Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 1 | Linux* Driver for Intel(R) Network Connection |
Jeff Kirsher | 072ed34f | 2010-12-09 23:44:42 -0800 | [diff] [blame] | 2 | ============================================= |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 3 | |
| 4 | Intel Gigabit Linux driver. |
| 5 | Copyright(c) 1999 - 2010 Intel Corporation. |
| 6 | |
| 7 | Contents |
| 8 | ======== |
| 9 | |
| 10 | - Identifying Your Adapter |
| 11 | - Command Line Parameters |
| 12 | - Additional Configurations |
| 13 | - Support |
| 14 | |
| 15 | Identifying Your Adapter |
| 16 | ======================== |
| 17 | |
| 18 | The e1000e driver supports all PCI Express Intel(R) Gigabit Network |
| 19 | Connections, except those that are 82575, 82576 and 82580-based*. |
| 20 | |
| 21 | * NOTE: The Intel(R) PRO/1000 P Dual Port Server Adapter is supported by |
| 22 | the e1000 driver, not the e1000e driver due to the 82546 part being used |
| 23 | behind a PCI Express bridge. |
| 24 | |
| 25 | For more information on how to identify your adapter, go to the Adapter & |
| 26 | Driver ID Guide at: |
| 27 | |
| 28 | http://support.intel.com/support/go/network/adapter/idguide.htm |
| 29 | |
| 30 | For the latest Intel network drivers for Linux, refer to the following |
| 31 | website. In the search field, enter your adapter name or type, or use the |
| 32 | networking link on the left to search for your adapter: |
| 33 | |
| 34 | http://support.intel.com/support/go/network/adapter/home.htm |
| 35 | |
| 36 | Command Line Parameters |
| 37 | ======================= |
| 38 | |
| 39 | The default value for each parameter is generally the recommended setting, |
| 40 | unless otherwise noted. |
| 41 | |
| 42 | NOTES: For more information about the InterruptThrottleRate, |
| 43 | RxIntDelay, TxIntDelay, RxAbsIntDelay, and TxAbsIntDelay |
| 44 | parameters, see the application note at: |
| 45 | http://www.intel.com/design/network/applnots/ap450.htm |
| 46 | |
| 47 | InterruptThrottleRate |
| 48 | --------------------- |
| 49 | Valid Range: 0,1,3,4,100-100000 (0=off, 1=dynamic, 3=dynamic conservative, |
| 50 | 4=simplified balancing) |
| 51 | Default Value: 3 |
| 52 | |
| 53 | The driver can limit the amount of interrupts per second that the adapter |
| 54 | will generate for incoming packets. It does this by writing a value to the |
| 55 | adapter that is based on the maximum amount of interrupts that the adapter |
| 56 | will generate per second. |
| 57 | |
| 58 | Setting InterruptThrottleRate to a value greater or equal to 100 |
| 59 | will program the adapter to send out a maximum of that many interrupts |
| 60 | per second, even if more packets have come in. This reduces interrupt |
| 61 | load on the system and can lower CPU utilization under heavy load, |
| 62 | but will increase latency as packets are not processed as quickly. |
| 63 | |
Jeff Kirsher | 072ed34f | 2010-12-09 23:44:42 -0800 | [diff] [blame] | 64 | The default behaviour of the driver previously assumed a static |
| 65 | InterruptThrottleRate value of 8000, providing a good fallback value for |
| 66 | all traffic types, but lacking in small packet performance and latency. |
| 67 | The hardware can handle many more small packets per second however, and |
| 68 | for this reason an adaptive interrupt moderation algorithm was implemented. |
| 69 | |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 70 | The driver has two adaptive modes (setting 1 or 3) in which |
| 71 | it dynamically adjusts the InterruptThrottleRate value based on the traffic |
| 72 | that it receives. After determining the type of incoming traffic in the last |
| 73 | timeframe, it will adjust the InterruptThrottleRate to an appropriate value |
| 74 | for that traffic. |
| 75 | |
| 76 | The algorithm classifies the incoming traffic every interval into |
| 77 | classes. Once the class is determined, the InterruptThrottleRate value is |
| 78 | adjusted to suit that traffic type the best. There are three classes defined: |
| 79 | "Bulk traffic", for large amounts of packets of normal size; "Low latency", |
| 80 | for small amounts of traffic and/or a significant percentage of small |
| 81 | packets; and "Lowest latency", for almost completely small packets or |
| 82 | minimal traffic. |
| 83 | |
| 84 | In dynamic conservative mode, the InterruptThrottleRate value is set to 4000 |
| 85 | for traffic that falls in class "Bulk traffic". If traffic falls in the "Low |
| 86 | latency" or "Lowest latency" class, the InterruptThrottleRate is increased |
| 87 | stepwise to 20000. This default mode is suitable for most applications. |
| 88 | |
| 89 | For situations where low latency is vital such as cluster or |
| 90 | grid computing, the algorithm can reduce latency even more when |
| 91 | InterruptThrottleRate is set to mode 1. In this mode, which operates |
| 92 | the same as mode 3, the InterruptThrottleRate will be increased stepwise to |
| 93 | 70000 for traffic in class "Lowest latency". |
| 94 | |
Jeff Kirsher | 072ed34f | 2010-12-09 23:44:42 -0800 | [diff] [blame] | 95 | In simplified mode the interrupt rate is based on the ratio of TX and |
| 96 | RX traffic. If the bytes per second rate is approximately equal, the |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 97 | interrupt rate will drop as low as 2000 interrupts per second. If the |
| 98 | traffic is mostly transmit or mostly receive, the interrupt rate could |
| 99 | be as high as 8000. |
| 100 | |
| 101 | Setting InterruptThrottleRate to 0 turns off any interrupt moderation |
| 102 | and may improve small packet latency, but is generally not suitable |
| 103 | for bulk throughput traffic. |
| 104 | |
| 105 | NOTE: InterruptThrottleRate takes precedence over the TxAbsIntDelay and |
| 106 | RxAbsIntDelay parameters. In other words, minimizing the receive |
| 107 | and/or transmit absolute delays does not force the controller to |
| 108 | generate more interrupts than what the Interrupt Throttle Rate |
| 109 | allows. |
| 110 | |
| 111 | NOTE: When e1000e is loaded with default settings and multiple adapters |
| 112 | are in use simultaneously, the CPU utilization may increase non- |
| 113 | linearly. In order to limit the CPU utilization without impacting |
| 114 | the overall throughput, we recommend that you load the driver as |
| 115 | follows: |
| 116 | |
| 117 | modprobe e1000e InterruptThrottleRate=3000,3000,3000 |
| 118 | |
| 119 | This sets the InterruptThrottleRate to 3000 interrupts/sec for |
| 120 | the first, second, and third instances of the driver. The range |
| 121 | of 2000 to 3000 interrupts per second works on a majority of |
| 122 | systems and is a good starting point, but the optimal value will |
| 123 | be platform-specific. If CPU utilization is not a concern, use |
| 124 | RX_POLLING (NAPI) and default driver settings. |
| 125 | |
| 126 | RxIntDelay |
| 127 | ---------- |
| 128 | Valid Range: 0-65535 (0=off) |
| 129 | Default Value: 0 |
| 130 | |
| 131 | This value delays the generation of receive interrupts in units of 1.024 |
| 132 | microseconds. Receive interrupt reduction can improve CPU efficiency if |
| 133 | properly tuned for specific network traffic. Increasing this value adds |
| 134 | extra latency to frame reception and can end up decreasing the throughput |
| 135 | of TCP traffic. If the system is reporting dropped receives, this value |
| 136 | may be set too high, causing the driver to run out of available receive |
| 137 | descriptors. |
| 138 | |
| 139 | CAUTION: When setting RxIntDelay to a value other than 0, adapters may |
| 140 | hang (stop transmitting) under certain network conditions. If |
| 141 | this occurs a NETDEV WATCHDOG message is logged in the system |
| 142 | event log. In addition, the controller is automatically reset, |
| 143 | restoring the network connection. To eliminate the potential |
| 144 | for the hang ensure that RxIntDelay is set to 0. |
| 145 | |
| 146 | RxAbsIntDelay |
| 147 | ------------- |
| 148 | Valid Range: 0-65535 (0=off) |
| 149 | Default Value: 8 |
| 150 | |
| 151 | This value, in units of 1.024 microseconds, limits the delay in which a |
| 152 | receive interrupt is generated. Useful only if RxIntDelay is non-zero, |
| 153 | this value ensures that an interrupt is generated after the initial |
| 154 | packet is received within the set amount of time. Proper tuning, |
| 155 | along with RxIntDelay, may improve traffic throughput in specific network |
| 156 | conditions. |
| 157 | |
| 158 | TxIntDelay |
| 159 | ---------- |
| 160 | Valid Range: 0-65535 (0=off) |
| 161 | Default Value: 8 |
| 162 | |
| 163 | This value delays the generation of transmit interrupts in units of |
| 164 | 1.024 microseconds. Transmit interrupt reduction can improve CPU |
| 165 | efficiency if properly tuned for specific network traffic. If the |
| 166 | system is reporting dropped transmits, this value may be set too high |
| 167 | causing the driver to run out of available transmit descriptors. |
| 168 | |
| 169 | TxAbsIntDelay |
| 170 | ------------- |
| 171 | Valid Range: 0-65535 (0=off) |
| 172 | Default Value: 32 |
| 173 | |
| 174 | This value, in units of 1.024 microseconds, limits the delay in which a |
| 175 | transmit interrupt is generated. Useful only if TxIntDelay is non-zero, |
| 176 | this value ensures that an interrupt is generated after the initial |
| 177 | packet is sent on the wire within the set amount of time. Proper tuning, |
| 178 | along with TxIntDelay, may improve traffic throughput in specific |
| 179 | network conditions. |
| 180 | |
| 181 | Copybreak |
| 182 | --------- |
| 183 | Valid Range: 0-xxxxxxx (0=off) |
| 184 | Default Value: 256 |
| 185 | |
Jeff Kirsher | 072ed34f | 2010-12-09 23:44:42 -0800 | [diff] [blame] | 186 | Driver copies all packets below or equaling this size to a fresh RX |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 187 | buffer before handing it up the stack. |
| 188 | |
| 189 | This parameter is different than other parameters, in that it is a |
| 190 | single (not 1,1,1 etc.) parameter applied to all driver instances and |
| 191 | it is also available during runtime at |
| 192 | /sys/module/e1000e/parameters/copybreak |
| 193 | |
| 194 | SmartPowerDownEnable |
| 195 | -------------------- |
| 196 | Valid Range: 0-1 |
| 197 | Default Value: 0 (disabled) |
| 198 | |
| 199 | Allows PHY to turn off in lower power states. The user can set this parameter |
| 200 | in supported chipsets. |
| 201 | |
| 202 | KumeranLockLoss |
| 203 | --------------- |
| 204 | Valid Range: 0-1 |
| 205 | Default Value: 1 (enabled) |
| 206 | |
| 207 | This workaround skips resetting the PHY at shutdown for the initial |
| 208 | silicon releases of ICH8 systems. |
| 209 | |
| 210 | IntMode |
| 211 | ------- |
| 212 | Valid Range: 0-2 (0=legacy, 1=MSI, 2=MSI-X) |
| 213 | Default Value: 2 |
| 214 | |
| 215 | Allows changing the interrupt mode at module load time, without requiring a |
| 216 | recompile. If the driver load fails to enable a specific interrupt mode, the |
| 217 | driver will try other interrupt modes, from least to most compatible. The |
| 218 | interrupt order is MSI-X, MSI, Legacy. If specifying MSI (IntMode=1) |
| 219 | interrupts, only MSI and Legacy will be attempted. |
| 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 | --------------- |
Jeff Kirsher | 072ed34f | 2010-12-09 23:44:42 -0800 | [diff] [blame] | 232 | Valid Range: 0,1 |
| 233 | Default Value: 1 |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 234 | |
Jeff Kirsher | 072ed34f | 2010-12-09 23:44:42 -0800 | [diff] [blame] | 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 | NOTE: The machine must be power cycled (full off/on) when enabling NVM writes |
| 240 | via setting the parameter to zero. Once the NVM has been locked (via the |
| 241 | parameter at 1 when the driver loads) it cannot be unlocked except via power |
| 242 | cycle. |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 243 | |
| 244 | Additional Configurations |
| 245 | ========================= |
| 246 | |
| 247 | Jumbo Frames |
| 248 | ------------ |
| 249 | Jumbo Frames support is enabled by changing the MTU to a value larger than |
| 250 | the default of 1500. Use the ifconfig command to increase the MTU size. |
| 251 | For example: |
| 252 | |
| 253 | ifconfig eth<x> mtu 9000 up |
| 254 | |
| 255 | This setting is not saved across reboots. |
| 256 | |
| 257 | Notes: |
| 258 | |
| 259 | - The maximum MTU setting for Jumbo Frames is 9216. This value coincides |
| 260 | with the maximum Jumbo Frames size of 9234 bytes. |
| 261 | |
| 262 | - Using Jumbo Frames at 10 or 100 Mbps is not supported and may result in |
| 263 | poor performance or loss of link. |
| 264 | |
| 265 | - Some adapters limit Jumbo Frames sized packets to a maximum of |
| 266 | 4096 bytes and some adapters do not support Jumbo Frames. |
| 267 | |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 268 | Ethtool |
| 269 | ------- |
| 270 | The driver utilizes the ethtool interface for driver configuration and |
| 271 | diagnostics, as well as displaying statistical information. We |
Jeff Kirsher | 68f20d9 | 2010-12-17 12:14:34 +0000 | [diff] [blame] | 272 | strongly recommend downloading the latest version of ethtool at: |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 273 | |
Jeff Kirsher | 68f20d9 | 2010-12-17 12:14:34 +0000 | [diff] [blame] | 274 | http://ftp.kernel.org/pub/software/network/ethtool/ |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 275 | |
| 276 | Speed and Duplex |
| 277 | ---------------- |
Jeff Kirsher | 68f20d9 | 2010-12-17 12:14:34 +0000 | [diff] [blame] | 278 | Speed and Duplex are configured through the ethtool* utility. For |
| 279 | instructions, refer to the ethtool man page. |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 280 | |
| 281 | Enabling Wake on LAN* (WoL) |
| 282 | --------------------------- |
Jeff Kirsher | 68f20d9 | 2010-12-17 12:14:34 +0000 | [diff] [blame] | 283 | WoL is configured through the ethtool* utility. For instructions on |
| 284 | enabling WoL with ethtool, refer to the ethtool man page. |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 285 | |
| 286 | WoL will be enabled on the system during the next shut down or reboot. |
| 287 | For this driver version, in order to enable WoL, the e1000e driver must be |
| 288 | loaded when shutting down or rebooting the system. |
| 289 | |
| 290 | In most cases Wake On LAN is only supported on port A for multiple port |
Jeff Kirsher | 68f20d9 | 2010-12-17 12:14:34 +0000 | [diff] [blame] | 291 | adapters. To verify if a port supports Wake on Lan run ethtool eth<X>. |
Jeff Kirsher | da8c01c | 2010-10-05 01:17:27 +0000 | [diff] [blame] | 292 | |
| 293 | Support |
| 294 | ======= |
| 295 | |
| 296 | For general information, go to the Intel support website at: |
| 297 | |
| 298 | www.intel.com/support/ |
| 299 | |
| 300 | or the Intel Wired Networking project hosted by Sourceforge at: |
| 301 | |
| 302 | http://sourceforge.net/projects/e1000 |
| 303 | |
| 304 | If an issue is identified with the released source code on the supported |
| 305 | kernel with a supported adapter, email the specific information related |
| 306 | to the issue to e1000-devel@lists.sf.net |