1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2018 Intel Corporation.
7 The ice PMD (librte_pmd_ice) provides poll mode driver support for
8 10/25/50/100 Gbps IntelĀ® Ethernet 810 Series Network Adapters based on
9 the Intel Ethernet Controller E810.
15 - Follow the DPDK :ref:`Getting Started Guide for Linux <linux_gsg>` to setup the basic DPDK environment.
17 - To get better performance on Intel platforms, please follow the "How to get best performance with NICs on Intel platforms"
18 section of the :ref:`Getting Started Guide for Linux <linux_gsg>`.
20 Pre-Installation Configuration
21 ------------------------------
26 The following options can be modified in the ``config`` file.
27 Please note that enabling debugging options may affect system performance.
29 - ``CONFIG_RTE_LIBRTE_ICE_PMD`` (default ``y``)
31 Toggle compilation of the ``librte_pmd_ice`` driver.
33 - ``CONFIG_RTE_LIBRTE_ICE_DEBUG_*`` (default ``n``)
35 Toggle display of generic debugging messages.
37 - ``CONFIG_RTE_LIBRTE_ICE_16BYTE_RX_DESC`` (default ``n``)
39 Toggle to use a 16-byte RX descriptor, by default the RX descriptor is 32 byte.
41 Runtime Config Options
42 ~~~~~~~~~~~~~~~~~~~~~~
44 - ``Safe Mode Support`` (default ``0``)
46 If driver failed to load OS package, by default driver's initialization failed.
47 But if user intend to use the device without OS package, user can take ``devargs``
48 parameter ``safe-mode-support``, for example::
50 -w 80:00.0,safe-mode-support=1
52 Then the driver will be initialized successfully and the device will enter Safe Mode.
53 NOTE: In Safe mode, only very limited features are available, features like RSS,
54 checksum, fdir, tunneling ... are all disabled.
56 - ``Generic Flow Pipeline Mode Support`` (default ``0``)
58 In pipeline mode, a flow can be set at one specific stage by setting parameter
59 ``priority``. Currently, we support two stages: priority = 0 or !0. Flows with
60 priority 0 located at the first pipeline stage which typically be used as a firewall
61 to drop the packet on a blacklist(we called it permission stage). At this stage,
62 flow rules are created for the device's exact match engine: switch. Flows with priority
63 !0 located at the second stage, typically packets are classified here and be steered to
64 specific queue or queue group (we called it distribution stage), At this stage, flow
65 rules are created for device's flow director engine.
66 For none-pipeline mode, ``priority`` is ignored, a flow rule can be created as a flow director
67 rule or a switch rule depends on its pattern/action and the resource allocation situation,
68 all flows are virtually at the same pipeline stage.
69 By default, generic flow API is enabled in none-pipeline mode, user can choose to
70 use pipeline mode by setting ``devargs`` parameter ``pipeline-mode-support``,
73 -w 80:00.0,pipeline-mode-support=1
75 - ``Flow Mark Support`` (default ``0``)
77 This is a hint to the driver to select the data path that supports flow mark extraction
79 NOTE: This is an experimental devarg, it will be removed when any of below conditions
81 1) all data paths support flow mark (currently vPMD does not)
82 2) a new offload like RTE_DEV_RX_OFFLOAD_FLOW_MARK be introduced as a standard way to hint.
85 -w 80:00.0,flow-mark-support=1
87 - ``Protocol extraction for per queue``
89 Configure the RX queues to do protocol extraction into mbuf for protocol
90 handling acceleration, like checking the TCP SYN packets quickly.
92 The argument format is::
94 -w 18:00.0,proto_xtr=<queues:protocol>[<queues:protocol>...]
95 -w 18:00.0,proto_xtr=<protocol>
97 Queues are grouped by ``(`` and ``)`` within the group. The ``-`` character
98 is used as a range separator and ``,`` is used as a single number separator.
99 The grouping ``()`` can be omitted for single element group. If no queues are
100 specified, PMD will use this protocol extraction type for all queues.
102 Protocol is : ``vlan, ipv4, ipv6, ipv6_flow, tcp, ip_offset``.
104 .. code-block:: console
106 testpmd -w 18:00.0,proto_xtr='[(1,2-3,8-9):tcp,10-13:vlan]'
108 This setting means queues 1, 2-3, 8-9 are TCP extraction, queues 10-13 are
109 VLAN extraction, other queues run with no protocol extraction.
111 .. code-block:: console
113 testpmd -w 18:00.0,proto_xtr=vlan,proto_xtr='[(1,2-3,8-9):tcp,10-23:ipv6]'
115 This setting means queues 1, 2-3, 8-9 are TCP extraction, queues 10-23 are
116 IPv6 extraction, other queues use the default VLAN extraction.
118 The extraction metadata is copied into the registered dynamic mbuf field, and
119 the related dynamic mbuf flags is set.
121 .. table:: Protocol extraction : ``vlan``
123 +----------------------------+----------------------------+
125 +======+===+=================+======+===+=================+
126 | PCP | D | VID | PCP | D | VID |
127 +------+---+-----------------+------+---+-----------------+
129 VLAN1 - single or EVLAN (first for QinQ).
131 VLAN2 - C-VLAN (second for QinQ).
133 .. table:: Protocol extraction : ``ipv4``
135 +----------------------------+----------------------------+
137 +======+=======+=============+==============+=============+
138 | Ver |Hdr Len| ToS | TTL | Protocol |
139 +------+-------+-------------+--------------+-------------+
141 IPHDR1 - IPv4 header word 4, "TTL" and "Protocol" fields.
143 IPHDR2 - IPv4 header word 0, "Ver", "Hdr Len" and "Type of Service" fields.
145 .. table:: Protocol extraction : ``ipv6``
147 +----------------------------+----------------------------+
149 +=====+=============+========+=============+==============+
150 | Ver |Traffic class| Flow | Next Header | Hop Limit |
151 +-----+-------------+--------+-------------+--------------+
153 IPHDR1 - IPv6 header word 3, "Next Header" and "Hop Limit" fields.
155 IPHDR2 - IPv6 header word 0, "Ver", "Traffic class" and high 4 bits of
158 .. table:: Protocol extraction : ``ipv6_flow``
160 +----------------------------+----------------------------+
162 +=====+=============+========+============================+
163 | Ver |Traffic class| Flow Label |
164 +-----+-------------+-------------------------------------+
166 IPHDR1 - IPv6 header word 1, 16 low bits of the "Flow Label" field.
168 IPHDR2 - IPv6 header word 0, "Ver", "Traffic class" and high 4 bits of
171 .. table:: Protocol extraction : ``tcp``
173 +----------------------------+----------------------------+
174 | TCPHDR2 | TCPHDR1 |
175 +============================+======+======+==============+
176 | Reserved |Offset| RSV | Flags |
177 +----------------------------+------+------+--------------+
179 TCPHDR1 - TCP header word 6, "Data Offset" and "Flags" fields.
183 .. table:: Protocol extraction : ``ip_offset``
185 +----------------------------+----------------------------+
187 +============================+============================+
188 | IPv6 HDR Offset | IPv4 HDR Offset |
189 +----------------------------+----------------------------+
191 IPHDR1 - Outer/Single IPv4 Header offset.
193 IPHDR2 - Outer/Single IPv6 Header offset.
195 Use ``rte_net_ice_dynf_proto_xtr_metadata_get`` to access the protocol
196 extraction metadata, and use ``RTE_PKT_RX_DYNF_PROTO_XTR_*`` to get the
197 metadata type of ``struct rte_mbuf::ol_flags``.
199 The ``rte_net_ice_dump_proto_xtr_metadata`` routine shows how to
200 access the protocol extraction result in ``struct rte_mbuf``.
202 Driver compilation and testing
203 ------------------------------
205 Refer to the document :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>`
214 Vector PMD for RX and TX path are selected automatically. The paths
215 are chosen based on 2 conditions.
218 On the X86 platform, the driver checks if the CPU supports AVX2.
219 If it's supported, AVX2 paths will be chosen. If not, SSE is chosen.
221 - ``Offload features``
222 The supported HW offload features are described in the document ice_vec.ini.
223 If any not supported features are used, ICE vector PMD is disabled and the
224 normal paths are chosen.
226 Malicious driver detection (MDD)
227 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
229 It's not appropriate to send a packet, if this packet's destination MAC address
230 is just this port's MAC address. If SW tries to send such packets, HW will
231 report a MDD event and drop the packets.
233 The APPs based on DPDK should avoid providing such packets.
235 Device Config Function (DCF)
236 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
238 This section demonstrates ICE DCF PMD, which shares the core module with ICE
241 A DCF (Device Config Function) PMD bounds to the device's trusted VF with ID 0,
242 it can act as a sole controlling entity to exercise advance functionality (such
243 as switch, ACL) for the rest VFs.
245 The DCF PMD needs to advertise and acquire DCF capability which allows DCF to
246 send AdminQ commands that it would like to execute over to the PF and receive
247 responses for the same from PF.
251 .. figure:: img/ice_dcf.*
253 DCF Communication flow.
257 echo 4 > /sys/bus/pci/devices/0000\:18\:00.0/sriov_numvfs
259 #. Enable the VF0 trust on::
261 ip link set dev enp24s0f0 vf 0 trust on
263 #. Bind the VF0, and run testpmd with 'cap=dcf' devarg::
265 testpmd -l 22-25 -n 4 -w 18:01.0,cap=dcf -- -i
267 #. Monitor the VF2 interface network traffic::
269 tcpdump -e -nn -i enp24s1f2
271 #. Create one flow to redirect the traffic to VF2 by DCF::
273 flow create 0 priority 0 ingress pattern eth / ipv4 src is 192.168.0.2 \
274 dst is 192.168.0.3 / end actions vf id 2 / end
276 #. Send the packet, and it should be displayed on tcpdump::
278 sendp(Ether(src='3c:fd:fe:aa:bb:78', dst='00:00:00:01:02:03')/IP(src=' \
279 192.168.0.2', dst="192.168.0.3")/TCP(flags='S')/Raw(load='XXXXXXXXXX'), \
280 iface="enp24s0f0", count=10)
282 Sample Application Notes
283 ------------------------
288 Vlan filter only works when Promiscuous mode is off.
290 To start ``testpmd``, and add vlan 10 to port 0:
292 .. code-block:: console
294 ./app/testpmd -l 0-15 -n 4 -- -i
297 testpmd> rx_vlan add 10 0
299 Limitations or Known issues
300 ---------------------------
302 The Intel E810 requires a programmable pipeline package be downloaded
303 by the driver to support normal operations. The E810 has a limited
304 functionality built in to allow PXE boot and other use cases, but the
305 driver must download a package file during the driver initialization
308 The default DDP package file name is ice.pkg. For a specific NIC, the
309 DDP package supposed to be loaded can have a filename: ice-xxxxxx.pkg,
310 where 'xxxxxx' is the 64-bit PCIe Device Serial Number of the NIC. For
311 example, if the NIC's device serial number is 00-CC-BB-FF-FF-AA-05-68,
312 the device-specific DDP package filename is ice-00ccbbffffaa0568.pkg
313 (in hex and all low case). During initialization, the driver searches
314 in the following paths in order: /lib/firmware/updates/intel/ice/ddp
315 and /lib/firmware/intel/ice/ddp. The corresponding device-specific DDP
316 package will be downloaded first if the file exists. If not, then the
317 driver tries to load the default package. The type of loaded package
318 is stored in ``ice_adapter->active_pkg_type``.
320 A symbolic link to the DDP package file is also ok. The same package
321 file is used by both the kernel driver and the DPDK PMD.