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 Depends on the PMD being used, a corresponding kernel driver should be load
13 and bind to the network ports.
18 A small kernel module to set up the device, map device memory to user-space and register interrupts.
19 In many cases, the standard ``uio_pci_generic`` module included in the Linux kernel
20 can provide the uio capability. This module can be loaded using the command:
22 .. code-block:: console
24 sudo modprobe uio_pci_generic
28 ``uio_pci_generic`` module doesn't support the creation of virtual functions.
30 As an alternative to the ``uio_pci_generic``, the DPDK also includes the igb_uio
31 module which can be found in the kernel/linux subdirectory referred to above. It can
32 be loaded as shown below:
34 .. code-block:: console
37 sudo insmod <build_dir>/kernel/linux/igb_uio/igb_uio.ko
41 Building DPDK Linux kernel modules is disabled by default starting from DPDK 20.02.
42 To enable them again, the config option "enable_kmods" needs to be set
43 in the meson build configuration.
44 See :ref:`adjusting_build_options` for details on how to set/clear build options.
45 It is planned to move ``igb_uio`` module to a different git repository.
49 For some devices which lack support for legacy interrupts, e.g. virtual function
50 (VF) devices, the ``igb_uio`` module may be needed in place of ``uio_pci_generic``.
54 If UEFI secure boot is enabled, the Linux kernel may disallow the use of
55 UIO on the system. Therefore, devices for use by DPDK should be bound to the
56 ``vfio-pci`` kernel module rather than ``igb_uio`` or ``uio_pci_generic``.
57 For more details see :ref:`linux_gsg_binding_kernel` below.
61 If the devices used for DPDK are bound to the ``uio_pci_generic`` kernel module,
62 please make sure that the IOMMU is disabled or passthrough. One can add
63 ``intel_iommu=off`` or ``amd_iommu=off`` or ``intel_iommu=on iommu=pt`` in GRUB
64 command line on x86_64 systems, or add ``iommu.passthrough=1`` on aarch64 system.
66 Since DPDK release 1.7 onward provides VFIO support, use of UIO is optional
67 for platforms that support using VFIO.
72 A more robust and secure driver in compare to the ``UIO``, relying on IOMMU protection.
73 To make use of VFIO, the ``vfio-pci`` module must be loaded:
75 .. code-block:: console
77 sudo modprobe vfio-pci
79 Note that in order to use VFIO, your kernel must support it.
80 VFIO kernel modules have been included in the Linux kernel since version 3.6.0 and are usually present by default,
81 however please consult your distributions documentation to make sure that is the case.
83 The ``vfio-pci`` module since Linux version 5.7 supports the creation of virtual
84 functions. After the PF is bound to vfio-pci module, the user can create the VFs
85 by sysfs interface, and these VFs are bound to vfio-pci module automatically.
87 When the PF is bound to vfio-pci, it has initial VF token generated by random. For
88 security reason, this token is write only, the user can't read it from the kernel
89 directly. To access the VF, the user needs to start the PF with token parameter to
90 setup a VF token in UUID format, then the VF can be accessed with this new token.
92 Since the ``vfio-pci`` module uses the VF token as internal data to provide the
93 collaboration between SR-IOV PF and VFs, so DPDK can use the same VF token for all
94 PF devices which bound to one application. This VF token can be specified by the EAL
95 parameter ``--vfio-vf-token``.
97 .. code-block:: console
99 1. Generate the VF token by uuid command
100 14d63f20-8445-11ea-8900-1f9ce7d5650d
102 2. sudo modprobe vfio-pci enable_sriov=1
104 2. ./usertools/dpdk-devbind.py -b vfio-pci 0000:86:00.0
106 3. echo 2 > /sys/bus/pci/devices/0000:86:00.0/sriov_numvfs
109 <build_dir>/app/dpdk-testpmd -l 22-25 -n 4 -w 86:00.0 \
110 --vfio-vf-token=14d63f20-8445-11ea-8900-1f9ce7d5650d --file-prefix=pf -- -i
113 <build_dir>/app/dpdk-testpmd -l 26-29 -n 4 -w 86:02.0 \
114 --vfio-vf-token=14d63f20-8445-11ea-8900-1f9ce7d5650d --file-prefix=vf0 -- -i
116 Also, to use VFIO, both kernel and BIOS must support and be configured to use IO virtualization (such as IntelĀ® VT-d).
120 ``vfio-pci`` module doesn't support the creation of virtual functions before Linux version 5.7.
122 For proper operation of VFIO when running DPDK applications as a non-privileged user, correct permissions should also be set up.
123 This can be done by using the DPDK setup script (called dpdk-setup.sh and located in the usertools directory).
127 VFIO can be used without IOMMU. While this is just as unsafe as using UIO, it does make it possible for the user to keep the degree of device access and programming that VFIO has, in situations where IOMMU is not available.
129 .. _bifurcated_driver:
134 PMDs which use the bifurcated driver co-exists with the device kernel driver.
135 On such model the NIC is controlled by the kernel, while the data
136 path is performed by the PMD directly on top of the device.
138 Such model has the following benefits:
140 - It is secure and robust, as the memory management and isolation
141 is done by the kernel.
142 - It enables the user to use legacy linux tools such as ``ethtool`` or
143 ``ifconfig`` while running DPDK application on the same network ports.
144 - It enables the DPDK application to filter only part of the traffic,
145 while the rest will be directed and handled by the kernel driver.
146 The flow bifurcation is performed by the NIC hardware.
147 As an example, using :ref:`flow_isolated_mode` allows to choose
148 strictly what is received in DPDK.
150 More about the bifurcated driver can be found in
151 `Mellanox Bifurcated DPDK PMD
152 <https://www.dpdk.org/wp-content/uploads/sites/35/2016/10/Day02-Session04-RonyEfraim-Userspace2016.pdf>`__.
154 .. _linux_gsg_binding_kernel:
156 Binding and Unbinding Network Ports to/from the Kernel Modules
157 --------------------------------------------------------------
161 PMDs Which use the bifurcated driver should not be unbind from their kernel drivers. this section is for PMDs which use the UIO or VFIO drivers.
163 As of release 1.4, DPDK applications no longer automatically unbind all supported network ports from the kernel driver in use.
164 Instead, in case the PMD being used use the UIO or VFIO drivers, all ports that are to be used by an DPDK application must be bound to the
165 ``uio_pci_generic``, ``igb_uio`` or ``vfio-pci`` module before the application is run.
166 For such PMDs, any network ports under Linux* control will be ignored and cannot be used by the application.
168 To bind ports to the ``uio_pci_generic``, ``igb_uio`` or ``vfio-pci`` module for DPDK use,
169 and then subsequently return ports to Linux* control,
170 a utility script called dpdk-devbind.py is provided in the usertools subdirectory.
171 This utility can be used to provide a view of the current state of the network ports on the system,
172 and to bind and unbind those ports from the different kernel modules, including the uio and vfio modules.
173 The following are some examples of how the script can be used.
174 A full description of the script and its parameters can be obtained by calling the script with the ``--help`` or ``--usage`` options.
175 Note that the uio or vfio kernel modules to be used, should be loaded into the kernel before
176 running the ``dpdk-devbind.py`` script.
180 Due to the way VFIO works, there are certain limitations to which devices can be used with VFIO.
181 Mainly it comes down to how IOMMU groups work.
182 Any Virtual Function device can be used with VFIO on its own, but physical devices will require either all ports bound to VFIO,
183 or some of them bound to VFIO while others not being bound to anything at all.
185 If your device is behind a PCI-to-PCI bridge, the bridge will then be part of the IOMMU group in which your device is in.
186 Therefore, the bridge driver should also be unbound from the bridge PCI device for VFIO to work with devices behind the bridge.
190 While any user can run the dpdk-devbind.py script to view the status of the network ports,
191 binding or unbinding network ports requires root privileges.
193 To see the status of all network ports on the system:
195 .. code-block:: console
197 ./usertools/dpdk-devbind.py --status
199 Network devices using DPDK-compatible driver
200 ============================================
201 0000:82:00.0 '82599EB 10-GbE NIC' drv=uio_pci_generic unused=ixgbe
202 0000:82:00.1 '82599EB 10-GbE NIC' drv=uio_pci_generic unused=ixgbe
204 Network devices using kernel driver
205 ===================================
206 0000:04:00.0 'I350 1-GbE NIC' if=em0 drv=igb unused=uio_pci_generic *Active*
207 0000:04:00.1 'I350 1-GbE NIC' if=eth1 drv=igb unused=uio_pci_generic
208 0000:04:00.2 'I350 1-GbE NIC' if=eth2 drv=igb unused=uio_pci_generic
209 0000:04:00.3 'I350 1-GbE NIC' if=eth3 drv=igb unused=uio_pci_generic
211 Other network devices
212 =====================
215 To bind device ``eth1``,``04:00.1``, to the ``uio_pci_generic`` driver:
217 .. code-block:: console
219 ./usertools/dpdk-devbind.py --bind=uio_pci_generic 04:00.1
223 .. code-block:: console
225 ./usertools/dpdk-devbind.py --bind=uio_pci_generic eth1
227 To restore device ``82:00.0`` to its original kernel binding:
229 .. code-block:: console
231 ./usertools/dpdk-devbind.py --bind=ixgbe 82:00.0