pveum: improve tfa section

[pve-docs.git] / qm-pci-passthrough.adoc

[[qm_pci_passthrough]]
PCI(e) Passthrough
------------------
ifdef::wiki[]
:pve-toplevel:
endif::wiki[]

PCI(e) passthrough is a mechanism to give a virtual machine control over
a PCI device from the host. This can have some advantages over using
virtualized hardware, for example lower latency, higher performance, or more
features (e.g., offloading).

But, if you pass through a device to a virtual machine, you cannot use that
device anymore on the host or in any other VM.

General Requirements
~~~~~~~~~~~~~~~~~~~~

Since passthrough is a feature which also needs hardware support, there are
some requirements to check and preparations to be done to make it work.


Hardware
^^^^^^^^
Your hardware needs to support `IOMMU` (*I*/*O* **M**emory **M**anagement
**U**nit) interrupt remapping, this includes the CPU and the mainboard.

Generally, Intel systems with VT-d, and AMD systems with AMD-Vi support this.
But it is not guaranteed that everything will work out of the box, due
to bad hardware implementation and missing or low quality drivers.

Further, server grade hardware has often better support than consumer grade
hardware, but even then, many modern system can support this.

Please refer to your hardware vendor to check if they support this feature
under Linux for your specific setup.


Configuration
^^^^^^^^^^^^^

Once you ensured that your hardware supports passthrough, you will need to do
some configuration to enable PCI(e) passthrough.


.IOMMU

The IOMMU has to be activated on the
xref:sysboot_edit_kernel_cmdline[kernel commandline].

The command line parameters are:

* for Intel CPUs:
+
----
 intel_iommu=on
----
* for AMD CPUs:
+
----
 amd_iommu=on
----


.Kernel Modules

You have to make sure the following modules are loaded. This can be achieved by
adding them to `'/etc/modules''

----
 vfio
 vfio_iommu_type1
 vfio_pci
 vfio_virqfd
----

[[qm_pci_passthrough_update_initramfs]]
After changing anything modules related, you need to refresh your
`initramfs`. On {pve} this can be done by executing:

----
# update-initramfs -u -k all
----

.Finish Configuration

Finally reboot to bring the changes into effect and check that it is indeed
enabled.

----
# dmesg | grep -e DMAR -e IOMMU -e AMD-Vi
----

should display that `IOMMU`, `Directed I/O` or `Interrupt Remapping` is
enabled, depending on hardware and kernel the exact message can vary.

It is also important that the device(s) you want to pass through
are in a *separate* `IOMMU` group. This can be checked with:

----
# find /sys/kernel/iommu_groups/ -type l
----

It is okay if the device is in an `IOMMU` group together with its functions
(e.g. a GPU with the HDMI Audio device) or with its root port or PCI(e) bridge.

.PCI(e) slots
[NOTE]
====
Some platforms handle their physical PCI(e) slots differently. So, sometimes
it can help to put the card in a another PCI(e) slot, if you do not get the
desired `IOMMU` group separation.
====

.Unsafe interrupts
[NOTE]
====
For some platforms, it may be necessary to allow unsafe interrupts.
For this add  the following line in a file ending with `.conf' file in
*/etc/modprobe.d/*:

----
 options vfio_iommu_type1 allow_unsafe_interrupts=1
----

Please be aware that this option can make your system unstable.
====

GPU Passthrough Notes
^^^^^^^^^^^^^^^^^^^^^

It is not possible to display the frame buffer of the GPU via NoVNC or SPICE on
the {pve} web interface.

When passing through a whole GPU or a vGPU and graphic output is wanted, one
has to either physically connect a monitor to the card, or configure a remote
desktop software (for example, VNC or RDP) inside the guest.

If you want to use the GPU as a hardware accelerator, for example, for
programs using OpenCL or CUDA, this is not required.

Host Device Passthrough
~~~~~~~~~~~~~~~~~~~~~~~

The most used variant of PCI(e) passthrough is to pass through a whole
PCI(e) card, for example a GPU or a network card.


Host Configuration
^^^^^^^^^^^^^^^^^^

In this case, the host must not use the card. There are two methods to achieve
this:

* pass the device IDs to the options of the 'vfio-pci' modules by adding
+
----
 options vfio-pci ids=1234:5678,4321:8765
----
+
to a .conf file in */etc/modprobe.d/* where `1234:5678` and `4321:8765` are
the vendor and device IDs obtained by:
+
----
# lspci -nn
----

* blacklist the driver completely on the host, ensuring that it is free to bind
for passthrough, with
+
----
 blacklist DRIVERNAME
----
+
in a .conf file in */etc/modprobe.d/*.

For both methods you need to
xref:qm_pci_passthrough_update_initramfs[update the `initramfs`] again and
reboot after that.

.Verify Configuration

To check if your changes were successful, you can use

----
# lspci -nnk
----

and check your device entry. If it says

----
Kernel driver in use: vfio-pci
----

or the 'in use' line is missing entirely, the device is ready to be used for
passthrough.

[[qm_pci_passthrough_vm_config]]
VM Configuration
^^^^^^^^^^^^^^^^
To pass through the device you need to set the *hostpciX* option in the VM
configuration, for example by executing:

----
# qm set VMID -hostpci0 00:02.0
----

If your device has multiple functions (e.g., ``00:02.0`' and ``00:02.1`' ),
you can pass them through all together with the shortened syntax ``00:02`'

There are some options to which may be necessary, depending on the device
and guest OS:

* *x-vga=on|off* marks the PCI(e) device as the primary GPU of the VM.
With this enabled the *vga* configuration option will be ignored.

* *pcie=on|off* tells {pve} to use a PCIe or PCI port. Some guests/device
combination require PCIe rather than PCI. PCIe is only available for 'q35'
machine types.

* *rombar=on|off* makes the firmware ROM visible for the guest. Default is on.
Some PCI(e) devices need this disabled.

* *romfile=<path>*, is an optional path to a ROM file for the device to use.
This is a relative path under */usr/share/kvm/*.

.Example

An example of PCIe passthrough with a GPU set to primary:

----
# qm set VMID -hostpci0 02:00,pcie=on,x-vga=on
----


Other considerations
^^^^^^^^^^^^^^^^^^^^

When passing through a GPU, the best compatibility is reached when using
'q35' as machine type, 'OVMF' ('EFI' for VMs) instead of SeaBIOS and PCIe
instead of PCI. Note that if you want to use 'OVMF' for GPU passthrough, the
GPU needs to have an EFI capable ROM, otherwise use SeaBIOS instead.

SR-IOV
~~~~~~

Another variant for passing through PCI(e) devices, is to use the hardware
virtualization features of your devices, if available.

'SR-IOV' (**S**ingle-**R**oot **I**nput/**O**utput **V**irtualization) enables
a single device to provide multiple 'VF' (**V**irtual **F**unctions) to the
system. Each of those 'VF' can be used in a different VM, with full hardware
features and also better performance and lower latency than software
virtualized devices.

Currently, the most common use case for this are NICs (**N**etwork
**I**nterface **C**ard) with SR-IOV support, which can provide multiple VFs per
physical port. This allows using features such as checksum offloading, etc. to
be used inside a VM, reducing the (host) CPU overhead.


Host Configuration
^^^^^^^^^^^^^^^^^^

Generally, there are two methods for enabling virtual functions on a device.

* sometimes there is an option for the driver module e.g. for some
Intel drivers
+
----
 max_vfs=4
----
+
which could be put file with '.conf' ending under */etc/modprobe.d/*.
(Do not forget to update your initramfs after that)
+
Please refer to your driver module documentation for the exact
parameters and options.

* The second, more generic, approach is using the `sysfs`.
If a device and driver supports this you can change the number of VFs on
the fly. For example, to setup 4 VFs on device 0000:01:00.0 execute:
+
----
# echo 4 > /sys/bus/pci/devices/0000:01:00.0/sriov_numvfs
----
+
To make this change persistent you can use the `sysfsutils` Debian package.
After installation configure it via */etc/sysfs.conf* or a `FILE.conf' in
*/etc/sysfs.d/*.

VM Configuration
^^^^^^^^^^^^^^^^

After creating VFs, you should see them as separate PCI(e) devices when
outputting them with `lspci`. Get their ID and pass them through like a
xref:qm_pci_passthrough_vm_config[normal PCI(e) device].

Other considerations
^^^^^^^^^^^^^^^^^^^^

For this feature, platform support is especially important. It may be necessary
to enable this feature in the BIOS/EFI first, or to use a specific PCI(e) port
for it to work. In doubt, consult the manual of the platform or contact its
vendor.

Mediated Devices (vGPU, GVT-g)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Mediated devices are another method to reuse features and performance from
physical hardware for virtualized hardware. These are found most common in
virtualized GPU setups such as Intels GVT-g and Nvidias vGPUs used in their
GRID technology.

With this, a physical Card is able to create virtual cards, similar to SR-IOV.
The difference is that mediated devices do not appear as PCI(e) devices in the
host, and are such only suited for using in virtual machines.


Host Configuration
^^^^^^^^^^^^^^^^^^

In general your card's driver must support that feature, otherwise it will
not work. So please refer to your vendor for compatible drivers and how to
configure them.

Intels drivers for GVT-g are integrated in the Kernel and should work
with 5th, 6th and 7th generation Intel Core Processors, as well as E3 v4, E3
v5 and E3 v6 Xeon Processors.

To enable it for Intel Graphics, you have to make sure to load the module
'kvmgt' (for example via `/etc/modules`) and to enable it on the
xref:sysboot_edit_kernel_cmdline[Kernel commandline] and add the following parameter:

----
 i915.enable_gvt=1
----

After that remember to
xref:qm_pci_passthrough_update_initramfs[update the `initramfs`],
and reboot your host.

VM Configuration
^^^^^^^^^^^^^^^^

To use a mediated device, simply specify the `mdev` property on a `hostpciX`
VM configuration option.

You can get the supported devices via the 'sysfs'. For example, to list the
supported types for the device '0000:00:02.0' you would simply execute:

----
# ls /sys/bus/pci/devices/0000:00:02.0/mdev_supported_types
----

Each entry is a directory which contains the following important files:

* 'available_instances' contains the amount of still available instances of
this type, each 'mdev' use in a VM reduces this.
* 'description' contains a short description about the capabilities of the type
* 'create' is the endpoint to create such a device, {pve} does this
automatically for you, if a 'hostpciX' option with `mdev` is configured.

Example configuration with an `Intel GVT-g vGPU` (`Intel Skylake 6700k`):

----
# qm set VMID -hostpci0 00:02.0,mdev=i915-GVTg_V5_4
----

With this set, {pve} automatically creates such a device on VM start, and
cleans it up again when the VM stops.

ifdef::wiki[]

See Also
~~~~~~~~

* link:/wiki/Pci_passthrough[PCI Passthrough Examples]

endif::wiki[]