1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2010-2015 Intel Corporation.
3 Copyright 2017 Mellanox Technologies, Ltd
6 .. _linux_gsg_linux_drivers:
11 Different PMDs may require different kernel drivers in order to work properly.
12 Depending on the PMD being used, a corresponding kernel driver should be loaded,
13 and network ports should be bound to that driver.
18 VFIO is a robust and secure driver that relies on IOMMU protection.
19 To make use of VFIO, the ``vfio-pci`` module must be loaded:
21 .. code-block:: console
23 sudo modprobe vfio-pci
25 VFIO kernel is usually present by default in all distributions,
26 however please consult your distributions documentation to make sure that is the case.
28 Since Linux version 5.7,
29 the ``vfio-pci`` module supports the creation of virtual functions.
30 After the PF is bound to ``vfio-pci`` module,
31 the user can create the VFs using the ``sysfs`` interface,
32 and these VFs will be bound to ``vfio-pci`` module automatically.
34 When the PF is bound to ``vfio-pci``,
35 by default it will have a randomly generated VF token.
36 For security reasons, this token is write only,
37 so the user cannot read it from the kernel directly.
38 To access the VFs, the user needs to create a new token,
39 and use it to initialize both VF and PF devices.
40 The tokens are in UUID format,
41 so any UUID generation tool can be used to create a new token.
43 This VF token can be passed to DPDK by using EAL parameter ``--vfio-vf-token``.
44 The token will be used for all PF and VF ports within the application.
46 #. Generate the VF token by uuid command
48 .. code-block:: console
50 14d63f20-8445-11ea-8900-1f9ce7d5650d
52 #. Load the ``vfio-pci`` module with ``enable_sriov`` parameter set
54 .. code-block:: console
56 sudo modprobe vfio-pci enable_sriov=1
58 #. Bind the PCI devices to ``vfio-pci`` driver
60 .. code-block:: console
62 ./usertools/dpdk-devbind.py -b vfio-pci 0000:86:00.0
64 #. Create the desired number of VF devices
66 .. code-block:: console
68 echo 2 > /sys/bus/pci/devices/0000:86:00.0/sriov_numvfs
70 #. Start the DPDK application that will manage the PF device
72 .. code-block:: console
74 <build_dir>/app/dpdk-testpmd -l 22-25 -n 4 -a 86:00.0 \
75 --vfio-vf-token=14d63f20-8445-11ea-8900-1f9ce7d5650d --file-prefix=pf -- -i
77 #. Start the DPDK application that will manage the VF device
79 .. code-block:: console
81 <build_dir>/app/dpdk-testpmd -l 26-29 -n 4 -a 86:02.0 \
82 --vfio-vf-token=14d63f20-8445-11ea-8900-1f9ce7d5650d --file-prefix=vf0 -- -i
84 To make use of full VFIO functionality,
85 both kernel and BIOS must support and be configured
86 to use IO virtualization (such as IntelĀ® VT-d).
90 Linux versions earlier than version 3.6 do not support VFIO.
94 Linux versions earlier than version 5.7 do not support the creation of
95 virtual functions within the VFIO framework.
99 In most cases, specifying "iommu=on" as kernel parameter should be enough to
100 configure the Linux kernel to use IOMMU.
102 For proper operation of VFIO when running DPDK applications as a non-privileged user, correct permissions should also be set up.
103 For more information, please refer to :ref:`Running_Without_Root_Privileges`.
108 If there is no IOMMU available on the system, VFIO can still be used,
109 but it has to be loaded with an additional module parameter:
111 .. code-block:: console
113 modprobe vfio enable_unsafe_noiommu_mode=1
115 Alternatively, one can also enable this option in an already loaded kernel module:
117 .. code-block:: console
119 echo 1 > /sys/module/vfio/parameters/enable_unsafe_noiommu_mode
121 After that, VFIO can be used with hardware devices as usual.
125 It may be required to unload all VFIO related-modules before probing
126 the module again with ``enable_unsafe_noiommu_mode=1`` parameter.
130 Since no-IOMMU mode forgoes IOMMU protection, it is inherently unsafe.
131 That said, it does make it possible for the user
132 to keep the degree of device access and programming that VFIO has,
133 in situations where IOMMU is not available.
138 In situations where using VFIO is not an option, there are alternative drivers one can use.
139 In many cases, the standard ``uio_pci_generic`` module included in the Linux kernel
140 can be used as a substitute for VFIO. This module can be loaded using the command:
142 .. code-block:: console
144 sudo modprobe uio_pci_generic
148 ``uio_pci_generic`` module doesn't support the creation of virtual functions.
150 As an alternative to the ``uio_pci_generic``, there is the ``igb_uio`` module
151 which can be found in the repository `dpdk-kmods <http://git.dpdk.org/dpdk-kmods>`_.
152 It can be loaded as shown below:
154 .. code-block:: console
157 sudo insmod igb_uio.ko
161 If UEFI secure boot is enabled,
162 the Linux kernel may disallow the use of UIO on the system.
163 Therefore, devices for use by DPDK should be bound to the ``vfio-pci`` kernel module
164 rather than any UIO-based module.
165 For more details see :ref:`linux_gsg_binding_kernel` below.
169 If the devices used for DPDK are bound to the ``uio_pci_generic`` kernel module,
170 please make sure that the IOMMU is disabled or is in passthrough mode.
171 One can add ``intel_iommu=off`` or ``amd_iommu=off`` or ``intel_iommu=on iommu=pt``
172 in GRUB command line on x86_64 systems,
173 or add ``iommu.passthrough=1`` on aarch64 systems.
177 Using UIO drivers is inherently unsafe due to this method lacking IOMMU protection,
178 and can only be done by root user.
180 .. _bifurcated_driver:
185 PMDs which use the bifurcated driver co-exists with the device kernel driver.
186 On such model the NIC is controlled by the kernel, while the data
187 path is performed by the PMD directly on top of the device.
189 Such model has the following benefits:
191 - It is secure and robust, as the memory management and isolation
192 is done by the kernel.
193 - It enables the user to use legacy linux tools such as ``ethtool`` or
194 ``ifconfig`` while running DPDK application on the same network ports.
195 - It enables the DPDK application to filter only part of the traffic,
196 while the rest will be directed and handled by the kernel driver.
197 The flow bifurcation is performed by the NIC hardware.
198 As an example, using :ref:`flow_isolated_mode` allows to choose
199 strictly what is received in DPDK.
201 More about the bifurcated driver can be found in
202 `Mellanox Bifurcated DPDK PMD
203 <https://www.dpdk.org/wp-content/uploads/sites/35/2016/10/Day02-Session04-RonyEfraim-Userspace2016.pdf>`__.
205 .. _linux_gsg_binding_kernel:
207 Binding and Unbinding Network Ports to/from the Kernel Modules
208 --------------------------------------------------------------
212 PMDs which use the bifurcated driver should not be unbound from their kernel drivers.
213 This section is for PMDs which use the UIO or VFIO drivers.
215 As of release 1.4, DPDK applications no longer automatically unbind all supported network ports from the kernel driver in use.
216 Instead, in case the PMD being used use the VFIO or UIO drivers,
217 all ports that are to be used by a DPDK application must be bound to
218 the ``vfio-pci``, ``uio_pci_generic``, or ``igb_uio`` module
219 before the application is run.
220 For such PMDs, any network ports under Linux* control will be ignored and cannot be used by the application.
222 To bind ports to the ``vfio-pci``, ``uio_pci_generic`` or ``igb_uio`` module
223 for DPDK use, or to return ports to Linux control,
224 a utility script called ``dpdk-devbind.py`` is provided in the ``usertools`` subdirectory.
225 This utility can be used to provide a view of the current state of the network ports on the system,
226 and to bind and unbind those ports from the different kernel modules,
227 including the VFIO and UIO modules.
228 The following are some examples of how the script can be used.
229 A full description of the script and its parameters can be obtained
230 by calling the script with the ``--help`` or ``--usage`` options.
231 Note that the UIO or VFIO kernel modules to be used,
232 should be loaded into the kernel before running the ``dpdk-devbind.py`` script.
236 Due to the way VFIO works, there are certain limitations
237 to which devices can be used with VFIO.
238 Mainly it comes down to how IOMMU groups work.
239 Any Virtual Function device can usually be used with VFIO on its own,
240 but physical devices may require either all ports bound to VFIO,
241 or some of them bound to VFIO while others not being bound to anything at all.
243 If your device is behind a PCI-to-PCI bridge,
244 the bridge will then be part of the IOMMU group in which your device is in.
245 Therefore, the bridge driver should also be unbound from the bridge PCI device
246 for VFIO to work with devices behind the bridge.
250 While any user can run the ``dpdk-devbind.py`` script
251 to view the status of the network ports,
252 binding or unbinding network ports requires root privileges.
254 To see the status of all network ports on the system:
256 .. code-block:: console
258 ./usertools/dpdk-devbind.py --status
260 Network devices using DPDK-compatible driver
261 ============================================
262 0000:82:00.0 '82599EB 10-GbE NIC' drv=uio_pci_generic unused=ixgbe
263 0000:82:00.1 '82599EB 10-GbE NIC' drv=uio_pci_generic unused=ixgbe
265 Network devices using kernel driver
266 ===================================
267 0000:04:00.0 'I350 1-GbE NIC' if=em0 drv=igb unused=uio_pci_generic *Active*
268 0000:04:00.1 'I350 1-GbE NIC' if=eth1 drv=igb unused=uio_pci_generic
269 0000:04:00.2 'I350 1-GbE NIC' if=eth2 drv=igb unused=uio_pci_generic
270 0000:04:00.3 'I350 1-GbE NIC' if=eth3 drv=igb unused=uio_pci_generic
272 Other network devices
273 =====================
276 To bind device ``eth1``,``04:00.1``, to the ``uio_pci_generic`` driver:
278 .. code-block:: console
280 ./usertools/dpdk-devbind.py --bind=uio_pci_generic 04:00.1
284 .. code-block:: console
286 ./usertools/dpdk-devbind.py --bind=uio_pci_generic eth1
288 To restore device ``82:00.0`` to its original kernel binding:
290 .. code-block:: console
292 ./usertools/dpdk-devbind.py --bind=ixgbe 82:00.0
297 In certain situations, using ``dpdk-devbind.py`` script
298 to bind a device to VFIO driver may fail.
299 The first place to check is the kernel messages:
301 .. code-block:: console
305 [ 1297.875090] vfio-pci: probe of 0000:31:00.0 failed with error -22
308 In most cases, the ``error -22`` indicates that the VFIO subsystem
309 could not be enabled because there is no IOMMU support.
311 To check whether the kernel has been booted with correct parameters,
312 one can check the kernel command-line:
314 .. code-block:: console
318 Please refer to earlier sections on how to configure kernel parameters
319 correctly for your system.
321 If the kernel is configured correctly, one also has to make sure that
322 the BIOS configuration has virtualization features (such as IntelĀ® VT-d).
323 There is no standard way to check if the platform is configured correctly,
324 so please check with your platform documentation to see if it has such features,
325 and how to enable them.
327 In certain distributions, default kernel configuration is such that
328 the no-IOMMU mode is disabled altogether at compile time.
329 This can be checked in the boot configuration of your system:
331 .. code-block:: console
333 cat /boot/config-$(uname -r) | grep NOIOMMU
334 # CONFIG_VFIO_NOIOMMU is not set
336 If ``CONFIG_VFIO_NOIOMMU`` is not enabled in the kernel configuration,
337 VFIO driver will not support the no-IOMMU mode,
338 and other alternatives (such as UIO drivers) will have to be used.