About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Documentation / networking / ixgbe.txt


Based on kernel version 4.16.1. Page generated on 2018-04-09 11:53 EST.

1	Linux* Base Driver for the Intel(R) Ethernet 10 Gigabit PCI Express Family of
2	Adapters
3	=============================================================================
4	
5	Intel 10 Gigabit Linux driver.
6	Copyright(c) 1999 - 2013 Intel Corporation.
7	
8	Contents
9	========
10	
11	- Identifying Your Adapter
12	- Additional Configurations
13	- Performance Tuning
14	- Known Issues
15	- Support
16	
17	Identifying Your Adapter
18	========================
19	
20	The driver in this release is compatible with 82598, 82599 and X540-based
21	Intel Network Connections.
22	
23	For more information on how to identify your adapter, go to the Adapter &
24	Driver ID Guide at:
25	
26	    http://support.intel.com/support/network/sb/CS-012904.htm
27	
28	SFP+ Devices with Pluggable Optics
29	----------------------------------
30	
31	82599-BASED ADAPTERS
32	
33	NOTES: If your 82599-based Intel(R) Network Adapter came with Intel optics, or
34	is an Intel(R) Ethernet Server Adapter X520-2, then it only supports Intel
35	optics and/or the direct attach cables listed below.
36	
37	When 82599-based SFP+ devices are connected back to back, they should be set to
38	the same Speed setting via ethtool. Results may vary if you mix speed settings.
39	82598-based adapters support all passive direct attach cables that comply
40	with SFF-8431 v4.1 and SFF-8472 v10.4 specifications. Active direct attach
41	cables are not supported.
42	
43	Supplier    Type                                             Part Numbers
44	
45	SR Modules
46	Intel       DUAL RATE 1G/10G SFP+ SR (bailed)                FTLX8571D3BCV-IT
47	Intel       DUAL RATE 1G/10G SFP+ SR (bailed)                AFBR-703SDDZ-IN1
48	Intel       DUAL RATE 1G/10G SFP+ SR (bailed)                AFBR-703SDZ-IN2
49	LR Modules
50	Intel       DUAL RATE 1G/10G SFP+ LR (bailed)                FTLX1471D3BCV-IT
51	Intel       DUAL RATE 1G/10G SFP+ LR (bailed)                AFCT-701SDDZ-IN1
52	Intel       DUAL RATE 1G/10G SFP+ LR (bailed)                AFCT-701SDZ-IN2
53	
54	The following is a list of 3rd party SFP+ modules and direct attach cables that
55	have received some testing. Not all modules are applicable to all devices.
56	
57	Supplier   Type                                              Part Numbers
58	
59	Finisar    SFP+ SR bailed, 10g single rate                   FTLX8571D3BCL
60	Avago      SFP+ SR bailed, 10g single rate                   AFBR-700SDZ
61	Finisar    SFP+ LR bailed, 10g single rate                   FTLX1471D3BCL
62	
63	Finisar    DUAL RATE 1G/10G SFP+ SR (No Bail)                FTLX8571D3QCV-IT
64	Avago      DUAL RATE 1G/10G SFP+ SR (No Bail)                AFBR-703SDZ-IN1
65	Finisar    DUAL RATE 1G/10G SFP+ LR (No Bail)                FTLX1471D3QCV-IT
66	Avago      DUAL RATE 1G/10G SFP+ LR (No Bail)                AFCT-701SDZ-IN1
67	Finistar   1000BASE-T SFP                                    FCLF8522P2BTL
68	Avago      1000BASE-T SFP                                    ABCU-5710RZ
69	
70	82599-based adapters support all passive and active limiting direct attach
71	cables that comply with SFF-8431 v4.1 and SFF-8472 v10.4 specifications.
72	
73	Laser turns off for SFP+ when device is down
74	-------------------------------------------
75	"ip link set down" turns off the laser for 82599-based SFP+ fiber adapters.
76	"ip link set up" turns on the laser.
77	
78	
79	82598-BASED ADAPTERS
80	
81	NOTES for 82598-Based Adapters:
82	- Intel(R) Network Adapters that support removable optical modules only support
83	  their original module type (i.e., the Intel(R) 10 Gigabit SR Dual Port
84	  Express Module only supports SR optical modules). If you plug in a different
85	  type of module, the driver will not load.
86	- Hot Swapping/hot plugging optical modules is not supported.
87	- Only single speed, 10 gigabit modules are supported.
88	- LAN on Motherboard (LOMs) may support DA, SR, or LR modules. Other module
89	  types are not supported. Please see your system documentation for details.
90	
91	The following is a list of 3rd party SFP+ modules and direct attach cables that
92	have received some testing. Not all modules are applicable to all devices.
93	
94	Supplier   Type                                              Part Numbers
95	
96	Finisar    SFP+ SR bailed, 10g single rate                   FTLX8571D3BCL
97	Avago      SFP+ SR bailed, 10g single rate                   AFBR-700SDZ
98	Finisar    SFP+ LR bailed, 10g single rate                   FTLX1471D3BCL
99	
100	82598-based adapters support all passive direct attach cables that comply
101	with SFF-8431 v4.1 and SFF-8472 v10.4 specifications. Active direct attach
102	cables are not supported.
103	
104	
105	Flow Control
106	------------
107	Ethernet Flow Control (IEEE 802.3x) can be configured with ethtool to enable
108	receiving and transmitting pause frames for ixgbe. When TX is enabled, PAUSE
109	frames are generated when the receive packet buffer crosses a predefined
110	threshold.  When rx is enabled, the transmit unit will halt for the time delay
111	specified when a PAUSE frame is received.
112	
113	Flow Control is enabled by default. If you want to disable a flow control
114	capable link partner, use ethtool:
115	
116	     ethtool -A eth? autoneg off RX off TX off
117	
118	NOTE: For 82598 backplane cards entering 1 gig mode, flow control default
119	behavior is changed to off.  Flow control in 1 gig mode on these devices can
120	lead to Tx hangs.
121	
122	Intel(R) Ethernet Flow Director
123	-------------------------------
124	Supports advanced filters that direct receive packets by their flows to
125	different queues. Enables tight control on routing a flow in the platform.
126	Matches flows and CPU cores for flow affinity. Supports multiple parameters
127	for flexible flow classification and load balancing.
128	
129	Flow director is enabled only if the kernel is multiple TX queue capable.
130	
131	An included script (set_irq_affinity.sh) automates setting the IRQ to CPU
132	affinity.
133	
134	You can verify that the driver is using Flow Director by looking at the counter
135	in ethtool: fdir_miss and fdir_match.
136	
137	Other ethtool Commands:
138	To enable Flow Director
139		ethtool -K ethX ntuple on
140	To add a filter
141		Use -U switch. e.g., ethtool -U ethX flow-type tcp4 src-ip 10.0.128.23
142	        action 1
143	To see the list of filters currently present:
144		ethtool -u ethX
145	
146	Perfect Filter: Perfect filter is an interface to load the filter table that
147	funnels all flow into queue_0 unless an alternative queue is specified using
148	"action". In that case, any flow that matches the filter criteria will be
149	directed to the appropriate queue.
150	
151	If the queue is defined as -1, filter will drop matching packets.
152	
153	To account for filter matches and misses, there are two stats in ethtool:
154	fdir_match and fdir_miss. In addition, rx_queue_N_packets shows the number of
155	packets processed by the Nth queue.
156	
157	NOTE: Receive Packet Steering (RPS) and Receive Flow Steering (RFS) are not
158	compatible with Flow Director. IF Flow Director is enabled, these will be
159	disabled.
160	
161	The following three parameters impact Flow Director.
162	
163	FdirMode
164	--------
165	Valid Range: 0-2 (0=off, 1=ATR, 2=Perfect filter mode)
166	Default Value: 1
167	
168	  Flow Director filtering modes.
169	
170	FdirPballoc
171	-----------
172	Valid Range: 0-2 (0=64k, 1=128k, 2=256k)
173	Default Value: 0
174	
175	  Flow Director allocated packet buffer size.
176	
177	AtrSampleRate
178	--------------
179	Valid Range: 1-100
180	Default Value: 20
181	
182	  Software ATR Tx packet sample rate. For example, when set to 20, every 20th
183	  packet, looks to see if the packet will create a new flow.
184	
185	Node
186	----
187	Valid Range:   0-n
188	Default Value: 1 (off)
189	
190	  0 - n: where n is the number of NUMA nodes (i.e. 0 - 3) currently online in
191	  your system
192	  1: turns this option off
193	
194	  The Node parameter will allow you to pick which NUMA node you want to have
195	  the adapter allocate memory on.
196	
197	max_vfs
198	-------
199	Valid Range:   1-63
200	Default Value: 0
201	
202	  If the value is greater than 0 it will also force the VMDq parameter to be 1
203	  or more.
204	
205	  This parameter adds support for SR-IOV.  It causes the driver to spawn up to
206	  max_vfs worth of virtual function.
207	
208	
209	Additional Configurations
210	=========================
211	
212	  Jumbo Frames
213	  ------------
214	  The driver supports Jumbo Frames for all adapters. Jumbo Frames support is
215	  enabled by changing the MTU to a value larger than the default of 1500.
216	  The maximum value for the MTU is 16110.  Use the ip command to
217	  increase the MTU size.  For example:
218	
219	        ip link set dev ethx mtu 9000
220	
221	  The maximum MTU setting for Jumbo Frames is 9710.  This value coincides
222	  with the maximum Jumbo Frames size of 9728.
223	
224	  Generic Receive Offload, aka GRO
225	  --------------------------------
226	  The driver supports the in-kernel software implementation of GRO.  GRO has
227	  shown that by coalescing Rx traffic into larger chunks of data, CPU
228	  utilization can be significantly reduced when under large Rx load.  GRO is an
229	  evolution of the previously-used LRO interface.  GRO is able to coalesce
230	  other protocols besides TCP.  It's also safe to use with configurations that
231	  are problematic for LRO, namely bridging and iSCSI.
232	
233	  Data Center Bridging, aka DCB
234	  -----------------------------
235	  DCB is a configuration Quality of Service implementation in hardware.
236	  It uses the VLAN priority tag (802.1p) to filter traffic.  That means
237	  that there are 8 different priorities that traffic can be filtered into.
238	  It also enables priority flow control which can limit or eliminate the
239	  number of dropped packets during network stress.  Bandwidth can be
240	  allocated to each of these priorities, which is enforced at the hardware
241	  level.
242	
243	  To enable DCB support in ixgbe, you must enable the DCB netlink layer to
244	  allow the userspace tools (see below) to communicate with the driver.
245	  This can be found in the kernel configuration here:
246	
247	        -> Networking support
248	          -> Networking options
249	            -> Data Center Bridging support
250	
251	  Once this is selected, DCB support must be selected for ixgbe.  This can
252	  be found here:
253	
254	        -> Device Drivers
255	          -> Network device support (NETDEVICES [=y])
256	            -> Ethernet (10000 Mbit) (NETDEV_10000 [=y])
257	              -> Intel(R) 10GbE PCI Express adapters support
258	                -> Data Center Bridging (DCB) Support
259	
260	  After these options are selected, you must rebuild your kernel and your
261	  modules.
262	
263	  In order to use DCB, userspace tools must be downloaded and installed.
264	  The dcbd tools can be found at:
265	
266	        http://e1000.sf.net
267	
268	  Ethtool
269	  -------
270	  The driver utilizes the ethtool interface for driver configuration and
271	  diagnostics, as well as displaying statistical information. The latest
272	  ethtool version is required for this functionality.
273	
274	  The latest release of ethtool can be found from
275	  https://www.kernel.org/pub/software/network/ethtool/
276	
277	  FCoE
278	  ----
279	  This release of the ixgbe driver contains new code to enable users to use
280	  Fiber Channel over Ethernet (FCoE) and Data Center Bridging (DCB)
281	  functionality that is supported by the 82598-based hardware.  This code has
282	  no default effect on the regular driver operation, and configuring DCB and
283	  FCoE is outside the scope of this driver README. Refer to
284	  http://www.open-fcoe.org/ for FCoE project information and contact
285	  e1000-eedc@lists.sourceforge.net for DCB information.
286	
287	  MAC and VLAN anti-spoofing feature
288	  ----------------------------------
289	  When a malicious driver attempts to send a spoofed packet, it is dropped by
290	  the hardware and not transmitted.  An interrupt is sent to the PF driver
291	  notifying it of the spoof attempt.
292	
293	  When a spoofed packet is detected the PF driver will send the following
294	  message to the system log (displayed by  the "dmesg" command):
295	
296	  Spoof event(s) detected on VF (n)
297	
298	  Where n=the VF that attempted to do the spoofing.
299	
300	
301	Performance Tuning
302	==================
303	
304	An excellent article on performance tuning can be found at:
305	
306	http://www.redhat.com/promo/summit/2008/downloads/pdf/Thursday/Mark_Wagner.pdf
307	
308	
309	Known Issues
310	============
311	
312	  Enabling SR-IOV in a 32-bit or 64-bit Microsoft* Windows* Server 2008/R2
313	  Guest OS using Intel (R) 82576-based GbE or Intel (R) 82599-based 10GbE
314	  controller under KVM
315	  ------------------------------------------------------------------------
316	  KVM Hypervisor/VMM supports direct assignment of a PCIe device to a VM.  This
317	  includes traditional PCIe devices, as well as SR-IOV-capable devices using
318	  Intel 82576-based and 82599-based controllers.
319	
320	  While direct assignment of a PCIe device or an SR-IOV Virtual Function (VF)
321	  to a Linux-based VM running 2.6.32 or later kernel works fine, there is a
322	  known issue with Microsoft Windows Server 2008 VM that results in a "yellow
323	  bang" error. This problem is within the KVM VMM itself, not the Intel driver,
324	  or the SR-IOV logic of the VMM, but rather that KVM emulates an older CPU
325	  model for the guests, and this older CPU model does not support MSI-X
326	  interrupts, which is a requirement for Intel SR-IOV.
327	
328	  If you wish to use the Intel 82576 or 82599-based controllers in SR-IOV mode
329	  with KVM and a Microsoft Windows Server 2008 guest try the following
330	  workaround. The workaround is to tell KVM to emulate a different model of CPU
331	  when using qemu to create the KVM guest:
332	
333	       "-cpu qemu64,model=13"
334	
335	
336	Support
337	=======
338	
339	For general information, go to the Intel support website at:
340	
341	    http://support.intel.com
342	
343	or the Intel Wired Networking project hosted by Sourceforge at:
344	
345	    http://e1000.sourceforge.net
346	
347	If an issue is identified with the released source code on the supported
348	kernel with a supported adapter, email the specific information related
349	to the issue to e1000-devel@lists.sf.net
Hide Line Numbers


About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog