Linux* Base Driver for the Intel(R) PRO/10GbE Family of Adapters
================================================================

November 17, 2004


Contents
========

- In This Release
- Identifying Your Adapter
- Command Line Parameters
- Improving Performance
- Support


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

This file describes the Linux* Base Driver for the Intel(R) PRO/10GbE Family 
of Adapters, version 1.0.x.  

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

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

To verify your Intel adapter is supported, find the board ID number on the 
adapter. Look for a label that has a barcode and a number in the format  
A12345-001. 

Use the above information and the Adapter & Driver ID Guide at:

  http://support.intel.com/support/network/adapter/pro100/21397.htm

For the latest Intel network drivers for Linux, go to:

    http://downloadfinder.intel.com/scripts-df/support_intel.asp

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

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

For example, with two PRO/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. Also, if the driver is statically built into the
kernel, the driver is loaded with the default values for all the parameters.
Ethtool can be used to change some of the parameters at runtime.

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 3
    This parameter controls the automatic generation(Tx) and response(Rx) to 
    Ethernet PAUSE frames.

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: 6
    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.

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

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

With the Intel PRO/10 GbE 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:
- 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 8086:1048 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"
# set mmrbc to 4k reads, modify only Intel 10GbE device IDs
setpci -d 8086:1048 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
net.ipv4.tcp_timestamps = 0 # turns TCP timestamp support off, default 1, reduces CPU use
net.ipv4.tcp_sack = 0 # turn SACK support off, default on
# on systems with a VERY fast bus -> memory interface this is the big gainer 
net.ipv4.tcp_rmem = 10000000 10000000 10000000 # sets min/default/max TCP read buffer, default 4096 87380 174760
net.ipv4.tcp_wmem = 10000000 10000000 10000000 # sets min/pressure/max TCP write buffer, default 4096 16384 131072
net.ipv4.tcp_mem = 10000000 10000000 10000000 # sets min/pressure/max TCP buffer space, default 31744 32256 32768

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

Edit the ixgb_perf.sh script if necessary to change eth1 to whatever interface 
your ixgb driver is using.

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.

Support
=======

For general information and support, 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.