory unit="KiB">13631488</memory> <currentMemory unit="KiB">13631488</currentMemory> <memoryBacking> <hugepages/> </memoryBacking> ... </domain>

The memory need to match your VM_MEMORY from your config (to convert KiB to MiB you need to divide by 1024).

CPU Governor

This performance tweak takes advantage of the CPU frequency scaling governor in Linux.

Create CPU Performance Script

vim /etc/libvirt/hooks/qemu.d/${KVM_NAME}/prepare/begin/cpu_mode_performance.sh
chmod +x /etc/libvirt/hooks/qemu.d/${KVM_NAME}/prepare/begin/cpu_mode_performance.sh

/etc/libvirt/hooks/qemu.d/VM_NAME/prepare/begin/cpu_mode_performance.sh
#!/bin/bash ## Enable CPU governor performance mode cat /sys/devices/system/cpu/cpu/cpufreq/scaling_governor for file in /sys/devices/system/cpu/cpu/cpufreq/scaling_governor; do echo "performance" > $file; done cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

/etc/libvirt/hooks/qemu.d/VM_NAME/prepare/begin/cpu_mode_performance.sh

#!/bin/bash

## Enable CPU governor performance mode
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
for file in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor; do echo "performance" > $file; done
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

Create CPU Ondemand Script

vim /etc/libvirt/hooks/qemu.d/${KVM_NAME}/release/end/cpu_mode_ondemand.sh
chmod +x /etc/libvirt/hooks/qemu.d/${KVM_NAME}/release/end/cpu_mode_ondemand.sh

/etc/libvirt/hooks/qemu.d/VM_NAME/release/end/cpu_mode_ondemand.sh
#!/bin/bash ## Enable CPU governor on-demand mode cat /sys/devices/system/cpu/cpu/cpufreq/scaling_governor for file in /sys/devices/system/cpu/cpu/cpufreq/scaling_governor; do echo "ondemand" > $file; done cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

/etc/libvirt/hooks/qemu.d/VM_NAME/release/end/cpu_mode_ondemand.sh

#!/bin/bash

## Enable CPU governor on-demand mode
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
for file in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor; do echo "ondemand" > $file; done
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

Update drivers on Windows

To get the network working properly you need to install the drivers.

In Device Manager update drivers with the local virtio iso /path/to/virtio-driver.

Enable Hyper-V on Windows

Enable Hyper-V using PowerShell:

Enable-WindowsOptionalFeature -Online -FeatureName Microsoft-Hyper-V -All

Enable the Hyper-V through Settings:

Search for Turn Windows Features on or off, select Hyper-V and click Ok.

Mageas/single-gpu-passthrough

Table Of Contents