Entries by Raymond P. Burkholder

Because ovsdb is a generic database tool, some one may have build other schemas. But for a native open vswitch install, this is the default state:

# ovsdb-client list-dbs
Open_vSwitch
_Server

It is possible to add a tcp connection to gain access to the database:

# ovs-appctl -t ovsdb-server ovsdb-server/add-remote ptcp:6640
# netstat -antp |grep ovs
tcp        0      0 0.0.0.0:6640            0.0.0.0:*               LISTEN      746/ovsdb-server
tcp        0      0 127.0.0.1:57424         127.0.0.1:6633          ESTABLISHED 397/ovs-vswitchd

But on linux, there is an existing linux stream available:

# lsof|grep ovsdb|grep STREAM
ovsdb-ser   746                   root    5u     unix 0x000000001740f722        0t0   59425581 /var/run/openvswitch/db.sock type=STREAM
ovsdb-ser   746                   root   16u     unix 0x0000000021edd23e        0t0      12914 /var/run/openvswitch/db.sock type=STREAM
ovsdb-ser   746                   root   19u     unix 0x000000001b392a81        0t0      20287 /var/run/openvswitch/ovsdb-server.746.ctl type=STREAM

# ovs-vsctl list open_vswitch
_uuid               : 8cfeb1c0-34ca-4f91-a162-7e9aef141344
bridges             : [e65f5e97-36ec-45e3-b83e-64129dae61d3]
cur_cfg             : 65
datapath_types      : [netdev, system]
db_version          : "7.16.1"
dpdk_initialized    : false
dpdk_version        : none
external_ids        : {hostname="nuc8i7hvk01.burkholder.net", rundir="/var/run/openvswitch", system-id="02700cd5-96f5-4b4e-90b4-5e67fe06abf7"}
iface_types         : [erspan, geneve, gre, internal, "ip6erspan", "ip6gre", lisp, patch, stt, system, tap, vxlan]
manager_options     : []
next_cfg            : 65
other_config        : {}
ovs_version         : "2.10.1"
ssl                 : []
statistics          : {}
system_type         : debian
system_version      : []

# ovs-appctl dpif/show
system@ovs-system: hit:50661581 missed:207452
  ovsbr0:
    eno1 5/3: (system)
    enp5s0 2/1: (system)
    ovsbr0 65534/2: (internal)
    veth-nvpn-v90 6/6: (system)
    vlan110 4/5: (internal)
    vlan90 3/4: (internal)

# ovs-appctl dpctl/show -s
system@ovs-system:
  lookups: hit:50661616 missed:207458 lost:0
  flows: 6
  masks: hit:211426886 total:10 hit/pkt:4.16
  port 0: ovs-system (internal)
    RX packets:0 errors:0 dropped:0 overruns:0 frame:0
    TX packets:1428 errors:0 dropped:0 aborted:0 carrier:0
    collisions:0
    RX bytes:0  TX bytes:177489 (173.3 KiB)
  port 1: enp5s0
    RX packets:19439356 errors:0 dropped:0 overruns:0 frame:0
    TX packets:30374161 errors:0 dropped:0 aborted:0 carrier:0
    collisions:0
    RX bytes:10608677166 (9.9 GiB)  TX bytes:29950322693 (27.9 GiB)
  port 2: ovsbr0 (internal)
    RX packets:333032 errors:0 dropped:29829 overruns:0 frame:0
    TX packets:1428 errors:0 dropped:0 aborted:0 carrier:0
    collisions:0
    RX bytes:23060625 (22.0 MiB)  TX bytes:177489 (173.3 KiB)
  port 3: eno1
    RX packets:6037565 errors:0 dropped:0 overruns:0 frame:0
    TX packets:5719808 errors:0 dropped:0 aborted:0 carrier:0
    collisions:0
    RX bytes:4726094728 (4.4 GiB)  TX bytes:3277870100 (3.1 GiB)
  port 4: vlan90 (internal)
    RX packets:0 errors:0 dropped:274558 overruns:0 frame:0
    TX packets:0 errors:0 dropped:0 aborted:0 carrier:0
    collisions:0
    RX bytes:0  TX bytes:0
  port 5: vlan110 (internal)
    RX packets:0 errors:0 dropped:274101 overruns:0 frame:0
    TX packets:0 errors:0 dropped:0 aborted:0 carrier:0
    collisions:0
    RX bytes:0  TX bytes:0
  port 6: veth-nvpn-v90
    RX packets:5000388 errors:0 dropped:0 overruns:0 frame:0
    TX packets:5788139 errors:0 dropped:0 aborted:0 carrier:0
    collisions:0
    RX bytes:3055006377 (2.8 GiB)  TX bytes:4647166781 (4.3 GiB)

• Melexis at Mouser
• Melexis MLX90640 Far Infrared Thermal Sensor is a fully calibrated 32 x 24 pixel thermal IR array in a compact
• Excelitas CoolEYE IR Imagers
• AMG88x3,4 Grid-EYE® Sensor Datasheet
• 16 x 4 Infra-Red Array by McLaren Technologies

Sometimes, you just don't know what is on your network. A couple ways of finding out include using nmap or arp-scan.

An easy install:

sudo apt install arp-scan nmap

And an easy use with some basic defaults (the Unknown ones are LXC containers with manual mac addresses, and Cadmus Computer Systems is a VirtualBox device):

$sudo arp-scan --interface=vlan90 --localnet
Interface: vlan90, datalink type: EN10MB (Ethernet)
Starting arp-scan 1.9.5 with 256 hosts (https://github.com/royhills/arp-scan)
10.55.90.1      00:13:3b:0f:59:24       Speed Dragon Multimedia Limited
10.55.90.14     08:00:27:de:fc:9d       Cadmus Computer Systems
10.55.90.21     0a:00:55:90:00:ab       (Unknown)
10.55.90.22     0a:00:55:90:00:0c       (Unknown)
10.55.90.23     0a:00:55:90:00:23       (Unknown)
10.55.90.31     f0:ad:4e:03:64:7f       Globalscale Technologies, Inc.

8 packets received by filter, 0 packets dropped by kernel
Ending arp-scan 1.9.5: 256 hosts scanned in 3.155 seconds (81.14 hosts/sec). 6 responded

With nmap, it knows about the mis-named mac:

$ sudo nmap -sn 10.55.90.0/24
Starting Nmap 7.70 ( https://nmap.org ) at 2018-09-29 20:25 MDT
Nmap scan report for 10.55.90.1
Host is up (0.00032s latency).
MAC Address: 00:13:3B:0F:59:24 (Speed Dragon Multimedia Limited)
Nmap scan report for 10.55.90.14
Host is up (-0.10s latency).
MAC Address: 08:00:27:DE:FC:9D (Oracle VirtualBox virtual NIC)
Nmap scan report for 10.55.90.21
Host is up (-0.100s latency).
MAC Address: 0A:00:55:90:00:AB (Unknown)
Nmap scan report for 10.55.90.22
Host is up (-0.100s latency).
MAC Address: 0A:00:55:90:00:0C (Unknown)
Nmap scan report for 10.55.90.23
Host is up (-0.087s latency).
MAC Address: 0A:00:55:90:00:23 (Unknown)
Nmap scan report for 10.55.90.31
Host is up (0.00074s latency).
MAC Address: F0:AD:4E:03:64:7F (Globalscale Technologies)
Nmap scan report for 10.55.90.10
Host is up.
Nmap done: 256 IP addresses (7 hosts up) scanned in 3.31 seconds

With these specific commands, arp-scan may see things which nmap may not, ie, firewall that are blocking pings. nmap probably has a mode for scanning similarly to arp-scan.

IPv6 Transition and Coexistence IPv6 - only and IPv4 as - a - Service

You can use Jool for both 464XLAT and just NAT64.

Question:

Thanks... that is what I don't understand: Why is NAT64 such a difficult concept to put into routers and firewalls? We already do NAT with IPv4 just fine.

I feel like IPv6 adoption would be much faster if there was a transition mechanism other than dual stacking.

Think: Corporate offices. Rather than renumbering everything inside, they just turn on NAT64 and now they can begin a slow and controlled transition.

Answer:

This document, which is already in the IESG review, may answer your question: draft-ietf-v6ops-transition-ipv4aas

Also take a look into this one: draft-ietf-v6ops-nat64-deployment.

Remember that if your enterprise network has apps that use literals, or they don't support IPv6, you still need dual-stack in the LANs, but access IPv6-only is just fine.

Other comments:

It’s not s difficult concept but you need to remember NAT44 breaks stuff and NAT64/NAT46 breaks more stuff.

Dual stacking is SIMPLE. REALLY. Turn on IPv6 with the M bit set and configure the DHCPv6 server. If you don’t need that level of control of address assignments leave the M bit off. 99.99% of your machines will just add a second address to the interface without you having to do anything more.

Getting to IPv6 resources from IPv4 address is a *much* harder problem that getting to IPv4 resources from IPv6 which is what you are describing here with the “no renumber everything as they already have a IPv4 address” requirement. NAT64 allows IPv6 devices to get to legacy IPv4 servers. To allow IPv4 devices to get to IPv6 servers you need to map the IPv6 addresses you want to talk to in to a pool of IPv4 addresses and push that mapping to a NAT46 (not NAT64) device.

Go dual stack then, once IPv6 is stable, turn off IPv4 if you want to be single stacked. You are then no longer dependent on the services you want want to access continuing to be offered over IPv4. 464XLAT will only work as a stop gap for IPv4 clients while services are offered over IPv4. After ~20 years of IPv6 being available (Windows XP had IPv6 support and it was not the first major OS to have it) just turn on IPv6.

Some ideas from a mailing list:

quick and dirty:

jens@screen:~$ dig nanog.org @8.8.8.8 | grep "Query time"
;; Query time: 16 msec
jens@screen:~$ dig nanog.org @1.1.1.1 | grep "Query time"
;; Query time: 3 msec

Indeed. For instance, the delay depends wether the cache it hot or cold (measuring response time for an authoritative server is easier).

also could use ripe atlas ...

Which embeds clients for ICMP Echo, DNS, NTP, TLS, arbitrary TCP (with some hacks), and, with serious limitations, HTTP.

use dig -u to get microsecond resolution, e.g.

$ dig -u @131.111.8.42 nanog.org | grep time:
;; Query time: 611 usec

I recommend that eyeball networks don't run any external recursive server for optimal CDN performance. Yes, some CDNs support other methods, but not all. If not all do, then the requirement remains. See On Firefox Moving DNS to a Third Party

freeware GRC tool for benchmarking DNS performance

Upcoming changes to the Linux kernel may include add flow_rule infrastructure capabilities in TC's cls_flower dissector. Some of the flow matching performed by Open vSwitch could be / will be offloaded to the kernel. This capability then allows the flow matching to be off-loaded to hardware for specific match / action pairs. I'm not sure how this would integrate into OpenFlows's complicated sequential table evaluations. They are using the bcm_sf2 driver for intial hardware acceleration testing.

Another recently released patch set is the MPTCP v0.94.1 maintenance release based on Linux v4.14.70 which is "the Multipath TCP Kernel v0.94.1 has been released, containing important bug-fixes. This release is based on Linux v4.14.70".

From user land is the util-linux v2.33-rc1 which includes things like lsblk, mount/umount, fdisk, wipefs, blkid, ...........

2018/11/09 - Three ports, even though they are labelled as wan, lan0, and lan1, all three exist on the same Topaz switch. Layer 3 functions on each port will probably work, but trying to make something like Open vSwitch handle packet switching between ports will not work (unless the hardware offload works, which I doubt). The Topaz switch takes care of all mac-learning and line forwarding. The three ports are not independent of each other.

2019/01/08 - I believe the Esspressobin has a Marvell® Link Street®-88E6341 switch. The switch has a Linux DSA configuration. This means that the Linux IPRoute2 utilities shouild be able to set VLAN parameters for each port. By setting each port to different VLAN, or set of VLANS, then all traffic should be forwarded to the CPU for routing/switching. If using open vswitch, ingress/egress vlans can be remapped to a common bridge if so desired.

Back to the original article:

Following along the Software HowTo information page, I was able to:

• Build system requirements - check
• Build from source - Toolchain - check
• Build from source - Kernel - check
• Build from source - Bootloader - check
• Build from source - Buildroot - build failure (due to SSL include issues), instead, downloaded pre-built from BuildRoot 2015.11
• Boot from MicroSD card/USB stick - Buildroot - check, with one caveat that I had to rename the dtb file and remove _v5_ I believe.

With those steps, I was able to get a successful boot to ensure the board was functional. I used 'screen /dev/ttyUSB0 115200' as a console connection command to the EspressoBIN.

Ultimately, I want a Debian bootable system. So, for experiment #2.... GitHub armbian/build has the instructions for building from scratch. But I opted for instant gratification with downloading and installing a pre-built image. Even though the instructions at armbian espressobin are terse, they are accurate in terms of what is needed to get a (as of this writing) Debian Stretch v4.18.y kernel installed on EspressoBIN. The FAQ describes how to burn an image to an SDCard (in my case Etcher on Linux). I disable NetworkManager.service, NetworkManager-wait-online.service, systemd-networkd and systemd-resolved. I then manually adjust /etc/network/interfaces. Armbian Documentation talks about special optimizations as part of the build.

• EspressoBIN Block Diagram
• GPIO as a character device
• libgpiod
• New GPIO Presentation
• GPIO Device driver along with a collection of other embedded topics
• Driving an LED from an 1.8V GPIO pin that defaults to “IN”
• 88F3710 and 88F3720 - ARMADA® 3700 Family Single/Dual CPU System-on-Chip Hardware Specifications
• GPIO Map
• EspressoBIN GPIO Pins

Other related:
• Debian Arm64Port/AArch64 - port of Debian to the 64-bit Arm platform (arm64), known in some other places as AArch64, info up to Stretch
• InstallingDebianOnMarvellESPRESSOBin
• Run UBOS on ESPRESSObin - is a Linux distro that aims to make it 10x easier to run server-side apps for yourself, or for your family, on hardware or cloud servers you own.
• build/config/bootscripts/boot-espressobin.cmd - u-boot boot script for armbian
• ondemand governor does not scale up the cpu frequencies as expected.
• armbian build with governor
• I/O Related testing parameters
• govenor settings
• marvell dts file
• firmware?
• reference document of using crypto-safexcel engine in 3720 and 7K/8K SOCs.
• Command to recompile script into .scr format, cmd - armada-3720-espressobin.dtb is for the mainline kernel
• openssl crypt testing - must be kernel compile, no crypto module shows up in lsmod
• checking for crypto
• booting from sata
• armbian-config menu to build from beta
• ptp build
• build with custom options
• through put on various interfaces
• Marvell Linux Porting Guide
• Linux Kernel Crypto API
• armbian boot and os downloads
• Armada A388, A3700 forums
• Add support for the ethernet switch on the ESPRESSObin
• Marvell EBU:Mainline Linux - status of the support in the official Linux kernel for Marvell EBU processors.

Commands to try:

• armbianmonitor -m
• iperf3
• lspc sysbench --test=cpu --cpu-max-prime=20000 run --num-threads=$(grep -c '^processor' /proc/cpuinfo)
• isysbench --test=cpu --cpu-max-prime=20000 run --num-threads=$(grep -c '^processor' /proc/cpuinfo
• cat /proc/cpuinfo
• cpufreq-info
• lsblk
• hdparm -tT /dev/sda
• ethtool --offload eth0 rx on tx on sg on tso on gso on gro on
• apt install armbian-config
• ls /sys/class/leds/
• cat /etc/default/cpufrequtils
• ls -al /sys/class/net
• udevadm info -q all -p /sys/class/net/wan
• cat /proc/version
• systemd --v
• cat /sys/devices/system/cpu/cpu1/cpufreq/stats/trans_table
• cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies

From Armbian Ubuntu / Debian, be aware that:

manpage for systemd.netdev reads, in the part concerning the MACAddress= entry in the [NETDEV] section: If none is given, one is generated based on the interface name and the machine-id(5)

manpage for dbus-uuidgen, it says that the D-Bus machine-id should not be changed on a running machine or as they write: it will probably result in bad things happening. The machine-id should remain constant at least until next reboot.

• remove the SD card from the EspressoBin,
• mount it on a different Linux system
• delete /the/mounting/point/var/lib/dbus/machine-id
• generate a random new machine id using dbus-uuidgen –ensure=/the/mounting/point/var/lib/dbus/machine-id
• copy /the/mounting/point/var/lib/dbus/machine-id to /the/mounting/point/etc/machine-id
• put the SD card back to the EspressoBin
• use serial console to get the newly generated MAC address for br0
• use this MAC address in the dnsmasq configuration of my LeDe router

A few special projects to consider:

• TV headend (IPTV server)
• Syncthing (personal cloud)
• SoftEther VPN server (VPN server)
• Transmission (torrent server)
• ISPConfig (WEB & MAIL server)
• Openmediavault NAS (NAS server)
• PI hole (ad blocker)
• MiniDLNA (media sharing)

ust for the record: We have a SoC talking via RGMII to the onboard switch (currently with 1GbE but maybe 2.5GbE possible). We use DSA to tell the switch to not act as a switch on layer 2 anymore but separate downstream ports to get 3 individual interfaces just to bridge them again at the kernel layer above. Doing this at the DSA layer (telling the switch to be a dumb layer 2 switch accessible as eth0) might save CPU resources and result in better performance if I'm not wrong?

so it looks like the topaz switch needs to be talking to soc via SGMII instead of RGMII to achieve 2.5G.

..and I think i'm officially convinced that the max bandwidth between topaz switch and cpu is in fact 1gigabit... The connection is using a seperate RGMII lane. (eth0)

eth1 -- would be the interface capable of of SGMII, but the 3 fast lanes are occupied with USB, MiniPCI and SATA

• MTD: where the firmware and the u-boot environment are stored. The u-boot-tools package contains some utilities to read/modify the u-boot environment which is much more comfortable then in the interactive u-boot shell.
• pass the mtd partitioning information via the kernel command

Use HELIOS LanTest for file testing across network.

I/O Related Testing: "keep in mind that we've tested there 'advanced' stuff with mPCIe SATA controllers. On the onboard SATA port EspressoBin should in a single disk configuration easily exceed 500 MB/s (or in other words: Fast enough for any HDD imaginable) "

finding bottlenecks

beta sources

bootlin (formerly Free Electrons) - it was confirmed that support for the Marvell 3720 security engine is available in the crypto tree of 4.16-rc1 and that some known bugs will be fixed until the final release of 4.16.

I had a quick look over the diff for 4.16-rc1 and it looks like there's progress on support for DVFS too

These problems seem to arise (after some time) if your board is powered simultaneously by two sources with different GND potentials (check DC and AC). In this case your board will be exposed to severe electrical and thermal strain causing hardware issues after some time. It can be avoided if you access the serial console using a laptop that is not connected to a power supply itself (see usb/main power ).

IO Crest SI-MPE15047 4 Port Serial Mini PCIe Controller Card (RS-232) - could be used as console server.

IO Crest 4 Port SATA III Mini Pci-E Controller Card Components Other SI-MPE40125

M.2 form factor

Inside Secure, a SafeXcel EIP-97/197 cryptographic engine designed by Inside Secure, installed with 'modprobe crypto_safexcel'. More info at in a forum posting. The Kernel has some docs. Intelectual property sheet for EIP-97 - References IPsec, MACsec, SRTP, SSL, TLS, and DTLS.

From Reverse-engineering the ESPRESSObin’s GPIO pins:

From the iovisor-dev mailing list:

>I am trying to collect IPFIX flow data from the linux host interface.

Why IPFIX and not sFlow or netflow ?

>Can someone guide me the best way to collect the data using XDP.

It depends a bit on you setup. Assuming you want to do this "inline" on the box receiving the traffic. Then you should know/learn, that XDP cannot allocate a new packet (that e.g. could be used sending IPFIX/sFlow info directly). Instead, I would use the perf-ringbuffer to store sampled-packets (via copy), and then code a userspace program that reads from this perf-ringbuffer, and it will communicate with the central IPFIX/sFlow server.

>Any samples for reference will be a great help.

From XDP howto use the perf-ringbuffer via bpf_perf_event_output, samples are avail here:

• https://github.com/torvalds/linux/blob/master/samples/bpf/xdp_sample_pkts_kern.c
• https://github.com/torvalds/linux/blob/master/samples/bpf/xdp_sample_pkts_user.c

Notice, there are also plenty of BCC examples using the perf-ringbuffer, look for BCC code with:

   BPF_PERF_OUTPUT(events);
   events.perf_submit(ctx, data, sizeof(struct data_t));

• OCC Batch Processing Links - reports, summary, and raw data [OCC, founded in 1973, is the world's largest equity derivatives clearing organization]

• Functional Data Structures: (a flânerie by Prabhakar Ragde) [not really C++ though][uses OCaml]
• What can C++ offer as far as functional programming? from Stack OverFlow

• Symmetry Electronics has Powered Equipment POE modules by SilverTel

• Vitesse VSC73xx DSA router driver for a now obsolete Enhanced 5 + 1-Port Integrated Gigabit Ethernet Switch with Transceivers.
• swconfig from OpenWrt, goal of the switch configuration API is to provide a common basis to build re-usable and extensible switch drivers, developed in parallel with the kernel's own:
• Distributed Switch Architecture
• LWN: The rise of Linux-based networking hardware

• Get-Console: "world's smallest device server", very inexpensive as well
• OpenGear: a bit more sophisticated, higher in price as well

I found that pictures I captured and edited were printing somewhat dark. Time to investigate how to get a consistent photographic workflow. I use darktable for photo management and preliminary colour management. So, some links I have to come back to:

• DisplayCAL - Open Source Display Calibration and Characterization powered by ArgyllCMS
• Loading/Applying profiles in Xfce in the displaycal forums.
• Display color management in darktable
• My Free and Open Source Photography Workflow - darktable, gimp, Geeqie, ArgyllCMS
• X-Rite i1Display Pro and ColorChecker Passport Bundle at B&H Photo Video

• SAMA5D3 Xplained is a fast prototyping and evaluation platform for microprocessor-based design based upon the SAMA5D36 processor. It has a 100mbps and a 1gbps interface. Linux4SAM supplies the tooling to support the card. Linux mainline 4.9 with specific contributions. Features KSZ9031 and KSZ8081 PHYs with connectors. Could maybe replace with something like KSZ9893R 3 port Gigabit Switch or KSZ9896 6 port Gigabit Switch or KSZ9897S 7 port Gigabit Switch? A 100 <-> 1000 Software Bridge will use 95% cpu.
• Cyclone V SoC Multi-Port Ethernet Aggregator Board - NetLeap - NovTech's NetLeap, an Industry 4.0 Multi-Protocol Ethernet Ports Aggregator Platform, allows the development and integration of a variety of Ethernet protocols including PROFINET, EtherCAT, EtherNet/IP, Ethernet Powerlink, Modbus TCP, SERCOS III and more. This solution is preproduction ready. A total of six 1G/100/10 Ethernet ports are present. Two are connected to the HPS (The ARM* Cortes-A9 core of the Cyclone® V SoC) and four are connected to the FPGA fabric. With its six ports, the NetLeapâ„¢ allows different protocols to reside on the same platform and can be used as a protocol bridge, a switch, or a router. The kit comes with a templated project example that allows the board to boot to Linux and have all six ports work as a standard Ethernet.
• Networking Products - SG-1000 microFirewall Security Appliance
• Newport GW6400 - High Performance Single Board Computers for Networking with 4x Mini-PCIe Sockets, 5x Gigabit Ethernet Ports, and USB 2.0/3.0
• NFV PicoPod s a Cluster of Six MACCHIATOBin Networking Boards for OPNFV, ODP, DPDK and OPF
• SolidRun ClearFog Base is a $90 Router/Networking Board with USB 3.0, M.2, mSATA, and Gigabit Ethernet Support
• Orange Pi R1 - dual 10/100 ports for $14
• EspressoBin offers Marvell’s dual-core, Cortex-A53 Armada 3720 SoC, plus 2x GbE, 1x WAN, SATA, and mini-PCIe. GlobalScale shopping cart. Hack-A-Day leads to Building my ideal router for $50 (which talks about wireless, netflow, and security). ArchLinux|ARM has an install page. reddit says " 4.17 in particular ? It has hardware offload for the Topaz switch and the nic chip in the Espressobin which might reduce cpu usage", using a four drive setup as an example in the article. armbian has various add-on cards for mPCI and USB. arch linux ARM setup.
• Sinovoip’s “Banana Pi BPI-R2” router SBC gives you 5x GbE, WiFi, BT, 2GB RAM, 8GB eMMC, SATA, and mini-PCIe, plus a quad-core -A7 MediaTek MT7623N. More info on the gitbooks and the wiki and forum. Among other sites, can be ordered from AliExpress. Other info at armbian. Lots of links at end of BPI-R2. aluminum cases. BPI-4G LTE module.
• SolidRun’s “Marvell MacchiatoBIN” is a $349, Mini-ITX networking SBC that runs Linux 4.4 on Marvell’s quad -A72 Armada 8040, and supports ODP, OFP, and NFV. Main Site. I just wish Marvell would be more forthcoming on documentation on their chipsets rather than hiding behind NDA all the time. Pico MacchiatoBIN - with case, power supply, fan. Provided link to Marvell Prestar DX Switches which have VxLAN support. Suggests it works with OpenFlow 1.4 (open vswitch enabled?). arch linux ARM setup.
• EPIC-KBS9 from Aaeon is an SBC with 6th or 7th Gen Core S-series chips, 4x GbE ports, up to 32GB DDR3, and mini-PCIe and PCIe x4 expansion.
• minnowboard two port boards in the $200 range.

Other boards:

• BeagleBone Blue - mucho I/O: Motor control: 8 6V servo out, 4 bidirectional DC motor out, 4 quadrature encoder in; Sensors: 9 axis IMU (accels, gyros, magnetometer), barometer, thermometer
• FireFly - interesting add-ons like cameras, ASTRA Pro Structured Light Camera, binocular camera, capacitive touch screens, remote controls, ..
• ODroid
• ZedBoard: community-based site is dedicated to helping you jump-start your design with the Xilinx Zynq®-7000 All Programmable SoCs and UltraScale+ MPSoCs

Canada vendors:

• Mouser Electronics
• Newark element14
• Digi-Key
• Lees Electronic - Vancouver - electronic and mechanical
• Buy a PI - various SBC's

Embedded systems reviews:

• Linux Gizmos
• CNXSoft – Embedded Systems News

Embedded Linux Wiki

search engines for electronic parts

• Octopart
• digipart

Evolution to 'build your own"

• mikroETH 100
• Microchip's Ethernet Portfolio
• KSZ9031 Gigabit Ethernet Transceiver with Power Saving Feature
• AN1120: Ethernet Theory of Operation
• Segger physical connectivity example
• Microchip High Throughput Ethernet Interface Solutions
• 32 bit ARM with 2xGE
• 1G Ethernet PHYs by Microsemi (some have MacSec capability

Other linux SoC

• Linux for Amlogic Meson
• BORA Xpress XILINX Zynq CPU module

AVB

• Audio Visual Bridging - AKA IEEE 1588 / 802.1AS Time Stamping
• KSZ9567 7-Port Gigabit Ethernet Switch with AVB, IEEE 1588, SGMII/RGMII/MII/RMII
• KSZ9477 7-Port Gigabit Ethernet Switch with Ring Redundancy, AVB, IEEE1588, SGMII/RGMII/MII/RMII

• Awesome Electronics - github curated links

2018/10/20

• PC Engines - an international dealer network, boards with 1, 2, 3, 4 ethernet ports
• board db - SoC search engine

2018/12/24 - not networking, but clustering - need to find a multiport switch

• Pine64