Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1.. SPDX-License-Identifier: GPL-2.0+
2
3===========================================================
4Linux* Base Driver for Intel(R) Ethernet Network Connection
5===========================================================
6
7Intel Gigabit Linux driver.
8Copyright(c) 1999 - 2013 Intel Corporation.
9
10Contents
11========
12
13- Identifying Your Adapter
14- Command Line Parameters
15- Speed and Duplex Configuration
16- Additional Configurations
17- Support
18
19Identifying Your Adapter
20========================
21
22For more information on how to identify your adapter, go to the Adapter &
23Driver ID Guide at:
24
25 http://support.intel.com/support/go/network/adapter/idguide.htm
26
27For the latest Intel network drivers for Linux, refer to the following
28website. In the search field, enter your adapter name or type, or use the
29networking link on the left to search for your adapter:
30
31 http://support.intel.com/support/go/network/adapter/home.htm
32
33Command Line Parameters
34=======================
35
36The default value for each parameter is generally the recommended setting,
37unless otherwise noted.
38
39NOTES:
40 For more information about the AutoNeg, Duplex, and Speed
41 parameters, see the "Speed and Duplex Configuration" section in
42 this document.
43
44 For more information about the InterruptThrottleRate,
45 RxIntDelay, TxIntDelay, RxAbsIntDelay, and TxAbsIntDelay
46 parameters, see the application note at:
47 http://www.intel.com/design/network/applnots/ap450.htm
48
49AutoNeg
50-------
51
52(Supported only on adapters with copper connections)
53
54:Valid Range: 0x01-0x0F, 0x20-0x2F
55:Default Value: 0x2F
56
57This parameter is a bit-mask that specifies the speed and duplex settings
58advertised by the adapter. When this parameter is used, the Speed and
59Duplex parameters must not be specified.
60
61NOTE:
62 Refer to the Speed and Duplex section of this readme for more
63 information on the AutoNeg parameter.
64
65Duplex
66------
67
68(Supported only on adapters with copper connections)
69
70:Valid Range: 0-2 (0=auto-negotiate, 1=half, 2=full)
71:Default Value: 0
72
73This defines the direction in which data is allowed to flow. Can be
74either one or two-directional. If both Duplex and the link partner are
75set to auto-negotiate, the board auto-detects the correct duplex. If the
76link partner is forced (either full or half), Duplex defaults to half-
77duplex.
78
79FlowControl
80-----------
81
82:Valid Range: 0-3 (0=none, 1=Rx only, 2=Tx only, 3=Rx&Tx)
83:Default Value: Reads flow control settings from the EEPROM
84
85This parameter controls the automatic generation(Tx) and response(Rx)
86to Ethernet PAUSE frames.
87
88InterruptThrottleRate
89---------------------
90
91(not supported on Intel(R) 82542, 82543 or 82544-based adapters)
92
93:Valid Range:
94 0,1,3,4,100-100000 (0=off, 1=dynamic, 3=dynamic conservative,
95 4=simplified balancing)
96:Default Value: 3
97
98The driver can limit the amount of interrupts per second that the adapter
99will generate for incoming packets. It does this by writing a value to the
100adapter that is based on the maximum amount of interrupts that the adapter
101will generate per second.
102
103Setting InterruptThrottleRate to a value greater or equal to 100
104will program the adapter to send out a maximum of that many interrupts
105per second, even if more packets have come in. This reduces interrupt
106load on the system and can lower CPU utilization under heavy load,
107but will increase latency as packets are not processed as quickly.
108
109The default behaviour of the driver previously assumed a static
110InterruptThrottleRate value of 8000, providing a good fallback value for
111all traffic types,but lacking in small packet performance and latency.
112The hardware can handle many more small packets per second however, and
113for this reason an adaptive interrupt moderation algorithm was implemented.
114
115Since 7.3.x, the driver has two adaptive modes (setting 1 or 3) in which
116it dynamically adjusts the InterruptThrottleRate value based on the traffic
117that it receives. After determining the type of incoming traffic in the last
118timeframe, it will adjust the InterruptThrottleRate to an appropriate value
119for that traffic.
120
121The algorithm classifies the incoming traffic every interval into
122classes. Once the class is determined, the InterruptThrottleRate value is
123adjusted to suit that traffic type the best. There are three classes defined:
124"Bulk traffic", for large amounts of packets of normal size; "Low latency",
125for small amounts of traffic and/or a significant percentage of small
126packets; and "Lowest latency", for almost completely small packets or
127minimal traffic.
128
129In dynamic conservative mode, the InterruptThrottleRate value is set to 4000
130for traffic that falls in class "Bulk traffic". If traffic falls in the "Low
131latency" or "Lowest latency" class, the InterruptThrottleRate is increased
132stepwise to 20000. This default mode is suitable for most applications.
133
134For situations where low latency is vital such as cluster or
135grid computing, the algorithm can reduce latency even more when
136InterruptThrottleRate is set to mode 1. In this mode, which operates
137the same as mode 3, the InterruptThrottleRate will be increased stepwise to
13870000 for traffic in class "Lowest latency".
139
140In simplified mode the interrupt rate is based on the ratio of TX and
141RX traffic. If the bytes per second rate is approximately equal, the
142interrupt rate will drop as low as 2000 interrupts per second. If the
143traffic is mostly transmit or mostly receive, the interrupt rate could
144be as high as 8000.
145
146Setting InterruptThrottleRate to 0 turns off any interrupt moderation
147and may improve small packet latency, but is generally not suitable
148for bulk throughput traffic.
149
150NOTE:
151 InterruptThrottleRate takes precedence over the TxAbsIntDelay and
152 RxAbsIntDelay parameters. In other words, minimizing the receive
153 and/or transmit absolute delays does not force the controller to
154 generate more interrupts than what the Interrupt Throttle Rate
155 allows.
156
157CAUTION:
158 If you are using the Intel(R) PRO/1000 CT Network Connection
159 (controller 82547), setting InterruptThrottleRate to a value
160 greater than 75,000, may hang (stop transmitting) adapters
161 under certain network conditions. If this occurs a NETDEV
162 WATCHDOG message is logged in the system event log. In
163 addition, the controller is automatically reset, restoring
164 the network connection. To eliminate the potential for the
165 hang, ensure that InterruptThrottleRate is set no greater
166 than 75,000 and is not set to 0.
167
168NOTE:
169 When e1000 is loaded with default settings and multiple adapters
170 are in use simultaneously, the CPU utilization may increase non-
171 linearly. In order to limit the CPU utilization without impacting
172 the overall throughput, we recommend that you load the driver as
173 follows::
174
175 modprobe e1000 InterruptThrottleRate=3000,3000,3000
176
177 This sets the InterruptThrottleRate to 3000 interrupts/sec for
178 the first, second, and third instances of the driver. The range
179 of 2000 to 3000 interrupts per second works on a majority of
180 systems and is a good starting point, but the optimal value will
181 be platform-specific. If CPU utilization is not a concern, use
182 RX_POLLING (NAPI) and default driver settings.
183
184RxDescriptors
185-------------
186
187:Valid Range:
188 - 48-256 for 82542 and 82543-based adapters
189 - 48-4096 for all other supported adapters
190:Default Value: 256
191
192This value specifies the number of receive buffer descriptors allocated
193by the driver. Increasing this value allows the driver to buffer more
194incoming packets, at the expense of increased system memory utilization.
195
196Each descriptor is 16 bytes. A receive buffer is also allocated for each
197descriptor and can be either 2048, 4096, 8192, or 16384 bytes, depending
198on the MTU setting. The maximum MTU size is 16110.
199
200NOTE:
201 MTU designates the frame size. It only needs to be set for Jumbo
202 Frames. Depending on the available system resources, the request
203 for a higher number of receive descriptors may be denied. In this
204 case, use a lower number.
205
206RxIntDelay
207----------
208
209:Valid Range: 0-65535 (0=off)
210:Default Value: 0
211
212This value delays the generation of receive interrupts in units of 1.024
213microseconds. Receive interrupt reduction can improve CPU efficiency if
214properly tuned for specific network traffic. Increasing this value adds
215extra latency to frame reception and can end up decreasing the throughput
216of TCP traffic. If the system is reporting dropped receives, this value
217may be set too high, causing the driver to run out of available receive
218descriptors.
219
220CAUTION:
221 When setting RxIntDelay to a value other than 0, adapters may
222 hang (stop transmitting) under certain network conditions. If
223 this occurs a NETDEV WATCHDOG message is logged in the system
224 event log. In addition, the controller is automatically reset,
225 restoring the network connection. To eliminate the potential
226 for the hang ensure that RxIntDelay is set to 0.
227
228RxAbsIntDelay
229-------------
230
231(This parameter is supported only on 82540, 82545 and later adapters.)
232
233:Valid Range: 0-65535 (0=off)
234:Default Value: 128
235
236This value, in units of 1.024 microseconds, limits the delay in which a
237receive interrupt is generated. Useful only if RxIntDelay is non-zero,
238this value ensures that an interrupt is generated after the initial
239packet is received within the set amount of time. Proper tuning,
240along with RxIntDelay, may improve traffic throughput in specific network
241conditions.
242
243Speed
244-----
245
246(This parameter is supported only on adapters with copper connections.)
247
248:Valid Settings: 0, 10, 100, 1000
249:Default Value: 0 (auto-negotiate at all supported speeds)
250
251Speed forces the line speed to the specified value in megabits per second
252(Mbps). If this parameter is not specified or is set to 0 and the link
253partner is set to auto-negotiate, the board will auto-detect the correct
254speed. Duplex should also be set when Speed is set to either 10 or 100.
255
256TxDescriptors
257-------------
258
259:Valid Range:
260 - 48-256 for 82542 and 82543-based adapters
261 - 48-4096 for all other supported adapters
262:Default Value: 256
263
264This value is the number of transmit descriptors allocated by the driver.
265Increasing this value allows the driver to queue more transmits. Each
266descriptor is 16 bytes.
267
268NOTE:
269 Depending on the available system resources, the request for a
270 higher number of transmit descriptors may be denied. In this case,
271 use a lower number.
272
273TxIntDelay
274----------
275
276:Valid Range: 0-65535 (0=off)
277:Default Value: 8
278
279This value delays the generation of transmit interrupts in units of
2801.024 microseconds. Transmit interrupt reduction can improve CPU
281efficiency if properly tuned for specific network traffic. If the
282system is reporting dropped transmits, this value may be set too high
283causing the driver to run out of available transmit descriptors.
284
285TxAbsIntDelay
286-------------
287
288(This parameter is supported only on 82540, 82545 and later adapters.)
289
290:Valid Range: 0-65535 (0=off)
291:Default Value: 32
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
303(This parameter is NOT supported on the 82542-based adapter.)
304
305:Valid Range: 0-1
306:Default Value: 1
307
308A value of '1' indicates that the driver should enable IP checksum
309offload for received packets (both UDP and TCP) to the adapter hardware.
310
311Copybreak
312---------
313
314:Valid Range: 0-xxxxxxx (0=off)
315:Default Value: 256
316:Usage: modprobe e1000.ko copybreak=128
317
318Driver copies all packets below or equaling this size to a fresh RX
319buffer before handing it up the stack.
320
321This parameter is different than other parameters, in that it is a
322single (not 1,1,1 etc.) parameter applied to all driver instances and
323it is also available during runtime at
324/sys/module/e1000/parameters/copybreak
325
326SmartPowerDownEnable
327--------------------
328
329:Valid Range: 0-1
330:Default Value: 0 (disabled)
331
332Allows PHY to turn off in lower power states. The user can turn off
333this parameter in supported chipsets.
334
335Speed and Duplex Configuration
336==============================
337
338Three keywords are used to control the speed and duplex configuration.
339These keywords are Speed, Duplex, and AutoNeg.
340
341If the board uses a fiber interface, these keywords are ignored, and the
342fiber interface board only links at 1000 Mbps full-duplex.
343
344For copper-based boards, the keywords interact as follows:
345
346- The default operation is auto-negotiate. The board advertises all
347 supported speed and duplex combinations, and it links at the highest
348 common speed and duplex mode IF the link partner is set to auto-negotiate.
349
350- If Speed = 1000, limited auto-negotiation is enabled and only 1000 Mbps
351 is advertised (The 1000BaseT spec requires auto-negotiation.)
352
353- If Speed = 10 or 100, then both Speed and Duplex should be set. Auto-
354 negotiation is disabled, and the AutoNeg parameter is ignored. Partner
355 SHOULD also be forced.
356
357The AutoNeg parameter is used when more control is required over the
358auto-negotiation process. It should be used when you wish to control which
359speed and duplex combinations are advertised during the auto-negotiation
360process.
361
362The parameter may be specified as either a decimal or hexadecimal value as
363determined by the bitmap below.
364
365============== ====== ====== ======= ======= ====== ====== ======= ======
366Bit position 7 6 5 4 3 2 1 0
367Decimal Value 128 64 32 16 8 4 2 1
368Hex value 80 40 20 10 8 4 2 1
369Speed (Mbps) N/A N/A 1000 N/A 100 100 10 10
370Duplex Full Full Half Full Half
371============== ====== ====== ======= ======= ====== ====== ======= ======
372
373Some examples of using AutoNeg::
374
375 modprobe e1000 AutoNeg=0x01 (Restricts autonegotiation to 10 Half)
376 modprobe e1000 AutoNeg=1 (Same as above)
377 modprobe e1000 AutoNeg=0x02 (Restricts autonegotiation to 10 Full)
378 modprobe e1000 AutoNeg=0x03 (Restricts autonegotiation to 10 Half or 10 Full)
379 modprobe e1000 AutoNeg=0x04 (Restricts autonegotiation to 100 Half)
380 modprobe e1000 AutoNeg=0x05 (Restricts autonegotiation to 10 Half or 100
381 Half)
382 modprobe e1000 AutoNeg=0x020 (Restricts autonegotiation to 1000 Full)
383 modprobe e1000 AutoNeg=32 (Same as above)
384
385Note that when this parameter is used, Speed and Duplex must not be specified.
386
387If the link partner is forced to a specific speed and duplex, then this
388parameter should not be used. Instead, use the Speed and Duplex parameters
389previously mentioned to force the adapter to the same speed and duplex.
390
391Additional Configurations
392=========================
393
394Jumbo Frames
395------------
396
397 Jumbo Frames support is enabled by changing the MTU to a value larger than
398 the default of 1500. Use the ifconfig command to increase the MTU size.
399 For example::
400
401 ifconfig eth<x> mtu 9000 up
402
403 This setting is not saved across reboots. It can be made permanent if
404 you add::
405
406 MTU=9000
407
408 to the file /etc/sysconfig/network-scripts/ifcfg-eth<x>. This example
409 applies to the Red Hat distributions; other distributions may store this
410 setting in a different location.
411
412Notes:
413 Degradation in throughput performance may be observed in some Jumbo frames
414 environments. If this is observed, increasing the application's socket buffer
415 size and/or increasing the /proc/sys/net/ipv4/tcp_*mem entry values may help.
416 See the specific application manual and /usr/src/linux*/Documentation/
417 networking/ip-sysctl.txt for more details.
418
419 - The maximum MTU setting for Jumbo Frames is 16110. This value coincides
420 with the maximum Jumbo Frames size of 16128.
421
422 - Using Jumbo frames at 10 or 100 Mbps is not supported and may result in
423 poor performance or loss of link.
424
425 - Adapters based on the Intel(R) 82542 and 82573V/E controller do not
426 support Jumbo Frames. These correspond to the following product names::
427
428 Intel(R) PRO/1000 Gigabit Server Adapter
429 Intel(R) PRO/1000 PM Network Connection
430
431ethtool
432-------
433
434 The driver utilizes the ethtool interface for driver configuration and
435 diagnostics, as well as displaying statistical information. The ethtool
436 version 1.6 or later is required for this functionality.
437
438 The latest release of ethtool can be found from
439 https://www.kernel.org/pub/software/network/ethtool/
440
441Enabling Wake on LAN* (WoL)
442---------------------------
443
444 WoL is configured through the ethtool* utility.
445
446 WoL will be enabled on the system during the next shut down or reboot.
447 For this driver version, in order to enable WoL, the e1000 driver must be
448 loaded when shutting down or rebooting the system.
449
450Support
451=======
452
453For general information, go to the Intel support website at:
454
455 http://support.intel.com
456
457or the Intel Wired Networking project hosted by Sourceforge at:
458
459 http://sourceforge.net/projects/e1000
460
461If an issue is identified with the released source code on the supported
462kernel with a supported adapter, email the specific information related
463to the issue to e1000-devel@lists.sf.net