Linux* Base Driver for 10 Gigabit Intel® Network Connection

NOTES:  This release includes the Linux* Base Driver for the Intel® 10 Gigabit Family of Adapters. All 82598-based 10 Gigabit network connections require the ixgbe driver. All other 10 Gigabit network connections require the ixgb driver.  First identify your adapter.  Then follow the appropriate steps for building, installing, and configuring the specified driver.

The ixgb driver can be downloaded from: http://sourceforge.net/projects/e1000

ixgbe Driver

ixgb Driver


Using the ixgbe base driver

Important Note

Overview

Building and Installation

Command Line Parameters

Improving Performance

Additional Configurations

Known Issues/Troubleshooting 


Important Note

Warning: The ixgbe driver compiles by default with the LRO (Large Receive Offload) feature enabled. This option offers the lowest CPU utilization for
receives, but is completely incompatible with *routing/ip fowarding* and *bridging*. If enabling ip forwarding or bridging is a requirement, it is necessary to disable LRO using compile time options as noted in the LRO section later in this document. The result of not disabling LRO when combined with ip forwarding or bridging can be low throughput or even a kernel PANIC.

Overview

The Linux* base driver supports the 2.6.x kernel, and includes support for any Linux supported system, including Itanium(R)2, x86_64, i686, and PPC.

These drivers are 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 driver. A version of the driver may already be included by your distribution and/or the kernel.org kernel. For questions related to hardware requirements, refer to the documentation supplied with your Intel adapter. All hardware requirements listed apply to use with Linux.

The following features are now available in supported kernels:

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

The driver information previously displayed in the /proc file system 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.

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

Controller Adapter Name Physical Layer
82598EB Intel® 10 Gigabit XF SR Server Adapter 10G Base -SR (850 nm optical fiber)
10G Base -LRM (850 nm optical fiber)
10G Base -LR (1310 nm optical fiber)
82598EB Intel® 10 Gigabit XF LR Server Adapter  
82598EB Intel® 10 Gigabit XF SR Dual Port Server Adapter  
82598EB Intel® 82598EB 10 Gigabit AF Dual Port Network Connection  
82598EB Intel® 82598EB 10 Gigabit AF Network Connection  
82598EB Intel® 82598EB 10 Gigabit AT CX4 Network Connection  
82598EB Intel® 10 Gigabit SR Dual Port Express Module  
82598EB Intel® 10 Gigabit AF DA Dual Port Server Adapter  
82598EB Intel® 10 Gigabit AT Server Adapter 10G Base -T

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

http://support.intel.com/support/go/network/adapter/proidguide.htm

For the latest Intel network drivers for Linux, refer to the following website. In the search field, enter your adapter name or type, or use the networking link on the left to search for your adapter:

http://downloadcenter.intel.com/scripts-df-external/Support_Intel.aspx


Building and Installation

To build a binary RPM* package of this driver, run 'rpmbuild -tb ixgbe.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 now.

RPM functionality has only been tested in RedHat distributions.

To manually build this driver:

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

  2. Untar/unzip the archive:

    tar zxf ixgbe-x.x.x.tar.gz

  3. Change to the driver src directory:

    cd ixgbe-x.x.x/src/

  4. Compile the driver module:

    make install

    The binary will be installed as:

    /lib/modules/[KERNEL_VERSION]/kernel/drivers/net/ixgbe/ixgbe.[k]o

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

  5. Load the module:

    For kernel 2.6.x, use the modprobe command -

          modprobe ixgbe <parameter>=<value>

    Note that for 2.6 kernels 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/ixgbe/ixgbe.ko

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

    rmmod ixgbe; modprobe ixgbe

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

    ifconfig ethx <IP_address> netmask <netmask>

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

    ping <IP_address>

To build ixgbe driver with DCA

This example assumes the ioatdma and ixgbe sources are in /usr/src

  1. Unpack the ioatdma source, build and install

        cd /usr/src
        tar zxf ioatdma-<ioat version>.tar.gz
        cd ioatdma-<ioat version>
        make
        make install

  2. Unpack with ixgbe driver, build with DCA support and install

        cd /usr/src
        tar zxf ixgbe-<ixgbe version>.tar.gz
        cd ixgbe-<ixgbe-version>/src
        make install CLFLAGS_EXTRA="-DIXGBE_DCA -I/path/to/ioatdma-<ioat-version>/include"


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 ixgbe [<option>=<VAL1>,<VAL2>,...]

For example:

modprobe ixgbe InterruptThrottleRate=16000,16000

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

Parameter Name Valid Range/Settings Default Description
RSS - Receive Side Scaling (or multiple queues for receives) 0 - 16

 

1 0 = disables RSS
1 = enables RSS and sets the descriptor queue count to 16 or the number of
online cpus, whichever is less.
2-16 = enables RSS, with 2-16 queues

RSS also effects the number of transmit queues allocated on 2.6.23 and
newer kernels with CONFIG_NETDEVICES_MULTIQUEUE set in the kernel .config file.

MQ - Multi Queue 0, 1

 

1 0 = Disables Multiple Queue support
1 = Enabled Multiple Queue support (a prerequisite for RSS)
DCA - Direct Cache Access 0, 1

 

1 (when IXGBE_DCA is enabled) 0 = Disables DCA support in the driver
1 = Enables DCA support in the driver

See the above instructions for enabling DCA. If the driver is enabled for
DCA this parameter allows load-time control of the feature.

RxBufferMode 0-2 2 0 = Driver will use single buffer for Rx packets.
1 = Driver will use packet split mode for Rx. Packet header will be received in header buffer and payload will be received in data buffer.
2. = Optimal mode. Driver will use single buffer mode for non-Jumbo configurations and packet split mode for Jumbo configurations.
InterruptType 0-2 0 = Legacy Int, 1 = MSI and 2 = MSIX 2 Interrupt type
InterruptThrottleRate 100-500,000 (0=off, 1=dynamic) 8000 Interrupt Throttle Rate (interrupts/sec). The ITR parameter controls how many interrupts each interrupt vector can generate per second. On MQ/RSS enabled kernels with MSI-X interrupts this means that each RX vector can generate (by default) 8000 interrupts per second and each TX vector can generate (by default) 4000 interrupts per second. Increasing ITR lowers latency at the cost of increased CPU utilization, though it may help throughput in some circumstances.

1 = Dynamic mode attempts to moderate interrupts per vector while maintaining very low latency. This can sometimes cause extra CPU utilization. If planning on deploying ixgbe in a latency sensitive environment please consider this parameter.

0 = Setting InterruptThrottleRate to 0 turns off any interrupt moderation and may improve small packet latency, but is generally not suitable for bulk throughput traffic due to the increased cpu utilization of the higher interrupt rate.

LLI (Low Latency Interrupts)

 
    LLI allows for immediate generation of an interrupt upon processing receive
packets that match certain criteria as set by the parameters described below.
LLI parameters are not enabled when Legacy interrupts are used. You must be
using MSI or MSI-X (see cat /proc/interrupts) to successfully use LLI.
LLIPort 0 - 65535 0 (disabled) LLI is configured with the LLIPort command-line parameter, which specifies
which TCP port should generate Low Latency Interrupts.

For example, using LLIPort=80 would cause the hardware to generate an
immediate interrupt upon receipt of any packet sent to TCP port 80 on the
local machine.
CAUTION: Enabling LLI can result in an excessive number of interrupts/second that may cause problems with the system and in some cases may cause a kernel panic.
    LLIPush       0-1 0 (disabled) LLIPush can be set to be enabled or disabled (default). It is most effective in an environment with many small transactions.
NOTE: Enabling LLIPush may allow a denial of service attack.
    LLISize       0-1500 0 (disabled) LLISize causes an immediate interrupt if the board receives a packet smaller
than the specified size.

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/modules.conf or 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 10 Gigabit Family of Adapters is ixgbe.

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 16110. Use the ifconfig command to increase the MTU size. For example, enter the following where <x> is the interface number:

ifconfig ethx mtu 9000 up

The maximum MTU setting for Jumbo Frames is 16110. This value coincides with the maximum Jumbo Frames size of 16128. 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.

Ethtool

The driver utilizes the ethtool interface for driver configuration and diagnostics, as well as displaying statistical information.  Ethtool version 3.0 or later is required for this functionality.

The latest release of ethtool can be found at: http://sourceforge.net/projects/gkernel

NAPI

NAPI (Rx polling mode) is supported in the ixgbe driver. NAPI is enabled or disabled based on the configuration of the kernel. To override the default, use the following compile-time flags.

To enable NAPI, compile the driver module, passing in a configuration option:

     make CFLAGS_EXTRA=-DIXGBE_NAPI install

NOTE: This will not do anything if NAPI is disabled in the kernel.

To disable NAPI, compile the driver module, passing in a configuration option:

     make CFLAGS_EXTRA=-DIXGBE_NO_NAPI install

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

LRO

Large Receive Offload (LRO) is a technique for increasing inbound throughput of high-bandwidth network connections by reducing CPU overhead. It works by aggregating multiple incoming packets from a single stream into a larger buffer before they are passed higher up the networking stack, thus reducing the number of packets that have to be processed. LRO combines multiple Ethernet frames into a single receive in the stack, thereby potentially decreasing CPU utilization for receives.

IXGBE_NO_LRO is a compile time flag. The user can enable it at compile time to remove support for LRO from the driver. The flag is used by adding
CFLAGS_EXTRA="-DIXGBE_NO_LRO" to the make file when it's being compiled.

make CFLAGS_EXTRA="-DIXGBE_NO_LRO" install

You can verify that the driver is using LRO by looking at these counters in Ethtool:

lro_flushed - the total number of receives using LRO.

lro_coal - counts the total number of Ethernet packets that were combined.

rx_dropped_backlog

When in a non-Napi (or Interrupt) mode, this counter indicates that the stack is dropping packets. There is an adjustable parameter in the stack that allows you to adjust the amount of backlog. We recommend increasing the netdev_max_backlog if the counter goes up.

# sysctl -a |grep netdev_max_backlog

net.core.netdev_max_backlog = 1000

# sysctl -e net.core.netdev_max_backlog=10000

net.core.netdev_max_backlog = 10000

Flow Control

Flow control is disabled by default. If you want to enable and have a flow control capable link partner, use Ethtool:

    ethtool -A eth? autoneg off rx on tx on

FCOE

This release of the ixgbe driver contains new code to enable users to use Fiber Channel over Ethernet (FCoE) and Data Center Bridging (DCB) functionality that is supported by the 82598-based hardware. This code has no default effect on the regular driver operation, and configuring DCB and FCoE is outside the scope of this driver README. Refer to http://www.open-fcoe.org/ for FCoE project information and contact e1000-eedc@lists.sourceforge.net for DCB information.


Known Issues/Troubleshooting

NOTE: After installing the driver, if your Intel Network Connection is not working, verify that you have installed the correct driver.

MSI-X Issues with 82598-based Intel(R)10GbE-LR/LRM/SR/AT Server Adapters

Kernel panics and instability may be observed on some platforms when running 82598-based Intel(R) 10GbE-LR/LRM/SR/AT Server Adapters with MSI-X in a stress environment. Symptoms of this issue include observing "APIC 40 Error" or "no irq handler for vector" error messages on the console or in "dmesg."

If such problems are encountered, you may disable the irqbalance daemon. If the problems persist, compile the driver in pin interrupt mode, do

        make CFLAGS_EXTRA=-DDISABLE_PCI_MSI

Or you can load the module with

modprobe ixgbe InterruptType=0

Driver Compilation

When trying to compile the driver by running make install, the following error may occur:  "Linux kernel source not configured - missing version.h"

To solve this issue, create the version.h file by going to the Linux source tree and entering:

make include/linux/version.h

Do Not Use LRO When Routing Packets

Due to a known general compatibility issue with LRO and routing, do not use LRO when routing packets.

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.

Multiple Interfaces on Same Ethernet Broadcast Network

Due to the 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, either turn on ARP filtering by entering:

     echo 1 > /proc/sys/net/ipv4/conf/all/arp_filter

(this only works if your kernel's version is higher than 2.4.5), or 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.

Or you can increase the kernel's default buffer sizes for UDP by changing the values in

/proc/sys/net/core/rmem_default and rmem_max

Unplugging network cable while ethtool -p is running

In kernel versions 2.5.50 and later (including 2.6 kernel), unplugging the network cable while ethtool -p is running will cause the system to become unresponsive to keyboard commands, except for control-alt-delete. Restarting the system appears to be the only remedy.

Cisco Catalyst 4948-10GE port resets may cause switch to shut down ports

82598-based hardware can re-establish link quickly and when connected to some switches, rapid resets within the driver may cause the switch port to become
isolated due to "link flap". This is typically indicated by a yellow instead of a green link light. Several operations may cause this problem, such as repeatedly running ethtool commands that cause a reset.

A potential workaround is to use the Cisco IOS command "no errdisable detect cause all" from the Global Configuration prompt which enables the switch to keep the interfaces up, regardless of errors.

Installing RedHat Enterprise Linux 5.2 with an Intel(R) 10 Gigabit AT Server Adapter may cause kernel panic

A known issue may cause a kernel panic or hang after installing an 82598AT-based Intel(R) 10 Gigabit AT Server Adapter in a RedHat Enterprise Linux 5.2 system. The ixgbe driver for both the install kernel and the runtime kernel can create this panic if the 82598AT adapter is installed. RedHat may release a security update that contains a fix for the panic that you can download using RHN (RedHat Network) or Intel recommends that you install the ixgbe-1.3.31.5 driver or newer BEFORE installing the hardware.


Using the ixgb base driver

Overview

Building and Installation

Command Line Parameters

Improving Performance

Additional Configurations

Known Issues/Troubleshooting 


Overview

The Linux* base driver supports the 2.6.x kernel, and includes support for any Linux supported system, including Itanium(R)2, x86_64, i686, and PPC.

NOTE: The ixgb driver is not included on this release. It can be downloaded from: http://sourceforge.net/projects/e1000

These drivers are 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 driver. A version of the driver may already be included by your distribution and/or the kernel.org kernel. For questions related to hardware requirements, refer to the documentation supplied with your Intel adapter. All hardware requirements listed apply to use with Linux.

The following features are now available in supported kernels:

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

The driver information previously displayed in the /proc file system 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.

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

Controller Adapter Name Physical Layer
82597EX Intel® PRO/10GbE SR Server Adapter 10GBase-SR (850 nm optical fiber)
82597EX Intel® PRO/10GbE LR Server Adapter 10GBase-LR (1310 nm optical fiber)
82597EX Intel® PRO/10GbE CX4 Server Adapter 10GBase-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/go/network/adapter/proidguide.htm

For the latest Intel network drivers for Linux, refer to the following website. In the search field, enter your adapter name or type, or use the networking link on the left to search for your adapter:

http://downloadcenter.intel.com/scripts-df-external/Support_Intel.aspx


Building and Installation

To build a binary RPM* package of this driver, run 'rpmbuild -tb ixgb.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 now.

RPM functionality has only been tested in RedHat distributions.

To manually build this driver:

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

  2. Untar/unzip the archive:

    tar zxf filename.tar.gz

  3. Change to the driver src directory:

    cd filename/src/

  4. Compile the driver module:

    make install

    The binary will be installed as:

    /lib/modules/[KERNEL_VERSION]/kernel/drivers/net/filename.[k]o

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

  5. Load the module:

     For kernel 2.4.x, use the insmod command -

         insmod filename <parameter>=<value>

    For kernel 2.6.x, use the modprobe command -

          modprobe filename <parameter>=<value>

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

    rmmod filename; modprobe ixgb

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

    ifconfig ethx <IP_address>

  7. Verify that the interface works. Enter the following, where <IP_address> is the IP address of 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 or insmod command using this syntax:

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

insmod filename [<option>=<VAL1>,<VAL2>,...]

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

insmod 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.

Parameter Name Valid Range/Settings Default Description
FlowControl 0-3 (0=none, 1=Rx only, 2=Tx only, 3=Rx&Tx) Read flow control settings from the EEPROM This parameter controls the automatic generation(Tx) and response(Rx) to Ethernet PAUSE frames.
RxDescriptors 64-512 512 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 0-65535 (0=off) 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 64-4096 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    0-1 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 adapter, the default Linux configuration will very likely limit the total available throughput artificially.  There is a set of things that when applied together 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:

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.
# 8086:1a48 is the Intel SR adapter
setpci -d 8086:1a48 e6.b=2e
# set the MTU (max transmission unit) - it requires your switch and clients to change too!
# 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, 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), use the commands: 

     sysctl -w net.core.rmem_max=262143

and

     sysctl -w net.core.rmem_default=262143

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/modules.conf or 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 10 Gigabit Family of Adapters is ixgb or ixgbe.

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 16110. Use the ifconfig command to increase the MTU size. For example, enter the following where <x> is the interface number:

ifconfig ethx mtu 9000 up

The maximum MTU setting for Jumbo Frames is 16110. 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.  Ethtool version 3.0 or later is required for this functionality.

The latest release of ethtool can be found at: http://sourceforge.net/projects/gkernel

NAPI

NAPI (Rx polling mode) is supported in the ixgb driver. NAPI is enabled or disabled based on the configuration of the kernel. To override the default, use the following compile-time flags.

To enable NAPI, compile the driver module, passing in a configuration option:

     make CFLAGS_EXTRA=-DIXGB_NAPI install

NOTE: This will not do anything if NAPI is disabled in the kernel.

To disable NAPI, compile the driver module, passing in a configuration option:

     make CFLAGS_EXTRA=-DIXGB_NO_NAPI install

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 that you have installed the correct driver.

Driver Compilation

When trying to compile the driver by running make install, the following error may occur:  "Linux kernel source not configured - missing version.h"

To solve this issue, create the version.h file by going to the Linux source tree and entering:

make include/linux/version.h

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 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, either turn on ARP filtering by entering:

     echo 1 > /proc/sys/net/ipv4/conf/all/arp_filter

(this only works if your kernel's version is higher than 2.4.5), or 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.

Or you can increase the kernel's default buffer sizes for UDP by changing the values in

/proc/sys/net/core/rmem_default and rmem_max

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

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 linux.nics@intel.com.


Last modified on 7/22/08 1:38p 10/22/04 9:45a 71