5.5.1. idpf Linux* Base Driver Readme for Infrastructure Data-Plane Function
****************************************************************************

September 17, 2024


On This Page
^^^^^^^^^^^^

* idpf Linux* Base Driver Readme for Infrastructure Data-Plane
  Function

  * Overview

  * Building and Installation

  * Command Line Parameters

  * Additional Features and Configurations

  * Performance Optimization

  * Known Issues/Troubleshooting


5.5.1.1. Overview
=================

This document provides the README for the out-of-tree idpf Linux
driver. The driver is compatible with devices based on the following:

* Infrastructure Data-Plane Function

The idpf driver serves as both the Physical Function (PF) and Virtual
Function (VF) driver for the Infrastructure Data-Plane Function.

Driver information can be obtained using ethtool, lspci, and ip.
Instructions on updating ethtool can be found in the Additional
Features and Configurations section later in this document.

This driver is only supported as a loadable module at this time. Intel
is not supplying patches against the kernel source to allow for static
linking of the drivers.

This driver supports XDP (Express Data Path) and AF_XDP zero-copy.
Note that XDP is blocked for frame sizes larger than 3KB.


5.5.1.2. Building and Installation
==================================


5.5.1. To manually build the driver
-----------------------------------

1. Move the base driver tar file to the directory of your choice. For
   example, use "/home/username/idpf" or "/usr/local/src/idpf".

2. Untar/unzip the archive, where "<x.x.x>" is the version number for
   the driver tar file:

      tar zxf idpf-<x.x.x>.tar.gz

3. Change to the driver src directory, where "<x.x.x>" is the version
   number for the driver tar:

      cd idpf-<x.x.x>

4. Compile the driver module:

      make install

   The binary will be installed as:

      /lib/modules/<KERNEL VER>/updates/drivers/net/ethernet/intel/idpf/idpf.ko

   The install location listed above is the default location. This may
   differ for various Linux distributions.

   Note:

     To gather and display additional statistics, use the
     "IDPF_ADD_PROBES" pre-processor macro:

        make CFLAGS_EXTRA=-DIDPF_ADD_PROBES

     Please note that this additional statistics gathering can impact
     performance.

5. Load the module using the modprobe command.

   To check the version of the driver and then load it:

      modinfo idpf
      modprobe idpf

   Alternately, make sure that any older idpf drivers are removed from
   the kernel before loading the new module:

      rmmod idpf; modprobe idpf

6. Assign an IP address to the interface by entering the following,
   where "<ethX>" is the interface name that was shown in dmesg after
   modprobe:

      ip address add <IP_address>/<netmask bits> dev <ethX>

7. Verify that the interface works. Enter the following, where
   IP_address is the IP address for another machine on the same subnet
   as the interface that is being tested:

      ping <IP_address>


5.5.1. To build a binary RPM package of this driver
---------------------------------------------------

Note:

  RPM functionality has only been tested in Red Hat distributions.

1. Run the following command, where "<x.x.x>" is the version number
   for the driver tar file:

      rpmbuild -tb idpf-<x.x.x>.tar.gz

   Note:

     For the build to work properly, the currently running kernel MUST
     match the version and configuration of the installed kernel
     sources. If you have just recompiled the kernel, reboot the
     system before building.

2. After building the RPM, the last few lines of the tool output
   contain the location of the RPM file that was built. Install the
   RPM with one of the following commands, where "<RPM>" is the
   location of the RPM file:

      rpm -Uvh <RPM>

   or:

      dnf/yum localinstall <RPM>

Note:

  * To compile the driver on some kernel/arch combinations, you may
    need to install a package with the development version of libelf
    (e.g. libelf-dev, libelf-devel, elfutils-libelf-devel).

  * When compiling an out-of-tree driver, details will vary by
    distribution. However, you will usually need a kernel-devel RPM or
    some RPM that provides the kernel headers at a minimum. The RPM
    kernel-devel will usually fill in the link at "/lib/modules/'uname
    -r'/build".


5.5.1.3. Command Line Parameters
================================

The idpf driver does not support any command line parameters.


5.5.1.4. Additional Features and Configurations
===============================================


5.5.1. Configuring SR-IOV for improved network security
-------------------------------------------------------

In a virtualized environment, on Intel(R) Ethernet Network Adapters
that support SR-IOV or Intel(R) Scalable I/O Virtualization (Intel(R)
Scalable IOV), the virtual function (VF) may be subject to malicious
behavior. Software-generated layer-two frames, like IEEE 802.3x (link
flow control), IEEE 802.1Qbb (priority-based flow control), and others
of this type are not expected and can throttle traffic between the
host and the virtual switch, reducing performance.

To resolve this issue, and to ensure isolation from unintended traffic
streams, configure all SR-IOV or Intel Scalable IOV enabled ports for
VLAN tagging from the administrative interface on the PF. This
configuration allows unexpected, and potentially malicious, frames to
be dropped.


5.5.1. ethtool
--------------

The driver utilizes the ethtool interface for driver configuration and
diagnostics, as well as displaying statistical information. The latest
ethtool version is required for this functionality. If you don't have
one yet, you can obtain it at  at:
https://kernel.org/pub/software/network/ethtool/.


5.5.1. Viewing Link Messages
----------------------------

Link messages will not be displayed to the console if the distribution
is restricting system messages. In order to see network driver link
messages on your console, set dmesg to eight by entering the
following:

   dmesg -n 8


5.5.1. Jumbo Frames
-------------------

Jumbo Frames support is enabled by changing the Maximum Transmission
Unit (MTU) to a value larger than the default value of 1500.

Use the ip command to increase the MTU size. For example, enter the
following where "<ethX>" is the name of the network interface:

   ip link set mtu 5120 dev <ethX>
   ip link set up dev <ethX>

This setting is not saved across reboots.

Note:

  * The supported maximum MTU setting for jumbo frames on Intel(R) IPU
    ASIC E2100 B1 Stepping is 7652 bytes as mentioned in the release
    notes for Release 1.3.0 Release Notes. This corresponds to the
    maximum frame size of 7678 bytes. Later revisions of the Intel IPU
    support the maximum MTU of 9188 bytes and maximum frame size of
    9216 bytes.

  * This driver will attempt to use multiple page sized buffers to
    receive each jumbo packet. This should help to avoid buffer
    starvation issues when allocating receive packets.

  * Packet loss may have a greater impact on throughput when you use
    jumbo frames. If you observe a drop in performance after enabling
    jumbo frames, enabling flow control may mitigate the issue.


5.5.1. Subfunctions
-------------------

Subfunctions are supported using the devlink interface to create,
activate or delete subfunction netdevs or dynamic vports.


5.5.1. Configuring Subfunctions
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

1. After loading the driver, run the following command to verify that
   the IDPF driver supports devlink based interface:

      devlink dev show

   In the output, look for the device's PCI address using its
   "Domain:Bus:Device.Function". The examples below use
   "pci/0000:b1:00.0". The following output indicates that the IDPF
   driver supports devlink:

      lspci | grep b1:00.0
      b1:00.0 Ethernet Controller: Intel Corporation Infrastructure Data Path Function (rev 10)

2. Run the following command to see if a dynamic vPort exists:

      devlink port show

   For example, the following output shows a dynamic vPort does not
   exist:

      pci/0000:4b:00.0/0: type eth netdev ens785f0 flavour physical port 0 splittable false
      pci/0000:4b:00.1/0: type eth netdev ens785f1 flavour physical port 1 splittable false

3. Issue the following command to prepare for a single vPort:

      devlink port add pci/0000:b1:00.0 flavour pcisf pfnum 0 sfnum 101

   Example output could be:

      pci/0000:b1:00.0/1: type notset flavour pcisf controller 0 pfnum 0 sfnum 101 splittable false
      [root@lo0-100 sbhatna1]# devlink port show
      pci/0000:4b:00.0/0: type eth netdev ens785f0 flavour physical port 0 splittable false
      pci/0000:4b:00.1/0: type eth netdev ens785f1 flavour physical port 1 splittable false
      pci/0000:b1:00.0/1: type notset flavour pcisf controller 0 pfnum 0 sfnum 101 splittable false ? type notset

4. Create and activate a single vPort:

      devlink port func set pci/0000:b1:00.0/1 state active
      dmesg -c

   Example output:

      [ 3186.229091] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready

   Run the following to display the vPort status:

      devlink port show

   Example output:

      pci/0000:4b:00.0/0: type eth netdev ens785f0 flavour physical port 0 splittable false
      pci/0000:4b:00.1/0: type eth netdev ens785f1 flavour physical port 1 splittable false
      pci/0000:b1:00.0/1: type eth netdev eth0 flavour pcisf controller 0 pfnum 0 sfnum 101 splittable false ? type is eth

5. Create another 15 vPorts and activate them:

      devlink port show | grep "pci/0000:b1:00" | wc -l

   The output shows the number of activated vPorts. For example:

       16

      [ 3424.563240] IPv6: ADDRCONF(NETDEV_CHANGE): eth4: link becomes ready
      [ 3425.739128] IPv6: ADDRCONF(NETDEV_CHANGE): eth5: link becomes ready
      [ 3425.739230] IPv6: ADDRCONF(NETDEV_CHANGE): eth6: link becomes ready
      [ 3426.947972] IPv6: ADDRCONF(NETDEV_CHANGE): eth7: link becomes ready
      [ 3426.948245] IPv6: ADDRCONF(NETDEV_CHANGE): eth8: link becomes ready
      [ 3428.220488] IPv6: ADDRCONF(NETDEV_CHANGE): eth9: link becomes ready
      [ 3428.220604] IPv6: ADDRCONF(NETDEV_CHANGE): eth10: link becomes ready
      [ 3429.500157] IPv6: ADDRCONF(NETDEV_CHANGE): eth11: link becomes ready
      [ 3429.500262] IPv6: ADDRCONF(NETDEV_CHANGE): eth12: link becomes ready
      [ 3430.825114] IPv6: ADDRCONF(NETDEV_CHANGE): eth13: link becomes ready
      [ 3430.825241] IPv6: ADDRCONF(NETDEV_CHANGE): eth14: link becomes ready
      [ 3432.195451] IPv6: ADDRCONF(NETDEV_CHANGE): eth15: link becomes ready
      [ 3432.195560] IPv6: ADDRCONF(NETDEV_CHANGE): eth16: link becomes ready
      [ 3433.420897] IPv6: ADDRCONF(NETDEV_CHANGE): eth17: link becomes ready
      [ 3433.421007] IPv6: ADDRCONF(NETDEV_CHANGE): eth18: link becomes ready

6. Randomly delete a few:

      devlink port del pci/0000:b1:00.0/10
      devlink port del pci/0000:b1:00.0/5
      devlink port show | grep "pci/0000:b1:00" | wc -l

   The output of the last command should show fewer vPorts. For
   example:

      14

7. Remove the IDPF driver:

      rmmod idpf

8. Run the following to show information about the devlink vPorts on
   the system:

      devlink port show

   Example output:

      pci/0000:4b:00.0/0: type eth netdev ens785f0 flavour physical port 0 splittable false
      pci/0000:4b:00.1/0: type eth netdev ens785f1 flavour physical port 1 splittable false


5.5.1.5. Performance Optimization
=================================

Driver defaults are meant to fit a wide variety of workloads, but if
further optimization is required, we recommend experimenting with the
following settings.


5.5.1. IRQ to Adapter Queue Alignment
-------------------------------------

Pin the adapter's IRQs to specific cores by disabling the irqbalance
service and using the included "set_irq_affinity" script. Please see
the script's help text for further options.

* The following settings will distribute the IRQs across all the cores
  evenly:

     scripts/set_irq_affinity -x all <interface1> , [ <interface2>, ... ]

* The following settings will distribute the IRQs across all the cores
  that are local to the adapter (same NUMA node):

     scripts/set_irq_affinity -x local <interface1> ,[ <interface2>, ... ]

* For very CPU-intensive workloads, we recommend pinning the IRQs to
  all cores.


5.5.1. Interrupt Rate Limiting
------------------------------

This driver supports an adaptive interrupt throttle rate (ITR)
mechanism that is tuned for general workloads. The user can customize
the interrupt rate control for specific workloads, via ethtool,
adjusting the number of microseconds between interrupts.

To set the interrupt rate manually, you must disable adaptive mode:

   ethtool -C <ethX> adaptive-rx off adaptive-tx off

For lower CPU utilization:

* Disable adaptive ITR and lower Rx and Tx interrupts. The examples,
  below, affect every queue of the specified interface.

* Setting "rx-usecs" and "tx-usecs" to 80 will limit interrupts to
  about 12,500 interrupts per second per queue:

     ethtool -C <ethX> adaptive-rx off adaptive-tx off rx-usecs 80 tx-usecs 80

For reduced latency:

* Disable adaptive ITR and ITR by setting "rx-usecs" and "tx-usecs" to
  0 using ethtool:

     ethtool -C <ethX> adaptive-rx off adaptive-tx off rx-usecs 0 tx-usecs 0

Per-queue interrupt rate settings:

* The following examples are for queues 1 and 3, but you can adjust
  other queues.

* To disable Rx adaptive ITR and set static Rx ITR to 10 microseconds
  or about 100,000 interrupts/second, for queues 1 and 3:

     ethtool --per-queue <ethX> queue_mask 0xa --coalesce adaptive-rx off rx-usecs 10

* To show the current coalesce settings for queues 1 and 3:

     ethtool --per-queue <ethX> queue_mask 0xa --show-coalesce


5.5.1. Transmit/Receive Queue Allocation
----------------------------------------

To set the number of symmetrical (Rx/Tx) or asymmetrical (mix of
combined and Tx or Rx) queues, use the "ethtool -L" option. For
example:

* To set 16 queue pairs for the interface:

     ethtool -L <interface> combined 16

* To set a pair of 12 Tx/Rx and 4 Tx queues:

     ethtool -L combined 12 tx 4

Note:

  Dedicated Tx and Rx queues are not supported.


5.5.1. Virtualized Environments
-------------------------------

The following methods may be helpful to optimize performance in
virtual machines (VMs):

* Using the appropriate mechanism (vcpupin) in the VM, pin the CPUs to
  individual LCPUs, making sure to use a set of CPUs included in the
  device's "local_cpulist":
  "/sys/class/net/<ethX>/device/local_cpulist"

* Configure as many Rx/Tx queues in the VM as available. For example:

     ethtool -L <virt_interface> combined <max>

  Note:

    Dedicated Tx and Rx queues are not supported.


5.5.1.6. Known Issues/Troubleshooting
=====================================


5.5.1. Receive Error Counts May Be Higher Than the Actual Packet Error Count
----------------------------------------------------------------------------

When a packet is received with more than one error, two bad packets
may be reported.


5.5.1. "ethtool -S" Does Not Display Tx/Rx Packet Statistics
------------------------------------------------------------

Issuing the command "ethtool -S" does not display Tx/Rx packet
statistics. This is by convention. Use other tools (such as the **ip**
command) that display standard netdev statistics such as Tx/Rx packet
statistics.


5.5.1. Changing Ring Size During Heavy Traffic is Unstable
----------------------------------------------------------

"ethtool -G" should not be used while the driver is being used to send
or receive heavy traffic. This can result in the interface going into
the no-carrier state.


5.5.1. Unexpected Issues When the Device Driver and DPDK Share a Device
-----------------------------------------------------------------------

Unexpected issues may result when an idpf device is in multi driver
mode and the kernel driver and DPDK driver are sharing the device.
This is because access to the global NIC resources is not synchronized
between multiple drivers. Any change to the global NIC configuration
(writing to a global register, setting global configuration by AQ, or
changing switch modes) will affect all ports and drivers on the
device. Loading DPDK with the "multi-driver" module parameter may
mitigate some of the issues.


5.5.1.7. Support
================

For general information, go to the Intel support website at:
http://www.intel.com/support/

If an issue is identified with the released source code on a supported
kernel with a supported adapter, email the specific information
related to the issue to intel-wired-lan@lists.osuosl.org.


5.5.1.8. License
================

This program is free software; you can redistribute it and/or modify
it under the terms and conditions of the GNU General Public License,
version 2, as published by the Free Software Foundation.

This program is distributed in the hope it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin St - Fifth Floor, Boston, MA 02110-1301
USA.

The full GNU General Public License is included in this distribution
in the file called "COPYING".

Copyright (c) 2019 - 2024 Intel Corporation.


5.5.1.9. Trademarks
===================

Intel is a trademark or registered trademark of Intel Corporation or
its subsidiaries in the United States and/or other countries.

* Other names and brands may be claimed as the property of others.