1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2015-2016 Intel Corporation.
7 The FM10K poll mode driver library provides support for the Intel FM10000
8 (FM10K) family of 40GbE/100GbE adapters.
10 FTAG Based Forwarding of FM10K
11 ------------------------------
13 FTAG Based Forwarding is a unique feature of FM10K. The FM10K family of NICs
14 support the addition of a Fabric Tag (FTAG) to carry special information.
15 The FTAG is placed at the beginning of the frame, it contains information
16 such as where the packet comes from and goes, and the vlan tag. In FTAG based
17 forwarding mode, the switch logic forwards packets according to glort (global
18 resource tag) information, rather than the mac and vlan table. Currently this
19 feature works only on PF.
21 To enable this feature, the user should pass a devargs parameter to the eal
22 like "-w 84:00.0,enable_ftag=1", and the application should make sure an
23 appropriate FTAG is inserted for every frame on TX side.
28 Vector PMD (vPMD) uses IntelĀ® SIMD instructions to optimize packet I/O.
29 It improves load/store bandwidth efficiency of L1 data cache by using a wider
30 SSE/AVX ''register (1)''.
31 The wider register gives space to hold multiple packet buffers so as to save
32 on the number of instructions when bulk processing packets.
34 There is no change to the PMD API. The RX/TX handlers are the only two entries for
35 vPMD packet I/O. They are transparently registered at runtime RX/TX execution
36 if all required conditions are met.
38 1. To date, only an SSE version of FM10K vPMD is available.
39 To ensure that vPMD is in the binary code, set
40 ``CONFIG_RTE_LIBRTE_FM10K_INC_VECTOR=y`` in the configure file.
42 Some constraints apply as pre-conditions for specific optimizations on bulk
43 packet transfers. The following sections explain RX and TX constraints in the
51 Prerequisites and Pre-conditions
52 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
54 For Vector RX it is assumed that the number of descriptor rings will be a power
55 of 2. With this pre-condition, the ring pointer can easily scroll back to the
56 head after hitting the tail without a conditional check. In addition Vector RX
57 can use this assumption to do a bit mask using ``ring_size - 1``.
60 Features not Supported by Vector RX PMD
61 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
63 Some features are not supported when trying to increase the throughput in
74 Other features are supported using optional MACRO configuration. They include:
80 To enable via ``RX_OLFLAGS`` use ``RTE_LIBRTE_FM10K_RX_OLFLAGS_ENABLE=y``.
82 To guarantee the constraint, the following capabilities in ``dev_conf.rxmode.offloads``
85 * ``DEV_RX_OFFLOAD_VLAN_EXTEND``
87 * ``DEV_RX_OFFLOAD_CHECKSUM``
89 * ``DEV_RX_OFFLOAD_HEADER_SPLIT``
97 As vPMD is focused on high throughput, it processes 4 packets at a time. So it assumes
98 that the RX burst should be greater than 4 packets per burst. It returns zero if using
99 ``nb_pkt`` < 4 in the receive handler. If ``nb_pkt`` is not a multiple of 4, a
100 floor alignment will be applied.
106 Features not Supported by TX Vector PMD
107 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
109 TX vPMD only works when offloads is set to 0
111 This means that it does not support any TX offload.
120 The Intel FM10000 family of NICs integrate a hardware switch and multiple host
121 interfaces. The FM10000 PMD driver only manages host interfaces. For the
122 switch component another switch driver has to be loaded prior to to the
123 FM10000 PMD driver. The switch driver can be acquired from Intel support.
124 Only Testpoint is validated with DPDK, the latest version that has been
125 validated with DPDK is 4.1.6.
127 Support for Switch Restart
128 ~~~~~~~~~~~~~~~~~~~~~~~~~~
130 For FM10000 multi host based design a DPDK app running in the VM or host needs
131 to be aware of the switch's state since it may undergo a quit-restart. When
132 the switch goes down the DPDK app will receive a LSC event indicating link
133 status down, and the app should stop the worker threads that are polling on
134 the Rx/Tx queues. When switch comes up, a LSC event indicating ``LINK_UP`` is
135 sent to the app, which can then restart the FM10000 port to resume network
141 The FM10000 family of NICs strip the CRC for every packets coming into the
142 host interface. So, CRC will be stripped even when ``DEV_RX_OFFLOAD_CRC_STRIP``
143 in ``rxmode.offloads`` is NOT set in ``struct rte_eth_conf``.
145 Maximum packet length
146 ~~~~~~~~~~~~~~~~~~~~~
148 The FM10000 family of NICS support a maximum of a 15K jumbo frame. The value
149 is fixed and cannot be changed. So, even when the ``rxmode.max_rx_pkt_len``
150 member of ``struct rte_eth_conf`` is set to a value lower than 15364, frames
151 up to 15364 bytes can still reach the host interface.
153 Statistic Polling Frequency
154 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
156 The FM10000 NICs expose a set of statistics via the PCI BARs. These statistics
157 are read from the hardware registers when ``rte_eth_stats_get()`` or
158 ``rte_eth_xstats_get()`` is called. The packet counting registers are 32 bits
159 while the byte counting registers are 48 bits. As a result, the statistics must
160 be polled regularly in order to ensure the consistency of the returned reads.
162 Given the PCIe Gen3 x8, about 50Gbps of traffic can occur. With 64 byte packets
163 this gives almost 100 million packets/second, causing 32 bit integer overflow
164 after approx 40 seconds. To ensure these overflows are detected and accounted
165 for in the statistics, it is necessary to read statistic regularly. It is
166 suggested to read stats every 20 seconds, which will ensure the statistics
173 The FM10000 family of NICS need one separate interrupt for mailbox. So only
174 drivers which support multiple interrupt vectors e.g. vfio-pci can work
175 for fm10k interrupt mode.