8 PCI(e) passthrough is a mechanism to give a virtual machine control over
9 a PCI device from the host. This can have some advantages over using
10 virtualized hardware, for example lower latency, higher performance, or more
11 features (e.g., offloading).
13 But, if you pass through a device to a virtual machine, you cannot use that
14 device anymore on the host or in any other VM.
19 Since passthrough is a feature which also needs hardware support, there are
20 some requirements to check and preparations to be done to make it work.
25 Your hardware needs to support `IOMMU` (*I*/*O* **M**emory **M**anagement
26 **U**nit) interrupt remapping, this includes the CPU and the mainboard.
28 Generally, Intel systems with VT-d, and AMD systems with AMD-Vi support this.
29 But it is not guaranteed that everything will work out of the box, due
30 to bad hardware implementation and missing or low quality drivers.
32 Further, server grade hardware has often better support than consumer grade
33 hardware, but even then, many modern system can support this.
35 Please refer to your hardware vendor to check if they support this feature
36 under Linux for your specific setup
42 Once you ensured that your hardware supports passthrough, you will need to do
43 some configuration to enable PCI(e) passthrough.
48 The IOMMU has to be activated on the kernel commandline. The easiest way is to
49 enable trough grub. Edit `'/etc/default/grub'' and add the following to the
50 'GRUB_CMDLINE_LINUX_DEFAULT' variable:
63 [[qm_pci_passthrough_update_grub]]
64 To bring this change in effect, make sure you run:
72 You have to make sure the following modules are loaded. This can be achieved by
73 adding them to `'/etc/modules''
82 [[qm_pci_passthrough_update_initramfs]]
83 After changing anything modules related, you need to refresh your
84 `initramfs`. On {pve} this can be done by executing:
87 # update-initramfs -u -k all
92 Finally reboot to bring the changes into effect and check that it is indeed
96 # dmesg | grep -e DMAR -e IOMMU -e AMD-Vi
99 should display that `IOMMU`, `Directed I/O` or `Interrupt Remapping` is
100 enabled, depending on hardware and kernel the exact message can vary.
102 It is also important that the device(s) you want to pass through
103 are in a *separate* `IOMMU` group. This can be checked with:
106 # find /sys/kernel/iommu_groups/ -type l
109 It is okay if the device is in an `IOMMU` group together with its functions
110 (e.g. a GPU with the HDMI Audio device) or with its root port or PCI(e) bridge.
115 Some platforms handle their physical PCI(e) slots differently. So, sometimes
116 it can help to put the card in a another PCI(e) slot, if you do not get the
117 desired `IOMMU` group separation.
123 For some platforms, it may be necessary to allow unsafe interrupts.
124 For this add the following line in a file ending with `.conf' file in
128 options vfio_iommu_type1 allow_unsafe_interrupts=1
131 Please be aware that this option can make your system unstable.
134 GPU Passthrough Notes
135 ^^^^^^^^^^^^^^^^^^^^^
137 It is not possible to display the frame buffer of the GPU via NoVNC or SPICE on
138 the {pve} web interface.
140 When passing through a whole GPU or a vGPU and graphic output is wanted, one
141 has to either physically connect a monitor to the card, or configure a remote
142 desktop software (for example, VNC or RDP) inside the guest.
144 If you want to use the GPU as a hardware accelerator, for example, for
145 programs using OpenCL or CUDA, this is not required.
147 Host Device Passthrough
148 ~~~~~~~~~~~~~~~~~~~~~~~
150 The most used variant of PCI(e) passthrough is to pass through a whole
151 PCI(e) card, for example a GPU or a network card.
157 In this case, the host cannot use the card. There are two methods to achieve
160 * pass the device IDs to the options of the 'vfio-pci' modules by adding
163 options vfio-pci ids=1234:5678,4321:8765
166 to a .conf file in */etc/modprobe.d/* where `1234:5678` and `4321:8765` are
167 the vendor and device IDs obtained by:
173 * blacklist the driver completely on the host, ensuring that it is free to bind
174 for passthrough, with
180 in a .conf file in */etc/modprobe.d/*.
182 For both methods you need to
183 xref:qm_pci_passthrough_update_initramfs[update the `initramfs`] again and
186 [[qm_pci_passthrough_vm_config]]
189 To pass through the device you need to set the *hostpciX* option in the VM
190 configuration, for example by executing:
193 # qm set VMID -hostpci0 00:02.0
196 If your device has multiple functions (e.g., ``00:02.0`' and ``00:02.1`' ),
197 you can pass them through all together with the shortened syntax ``00:02`'
199 There are some options to which may be necessary, depending on the device
202 * *x-vga=on|off* marks the PCI(e) device as the primary GPU of the VM.
203 With this enabled the *vga* configuration option will be ignored.
205 * *pcie=on|off* tells {pve} to use a PCIe or PCI port. Some guests/device
206 combination require PCIe rather than PCI. PCIe is only available for 'q35'
209 * *rombar=on|off* makes the firmware ROM visible for the guest. Default is on.
210 Some PCI(e) devices need this disabled.
212 * *romfile=<path>*, is an optional path to a ROM file for the device to use.
213 This is a relative path under */usr/share/kvm/*.
217 An example of PCIe passthrough with a GPU set to primary:
220 # qm set VMID -hostpci0 02:00,pcie=on,x-vga=on
227 When passing through a GPU, the best compatibility is reached when using
228 'q35' as machine type, 'OVMF' ('EFI' for VMs) instead of SeaBIOS and PCIe
229 instead of PCI. Note that if you want to use 'OVMF' for GPU passthrough, the
230 GPU needs to have an EFI capable ROM, otherwise use SeaBIOS instead.
235 Another variant for passing through PCI(e) devices, is to use the hardware
236 virtualization features of your devices, if available.
238 'SR-IOV' (**S**ingle-**R**oot **I**nput/**O**utput **V**irtualization) enables
239 a single device to provide multiple 'VF' (**V**irtual **F**unctions) to the
240 system. Each of those 'VF' can be used in a different VM, with full hardware
241 features and also better performance and lower latency than software
244 Currently, the most common use case for this are NICs (**N**etwork
245 **I**nterface **C**ard) with SR-IOV support, which can provide multiple VFs per
246 physical port. This allows using features such as checksum offloading, etc. to
247 be used inside a VM, reducing the (host) CPU overhead.
253 Generally, there are two methods for enabling virtual functions on a device.
255 * sometimes there is an option for the driver module e.g. for some
262 which could be put file with '.conf' ending under */etc/modprobe.d/*.
263 (Do not forget to update your initramfs after that)
265 Please refer to your driver module documentation for the exact
266 parameters and options.
268 * The second, more generic, approach is using the `sysfs`.
269 If a device and driver supports this you can change the number of VFs on
270 the fly. For example, to setup 4 VFs on device 0000:01:00.0 execute:
273 # echo 4 > /sys/bus/pci/devices/0000:01:00.0/sriov_numvfs
276 To make this change persistent you can use the `sysfsutils` Debian package.
277 After installation configure it via */etc/sysfs.conf* or a `FILE.conf' in
283 After creating VFs, you should see them as separate PCI(e) devices when
284 outputting them with `lspci`. Get their ID and pass them through like a
285 xref:qm_pci_passthrough_vm_config[normal PCI(e) device].
290 For this feature, platform support is especially important. It may be necessary
291 to enable this feature in the BIOS/EFI first, or to use a specific PCI(e) port
292 for it to work. In doubt, consult the manual of the platform or contact its
295 Mediated Devices (vGPU, GVT-g)
296 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
298 Mediated devices are another method to use reuse features and performance from
299 physical hardware for virtualized hardware. These are found most common in
300 virtualized GPU setups such as Intels GVT-g and Nvidias vGPUs used in their
303 With this, a physical Card is able to create virtual cards, similar to SR-IOV.
304 The difference is that mediated devices do not appear as PCI(e) devices in the
305 host, and are such only suited for using in virtual machines.
311 In general your card's driver must support that feature, otherwise it will
312 not work. So please refer to your vendor for compatbile drivers and how to
315 Intels drivers for GVT-g are integraded in the Kernel and should work
316 with the 5th, 6th and 7th generation Intel Core Processors, further E3 v4, E3
317 v5 and E3 v6 Xeon Processors are supported.
319 To enable it for Intel Graphcs, you have to make sure to load the module
320 'kvmgt' (for example via `/etc/modules`) and to enable it on the Kernel
321 commandline. For this you can edit `'/etc/default/grub'' and add the following
322 to the 'GRUB_CMDLINE_LINUX_DEFAULT' variable:
328 After that remember to
329 xref:qm_pci_passthrough_update_initramfs[update the `initramfs`],
330 xref:qm_pci_passthrough_update_grub[update grub] and
336 To use a mediated device, simply specify the `mdev` property on a `hostpciX`
337 VM configuration option.
339 You can get the supported devices via the 'sysfs'. For example, to list the
340 supported types for the device '0000:00:02.0' you would simply execute:
343 # ls /sys/bus/pci/devices/0000:00:02.0/mdev_supported_types
346 Each entry is a directory which contains the following important files:
348 * 'available_instances' contains the amount of still available instances of
349 this type, each 'mdev' use in a VM reduces this.
350 * 'description' contains a short description about the capabilities of the type
351 * 'create' is the endpoint to create such a device, {pve} does this
352 automatically for you, if a 'hostpciX' option with `mdev` is configured.
354 Example configuration with an `Intel GVT-g vGPU` (`Intel Skylake 6700k`):
357 # qm set VMID -hostpci0 00:02.0,mdev=i915-GVTg_V5_4
360 With this set, {pve} automatically creates such a device on VM start, and
361 cleans it up again when the VM stops.
368 * link:/wiki/Pci_passthrough[PCI Passthrough Examples]