1b4081870e
Rewrite the RX timestamp setup code to use standard offload flag. Signed-off-by: Martin Spinler <spinler@cesnet.cz>
141 lines
4.8 KiB
ReStructuredText
141 lines
4.8 KiB
ReStructuredText
.. SPDX-License-Identifier: BSD-3-Clause
|
|
Copyright 2019 Cesnet
|
|
Copyright 2019 Netcope Technologies
|
|
|
|
NFB poll mode driver library
|
|
=================================
|
|
|
|
The NFB poll mode driver library implements support for the Netcope
|
|
FPGA Boards (**NFB-40G2, NFB-100G2, NFB-200G2QL**) and Silicom **FB2CGG3** card,
|
|
FPGA-based programmable NICs. The NFB PMD uses interface provided by the libnfb
|
|
library to communicate with these cards over the nfb layer.
|
|
|
|
More information about the
|
|
`NFB cards <https://www.liberouter.org/technologies/cards/>`_
|
|
and used technology
|
|
(`Network Development Kit <https://www.liberouter.org/ndk/>`_)
|
|
can be found on the `Liberouter website <http://www.liberouter.org/>`_.
|
|
|
|
.. note::
|
|
|
|
Currently the driver is supported only on x86_64 architectures.
|
|
Only x86_64 versions of the external libraries are provided.
|
|
|
|
Prerequisites
|
|
-------------
|
|
|
|
This PMD requires kernel modules which are responsible for initialization and
|
|
allocation of resources needed for nfb layer function.
|
|
Communication between PMD and kernel modules is mediated by libnfb library.
|
|
These kernel modules and library are not part of DPDK and must be installed
|
|
separately:
|
|
|
|
* **libnfb library**
|
|
|
|
The library provides API for initialization of nfb transfers, receiving and
|
|
transmitting data segments.
|
|
|
|
* **Kernel modules**
|
|
|
|
* nfb
|
|
|
|
Kernel modules manage initialization of hardware, allocation and
|
|
sharing of resources for user space applications.
|
|
|
|
Dependencies can be found here:
|
|
`Netcope common <https://github.com/CESNET/ndk-sw>`_.
|
|
|
|
Versions of the packages
|
|
~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
The minimum version of the provided packages:
|
|
|
|
* for DPDK from 19.05
|
|
|
|
Configuration
|
|
-------------
|
|
|
|
Timestamps
|
|
|
|
The PMD supports hardware timestamps of frame receipt on physical network interface. In order to use
|
|
the timestamps, the hardware timestamping unit must be enabled (follow the documentation of the NFB
|
|
products). The standard `RTE_ETH_RX_OFFLOAD_TIMESTAMP` flag can be used for this feature.
|
|
|
|
When the timestamps are enabled, a timestamp validity flag is set in the MBUFs
|
|
containing received frames and timestamp is inserted into the `rte_mbuf` struct.
|
|
|
|
The timestamp is an `uint64_t` field. Its lower 32 bits represent *seconds* portion of the timestamp
|
|
(number of seconds elapsed since 1.1.1970 00:00:00 UTC) and its higher 32 bits represent
|
|
*nanosecond* portion of the timestamp (number of nanoseconds elapsed since the beginning of the
|
|
second in the *seconds* portion.
|
|
|
|
|
|
Using the NFB PMD
|
|
----------------------
|
|
|
|
Kernel modules have to be loaded before running the DPDK application.
|
|
|
|
NFB card architecture
|
|
---------------------
|
|
|
|
The NFB cards are multi-port multi-queue cards, where (generally) data from any
|
|
Ethernet port may be sent to any queue.
|
|
They are represented in DPDK as a single port.
|
|
|
|
NFB-200G2QL card employs an add-on cable which allows to connect it to two
|
|
physical PCI-E slots at the same time (see the diagram below).
|
|
This is done to allow 200 Gbps of traffic to be transferred through the PCI-E
|
|
bus (note that a single PCI-E 3.0 x16 slot provides only 125 Gbps theoretical
|
|
throughput).
|
|
|
|
Although each slot may be connected to a different CPU and therefore to a different
|
|
NUMA node, the card is represented as a single port in DPDK. To work with data
|
|
from the individual queues on the right NUMA node, connection of NUMA nodes on
|
|
first and last queue (each NUMA node has half of the queues) need to be checked.
|
|
|
|
Limitations
|
|
-----------
|
|
|
|
Driver is usable only on Linux architecture, namely on CentOS.
|
|
|
|
Since a card is always represented as a single port, but can be connected to two
|
|
NUMA nodes, there is need for manual check where master/slave is connected.
|
|
|
|
Example of usage
|
|
----------------
|
|
|
|
Read packets from 0. and 1. receive queue and write them to 0. and 1.
|
|
transmit queue:
|
|
|
|
.. code-block:: console
|
|
|
|
./<build_dir>/app/dpdk-testpmd -l 0-3 -n 2 \
|
|
-- --port-topology=chained --rxq=2 --txq=2 --nb-cores=2 -i -a
|
|
|
|
Example output:
|
|
|
|
.. code-block:: console
|
|
|
|
[...]
|
|
EAL: PCI device 0000:06:00.0 on NUMA socket -1
|
|
EAL: probe driver: 1b26:c1c1 net_nfb
|
|
PMD: Initializing NFB device (0000:06:00.0)
|
|
PMD: Available DMA queues RX: 8 TX: 8
|
|
PMD: NFB device (0000:06:00.0) successfully initialized
|
|
Interactive-mode selected
|
|
Auto-start selected
|
|
Configuring Port 0 (socket 0)
|
|
Port 0: 00:11:17:00:00:00
|
|
Checking link statuses...
|
|
Port 0 Link Up - speed 10000 Mbps - full-duplex
|
|
Done
|
|
Start automatic packet forwarding
|
|
io packet forwarding - CRC stripping disabled - packets/burst=32
|
|
nb forwarding cores=2 - nb forwarding ports=1
|
|
RX queues=2 - RX desc=128 - RX free threshold=0
|
|
RX threshold registers: pthresh=0 hthresh=0 wthresh=0
|
|
TX queues=2 - TX desc=512 - TX free threshold=0
|
|
TX threshold registers: pthresh=0 hthresh=0 wthresh=0
|
|
TX RS bit threshold=0 - TXQ flags=0x0
|
|
testpmd>
|