| 1 |
.. SPDX-License-Identifier: GPL-2.0+ |
| 2 |
|
| 3 |
===================================================== |
| 4 |
Linux Driver for Intel(R) Ethernet Network Connection |
| 5 |
===================================================== |
| 6 |
|
| 7 |
Intel Gigabit Linux driver. |
| 8 |
Copyright(c) 2008-2018 Intel Corporation. |
| 9 |
|
| 10 |
Contents |
| 11 |
======== |
| 12 |
|
| 13 |
- Identifying Your Adapter |
| 14 |
- Command Line Parameters |
| 15 |
- Additional Configurations |
| 16 |
- Support |
| 17 |
|
| 18 |
|
| 19 |
Identifying Your Adapter |
| 20 |
======================== |
| 21 |
For information on how to identify your adapter, and for the latest Intel |
| 22 |
network drivers, refer to the Intel Support website: |
| 23 |
https://www.intel.com/support |
| 24 |
|
| 25 |
|
| 26 |
Command Line Parameters |
| 27 |
======================= |
| 28 |
If the driver is built as a module, the following optional parameters are used |
| 29 |
by entering them on the command line with the modprobe command using this |
| 30 |
syntax:: |
| 31 |
|
| 32 |
modprobe e1000e [<option>=<VAL1>,<VAL2>,...] |
| 33 |
|
| 34 |
There needs to be a <VAL#> for each network port in the system supported by |
| 35 |
this driver. The values will be applied to each instance, in function order. |
| 36 |
For example:: |
| 37 |
|
| 38 |
modprobe e1000e InterruptThrottleRate=16000,16000 |
| 39 |
|
| 40 |
In this case, there are two network ports supported by e1000e in the system. |
| 41 |
The default value for each parameter is generally the recommended setting, |
| 42 |
unless otherwise noted. |
| 43 |
|
| 44 |
NOTE: A descriptor describes a data buffer and attributes related to the data |
| 45 |
buffer. This information is accessed by the hardware. |
| 46 |
|
| 47 |
InterruptThrottleRate |
| 48 |
--------------------- |
| 49 |
:Valid Range: 0,1,3,4,100-100000 |
| 50 |
:Default Value: 3 |
| 51 |
|
| 52 |
Interrupt Throttle Rate controls the number of interrupts each interrupt |
| 53 |
vector can generate per second. Increasing ITR lowers latency at the cost of |
| 54 |
increased CPU utilization, though it may help throughput in some circumstances. |
| 55 |
|
| 56 |
Setting InterruptThrottleRate to a value greater or equal to 100 |
| 57 |
will program the adapter to send out a maximum of that many interrupts |
| 58 |
per second, even if more packets have come in. This reduces interrupt |
| 59 |
load on the system and can lower CPU utilization under heavy load, |
| 60 |
but will increase latency as packets are not processed as quickly. |
| 61 |
|
| 62 |
The default behaviour of the driver previously assumed a static |
| 63 |
InterruptThrottleRate value of 8000, providing a good fallback value for |
| 64 |
all traffic types, but lacking in small packet performance and latency. |
| 65 |
The hardware can handle many more small packets per second however, and |
| 66 |
for this reason an adaptive interrupt moderation algorithm was implemented. |
| 67 |
|
| 68 |
The driver has two adaptive modes (setting 1 or 3) in which |
| 69 |
it dynamically adjusts the InterruptThrottleRate value based on the traffic |
| 70 |
that it receives. After determining the type of incoming traffic in the last |
| 71 |
timeframe, it will adjust the InterruptThrottleRate to an appropriate value |
| 72 |
for that traffic. |
| 73 |
|
| 74 |
The algorithm classifies the incoming traffic every interval into |
| 75 |
classes. Once the class is determined, the InterruptThrottleRate value is |
| 76 |
adjusted to suit that traffic type the best. There are three classes defined: |
| 77 |
"Bulk traffic", for large amounts of packets of normal size; "Low latency", |
| 78 |
for small amounts of traffic and/or a significant percentage of small |
| 79 |
packets; and "Lowest latency", for almost completely small packets or |
| 80 |
minimal traffic. |
| 81 |
|
| 82 |
- 0: Off |
| 83 |
Turns off any interrupt moderation and may improve small packet latency. |
| 84 |
However, this is generally not suitable for bulk throughput traffic due |
| 85 |
to the increased CPU utilization of the higher interrupt rate. |
| 86 |
- 1: Dynamic mode |
| 87 |
This mode attempts to moderate interrupts per vector while maintaining |
| 88 |
very low latency. This can sometimes cause extra CPU utilization. If |
| 89 |
planning on deploying e1000e in a latency sensitive environment, this |
| 90 |
parameter should be considered. |
| 91 |
- 3: Dynamic Conservative mode (default) |
| 92 |
In dynamic conservative mode, the InterruptThrottleRate value is set to |
| 93 |
4000 for traffic that falls in class "Bulk traffic". If traffic falls in |
| 94 |
the "Low latency" or "Lowest latency" class, the InterruptThrottleRate is |
| 95 |
increased stepwise to 20000. This default mode is suitable for most |
| 96 |
applications. |
| 97 |
- 4: Simplified Balancing mode |
| 98 |
In simplified mode the interrupt rate is based on the ratio of TX and |
| 99 |
RX traffic. If the bytes per second rate is approximately equal, the |
| 100 |
interrupt rate will drop as low as 2000 interrupts per second. If the |
| 101 |
traffic is mostly transmit or mostly receive, the interrupt rate could |
| 102 |
be as high as 8000. |
| 103 |
- 100-100000: |
| 104 |
Setting InterruptThrottleRate to a value greater or equal to 100 |
| 105 |
will program the adapter to send at most that many interrupts per second, |
| 106 |
even if more packets have come in. This reduces interrupt load on the |
| 107 |
system and can lower CPU utilization under heavy load, but will increase |
| 108 |
latency as packets are not processed as quickly. |
| 109 |
|
| 110 |
NOTE: InterruptThrottleRate takes precedence over the TxAbsIntDelay and |
| 111 |
RxAbsIntDelay parameters. In other words, minimizing the receive and/or |
| 112 |
transmit absolute delays does not force the controller to generate more |
| 113 |
interrupts than what the Interrupt Throttle Rate allows. |
| 114 |
|
| 115 |
RxIntDelay |
| 116 |
---------- |
| 117 |
:Valid Range: 0-65535 (0=off) |
| 118 |
:Default Value: 0 |
| 119 |
|
| 120 |
This value delays the generation of receive interrupts in units of 1.024 |
| 121 |
microseconds. Receive interrupt reduction can improve CPU efficiency if |
| 122 |
properly tuned for specific network traffic. Increasing this value adds extra |
| 123 |
latency to frame reception and can end up decreasing the throughput of TCP |
| 124 |
traffic. If the system is reporting dropped receives, this value may be set |
| 125 |
too high, causing the driver to run out of available receive descriptors. |
| 126 |
|
| 127 |
CAUTION: When setting RxIntDelay to a value other than 0, adapters may hang |
| 128 |
(stop transmitting) under certain network conditions. If this occurs a NETDEV |
| 129 |
WATCHDOG message is logged in the system event log. In addition, the |
| 130 |
controller is automatically reset, restoring the network connection. To |
| 131 |
eliminate the potential for the hang ensure that RxIntDelay is set to 0. |
| 132 |
|
| 133 |
RxAbsIntDelay |
| 134 |
------------- |
| 135 |
:Valid Range: 0-65535 (0=off) |
| 136 |
:Default Value: 8 |
| 137 |
|
| 138 |
This value, in units of 1.024 microseconds, limits the delay in which a |
| 139 |
receive interrupt is generated. This value ensures that an interrupt is |
| 140 |
generated after the initial packet is received within the set amount of time, |
| 141 |
which is useful only if RxIntDelay is non-zero. Proper tuning, along with |
| 142 |
RxIntDelay, may improve traffic throughput in specific network conditions. |
| 143 |
|
| 144 |
TxIntDelay |
| 145 |
---------- |
| 146 |
:Valid Range: 0-65535 (0=off) |
| 147 |
:Default Value: 8 |
| 148 |
|
| 149 |
This value delays the generation of transmit interrupts in units of 1.024 |
| 150 |
microseconds. Transmit interrupt reduction can improve CPU efficiency if |
| 151 |
properly tuned for specific network traffic. If the system is reporting |
| 152 |
dropped transmits, this value may be set too high causing the driver to run |
| 153 |
out of available transmit descriptors. |
| 154 |
|
| 155 |
TxAbsIntDelay |
| 156 |
------------- |
| 157 |
:Valid Range: 0-65535 (0=off) |
| 158 |
:Default Value: 32 |
| 159 |
|
| 160 |
This value, in units of 1.024 microseconds, limits the delay in which a |
| 161 |
transmit interrupt is generated. It is useful only if TxIntDelay is non-zero. |
| 162 |
It ensures that an interrupt is generated after the initial Packet is sent on |
| 163 |
the wire within the set amount of time. Proper tuning, along with TxIntDelay, |
| 164 |
may improve traffic throughput in specific network conditions. |
| 165 |
|
| 166 |
copybreak |
| 167 |
--------- |
| 168 |
:Valid Range: 0-xxxxxxx (0=off) |
| 169 |
:Default Value: 256 |
| 170 |
|
| 171 |
The driver copies all packets below or equaling this size to a fresh receive |
| 172 |
buffer before handing it up the stack. |
| 173 |
This parameter differs from other parameters because it is a single (not 1,1,1 |
| 174 |
etc.) parameter applied to all driver instances and it is also available |
| 175 |
during runtime at /sys/module/e1000e/parameters/copybreak. |
| 176 |
|
| 177 |
To use copybreak, type:: |
| 178 |
|
| 179 |
modprobe e1000e.ko copybreak=128 |
| 180 |
|
| 181 |
SmartPowerDownEnable |
| 182 |
-------------------- |
| 183 |
:Valid Range: 0,1 |
| 184 |
:Default Value: 0 (disabled) |
| 185 |
|
| 186 |
Allows the PHY to turn off in lower power states. The user can turn off this |
| 187 |
parameter in supported chipsets. |
| 188 |
|
| 189 |
KumeranLockLoss |
| 190 |
--------------- |
| 191 |
:Valid Range: 0,1 |
| 192 |
:Default Value: 1 (enabled) |
| 193 |
|
| 194 |
This workaround skips resetting the PHY at shutdown for the initial silicon |
| 195 |
releases of ICH8 systems. |
| 196 |
|
| 197 |
IntMode |
| 198 |
------- |
| 199 |
:Valid Range: 0-2 |
| 200 |
:Default Value: 0 |
| 201 |
|
| 202 |
+-------+----------------+ |
| 203 |
| Value | Interrupt Mode | |
| 204 |
+=======+================+ |
| 205 |
| 0 | Legacy | |
| 206 |
+-------+----------------+ |
| 207 |
| 1 | MSI | |
| 208 |
+-------+----------------+ |
| 209 |
| 2 | MSI-X | |
| 210 |
+-------+----------------+ |
| 211 |
|
| 212 |
IntMode allows load time control over the type of interrupt registered for by |
| 213 |
the driver. MSI-X is required for multiple queue support, and some kernels and |
| 214 |
combinations of kernel .config options will force a lower level of interrupt |
| 215 |
support. |
| 216 |
|
| 217 |
This command will show different values for each type of interrupt:: |
| 218 |
|
| 219 |
cat /proc/interrupts |
| 220 |
|
| 221 |
CrcStripping |
| 222 |
------------ |
| 223 |
:Valid Range: 0,1 |
| 224 |
:Default Value: 1 (enabled) |
| 225 |
|
| 226 |
Strip the CRC from received packets before sending up the network stack. If |
| 227 |
you have a machine with a BMC enabled but cannot receive IPMI traffic after |
| 228 |
loading or enabling the driver, try disabling this feature. |
| 229 |
|
| 230 |
WriteProtectNVM |
| 231 |
--------------- |
| 232 |
:Valid Range: 0,1 |
| 233 |
:Default Value: 1 (enabled) |
| 234 |
|
| 235 |
If set to 1, configure the hardware to ignore all write/erase cycles to the |
| 236 |
GbE region in the ICHx NVM (in order to prevent accidental corruption of the |
| 237 |
NVM). This feature can be disabled by setting the parameter to 0 during initial |
| 238 |
driver load. |
| 239 |
|
| 240 |
NOTE: The machine must be power cycled (full off/on) when enabling NVM writes |
| 241 |
via setting the parameter to zero. Once the NVM has been locked (via the |
| 242 |
parameter at 1 when the driver loads) it cannot be unlocked except via power |
| 243 |
cycle. |
| 244 |
|
| 245 |
Debug |
| 246 |
----- |
| 247 |
:Valid Range: 0-16 (0=none,...,16=all) |
| 248 |
:Default Value: 0 |
| 249 |
|
| 250 |
This parameter adjusts the level of debug messages displayed in the system logs. |
| 251 |
|
| 252 |
|
| 253 |
Additional Features and Configurations |
| 254 |
====================================== |
| 255 |
|
| 256 |
Jumbo Frames |
| 257 |
------------ |
| 258 |
Jumbo Frames support is enabled by changing the Maximum Transmission Unit (MTU) |
| 259 |
to a value larger than the default value of 1500. |
| 260 |
|
| 261 |
Use the ifconfig command to increase the MTU size. For example, enter the |
| 262 |
following where <x> is the interface number:: |
| 263 |
|
| 264 |
ifconfig eth<x> mtu 9000 up |
| 265 |
|
| 266 |
Alternatively, you can use the ip command as follows:: |
| 267 |
|
| 268 |
ip link set mtu 9000 dev eth<x> |
| 269 |
ip link set up dev eth<x> |
| 270 |
|
| 271 |
This setting is not saved across reboots. The setting change can be made |
| 272 |
permanent by adding 'MTU=9000' to the file: |
| 273 |
|
| 274 |
- For RHEL: /etc/sysconfig/network-scripts/ifcfg-eth<x> |
| 275 |
- For SLES: /etc/sysconfig/network/<config_file> |
| 276 |
|
| 277 |
NOTE: The maximum MTU setting for Jumbo Frames is 8996. This value coincides |
| 278 |
with the maximum Jumbo Frames size of 9018 bytes. |
| 279 |
|
| 280 |
NOTE: Using Jumbo frames at 10 or 100 Mbps is not supported and may result in |
| 281 |
poor performance or loss of link. |
| 282 |
|
| 283 |
NOTE: The following adapters limit Jumbo Frames sized packets to a maximum of |
| 284 |
4088 bytes: |
| 285 |
|
| 286 |
- Intel(R) 82578DM Gigabit Network Connection |
| 287 |
- Intel(R) 82577LM Gigabit Network Connection |
| 288 |
|
| 289 |
The following adapters do not support Jumbo Frames: |
| 290 |
|
| 291 |
- Intel(R) PRO/1000 Gigabit Server Adapter |
| 292 |
- Intel(R) PRO/1000 PM Network Connection |
| 293 |
- Intel(R) 82562G 10/100 Network Connection |
| 294 |
- Intel(R) 82562G-2 10/100 Network Connection |
| 295 |
- Intel(R) 82562GT 10/100 Network Connection |
| 296 |
- Intel(R) 82562GT-2 10/100 Network Connection |
| 297 |
- Intel(R) 82562V 10/100 Network Connection |
| 298 |
- Intel(R) 82562V-2 10/100 Network Connection |
| 299 |
- Intel(R) 82566DC Gigabit Network Connection |
| 300 |
- Intel(R) 82566DC-2 Gigabit Network Connection |
| 301 |
- Intel(R) 82566DM Gigabit Network Connection |
| 302 |
- Intel(R) 82566MC Gigabit Network Connection |
| 303 |
- Intel(R) 82566MM Gigabit Network Connection |
| 304 |
- Intel(R) 82567V-3 Gigabit Network Connection |
| 305 |
- Intel(R) 82577LC Gigabit Network Connection |
| 306 |
- Intel(R) 82578DC Gigabit Network Connection |
| 307 |
|
| 308 |
NOTE: Jumbo Frames cannot be configured on an 82579-based Network device if |
| 309 |
MACSec is enabled on the system. |
| 310 |
|
| 311 |
|
| 312 |
ethtool |
| 313 |
------- |
| 314 |
The driver utilizes the ethtool interface for driver configuration and |
| 315 |
diagnostics, as well as displaying statistical information. The latest ethtool |
| 316 |
version is required for this functionality. Download it at: |
| 317 |
|
| 318 |
https://www.kernel.org/pub/software/network/ethtool/ |
| 319 |
|
| 320 |
NOTE: When validating enable/disable tests on some parts (for example, 82578), |
| 321 |
it is necessary to add a few seconds between tests when working with ethtool. |
| 322 |
|
| 323 |
|
| 324 |
Speed and Duplex Configuration |
| 325 |
------------------------------ |
| 326 |
In addressing speed and duplex configuration issues, you need to distinguish |
| 327 |
between copper-based adapters and fiber-based adapters. |
| 328 |
|
| 329 |
In the default mode, an Intel(R) Ethernet Network Adapter using copper |
| 330 |
connections will attempt to auto-negotiate with its link partner to determine |
| 331 |
the best setting. If the adapter cannot establish link with the link partner |
| 332 |
using auto-negotiation, you may need to manually configure the adapter and link |
| 333 |
partner to identical settings to establish link and pass packets. This should |
| 334 |
only be needed when attempting to link with an older switch that does not |
| 335 |
support auto-negotiation or one that has been forced to a specific speed or |
| 336 |
duplex mode. Your link partner must match the setting you choose. 1 Gbps speeds |
| 337 |
and higher cannot be forced. Use the autonegotiation advertising setting to |
| 338 |
manually set devices for 1 Gbps and higher. |
| 339 |
|
| 340 |
Speed, duplex, and autonegotiation advertising are configured through the |
| 341 |
ethtool utility. |
| 342 |
|
| 343 |
Caution: Only experienced network administrators should force speed and duplex |
| 344 |
or change autonegotiation advertising manually. The settings at the switch must |
| 345 |
always match the adapter settings. Adapter performance may suffer or your |
| 346 |
adapter may not operate if you configure the adapter differently from your |
| 347 |
switch. |
| 348 |
|
| 349 |
An Intel(R) Ethernet Network Adapter using fiber-based connections, however, |
| 350 |
will not attempt to auto-negotiate with its link partner since those adapters |
| 351 |
operate only in full duplex and only at their native speed. |
| 352 |
|
| 353 |
|
| 354 |
Enabling Wake on LAN (WoL) |
| 355 |
-------------------------- |
| 356 |
WoL is configured through the ethtool utility. |
| 357 |
|
| 358 |
WoL will be enabled on the system during the next shut down or reboot. For |
| 359 |
this driver version, in order to enable WoL, the e1000e driver must be loaded |
| 360 |
prior to shutting down or suspending the system. |
| 361 |
|
| 362 |
NOTE: Wake on LAN is only supported on port A for the following devices: |
| 363 |
- Intel(R) PRO/1000 PT Dual Port Network Connection |
| 364 |
- Intel(R) PRO/1000 PT Dual Port Server Connection |
| 365 |
- Intel(R) PRO/1000 PT Dual Port Server Adapter |
| 366 |
- Intel(R) PRO/1000 PF Dual Port Server Adapter |
| 367 |
- Intel(R) PRO/1000 PT Quad Port Server Adapter |
| 368 |
- Intel(R) Gigabit PT Quad Port Server ExpressModule |
| 369 |
|
| 370 |
|
| 371 |
Support |
| 372 |
======= |
| 373 |
For general information, go to the Intel support website at: |
| 374 |
https://www.intel.com/support/ |
| 375 |
|
| 376 |
If an issue is identified with the released source code on a supported kernel |
| 377 |
with a supported adapter, email the specific information related to the issue |
| 378 |
to [email protected]. |