linux/Documentation/networking/e1000.txt
<<
>>
Prefs
   1Linux* Base Driver for the Intel(R) PRO/1000 Family of Adapters
   2===============================================================
   3
   4Intel Gigabit Linux driver.
   5Copyright(c) 1999 - 2010 Intel Corporation.
   6
   7Contents
   8========
   9
  10- Identifying Your Adapter
  11- Command Line Parameters
  12- Speed and Duplex Configuration
  13- Additional Configurations
  14- Support
  15
  16Identifying Your Adapter
  17========================
  18
  19For more information on how to identify your adapter, go to the Adapter &
  20Driver ID Guide at:
  21
  22    http://support.intel.com/support/go/network/adapter/idguide.htm
  23
  24For the latest Intel network drivers for Linux, refer to the following
  25website.  In the search field, enter your adapter name or type, or use the
  26networking link on the left to search for your adapter:
  27
  28    http://support.intel.com/support/go/network/adapter/home.htm
  29
  30Command Line Parameters
  31=======================
  32
  33The default value for each parameter is generally the recommended setting,
  34unless otherwise noted.
  35
  36NOTES:  For more information about the AutoNeg, Duplex, and Speed
  37        parameters, see the "Speed and Duplex Configuration" section in
  38        this document.
  39
  40        For more information about the InterruptThrottleRate,
  41        RxIntDelay, TxIntDelay, RxAbsIntDelay, and TxAbsIntDelay
  42        parameters, see the application note at:
  43        http://www.intel.com/design/network/applnots/ap450.htm
  44
  45AutoNeg
  46-------
  47(Supported only on adapters with copper connections)
  48Valid Range:   0x01-0x0F, 0x20-0x2F
  49Default Value: 0x2F
  50
  51This parameter is a bit-mask that specifies the speed and duplex settings
  52advertised by the adapter.  When this parameter is used, the Speed and
  53Duplex parameters must not be specified.
  54
  55NOTE:  Refer to the Speed and Duplex section of this readme for more
  56       information on the AutoNeg parameter.
  57
  58Duplex
  59------
  60(Supported only on adapters with copper connections)
  61Valid Range:   0-2 (0=auto-negotiate, 1=half, 2=full)
  62Default Value: 0
  63
  64This defines the direction in which data is allowed to flow.  Can be
  65either one or two-directional.  If both Duplex and the link partner are
  66set to auto-negotiate, the board auto-detects the correct duplex.  If the
  67link partner is forced (either full or half), Duplex defaults to half-
  68duplex.
  69
  70FlowControl
  71-----------
  72Valid Range:   0-3 (0=none, 1=Rx only, 2=Tx only, 3=Rx&Tx)
  73Default Value: Reads flow control settings from the EEPROM
  74
  75This parameter controls the automatic generation(Tx) and response(Rx)
  76to Ethernet PAUSE frames.
  77
  78InterruptThrottleRate
  79---------------------
  80(not supported on Intel(R) 82542, 82543 or 82544-based adapters)
  81Valid Range:   0,1,3,4,100-100000 (0=off, 1=dynamic, 3=dynamic conservative,
  82                                 4=simplified balancing)
  83Default Value: 3
  84
  85The driver can limit the amount of interrupts per second that the adapter
  86will generate for incoming packets. It does this by writing a value to the
  87adapter that is based on the maximum amount of interrupts that the adapter
  88will generate per second.
  89
  90Setting InterruptThrottleRate to a value greater or equal to 100
  91will program the adapter to send out a maximum of that many interrupts
  92per second, even if more packets have come in. This reduces interrupt
  93load on the system and can lower CPU utilization under heavy load,
  94but will increase latency as packets are not processed as quickly.
  95
  96The default behaviour of the driver previously assumed a static
  97InterruptThrottleRate value of 8000, providing a good fallback value for
  98all traffic types,but lacking in small packet performance and latency.
  99The hardware can handle many more small packets per second however, and
 100for this reason an adaptive interrupt moderation algorithm was implemented.
 101
 102Since 7.3.x, the driver has two adaptive modes (setting 1 or 3) in which
 103it dynamically adjusts the InterruptThrottleRate value based on the traffic
 104that it receives. After determining the type of incoming traffic in the last
 105timeframe, it will adjust the InterruptThrottleRate to an appropriate value
 106for that traffic.
 107
 108The algorithm classifies the incoming traffic every interval into
 109classes.  Once the class is determined, the InterruptThrottleRate value is
 110adjusted to suit that traffic type the best. There are three classes defined:
 111"Bulk traffic", for large amounts of packets of normal size; "Low latency",
 112for small amounts of traffic and/or a significant percentage of small
 113packets; and "Lowest latency", for almost completely small packets or
 114minimal traffic.
 115
 116In dynamic conservative mode, the InterruptThrottleRate value is set to 4000
 117for traffic that falls in class "Bulk traffic". If traffic falls in the "Low
 118latency" or "Lowest latency" class, the InterruptThrottleRate is increased
 119stepwise to 20000. This default mode is suitable for most applications.
 120
 121For situations where low latency is vital such as cluster or
 122grid computing, the algorithm can reduce latency even more when
 123InterruptThrottleRate is set to mode 1. In this mode, which operates
 124the same as mode 3, the InterruptThrottleRate will be increased stepwise to
 12570000 for traffic in class "Lowest latency".
 126
 127In simplified mode the interrupt rate is based on the ratio of TX and
 128RX traffic.  If the bytes per second rate is approximately equal, the
 129interrupt rate will drop as low as 2000 interrupts per second.  If the
 130traffic is mostly transmit or mostly receive, the interrupt rate could
 131be as high as 8000.
 132
 133Setting InterruptThrottleRate to 0 turns off any interrupt moderation
 134and may improve small packet latency, but is generally not suitable
 135for bulk throughput traffic.
 136
 137NOTE:  InterruptThrottleRate takes precedence over the TxAbsIntDelay and
 138       RxAbsIntDelay parameters.  In other words, minimizing the receive
 139       and/or transmit absolute delays does not force the controller to
 140       generate more interrupts than what the Interrupt Throttle Rate
 141       allows.
 142
 143CAUTION:  If you are using the Intel(R) PRO/1000 CT Network Connection
 144          (controller 82547), setting InterruptThrottleRate to a value
 145          greater than 75,000, may hang (stop transmitting) adapters
 146          under certain network conditions.  If this occurs a NETDEV
 147          WATCHDOG message is logged in the system event log.  In
 148          addition, the controller is automatically reset, restoring
 149          the network connection.  To eliminate the potential for the
 150          hang, ensure that InterruptThrottleRate is set no greater
 151          than 75,000 and is not set to 0.
 152
 153NOTE:  When e1000 is loaded with default settings and multiple adapters
 154       are in use simultaneously, the CPU utilization may increase non-
 155       linearly.  In order to limit the CPU utilization without impacting
 156       the overall throughput, we recommend that you load the driver as
 157       follows:
 158
 159           modprobe e1000 InterruptThrottleRate=3000,3000,3000
 160
 161       This sets the InterruptThrottleRate to 3000 interrupts/sec for
 162       the first, second, and third instances of the driver.  The range
 163       of 2000 to 3000 interrupts per second works on a majority of
 164       systems and is a good starting point, but the optimal value will
 165       be platform-specific.  If CPU utilization is not a concern, use
 166       RX_POLLING (NAPI) and default driver settings.
 167
 168RxDescriptors
 169-------------
 170Valid Range:   80-256 for 82542 and 82543-based adapters
 171               80-4096 for all other supported adapters
 172Default Value: 256
 173
 174This value specifies the number of receive buffer descriptors allocated
 175by the driver.  Increasing this value allows the driver to buffer more
 176incoming packets, at the expense of increased system memory utilization.
 177
 178Each descriptor is 16 bytes.  A receive buffer is also allocated for each
 179descriptor and can be either 2048, 4096, 8192, or 16384 bytes, depending
 180on the MTU setting. The maximum MTU size is 16110.
 181
 182NOTE:  MTU designates the frame size.  It only needs to be set for Jumbo
 183       Frames.  Depending on the available system resources, the request
 184       for a higher number of receive descriptors may be denied.  In this
 185       case, use a lower number.
 186
 187RxIntDelay
 188----------
 189Valid Range:   0-65535 (0=off)
 190Default Value: 0
 191
 192This value delays the generation of receive interrupts in units of 1.024
 193microseconds.  Receive interrupt reduction can improve CPU efficiency if
 194properly tuned for specific network traffic.  Increasing this value adds
 195extra latency to frame reception and can end up decreasing the throughput
 196of TCP traffic.  If the system is reporting dropped receives, this value
 197may be set too high, causing the driver to run out of available receive
 198descriptors.
 199
 200CAUTION:  When setting RxIntDelay to a value other than 0, adapters may
 201          hang (stop transmitting) under certain network conditions.  If
 202          this occurs a NETDEV WATCHDOG message is logged in the system
 203          event log.  In addition, the controller is automatically reset,
 204          restoring the network connection.  To eliminate the potential
 205          for the hang ensure that RxIntDelay is set to 0.
 206
 207RxAbsIntDelay
 208-------------
 209(This parameter is supported only on 82540, 82545 and later adapters.)
 210Valid Range:   0-65535 (0=off)
 211Default Value: 128
 212
 213This value, in units of 1.024 microseconds, limits the delay in which a
 214receive interrupt is generated.  Useful only if RxIntDelay is non-zero,
 215this value ensures that an interrupt is generated after the initial
 216packet is received within the set amount of time.  Proper tuning,
 217along with RxIntDelay, may improve traffic throughput in specific network
 218conditions.
 219
 220Speed
 221-----
 222(This parameter is supported only on adapters with copper connections.)
 223Valid Settings: 0, 10, 100, 1000
 224Default Value:  0 (auto-negotiate at all supported speeds)
 225
 226Speed forces the line speed to the specified value in megabits per second
 227(Mbps).  If this parameter is not specified or is set to 0 and the link
 228partner is set to auto-negotiate, the board will auto-detect the correct
 229speed.  Duplex should also be set when Speed is set to either 10 or 100.
 230
 231TxDescriptors
 232-------------
 233Valid Range:   80-256 for 82542 and 82543-based adapters
 234               80-4096 for all other supported adapters
 235Default Value: 256
 236
 237This value is the number of transmit descriptors allocated by the driver.
 238Increasing this value allows the driver to queue more transmits.  Each
 239descriptor is 16 bytes.
 240
 241NOTE:  Depending on the available system resources, the request for a
 242       higher number of transmit descriptors may be denied.  In this case,
 243       use a lower number.
 244
 245TxDescriptorStep
 246----------------
 247Valid Range:    1 (use every Tx Descriptor)
 248                4 (use every 4th Tx Descriptor)
 249
 250Default Value:  1 (use every Tx Descriptor)
 251
 252On certain non-Intel architectures, it has been observed that intense TX
 253traffic bursts of short packets may result in an improper descriptor
 254writeback. If this occurs, the driver will report a "TX Timeout" and reset
 255the adapter, after which the transmit flow will restart, though data may
 256have stalled for as much as 10 seconds before it resumes.
 257
 258The improper writeback does not occur on the first descriptor in a system
 259memory cache-line, which is typically 32 bytes, or 4 descriptors long.
 260
 261Setting TxDescriptorStep to a value of 4 will ensure that all TX descriptors
 262are aligned to the start of a system memory cache line, and so this problem
 263will not occur.
 264
 265NOTES: Setting TxDescriptorStep to 4 effectively reduces the number of
 266       TxDescriptors available for transmits to 1/4 of the normal allocation.
 267       This has a possible negative performance impact, which may be
 268       compensated for by allocating more descriptors using the TxDescriptors
 269       module parameter.
 270
 271       There are other conditions which may result in "TX Timeout", which will
 272       not be resolved by the use of the TxDescriptorStep parameter. As the
 273       issue addressed by this parameter has never been observed on Intel
 274       Architecture platforms, it should not be used on Intel platforms.
 275
 276TxIntDelay
 277----------
 278Valid Range:   0-65535 (0=off)
 279Default Value: 64
 280
 281This value delays the generation of transmit interrupts in units of
 2821.024 microseconds.  Transmit interrupt reduction can improve CPU
 283efficiency if properly tuned for specific network traffic.  If the
 284system is reporting dropped transmits, this value may be set too high
 285causing the driver to run out of available transmit descriptors.
 286
 287TxAbsIntDelay
 288-------------
 289(This parameter is supported only on 82540, 82545 and later adapters.)
 290Valid Range:   0-65535 (0=off)
 291Default Value: 64
 292
 293This value, in units of 1.024 microseconds, limits the delay in which a
 294transmit interrupt is generated.  Useful only if TxIntDelay is non-zero,
 295this value ensures that an interrupt is generated after the initial
 296packet is sent on the wire within the set amount of time.  Proper tuning,
 297along with TxIntDelay, may improve traffic throughput in specific
 298network conditions.
 299
 300XsumRX
 301------
 302(This parameter is NOT supported on the 82542-based adapter.)
 303Valid Range:   0-1
 304Default Value: 1
 305
 306A value of '1' indicates that the driver should enable IP checksum
 307offload for received packets (both UDP and TCP) to the adapter hardware.
 308
 309Copybreak
 310---------
 311Valid Range:   0-xxxxxxx (0=off)
 312Default Value: 256
 313Usage: insmod e1000.ko copybreak=128
 314
 315Driver copies all packets below or equaling this size to a fresh RX
 316buffer before handing it up the stack.
 317
 318This parameter is different than other parameters, in that it is a
 319single (not 1,1,1 etc.) parameter applied to all driver instances and
 320it is also available during runtime at
 321/sys/module/e1000/parameters/copybreak
 322
 323SmartPowerDownEnable
 324--------------------
 325Valid Range: 0-1
 326Default Value:  0 (disabled)
 327
 328Allows PHY to turn off in lower power states. The user can turn off
 329this parameter in supported chipsets.
 330
 331KumeranLockLoss
 332---------------
 333Valid Range: 0-1
 334Default Value: 1 (enabled)
 335
 336This workaround skips resetting the PHY at shutdown for the initial
 337silicon releases of ICH8 systems.
 338
 339Speed and Duplex Configuration
 340==============================
 341
 342Three keywords are used to control the speed and duplex configuration.
 343These keywords are Speed, Duplex, and AutoNeg.
 344
 345If the board uses a fiber interface, these keywords are ignored, and the
 346fiber interface board only links at 1000 Mbps full-duplex.
 347
 348For copper-based boards, the keywords interact as follows:
 349
 350  The default operation is auto-negotiate.  The board advertises all
 351  supported speed and duplex combinations, and it links at the highest
 352  common speed and duplex mode IF the link partner is set to auto-negotiate.
 353
 354  If Speed = 1000, limited auto-negotiation is enabled and only 1000 Mbps
 355  is advertised (The 1000BaseT spec requires auto-negotiation.)
 356
 357  If Speed = 10 or 100, then both Speed and Duplex should be set.  Auto-
 358  negotiation is disabled, and the AutoNeg parameter is ignored.  Partner
 359  SHOULD also be forced.
 360
 361The AutoNeg parameter is used when more control is required over the
 362auto-negotiation process.  It should be used when you wish to control which
 363speed and duplex combinations are advertised during the auto-negotiation
 364process.
 365
 366The parameter may be specified as either a decimal or hexadecimal value as
 367determined by the bitmap below.
 368
 369Bit position   7      6      5       4       3      2      1       0
 370Decimal Value  128    64     32      16      8      4      2       1
 371Hex value      80     40     20      10      8      4      2       1
 372Speed (Mbps)   N/A    N/A    1000    N/A     100    100    10      10
 373Duplex                       Full            Full   Half   Full    Half
 374
 375Some examples of using AutoNeg:
 376
 377  modprobe e1000 AutoNeg=0x01 (Restricts autonegotiation to 10 Half)
 378  modprobe e1000 AutoNeg=1 (Same as above)
 379  modprobe e1000 AutoNeg=0x02 (Restricts autonegotiation to 10 Full)
 380  modprobe e1000 AutoNeg=0x03 (Restricts autonegotiation to 10 Half or 10 Full)
 381  modprobe e1000 AutoNeg=0x04 (Restricts autonegotiation to 100 Half)
 382  modprobe e1000 AutoNeg=0x05 (Restricts autonegotiation to 10 Half or 100
 383  Half)
 384  modprobe e1000 AutoNeg=0x020 (Restricts autonegotiation to 1000 Full)
 385  modprobe e1000 AutoNeg=32 (Same as above)
 386
 387Note that when this parameter is used, Speed and Duplex must not be specified.
 388
 389If the link partner is forced to a specific speed and duplex, then this
 390parameter should not be used.  Instead, use the Speed and Duplex parameters
 391previously mentioned to force the adapter to the same speed and duplex.
 392
 393Additional Configurations
 394=========================
 395
 396  Jumbo Frames
 397  ------------
 398  Jumbo Frames support is enabled by changing the MTU to a value larger than
 399  the default of 1500.  Use the ifconfig command to increase the MTU size.
 400  For example:
 401
 402       ifconfig eth<x> mtu 9000 up
 403
 404  This setting is not saved across reboots.  It can be made permanent if
 405  you add:
 406
 407       MTU=9000
 408
 409   to the file /etc/sysconfig/network-scripts/ifcfg-eth<x>.  This example
 410   applies to the Red Hat distributions; other distributions may store this
 411   setting in a different location.
 412
 413  Notes:
 414  Degradation in throughput performance may be observed in some Jumbo frames
 415  environments. If this is observed, increasing the application's socket buffer
 416  size and/or increasing the /proc/sys/net/ipv4/tcp_*mem entry values may help.
 417  See the specific application manual and /usr/src/linux*/Documentation/
 418  networking/ip-sysctl.txt for more details.
 419
 420  - The maximum MTU setting for Jumbo Frames is 16110.  This value coincides
 421    with the maximum Jumbo Frames size of 16128.
 422
 423  - Using Jumbo Frames at 10 or 100 Mbps may result in poor performance or
 424    loss of link.
 425
 426  - Adapters based on the Intel(R) 82542 and 82573V/E controller do not
 427    support Jumbo Frames. These correspond to the following product names:
 428     Intel(R) PRO/1000 Gigabit Server Adapter
 429     Intel(R) PRO/1000 PM Network Connection
 430
 431  Ethtool
 432  -------
 433  The driver utilizes the ethtool interface for driver configuration and
 434  diagnostics, as well as displaying statistical information.  The ethtool
 435  version 1.6 or later is required for this functionality.
 436
 437  The latest release of ethtool can be found from
 438  http://ftp.kernel.org/pub/software/network/ethtool/
 439
 440  Enabling Wake on LAN* (WoL)
 441  ---------------------------
 442  WoL is configured through the ethtool* utility.
 443
 444  WoL will be enabled on the system during the next shut down or reboot.
 445  For this driver version, in order to enable WoL, the e1000 driver must be
 446  loaded when shutting down or rebooting the system.
 447
 448Support
 449=======
 450
 451For general information, go to the Intel support website at:
 452
 453    http://support.intel.com
 454
 455or the Intel Wired Networking project hosted by Sourceforge at:
 456
 457    http://sourceforge.net/projects/e1000
 458
 459If an issue is identified with the released source code on the supported
 460kernel with a supported adapter, email the specific information related
 461to the issue to e1000-devel@lists.sf.net
 462
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.