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
49 xref:sysboot_edit_kernel_cmdline[kernel commandline].
51 The command line parameters are:
67 You have to make sure the following modules are loaded. This can be achieved by
68 adding them to `'/etc/modules''
77 [[qm_pci_passthrough_update_initramfs]]
78 After changing anything modules related, you need to refresh your
79 `initramfs`. On {pve} this can be done by executing:
82 # update-initramfs -u -k all
87 Finally reboot to bring the changes into effect and check that it is indeed
91 # dmesg | grep -e DMAR -e IOMMU -e AMD-Vi
94 should display that `IOMMU`, `Directed I/O` or `Interrupt Remapping` is
95 enabled, depending on hardware and kernel the exact message can vary.
97 It is also important that the device(s) you want to pass through
98 are in a *separate* `IOMMU` group. This can be checked with:
101 # find /sys/kernel/iommu_groups/ -type l
104 It is okay if the device is in an `IOMMU` group together with its functions
105 (e.g. a GPU with the HDMI Audio device) or with its root port or PCI(e) bridge.
110 Some platforms handle their physical PCI(e) slots differently. So, sometimes
111 it can help to put the card in a another PCI(e) slot, if you do not get the
112 desired `IOMMU` group separation.
118 For some platforms, it may be necessary to allow unsafe interrupts.
119 For this add the following line in a file ending with `.conf' file in
123 options vfio_iommu_type1 allow_unsafe_interrupts=1
126 Please be aware that this option can make your system unstable.
129 GPU Passthrough Notes
130 ^^^^^^^^^^^^^^^^^^^^^
132 It is not possible to display the frame buffer of the GPU via NoVNC or SPICE on
133 the {pve} web interface.
135 When passing through a whole GPU or a vGPU and graphic output is wanted, one
136 has to either physically connect a monitor to the card, or configure a remote
137 desktop software (for example, VNC or RDP) inside the guest.
139 If you want to use the GPU as a hardware accelerator, for example, for
140 programs using OpenCL or CUDA, this is not required.
142 Host Device Passthrough
143 ~~~~~~~~~~~~~~~~~~~~~~~
145 The most used variant of PCI(e) passthrough is to pass through a whole
146 PCI(e) card, for example a GPU or a network card.
152 In this case, the host must not use the card. There are two methods to achieve
155 * pass the device IDs to the options of the 'vfio-pci' modules by adding
158 options vfio-pci ids=1234:5678,4321:8765
161 to a .conf file in */etc/modprobe.d/* where `1234:5678` and `4321:8765` are
162 the vendor and device IDs obtained by:
168 * blacklist the driver completely on the host, ensuring that it is free to bind
169 for passthrough, with
175 in a .conf file in */etc/modprobe.d/*.
177 For both methods you need to
178 xref:qm_pci_passthrough_update_initramfs[update the `initramfs`] again and
181 .Verify Configuration
183 To check if your changes were successful, you can use
189 and check your device entry. If it says
192 Kernel driver in use: vfio-pci
195 or the 'in use' line is missing entirely, the device is ready to be used for
198 [[qm_pci_passthrough_vm_config]]
201 To pass through the device you need to set the *hostpciX* option in the VM
202 configuration, for example by executing:
205 # qm set VMID -hostpci0 00:02.0
208 If your device has multiple functions (e.g., ``00:02.0`' and ``00:02.1`' ),
209 you can pass them through all together with the shortened syntax ``00:02`'
211 There are some options to which may be necessary, depending on the device
214 * *x-vga=on|off* marks the PCI(e) device as the primary GPU of the VM.
215 With this enabled the *vga* configuration option will be ignored.
217 * *pcie=on|off* tells {pve} to use a PCIe or PCI port. Some guests/device
218 combination require PCIe rather than PCI. PCIe is only available for 'q35'
221 * *rombar=on|off* makes the firmware ROM visible for the guest. Default is on.
222 Some PCI(e) devices need this disabled.
224 * *romfile=<path>*, is an optional path to a ROM file for the device to use.
225 This is a relative path under */usr/share/kvm/*.
229 An example of PCIe passthrough with a GPU set to primary:
232 # qm set VMID -hostpci0 02:00,pcie=on,x-vga=on
239 When passing through a GPU, the best compatibility is reached when using
240 'q35' as machine type, 'OVMF' ('EFI' for VMs) instead of SeaBIOS and PCIe
241 instead of PCI. Note that if you want to use 'OVMF' for GPU passthrough, the
242 GPU needs to have an EFI capable ROM, otherwise use SeaBIOS instead.
247 Another variant for passing through PCI(e) devices, is to use the hardware
248 virtualization features of your devices, if available.
250 'SR-IOV' (**S**ingle-**R**oot **I**nput/**O**utput **V**irtualization) enables
251 a single device to provide multiple 'VF' (**V**irtual **F**unctions) to the
252 system. Each of those 'VF' can be used in a different VM, with full hardware
253 features and also better performance and lower latency than software
256 Currently, the most common use case for this are NICs (**N**etwork
257 **I**nterface **C**ard) with SR-IOV support, which can provide multiple VFs per
258 physical port. This allows using features such as checksum offloading, etc. to
259 be used inside a VM, reducing the (host) CPU overhead.
265 Generally, there are two methods for enabling virtual functions on a device.
267 * sometimes there is an option for the driver module e.g. for some
274 which could be put file with '.conf' ending under */etc/modprobe.d/*.
275 (Do not forget to update your initramfs after that)
277 Please refer to your driver module documentation for the exact
278 parameters and options.
280 * The second, more generic, approach is using the `sysfs`.
281 If a device and driver supports this you can change the number of VFs on
282 the fly. For example, to setup 4 VFs on device 0000:01:00.0 execute:
285 # echo 4 > /sys/bus/pci/devices/0000:01:00.0/sriov_numvfs
288 To make this change persistent you can use the `sysfsutils` Debian package.
289 After installation configure it via */etc/sysfs.conf* or a `FILE.conf' in
295 After creating VFs, you should see them as separate PCI(e) devices when
296 outputting them with `lspci`. Get their ID and pass them through like a
297 xref:qm_pci_passthrough_vm_config[normal PCI(e) device].
302 For this feature, platform support is especially important. It may be necessary
303 to enable this feature in the BIOS/EFI first, or to use a specific PCI(e) port
304 for it to work. In doubt, consult the manual of the platform or contact its
307 Mediated Devices (vGPU, GVT-g)
308 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
310 Mediated devices are another method to reuse features and performance from
311 physical hardware for virtualized hardware. These are found most common in
312 virtualized GPU setups such as Intels GVT-g and Nvidias vGPUs used in their
315 With this, a physical Card is able to create virtual cards, similar to SR-IOV.
316 The difference is that mediated devices do not appear as PCI(e) devices in the
317 host, and are such only suited for using in virtual machines.
323 In general your card's driver must support that feature, otherwise it will
324 not work. So please refer to your vendor for compatible drivers and how to
327 Intels drivers for GVT-g are integrated in the Kernel and should work
328 with 5th, 6th and 7th generation Intel Core Processors, as well as E3 v4, E3
329 v5 and E3 v6 Xeon Processors.
331 To enable it for Intel Graphics, you have to make sure to load the module
332 'kvmgt' (for example via `/etc/modules`) and to enable it on the
333 xref:sysboot_edit_kernel_cmdline[Kernel commandline] and add the following parameter:
339 After that remember to
340 xref:qm_pci_passthrough_update_initramfs[update the `initramfs`],
341 and reboot your host.
346 To use a mediated device, simply specify the `mdev` property on a `hostpciX`
347 VM configuration option.
349 You can get the supported devices via the 'sysfs'. For example, to list the
350 supported types for the device '0000:00:02.0' you would simply execute:
353 # ls /sys/bus/pci/devices/0000:00:02.0/mdev_supported_types
356 Each entry is a directory which contains the following important files:
358 * 'available_instances' contains the amount of still available instances of
359 this type, each 'mdev' use in a VM reduces this.
360 * 'description' contains a short description about the capabilities of the type
361 * 'create' is the endpoint to create such a device, {pve} does this
362 automatically for you, if a 'hostpciX' option with `mdev` is configured.
364 Example configuration with an `Intel GVT-g vGPU` (`Intel Skylake 6700k`):
367 # qm set VMID -hostpci0 00:02.0,mdev=i915-GVTg_V5_4
370 With this set, {pve} automatically creates such a device on VM start, and
371 cleans it up again when the VM stops.
378 * link:/wiki/Pci_passthrough[PCI Passthrough Examples]