Merge tag 'nfs-rdma-for-4.0-3' of git://git.linux-nfs.org/projects/anna/nfs-rdma

[karo-tx-linux.git] / Documentation / networking / ixgb.txt

Linux Base Driver for 10 Gigabit Intel(R) Ethernet Network Connection
=====================================================================

March 14, 2011


Contents
========

- In This Release
- Identifying Your Adapter
- Building and Installation
- Command Line Parameters
- Improving Performance
- Additional Configurations
- Known Issues/Troubleshooting
- Support



In This Release
===============

This file describes the ixgb Linux Base Driver for the 10 Gigabit Intel(R)
Network Connection.  This driver includes support for Itanium(R)2-based
systems.

For questions related to hardware requirements, refer to the documentation
supplied with your 10 Gigabit adapter.  All hardware requirements listed apply
to use with Linux.

The following features are available in this kernel:
 - Native VLANs
 - Channel Bonding (teaming)
 - SNMP

Channel Bonding documentation can be found in the Linux kernel source:
/Documentation/networking/bonding.txt

The driver information previously displayed in the /proc filesystem is not
supported in this release.  Alternatively, you can use ethtool (version 1.6
or later), lspci, and ifconfig to obtain the same information.

Instructions on updating ethtool can be found in the section "Additional
Configurations" later in this document.


Identifying Your Adapter
========================

The following Intel network adapters are compatible with the drivers in this
release:

Controller  Adapter Name                 Physical Layer
----------  ------------                 --------------
82597EX     Intel(R) PRO/10GbE LR/SR/CX4 10G Base-LR (1310 nm optical fiber)
            Server Adapters              10G Base-SR (850 nm optical fiber)
                                         10G Base-CX4(twin-axial copper cabling)

For more information on how to identify your adapter, go to the Adapter &
Driver ID Guide at:

    http://support.intel.com/support/network/sb/CS-012904.htm


Building and Installation
=========================

select m for "Intel(R) PRO/10GbE support" located at:
      Location:
        -> Device Drivers
          -> Network device support (NETDEVICES [=y])
            -> Ethernet (10000 Mbit) (NETDEV_10000 [=y])
1. make modules && make modules_install

2. Load the module:

    modprobe ixgb <parameter>=<value>

   The insmod command can be used if the full
   path to the driver module is specified.  For example:

     insmod /lib/modules/<KERNEL VERSION>/kernel/drivers/net/ixgb/ixgb.ko

   With 2.6 based kernels also make sure that older ixgb drivers are
   removed from the kernel, before loading the new module:

     rmmod ixgb; modprobe ixgb

3. Assign an IP address to the interface by entering the following, where
   x is the interface number:

     ifconfig ethx <IP_address>

4. 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>


Command Line Parameters
=======================

If the driver is built as a module, the  following optional parameters are
used by entering them on the command line with the modprobe command using
this syntax:

     modprobe ixgb [<option>=<VAL1>,<VAL2>,...]

For example, with two 10GbE PCI adapters, entering:

     modprobe ixgb TxDescriptors=80,128

loads the ixgb driver with 80 TX resources for the first adapter and 128 TX
resources for the second adapter.

The default value for each parameter is generally the recommended setting,
unless otherwise noted.

FlowControl
Valid Range: 0-3 (0=none, 1=Rx only, 2=Tx only, 3=Rx&Tx)
Default: Read from the EEPROM
         If EEPROM is not detected, default is 1
    This parameter controls the automatic generation(Tx) and response(Rx) to
    Ethernet PAUSE frames.  There are hardware bugs associated with enabling
    Tx flow control so beware.

RxDescriptors
Valid Range: 64-512
Default Value: 512
    This value is the number of receive descriptors allocated by the driver.
    Increasing this value allows the driver to buffer more incoming packets.
    Each descriptor is 16 bytes.  A receive buffer is also allocated for
    each descriptor and can be either 2048, 4056, 8192, or 16384 bytes,
    depending on the MTU setting.  When the MTU size is 1500 or less, the
    receive buffer size is 2048 bytes. When the MTU is greater than 1500 the
    receive buffer size will be either 4056, 8192, or 16384 bytes.  The
    maximum MTU size is 16114.

RxIntDelay
Valid Range: 0-65535 (0=off)
Default Value: 72
    This value delays the generation of receive interrupts in units of
    0.8192 microseconds.  Receive interrupt reduction can improve CPU
    efficiency if properly tuned for specific network traffic.  Increasing
    this value adds extra latency to frame reception and can end up
    decreasing the throughput of TCP traffic.  If the system is reporting
    dropped receives, this value may be set too high, causing the driver to
    run out of available receive descriptors.

TxDescriptors
Valid Range: 64-4096
Default Value: 256
    This value is the number of transmit descriptors allocated by the driver.
    Increasing this value allows the driver to queue more transmits.  Each
    descriptor is 16 bytes.

XsumRX
Valid Range: 0-1
Default Value: 1
    A value of '1' indicates that the driver should enable IP checksum
    offload for received packets (both UDP and TCP) to the adapter hardware.


Improving Performance
=====================

With the 10 Gigabit server adapters, the default Linux configuration will
very likely limit the total available throughput artificially.  There is a set
of configuration changes that, when applied together, will increase the ability
of Linux to transmit and receive data.  The following enhancements were
originally acquired from settings published at http://www.spec.org/web99/ for
various submitted results using Linux.

NOTE: These changes are only suggestions, and serve as a starting point for
      tuning your network performance.

The changes are made in three major ways, listed in order of greatest effect:
- Use ifconfig to modify the mtu (maximum transmission unit) and the txqueuelen
  parameter.
- Use sysctl to modify /proc parameters (essentially kernel tuning)
- Use setpci to modify the MMRBC field in PCI-X configuration space to increase
  transmit burst lengths on the bus.

NOTE: setpci modifies the adapter's configuration registers to allow it to read
up to 4k bytes at a time (for transmits).  However, for some systems the
behavior after modifying this register may be undefined (possibly errors of
some kind).  A power-cycle, hard reset or explicitly setting the e6 register
back to 22 (setpci -d 8086:1a48 e6.b=22) may be required to get back to a
stable configuration.

- COPY these lines and paste them into ixgb_perf.sh:
#!/bin/bash
echo "configuring network performance , edit this file to change the interface
or device ID of 10GbE card"
# set mmrbc to 4k reads, modify only Intel 10GbE device IDs
# replace 1a48 with appropriate 10GbE device's ID installed on the system,
# if needed.
setpci -d 8086:1a48 e6.b=2e
# set the MTU (max transmission unit) - it requires your switch and clients
# to change as well.
# set the txqueuelen
# your ixgb adapter should be loaded as eth1 for this to work, change if needed
ifconfig eth1 mtu 9000 txqueuelen 1000 up
# call the sysctl utility to modify /proc/sys entries
sysctl -p ./sysctl_ixgb.conf
- END ixgb_perf.sh

- COPY these lines and paste them into sysctl_ixgb.conf:
# some of the defaults may be different for your kernel
# call this file with sysctl -p <this file>
# these are just suggested values that worked well to increase throughput in
# several network benchmark tests, your mileage may vary

### IPV4 specific settings
# turn TCP timestamp support off, default 1, reduces CPU use
net.ipv4.tcp_timestamps = 0
# turn SACK support off, default on
# on systems with a VERY fast bus -> memory interface this is the big gainer
net.ipv4.tcp_sack = 0
# set min/default/max TCP read buffer, default 4096 87380 174760
net.ipv4.tcp_rmem = 10000000 10000000 10000000
# set min/pressure/max TCP write buffer, default 4096 16384 131072
net.ipv4.tcp_wmem = 10000000 10000000 10000000
# set min/pressure/max TCP buffer space, default 31744 32256 32768
net.ipv4.tcp_mem = 10000000 10000000 10000000

### CORE settings (mostly for socket and UDP effect)
# set maximum receive socket buffer size, default 131071
net.core.rmem_max = 524287
# set maximum send socket buffer size, default 131071
net.core.wmem_max = 524287
# set default receive socket buffer size, default 65535
net.core.rmem_default = 524287
# set default send socket buffer size, default 65535
net.core.wmem_default = 524287
# set maximum amount of option memory buffers, default 10240
net.core.optmem_max = 524287
# set number of unprocessed input packets before kernel starts dropping them; default 300
net.core.netdev_max_backlog = 300000
- END sysctl_ixgb.conf

Edit the ixgb_perf.sh script if necessary to change eth1 to whatever interface
your ixgb driver is using and/or replace '1a48' with appropriate 10GbE device's
ID installed on the system.

NOTE: Unless these scripts are added to the boot process, these changes will
      only last only until the next system reboot.


Resolving Slow UDP Traffic
--------------------------
If your server does not seem to be able to receive UDP traffic as fast as it
can receive TCP traffic, it could be because Linux, by default, does not set
the network stack buffers as large as they need to be to support high UDP
transfer rates.  One way to alleviate this problem is to allow more memory to
be used by the IP stack to store incoming data.

For instance, use the commands:
    sysctl -w net.core.rmem_max=262143
and
    sysctl -w net.core.rmem_default=262143
to increase the read buffer memory max and default to 262143 (256k - 1) from
defaults of max=131071 (128k - 1) and default=65535 (64k - 1).  These variables
will increase the amount of memory used by the network stack for receives, and
can be increased significantly more if necessary for your application.


Additional Configurations
=========================

  Configuring the Driver on Different Distributions
  -------------------------------------------------
  Configuring a network driver to load properly when the system is started is
  distribution dependent. Typically, the configuration process involves adding
  an alias line to /etc/modprobe.conf as well as editing other system startup
  scripts and/or configuration files.  Many popular Linux distributions ship
  with tools to make these changes for you.  To learn the proper way to
  configure a network device for your system, refer to your distribution
  documentation.  If during this process you are asked for the driver or module
  name, the name for the Linux Base Driver for the Intel 10GbE Family of
  Adapters is ixgb.

  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

  NOTE: This setting is not saved across reboots.


  Jumbo Frames
  ------------
  The driver supports Jumbo Frames for all adapters. Jumbo Frames support is
  enabled by changing the MTU to a value larger than the default of 1500.
  The maximum value for the MTU is 16114.  Use the ifconfig command to
  increase the MTU size.  For example:

        ifconfig ethx mtu 9000 up

  The maximum MTU setting for Jumbo Frames is 16114.  This value coincides
  with the maximum Jumbo Frames size of 16128.


  ethtool
  -------
  The driver utilizes the ethtool interface for driver configuration and
  diagnostics, as well as displaying statistical information.  The ethtool
  version 1.6 or later is required for this functionality.

  The latest release of ethtool can be found from
  http://ftp.kernel.org/pub/software/network/ethtool/

  NOTE: The ethtool version 1.6 only supports a limited set of ethtool options.
        Support for a more complete ethtool feature set can be enabled by
        upgrading to the latest version.


  NAPI
  ----

  NAPI (Rx polling mode) is supported in the ixgb driver.  NAPI is enabled
  or disabled based on the configuration of the kernel.  see CONFIG_IXGB_NAPI

  See www.cyberus.ca/~hadi/usenix-paper.tgz for more information on NAPI.


Known Issues/Troubleshooting
============================

  NOTE: After installing the driver, if your Intel Network Connection is not
  working, verify in the "In This Release" section of the readme that you have
  installed the correct driver.

  Intel(R) PRO/10GbE CX4 Server Adapter Cable Interoperability Issue with
  Fujitsu XENPAK Module in SmartBits Chassis
  ---------------------------------------------------------------------
  Excessive CRC errors may be observed if the Intel(R) PRO/10GbE CX4
  Server adapter is connected to a Fujitsu XENPAK CX4 module in a SmartBits
  chassis using 15 m/24AWG cable assemblies manufactured by Fujitsu or Leoni.
  The CRC errors may be received either by the Intel(R) PRO/10GbE CX4
  Server adapter or the SmartBits. If this situation occurs using a different
  cable assembly may resolve the issue.

  CX4 Server Adapter Cable Interoperability Issues with HP Procurve 3400cl
  Switch Port
  ------------------------------------------------------------------------
  Excessive CRC errors may be observed if the Intel(R) PRO/10GbE CX4 Server
  adapter is connected to an HP Procurve 3400cl switch port using short cables
  (1 m or shorter). If this situation occurs, using a longer cable may resolve
  the issue.

  Excessive CRC errors may be observed using Fujitsu 24AWG cable assemblies that
  Are 10 m or longer or where using a Leoni 15 m/24AWG cable assembly. The CRC
  errors may be received either by the CX4 Server adapter or at the switch. If
  this situation occurs, using a different cable assembly may resolve the issue.


  Jumbo Frames System Requirement
  -------------------------------
  Memory allocation failures have been observed on Linux systems with 64 MB
  of RAM or less that are running Jumbo Frames.  If you are using Jumbo
  Frames, your system may require more than the advertised minimum
  requirement of 64 MB of system memory.


  Performance Degradation with Jumbo Frames
  -----------------------------------------
  Degradation in throughput performance may be observed in some Jumbo frames
  environments.  If this is observed, increasing the application's socket buffer
  size and/or increasing the /proc/sys/net/ipv4/tcp_*mem entry values may help.
  See the specific application manual and /usr/src/linux*/Documentation/
  networking/ip-sysctl.txt for more details.


  Allocating Rx Buffers when Using Jumbo Frames
  ---------------------------------------------
  Allocating Rx buffers when using Jumbo Frames on 2.6.x kernels may fail if
  the available memory is heavily fragmented. This issue may be seen with PCI-X
  adapters or with packet split disabled. This can be reduced or eliminated
  by changing the amount of available memory for receive buffer allocation, by
  increasing /proc/sys/vm/min_free_kbytes.


  Multiple Interfaces on Same Ethernet Broadcast Network
  ------------------------------------------------------
  Due to the default ARP behavior on Linux, it is not possible to have
  one system on two IP networks in the same Ethernet broadcast domain
  (non-partitioned switch) behave as expected.  All Ethernet interfaces
  will respond to IP traffic for any IP address assigned to the system.
  This results in unbalanced receive traffic.

  If you have multiple interfaces in a server, do either of the following:

  - Turn on ARP filtering by entering:
      echo 1 > /proc/sys/net/ipv4/conf/all/arp_filter

  - Install the interfaces in separate broadcast domains - either in
    different switches or in a switch partitioned to VLANs.


  UDP Stress Test Dropped Packet Issue
  --------------------------------------
  Under small packets UDP stress test with 10GbE driver, the Linux system
  may drop UDP packets due to the fullness of socket buffers. You may want
  to change the driver's Flow Control variables to the minimum value for
  controlling packet reception.


  Tx Hangs Possible Under Stress
  ------------------------------
  Under stress conditions, if TX hangs occur, turning off TSO
  "ethtool -K eth0 tso off" may resolve the problem.


Support
=======

For general information, go to the Intel support website at:

    http://support.intel.com

or the Intel Wired Networking project hosted by Sourceforge at:

    http://sourceforge.net/projects/e1000

If an issue is identified with the released source code on the supported
kernel with a supported adapter, email the specific information related
to the issue to e1000-devel@lists.sf.net