QEMU/KVM and GPU Passthrough in Details

Vir­tu­al Ma­chines

This is a doc­u­men­ta­tion of my ex­pe­ri­ence in cre­at­ing a vir­tu­al ma­chine ca­pa­ble to run Win­dows 10 guest OS (for desk­top op­er­a­tions) with­in my home serv­er which op­er­at­ing sys­tem is Ubun­tu Serv­er. The Win­dows 10 guest it­self must be ca­pa­ble to run vir­tu­al­iza­tion in or­der to use WSL2 in­side. The passthrough op­tion is not manda­to­ry for my user case but I de­cid­ed to try it.

The Host Sys­tem

The host op­er­at­ing sys­tem is Ubun­tu Serv­er 20.04 with ker­nel 5.4. Al­so Prox­moxVE 7.2 with ker­nel 5.15 is valid test­ed host. The host CPU is In­tel Xeon but in the this man­u­al will be pro­vid­ed al­so AMD spe­cif­ic pa­ra­me­ters and com­mands.

Host Hard­ware

• Leno­vo ThinkServ­er TD350 with Xeon E5-2673 v3 » with the lat­est firmware
  • 2x Gi­ga­bit Eth­er­net Adapter In­tel i210
  • 2x 8GB DDR4-2133MHz (1Rx4) RDIMM
• NVIDIA NVS 315 at PCIE Slot 1 » FOR PASSTHROUGH
• On­board As­peed AST2400 with 16MB mem­o­ry, 1920x1200@60Hz max. (Roms and Dri­vers)

#Out­put | #CLI | #Hard­ware de­tails

lspci | grep VGA

#Out­put

02:00.0 VGA compatible controller: NVIDIA Corporation GF119 [NVS 315] (rev a1)
07:00.0 VGA compatible controller: ASPEED Technology, Inc. ASPEED Graphics Family (rev 30)

sudo lshw -class memory | sed -n -e '/bank:0/,/bank:1/p' -e '/bank:2/,/bank:3/p'| sed -e 's/^.*bank.*$//'

#Out­put

description: DIMM DDR4 Synchronous 2133 MHz (0.5 ns)
product: HMA41GR7AFR4N-TF
vendor: Hynix Semiconductor
physical id: 0
serial: 517692CB
slot: CPU1 DIMM A1
size: 8GiB
width: 64 bits
clock: 2133MHz (0.5ns)

description: DIMM DDR4 Synchronous 2133 MHz (0.5 ns)
product: HMA41GR7AFR4N-TF
vendor: Hynix Semiconductor
physical id: 2
serial: 51769065
slot: CPU1 DIMM B1
size: 8GiB
width: 64 bits
clock: 2133MHz (0.5ns)

lscpu | sed -nr '/Model name/ s/.*:\s*(.*) @ .*/\1/p'

#Out­put

Intel(R) Xeon(R) CPU E5-2673 v3

sudo lscpu

#Out­put

Architecture:                    x86_64
CPU op-mode(s):                  32-bit, 64-bit
Byte Order:                      Little Endian
Address sizes:                   46 bits physical, 48 bits virtual
CPU(s):                          24
On-line CPU(s) list:             0-23
Thread(s) per core:              2
Core(s) per socket:              12
Socket(s):                       1
NUMA node(s):                    1
Vendor ID:                       GenuineIntel
CPU family:                      6
Model:                           63
Model name:                      Intel(R) Xeon(R) CPU E5-2673 v3 @ 2.40GHz
Stepping:                        2
CPU MHz:                         1197.645
CPU max MHz:                     3100.0000
CPU min MHz:                     1200.0000
BogoMIPS:                        4789.15
Virtualization:                  VT-x
L1d cache:                       384 KiB
L1i cache:                       384 KiB
L2 cache:                        3 MiB
L3 cache:                        30 MiB
NUMA node0 CPU(s):               0-23
Vulnerability Itlb multihit:     KVM: Mitigation: Split huge pages
Vulnerability L1tf:              Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds:               Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown:          Mitigation; PTI
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1:        Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:        Mitigation; Full generic retpoline, IBPB conditional, IBRS_FW, STIBP conditional, RSB filling
Vulnerability Srbds:             Not affected
Vulnerability Tsx async abort:   Not affected
Flags:                           fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constan
                                 t_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtp
                                 r pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm cpuid_fault epb invpcid_single pti intel_ppin ssb
                                 d ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm xsaveopt cqm_llc cqm_occup_llc dther
                                 m ida arat pln pts md_clear flush_l1d

Test the Vir­tu­al­iza­tion Ca­pa­bil­i­ties of the Sys­tem

Check weath­er the sys­tem sup­ports vir­tu­al­i­sa­tion. The fol­low­ing com­mand must re­turn at least 1:

egrep -c '(vmx|svm)' /proc/cpuinfo

#Ref | #Out­put | #CLI | #Out­put

24

An­oth­er ap­proach is to use the com­mand kvm-ok from the pack­age cpu-check­er:

sudo apt install cpu-checker && kvm-ok

INFO: /dev/kvm exists
KVM acceleration can be used

En­able the rel­e­vant vir­tu­al­i­sa­tion set­tings (VT‑x/​​​AMD‑V) in the UEFI/BIOS:

• 2N: Vir­tu­al­iza­tion (VT‑x/​​​AMD‑V) – En­abling vir­tu­al­iza­tion on your com­put­er for run­ning 2N® Ac­cess Com­man­der
• Fe­do­ra Doc­u­men­ta­tion: En­abling In­tel VT and AMD‑V vir­tu­al­iza­tion hard­ware ex­ten­sions in BIOS

Set­ting-up the PCI Passthrough

This sec­tion is pri­ma­ry based on the great Math­ias Hueber's man­u­al, so not on­ly the com­mands, but some of the sen­tences are copy/​​​paste from there.

En­abling IOM­MU

In or­der to en­abling the IOM­MU fea­ture we must ed­it the con­fig­u­ra­tion file /​​​etc/​​​default/​​​grub, as fol­low:

sudo nano /etc/default/grub # cat /etc/default/grub | grep 'GRUB_CMDLINE_LINUX_DEFAULT'

GRUB_CMDLINE_LINUX_DEFAULT="intel_iommu=on iommu=pt"   # For Intel CPU (current case)

GRUB_CMDLINE_LINUX_DEFAULT="amd_iommu=on iommu=pt"     # For AMD CPU

#Ref | #Ex­pla­na­tions | #Ap­ply

Short ex­pla­na­tions:

• In com­put­ing, an input–output mem­o­ry man­age­ment unit (IOM­MU) is a mem­o­ry man­age­ment unit (MMU) that con­nects a direct-memory-access–capable (DMA-ca­pa­ble) I/O bus to the main mem­o­ry. Like a tra­di­tion­al MMU, which trans­lates CPU-vis­i­ble vir­tu­al ad­dress­es to phys­i­cal ad­dress­es, the IOM­MU maps de­vice-vis­i­ble vir­tu­al ad­dress­es (al­so called de­vice ad­dress­es or I/O ad­dress­es in this con­text) to phys­i­cal ad­dress­es. Some units al­so pro­vide mem­o­ry pro­tec­tion from faulty or ma­li­cious de­vices.
• To en­able sin­gle-root input/​​​output vir­tu­al­iza­tion (SR-IOV) in the ker­nel, con­fig­ure intel_iommu=on in the grub file. To get the best per­for­mance, add iommu=pt (pass-through) to the grub file when us­ing SR-IOV. When in pass-through mode, the adapter does not need to use DMA trans­la­tion to the mem­o­ry, and this im­proves the per­for­mance. iommu=pt is need­ed main­ly with hy­per­vi­sor per­for­mance is need­ed.
• The Open Vir­tu­al Ma­chine Firmware (OVMF) is a project to en­able UE­FI sup­port for vir­tu­al ma­chines. Start­ing with Lin­ux 3.9 and re­cent ver­sions of QE­MU, it is now pos­si­ble to passthrough a graph­ics card, of­fer­ing the VM na­tive graph­ics per­for­mance which is use­ful for graph­ic-in­ten­sive tasks.

Ref­er­ences:

• Mel­lanox Tech: Un­der­stand­ing the iom­mu Lin­ux grub File Con­fig­u­ra­tion
• Wikipedia: Sin­gle-root in­put-out­put vir­tu­al­iza­tion
• Wikipedia: Input–output mem­o­ry man­age­ment unit

Up­date the boot man­ag­er con­fig­u­ra­tion and re­boot the sys­tem.

sudo update-grub
sudo systemctl reboot

For sys­temd boot man­ag­er as used in Pop!_OS.

#Pop!_OS

One can use the ker­nel­stub mod­ule, on sys­temd boot­ing op­er­at­ing sys­tems, in or­der to pro­vide boot pa­ra­me­ters. Use it like so:

sudo kernelstub -o "amd_iommu=on amd_iommu=pt"

And lat­er to do the iso­la­tion use (with cor­rect ids):

sudo kernelstub --add-options "vfio-pci.ids=10de:1b80,10de:10f0,8086:1533"

Ref­er­ences:

• Math­i­asHue­ber: Vir­tu­al ma­chines with PCI passthrough on Ubun­tu 20.04, straight­for­ward guide for gam­ing on a vir­tu­al ma­chine

Af­ter the re­boot ver­i­fy does IOM­MU is en­abled:

sudo dmesg | grep -i 'IOMMU'

#Out­put

[0.000000] Command line: BOOT_IMAGE=/boot/vmlinuz-5.4.0-77-generic root=UUID=09e7c...14 ro intel_iommu=on iommu=pt
[0.068270] Kernel command line: BOOT_IMAGE=/boot/vmlinuz-5.4.0-77-generic root=UUID=09e7c...14 ro intel_iommu=on iommu=pt
[0.068326] DMAR: IOMMU enabled
[0.140127] DMAR-IR: IOAPIC id 1 under DRHD base  0xfbffc000 IOMMU 0
[0.140129] DMAR-IR: IOAPIC id 2 under DRHD base  0xfbffc000 IOMMU 0
[0.480605] iommu: Default domain type: Passthrough (set via kernel command line)
[0.764139] pci 0000:ff:0b.0: Adding to iommu group 0
[0.764165] pci 0000:ff:0b.1: Adding to iommu group 0
[0.764188] pci 0000:ff:0b.2: Adding to iommu group 0
[0.764348] pci 0000:ff:0c.0: Adding to iommu group 1
...

sudo dmesg | grep -i 'vfio'   # For Intel CPU (current case)

#Out­put

[    0.000000] Command line: BOOT_IMAGE=/boot/vmlinuz-5.4.0-77-generic root=UUID=09e7c8ed-fb55-4a44-8be4-18b1696fc714 ro intel_iommu=on iommu=pt kvm.ignore_msrs=1 irqpoll vfio-pci.ids=10de:107c,10de:0e08 vfio-pci.disable_vga=1
[    0.068558] Kernel command line: BOOT_IMAGE=/boot/vmlinuz-5.4.0-77-generic root=UUID=09e7c8ed-fb55-4a44-8be4-18b1696fc714 ro intel_iommu=on iommu=pt kvm.ignore_msrs=1 irqpoll vfio-pci.ids=10de:107c,10de:0e08 vfio-pci.disable_vga=1
[    0.862286] VFIO - User Level meta-driver version: 0.3
[    0.862594] vfio-pci 0000:02:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=none:owns=none
[    0.880568] vfio_pci: add [10de:107c[ffffffff:ffffffff]] class 0x000000/00000000
[    0.900583] vfio_pci: add [10de:0e08[ffffffff:ffffffff]] class 0x000000/00000000

sudo dmesg |grep 'AMD-Vi'     # For AMD CPU

#Out­put

[    0.607751] pci 0000:00:00.2: AMD-Vi: IOMMU performance counters supported
[    0.608569] pci 0000:00:00.2: AMD-Vi: Found IOMMU cap 0x40
[    0.608569] AMD-Vi: Extended features (0x58f77ef22294a5a): PPR NX GT IA PC GA_vAPIC
[    0.608572] AMD-Vi: Interrupt remapping enabled
[    0.890747] AMD-Vi: AMD IOMMUv2 loaded and initialized

Iden­ti­fi­ca­tion of the Group Con­trollers

In or­der to gen­er­ate a tidy list of your grouped de­vices cre­ate a script as the fol­low.

nano ~/bin/get_iommu_groups.sh && chmod +x ~/bin/get_iommu_groups.sh

#Script

#!/bin/bash
# https://mathiashueber.com/pci-passthrough-ubuntu-2004-virtual-machine/
# change the 9999 if needed
shopt -s nullglob
for d in /sys/kernel/iommu_groups/{0..9999}/devices/*; do
    n=${d#*/iommu_groups/*}; n=${n%%/*}
    printf 'IOMMU Group %s ' "$n"
    lspci -nns "${d##*/}"
done

Run the script and fil­ter the out­put.

get_iommu_groups.sh | grep -iP 'VGA compatible controller|Ethernet controller|SATA controller|USB controller|NVIDIA'

#Out­put

IOMMU Group 30 00:11.4 SATA controller [0106]: Intel Corporation C610/X99 series chipset sSATA Controller [AHCI mode] [8086:8d62] (rev 05)
IOMMU Group 31 00:14.0 USB controller [0c03]: Intel Corporation C610/X99 series chipset USB xHCI Host Controller [8086:8d31] (rev 05)
IOMMU Group 33 00:1a.0 USB controller [0c03]: Intel Corporation C610/X99 series chipset USB Enhanced Host Controller #2 [8086:8d2d] (rev 05)
IOMMU Group 38 00:1d.0 USB controller [0c03]: Intel Corporation C610/X99 series chipset USB Enhanced Host Controller #1 [8086:8d26] (rev 05)
IOMMU Group 39 00:1f.2 SATA controller [0106]: Intel Corporation C610/X99 series chipset 6-Port SATA Controller [AHCI mode] [8086:8d02] (rev 05)
IOMMU Group 40 02:00.0 VGA compatible controller [0300]: NVIDIA Corporation GF119 [NVS 315] [10de:107c] (rev a1)
IOMMU Group 40 02:00.1 Audio device [0403]: NVIDIA Corporation GF119 HDMI Audio Controller [10de:0e08] (rev a1)
IOMMU Group 42 07:00.0 VGA compatible controller [0300]: ASPEED Technology, Inc. ASPEED Graphics Family [1a03:2000] (rev 30)
IOMMU Group 43 08:00.0 Ethernet controller [0200]: Intel Corporation I210 Gigabit Network Connection [8086:1533] (rev 03)
IOMMU Group 44 09:00.0 Ethernet controller [0200]: Intel Corporation I210 Gigabit Network Connection [8086:1533] (rev 03)

We will iso­late IOM­MU Group 40 that con­tains PCI-bus 02:00.0 [de­vice ID 10de:107c] and 02:00.1 [de­vice ID 10de:0e08].

Iso­la­tion of the Guest GPU

In or­der to iso­late the GPU we have two op­tions. Se­lect the de­vices by PCI bus ad­dress or by de­vice ID. Both op­tions have pros and cons. Here we will iso­late VFIO-pci dri­ver by de­vice id. This op­tion should on­ly be used, in case the graph­ic cards (or oth­er de­vices that will be iso­lat­ed) in the sys­tem are not ex­act­ly the same mod­el, oth­er­wise we need to use iso­la­tion by PCI bus, be­cause the de­vices will have an iden­ti­cal IDs.

sudo nano /etc/default/grub # cat /etc/default/grub | grep 'GRUB_CMDLINE_LINUX_DEFAULT'

GRUB_CMDLINE_LINUX_DEFAULT="intel_iommu=on iommu=pt kvm.ignore_msrs=1 kvm.report_ignored_msrs=0 irqpoll vfio-pci.ids=10de:107c,10de:0e08 vfio-pci.disable_vga=1"

#Ref | #Ex­pla­na­tions | #Ap­ply

Short ex­pla­na­tions:

• The com­mand ignore_​​​msrs is on­ly nec­es­sary for Win­dows 10 ver­sions high­er 1803 (oth­er­wise BSOD).
• The com­mand irqpoll is a work around for an er­ror like irq XX: no­body cared (try boot­ing with the "irqpool" op­tion)…, pos­si­bly hang and restart of the host. Ac­tu­al­ly, I think, this prob­lem was solved by en­abling the MSI (Mes­sage Sig­naled In­ter­rupts) op­tion in the guest OS, as it is de­scribed be­low.
• The com­mands vfio-pci.disable_vga=1 is an at­tempt for workaround for my sys­tem, which hangs dur­ing the boot while a mon­i­tor is con­nect­ed to the guest GPU. But ac­tu­al­ly it doesn't change any­thing in my par­tic­u­lar case.

Ref­er­ences:

• Ubun­tu Co­mu­ni­ty Help Wi­ki: De­bug­ging IRQ Prob­lems
• Ubun­tu Blog: Hard­ware dis­cov­ery and ker­nel au­to-con­fig­u­ra­tion in MAAS.
• Google Search.
• Hackmd​.io: Prox­mox 6.2 GPU Passthrough Tu­to­r­i­al

Up­date the boot man­ag­er con­fig­u­ra­tion and re­boot the sys­tem.

#CLI

sudo update-grub
sudo systemctl reboot

Af­ter this re­boot the iso­lat­ed GPU will be ig­nored by the host OS. Now, you have to use the oth­er GPU for the host OS. Af­ter the re­boot, ver­i­fy the Iso­la­tion of the guest GPU by an­a­lyze the out­put of the fol­low­ing com­mand:

sudo lspci -nnv -s 02:00

• Note the lines Ker­nel dri­ver in use: vfio-pci

#Out­put

02:00.0 VGA compatible controller [0300]: NVIDIA Corporation GF119 [NVS 315] [10de:107c] (rev a1) (prog-if 00 [VGA controller])
        Subsystem: NVIDIA Corporation GF119 [NVS 315] [10de:102f]
        Physical Slot: 1
        Flags: bus master, fast devsel, latency 0, IRQ 255, NUMA node 0
        Memory at fa000000 (32-bit, non-prefetchable) [size=16M]
        Memory at 23ff0000000 (64-bit, prefetchable) [size=128M]
        Memory at 23ff8000000 (64-bit, prefetchable) [size=32M]
        I/O ports at d000 [size=128]
        Expansion ROM at fb000000 [disabled] [size=512K]
        Capabilities: [60] Power Management version 3
        Capabilities: [68] MSI: Enable- Count=1/1 Maskable- 64bit+
        Capabilities: [78] Express Endpoint, MSI 00
        Capabilities: [b4] Vendor Specific Information: Len=14 <?>
        Capabilities: [100] Virtual Channel
        Capabilities: [128] Power Budgeting <?>
        Capabilities: [600] Vendor Specific Information: ID=0001 Rev=1 Len=024 <?>
        Kernel driver in use: vfio-pci
        Kernel modules: nvidiafb, nouveau, nvidia_drm, nvidia

02:00.1 Audio device [0403]: NVIDIA Corporation GF119 HDMI Audio Controller [10de:0e08] (rev a1)
        Subsystem: NVIDIA Corporation GF119 HDMI Audio Controller [10de:102f]
        Physical Slot: 1
        Flags: fast devsel, IRQ 255, NUMA node 0
        Memory at fb080000 (32-bit, non-prefetchable) [disabled] [size=16K]
        Capabilities: [60] Power Management version 3
        Capabilities: [68] MSI: Enable- Count=1/1 Maskable- 64bit+
        Capabilities: [78] Express Endpoint, MSI 00
        Kernel driver in use: vfio-pci
        Kernel modules: snd_hda_intel

Con­grat­u­la­tions, the hard­est part is done!

Set­ting-up the Soft­ware En­vi­ron­ment

There is a plen­ty of man­u­als how to in­stall and man­age KVM and cre­ate VMs. So I will post few ref­er­ences that I read and will ex­plain just few things about this process.

In­stall QE­MU, KVM, LIB­VIRT

sudo apt install qemu-system-x86 libvirt-daemon bridge-utils
sudo apt install libvirt-clients virtinst libosinfo-bin ovmf
sudo apt install virt-manager virt-viewer remmina # For desktop user

#Ex­pla­na­tions | #Ap­ply

The above set of pack­ages is for the lat­est De­bian based dis­tri­b­u­tions. For lit­tle bit old­er like Ubun­tu 20.04 try with the fol­low­ing set.

sudo apt install qemu qemu-kvm bridge-utils
sudo apt install libvirt-daemon libvirt-clients virtinst libosinfo-bin ovmf
sudo apt install virt-manager virt-viewer remmina # For desktop user

Ex­pla­na­tion about the pack­ages:

• The qe­mu pack­age (quick em­u­la­tor) is an ap­pli­ca­tion that al­lows you to per­form hard­ware vir­tu­al­iza­tion.
• The qe­mu-kvm pack­age is the main KVM pack­age.
• The lib­vrit-dae­mon is the vir­tu­al­iza­tion dae­mon.
• The bridge-utils pack­age helps you cre­ate a bridge con­nec­tion to al­low oth­er users to ac­cess a vir­tu­al ma­chine oth­er than the host sys­tem.
• The virtinst pack­age con­tains pro­grams to cre­ate and clone vir­tu­al ma­chines. It is a set of com­mand-line tools to cre­ate vir­tu­al ma­chines us­ing lib­virt.
• The li­bosin­fo-bin pack­age con­tains tools for query­ing the os­in­fo data­base via li­bosin­fo… It in­cludes a data­base con­tain­ing de­vice meta­da­ta and pro­vides APIs to match/​​​identify op­ti­mal de­vices for de­ploy­ing an op­er­at­ing sys­tem on a hy­per­vi­sor.
• The ovmf pack­age is UE­FI firmware for 64-bit x86 vir­tu­al ma­chines. Open Vir­tu­al Ma­chine Firmware is a build of EDK II for 64-bit x86 vir­tu­al ma­chines. It in­cludes full sup­port for UE­FI, in­clud­ing Se­cure Boot, al­low­ing use of UE­FI in place of a tra­di­tion­al BIOS in your VM.
• The virt-man­ag­er is an ap­pli­ca­tion for man­ag­ing vir­tu­al ma­chines through a graph­i­cal user in­ter­face.
• The virt-view­er is a SPICE GUI client.
• The rem­mi­na pack­age is re­mote ses­sion man­ag­er it sup­ports RDP, VNC, SSH and SFTP pro­to­cols.

Ver­i­fy whether KVM mod­ule is loaded in­to the loaded and ver­i­fy whether lib­virt dae­mon will start au­to­mat­i­cal­ly:

sudo systemctl is-active libvirtd
lsmod | grep -i kvm

Ba­sic man­age­ment – en­able, start, get the sta­tus or stop and dis­able libvirtd.service:

sudo systemctl (enable|start|status/stop|disable) libvirtd.service

Add your user to the lib­virt and kvm groups in or­der to ex­e­cute re­lat­ed com­mand with­out su­do:

sudo usermod -aG libvirt $USER
sudo usermod -aG kvm $USER
grep "$USER" /etc/group

NVIDIA Ker­nel Mod­ules and Dri­vers at the Host Lev­el

Note this guide cov­ers in­stal­la­tion on Lin­ux serv­er where the host wont use the GPU or any oth­er NVIDIA GPU.

#Ref­er­ences | #Ap­ply

Re­move pre­vi­ous­ly in­stalled NVIDIA dri­vers. And an­a­lyze the out­put of lsmod, dmesg and lsp­ci ‑nnv in or­der to find which mod­ules are re­lat­ed to the guest GPU and black­list them by cre­at­ing a new sec­tion at the bot­tom of the file /etc/modprobe.d/blacklist.conf.

sudo apt remove --purge nvidia-headless-390 && sudo apt autoremove && sudo apt autoclean

dpkg -l | grep -i nvidia

sudo apt-get remove --purge '^nvidia-.*'
sudo apt autoremove && sudo apt autoclean

sudo nano /etc/modprobe.d/blacklist.conf

# Blacklist NVIDIA Modules:
# 'lsmod', 'dmesg' and 'lspci -nnv'
blacklist nvidiafb
blacklist nouveau
blacklist nvidia_drm
blacklist nvidia
blacklist rivafb
blacklist rivatv
blacklist snd_hda_intel

Ref­er­ences:

• Ubun­tu Man­u­als: modprobe.d, modprobe.conf
• Lin­ux­Con­fig: How to black­list a mod­ule on Ubuntu/​​​Debian Lin­ux
• AskUbun­tu: Black­list a Nvidia gpu for qemu/​​​kvm passthrough (black­list one of two NVIDIA cards)

De­ploy fresh OVMF Firmware for VMs

This step is no longer re­quired. The UE­FI im­ages from the lat­est ver­sions of the ovmf pack­age in­stalled above are ro­bust, sta­ble and fast.

#Ref­er­ences | #CLI | #Ap­ply

The de­fault OVMF im­ages that was in­stalled in the pre­vi­ous sec­tion are rel­a­tive­ly old. In ad­di­tion I didn't suc­ceed to set­up WSL2 with these UE­FI im­ages. Al­so with the im­ages pro­vid­ed be­low the VM per­for­mance is about 5% faster.

Down­load and de­ploy an ap­pro­pri­ate pack­age with fresh OVMF im­ages from kraxel​.org/​r​e​p​o​s​/​j​e​n​k​i​n​s​/​edk2/:

mkdir ~/Downloads/kvm-ovmf-kraxel/
cd ~/Downloads/kvm-ovmf-kraxel/
wget https://www.kraxel.org/repos/jenkins/edk2/edk2.git-ovmf-x64-0-20210421.18.g15ee7b7689.noarch.rpm

Ex­tract the pack­age and copy the im­ages at their places:

rpm2cpio edk2.git-ovmf-x64-0-20210421.18.g15ee7b7689.noarch.rpm | cpio -idmv
sudo cp -R usr/share/edk2.git /usr/share/
sudo cp usr/share/qemu/firmware/* /usr/share/qemu/firmware/

Cur­rent­ly I'm us­ing OVMF_VARS-with-csm.fd.

• Fabi­an Lee: KVM: Build­ing the lat­est OVMF firmware for vir­tu­al ma­chines
• Krax­el: Repos­i­to­ry with fresh pre­com­piled ovmf im­ages
• De­bian Wi­ki: Se­cure Boot » Vir­tu­al Ma­chine » Get Se­cure Boot Firmware from Fe­do­ra (ed­k2-ovmf)
• Wikipedia: UE­FI and SMM | Wikipedia: UE­FI and CSM

Ad­di­tion­al Set­up, Is­sues, So­lu­tions and Tips

• Lin­ux Net­work Bridge (refs)
• Virt-man­ag­er Set­ting-up Win­dows Vir­tu­al Ma­chines
• QEMU/KVM and GPU Passthrough Trou­bleshoot­ing
• QEMU/KVM Guest tools
• QEMU/KVM Vir­tu­al ma­chine Man­age­ment

Gen­er­al Ref­er­ences

• Math­ias Hue­ber: VMs with PCI passthrough on Ubun­tu 20.04, straight­for­ward guide for gam­ing on a vir­tu­al ma­chine [the main ref­er­ence]
• Serv­er world: KVM: In­stall, Cre­ate VM, Lib­virt ba­sics, Man­age­ment tools, SPICE, Nest­ed, Mi­gra­tion [sim­ple ex­am­ples]
• Adam Gradzki's Blog: Faster Vir­tu­al Ma­chines on Lin­ux Hosts with GPU Ac­cel­er­a­tion [in­ter­est­ing ad­vanced man­u­al]
• Arch­Lin­ux: PCI passthrough via OVMF | Arch­Lin­ux: PCI passthrough via OVMF » Per­for­mance tun­ing
• Red Hat Cus­tomer Por­tal: Con­fig­ure a host for PCI Passthrough
• KVM on Z: Disk Per­for­mance Hints & Tips [not so use­ful]
• Ubun­tu Com­mu­ni­ty Wi­ki: KVM/​​​Networking
• Red­Hat Open­Shift: In­stalling the QE­MU guest agent on vir­tu­al ma­chines | Prox­mox: Qe­mu-guest-agent
• Get Labs Done: How To In­stall Win­dows 10 on Ubun­tu KVM?
• Get Labs Done: How To In­stall Ubun­tu Desk­top On KVM?
• Ubun­tu Down­loads: In­stall Ubun­tu Core on KVM
• How to In­stall Nvidia Dri­vers on Ubun­tu 20.04
• Is there some­thing like "Vir­tu­al­Box Guest Ad­di­tions" for QEMU/KVM?
• Mouse cur­sor in KVM Ubun­tu Guest! [SOLVED]

Reg­is­ter: Vir­tu­al Ma­chines

Идентифициране

Разработване

Авторизиране

Утвърждаване

Spas

Spas

Spas

Spas

2.07.2022

21.08.2022

23.08.2022

26.09.2022

27120

30514

30588 30475

31908 30477