Networking with Torizon OS
Introduction
Networking with Torizon OS can refer to different topics:
- Configuration of the host network, not directly related to containers.
- Configuration of networking on a container, and the relationship between the container and the host networks.
- Configuration of inter-container networking, often with the purpose of multi-process communication using the network stack (e.g. REST API).
The first part of this article explains about host network configuration: the Torizon OS image currently provides NetworkManager, a program that provides detection and configuration for the system to automatically connect to networks.
The second part of this article explains about container network configuration and how to share a network between containers using docker-compose.
Ethernet Interface Naming on Torizon OS
Torizon OS Ethernet interfaces are always named ethernetX
, being X a number starting from 0, for instance, ethernet0
, ethernet1
and so on.
This article complies with the Typographic Conventions for Torizon Documentation.
Prerequisites
In order to take full advantage of networking with Torizon OS, see the articles:
- Configure Build Environment for Torizon Containers
- Operating Toradex Wi-Fi/BT Capable Modules Using Dual and Single Antenna Configuration (to set the antenna and use Wi-Fi properly)
Host Configuration: NetworkManager
The nmcli is a command-line client for NetworkManager. You can show the status of your network devices, detected by NetworkManager:
# nmcli device
Show our available connections and devices, on which the active connection is applied to:
# nmcli connection show
To disconnect from a network:
# nmcli con down id '<Connection_name>'
To delete a connection:
# nmcli con delete '<Connection_name>'
Static Network Configuration
If you looking for a way to configure a Static Network Configuration, nmcli provides the following commands:
# nmcli con mod '<Connection_name>' ipv4.addresses "<desired IP/mask>"
# nmcli con mod '<Connection_name>' ipv4.gateway "<desired gateway>"
# nmcli con mod '<Connection_name>' ipv4.dns "<DNS server 1>,<DNS server 2>"
# nmcli con mod '<Connection_name>' ipv4.method "manual"
After running the commands above, you can visualize your entire network configuration by opening the <connection-name>.nmconnection
file:
# cd /etc/NetworkManager/system-connections/
# sudo cat <connection-name>.nmconnection
Expected file output:
[connection]
id=<connection-name>
uuid=a690e7e8-a413-331d-830d-d0df5bad3983
type=ethernet
autoconnect-priority=-999
permissions=
timestamp=1581530428
[ethernet]
mac-address=00:14:2D:63:47:64
mac-address-blacklist=
[ipv4]
address1=<board-ip>,10.0.0.1
dns-search=
method=manual
[ipv6]
addr-gen-mode=stable-privacy
dns-search=
method=auto
After the changes were made, do not forget to reload the configuration file:
# sudo nmcli connection reload
Dynamic Network Configuration
Along with Static Network Configuration, ncmli provides a way to configure a dynamic connection:
# nmcli con mod '<Connection_name>' ipv4.method "auto"
Other nmcli Commands
You must read the nmcli man page, either running man nmcli
on a computer with nmcli installed or Googling after it. For quick reference, man --help
is also useful.
Wi-Fi
Torizon OS supports two Wi-Fi modes: client mode and access point (AP) mode.
Wi-Fi Client Mode
This mode is used when you want a Torizon OS Device to connect to a Wi-Fi access point.
To see a list of available Wi-Fi access points:
# nmcli device wifi list
To connect to a Wi-Fi access point:
# nmcli -a device wifi connect <WIFI_NAME>
Share internet over ethernet
Now that your device is connected to a Wi-Fi access point, you might want to share the internet to other devices over ethernet.
Note the Wi-Fi interface name(
<WIFI_IFACE_NAME>
) that NetworkManager used to connect to the Wi-Fi.# nmcli connection show --active | grep wifi | awk '{print $4}'
Activate a DHCP server on the ethernet interface(
<ETH_IFACE_NAME>
). To do that, you can leverage systemd's built-in DHCP server support, creating the file/etc/systemd/network/42-ethernet0.network
with the following content (substitute<ETH_IFACE_NAME>
,<IPV4_ADDR>
,<IPV4_ADDR_NETMASK>
,<DHCPD_POOL_OFFSET>
,<DHCPD_POOL_SIZE>
and<STATIC_DNS_ADDRESS>
accordingly):infoYou can refer to Systemd manpages for more configuration possibilities and details.
/etc/systemd/network/42-ethernet0.network[Match]
Name=<ETH_IFACE_NAME>
[Network]
Address=<IPV4_ADDR>/<IPV4_ADDR_NETMASK>
DHCPServer=yes
[DHCPServer]
PoolOffset=<DHCPD_POOL_OFFSET>
PoolSize=<DHCPD_POOL_SIZE>
EmitDNS=yes
DNS=<STATIC_DNS_ADRESS>Now just restart the
systemd-networkd
service:# sudo systemctl restart systemd-networkd
At this point, any device connected to the ethernet will be able to lease an IP address from the DHCP server.
Setup Packet Forwarding and NAT to give internet access to the device connected over the ethernet.
As this is a reverse scenario of Wi-Fi hotspot mode, we can follow the same steps described under the Wi-Fi hotspot mode and just reverse the NAT interfaces like below:(substitute
<ETH_IFACE_NAME>
and<WIFI_IFACE_NAME>
accordingly):# sudo iptables -t nat -A POSTROUTING -o <WIFI_IFACE_NAME> -j MASQUERADE
# sudo iptables -A FORWARD -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT
# sudo iptables -A FORWARD -i <ETH_IFACE_NAME> -o <WIFI_IFACE_NAME> -j ACCEPTNow the device connected over ethernet should be able to access the internet.
Wi-Fi Access Point Mode
This mode is used when you want a Torizon OS Device to act as a Wi-Fi access point.
This section is based on the BSP Access Point mode.
- Verify that NetworkManager is not managing the uap interface. Check for the file
/etc/NetworkManager/conf.d/99-disable-uap.conf
with the following content:
[main]
plugins=keyfile
[keyfile]
unmanaged-devices=interface-name:uap*
- Set the Access Point interface IP and SSID in Hostapd configuration file. Using 2.4GHz WLAN whit WPA2 security protocol. Change the following variables in
/etc/hostapd.conf
:
interface=uap0
ssid=<WIFI_SSID>
hw_mode=g
channel=1
own_ip_addr=<IP_ADDR>
wpa=2
wpa_passphrase=<WIFI_PASSWORD>
An alternate hostapd.conf
file that creates a WLan in the 5GHz range, protected with WPA2:
interface=uap0
ssid=<WIFI_SSID>
hw_mode=a
channel=40
ieee80211n=1
own_ip_addr=<IP_ADDR>
wpa=2
wpa_passphrase=<WIFI_PASSWORD>
country_code=<CC>
Replace <WIFI_SSID>
, <IP_ADDR>
, <WIFI_PASSWORD>
and <CC>
accordingly. In order to complete your file with the correct country code, please take a look at Country Codes. The hostapd.conf
is documented in the default hostapd.conf
file that is deployed, as well as here: https://w1.fi/cgit/hostap/plain/hostapd/hostapd.conf
To provide the Access Point interface with a static IP address and DHCP, it is necessary to create the file
/etc/systemd/network/80-wifi-ap.network
:/etc/systemd/network/80-wifi-ap.network[Match]
Name=uap0
Type=wlan
WLANInterfaceType=ap
[Network]
Address=<IPV4_ADDR>/<IPV4_ADDR_NETMASK>
DHCPServer=yes
[DHCPServer]
PoolOffset=10
PoolSize=30
EmitDNS=yes
DNS=<STATIC_DNS_ADRESS>Replace
<IPV4_ADDR>
,<IPV4_ADDR_NETMASK>
and<STATIC_DNS_ADRESS>
accordingly. This also automatically sets up forwarding and the iptables masquerading rules, also including the setup of the AP sharing interface with a DHCP server. You can find more information and examples on the systemd-networkd documentation.Enable and start hostapd:
# sudo systemctl enable hostapd
# sudo systemctl start hostapdChange the hostapd permissions:
# sudo mount -o remount, rw /usr/
AP will not survive a reboot.
The Hostapd service starts by default after the network target. This target is loosely defined and means "start after the network stack is up". That service will fail because the network device (uap0 by default) is not up yet, or not every time. To fix this we can tell systemd that the hostapd.service file depends on a device, named sys-subsystem-net-devices-uap0.device on Torizon OS. To make it persist do the following steps:
Edit
hostapd.service
located in/usr/lib/systemd/system/
to execute after the desired interface (uap) is created and add these lines:hostapd.service[Unit]
Description=Hostapd IEEE 802.11 AP, IEEE 802.1X/WPA/WPA2/EAP/RADIUS Authenticator
BindsTo=sys-subsystem-net-devices-uap0.device
After=sys-subsystem-net-devices-uap0.device
Now you can connect on your access point. If you are willing to have internet access through internet sharing, keep following this guide.
To establish an internet connection, create the
/etc/sysctl.d/30-ipforward.conf
file, as follows:/etc/sysctl.d/30-ipforward.confnet.ipv4.ip_forward=1
net.ipv6.conf.default.forwarding=1
net.ipv6.conf.all.forwarding=1Load the kernel modules to deal with IP tables, NAT and COnntrack:
# modprobe ip_tables
# modprobe nf_nat
# modprobe nf_conntrackEnable NAT using the
iptables
command.# sudo iptables -t nat -A POSTROUTING -o ethernet0 -j MASQUERADE
# sudo iptables -A FORWARD -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT
# sudo iptables -A FORWARD -i uap0 -o ethernet0 -j ACCEPTSave the current configuration on the module. That requires root permissions.
# sudo su
# cd /etc/iptables
# iptables-save > /etc/iptables/iptables.rules
#exitCreate the
etc/systemd/iptables.service
file to manage the iptables.etc/systemd/iptables.service[Unit]
Description=IPv4 Packet Filtering Framework
Before=network-pre.target
Wants=network-pre.target
[Service]
Type=oneshot
ExecStart=/usr/sbin/iptables-restore /etc/iptables/iptables.rules
ExecReload=/usr/sbin/iptables-restore /etc/iptables/iptables.rules
RemainAfterExit=yes
[Install]
WantedBy=multi-user.targetReboot the module to apply the changes.
# sudo reboot
Reload the systemd services and enable iptables:
# sudo systemctl --system daemon-reload
# sudo systemctl enable iptables
Wi-Fi Hotspot Mode
To configure a hotspot, first make sure to follow the steps at the section Wi-Fi Access Point Mode. Then, proceed with the next instructions.
Once AP Mode is configured, it is a common use-case to share an internet connection from another interface (hotspot), when you want your system-on-module with Torizon OS to act as a Wi-Fi hotspot. This section relies on the Internet Sharing article from the Arch Wiki and on BSP internet sharing mode.
Enable the Packet Forwarding settings by editing the
30-ipforward.conf
with the following content:/etc/sysctl.d/30-ipforward.confnet.ipv4.ip_forward=1
net.ipv6.conf.default.forwarding=1
net.ipv6.conf.all.forwarding=1
The changes will take effect after a reboot.
Use
modprobe
to load the kernel modules related to IP tables, NAT and Conntrack.# modprobe ip_tables
# modprobe nf_nat
# modprobe nf_conntrackUse
iptables
to enable NAT:# sudo iptables -t nat -A POSTROUTING -o ethernet0 -j MASQUERADE
# sudo iptables -A FORWARD -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT
# sudo iptables -A FORWARD -i uap0 -o ethernet0 -j ACCEPTSave the current config to
/etc/iptables/iptables.rules
:# sudo su
# mkdir /etc/iptables/
# iptables-save > /etc/iptables/iptables.rules
# exitCreate a systemd service file iptables.service. The example is modified from the iptables package from Arch Linux:
/etc/systemd/iptables.service[Unit]
Description=IPv4 Packet Filtering Framework
Before=network-pre.target
Wants=network-pre.target
[Service]
Type=oneshot
ExecStart=/usr/sbin/iptables-restore /etc/iptables/iptables.rules
ExecReload=/usr/sbin/iptables-restore /etc/iptables/iptables.rules
RemainAfterExit=yes
[Install]
WantedBy=multi-user.targetReload systemd services and enable the new service:
$ sudo systemctl --system daemon-reload
$ sudo systemctl enable iptables
Extra Resources
VPN
It's possible to configure a VPN tunnel in Torizon OS using WireGuard. In order to do this, please follow the instructions described in How to Use VPN on Torizon OS.
ifupdown Plugin
Torizon OS supports the NetworkManager's ifupdown
plugin. This plugin makes it possible to configure the network using a /etc/network/interfaces
file. For more information on how to use this plugin, please check the official NetworkManager documentation and the NetworkManager.conf manpage.
Production Release
After you make the changes to your device, you can use the command isolate
from the TorizonCore Builder Tool to generate your custom Torizon OS image for the Toradex Easy Installer. To learn how to do it, please refer to the article Capturing Changes in the Configuration of a Board on Torizon OS.
Networking Inside Docker Container
This section is a brief introduction on how to use different network configurations inside a Docker container. You must also refer to the Docker Networking documentation, which is a comprehensive source of information.
Show the list of networks:
# docker network ls
Inspect network to see what containers are connected to it:
# docker network inspect <NETWORK_NAME>
Network drivers:
Bridge (containers communicate on the same Docker host)
Host (uses the host's networking directly)
Overlay (when containers running on different Docker hosts to communicate)
Macvlan (when you need your containers to look like physical hosts )
None
3rd-party- network plugins
Bridge
When you run a new container, it automatically connects to the bridge network. A private network internal to the host is created in order to provide communication to the containers.
Create a user-defined bridge network:
# docker network create --subnet=<172.18.0.0/16> <NETWORK_NAME>
Create a container connected to our user-defined network:
# docker run --name <CONTAINER_NAME> -d --net <NETWORK_NAME> <IMAGE_NAME>
Specify the IP to a container and publish port 80 in the container to port 8080 to allow connections from other machine on the network :
# docker run --name <CONTAINER_NAME> -d --net <NETWORK_NAME> --ip <172.18.0.5> --publish <8080>:<80> <IMAGE_NAME>
Connect a running container to a network:
# docker network connect <NETWORK_NAME> <CONTAINER_NAME>
Macvlan
Macvlan driver can be configured in different ways. The advantage is to use the newest built-in and a lightweight driver, allowing the container to connect directly to host interfaces.
Create a macvlan network:
# docker network create -d macvlan --subnet=<172.16.86.0/24> \
--gateway=<172.16.86.1> -o parent=<ETHERNET_INTERFACE> \
<NETWORK_NAME>
Attach the container to the macvlan network:
# docker run -dit --network <NETWORK_NAME> \
--name <CONTAINER_NAME> <IMAGE_NAME> /bin/bash
Docker Networking Drivers Use Cases
To understand more about Docker networking drivers and which one is more advised to use on your application, please take a look at Understanding Docker Networking Driver Use Cases (archived).
Docker Network Using Docker-Compose
When you start your application, Docker Compose sets up a bridge network by default. Each service connects to the network, which makes them reachable with each other.
You can create your own networks to provide isolation and more options:
services:
app1:
image: app
networks:
- frontend
app2:
image: app
networks:
- frontend
- backend
app3:
image: app
networks:
- backend
networks:
backend:
# here you can configure your network
frontend:
App2 is connected to frontend and backend network, so it can communicate with app1 and app3. App1 and app3 can't communicate with each other, because they are on separate networks.
Connect to the external network:
networks:
default:
external:
name: <pre-existing-network>
Docker compose looks for the pre-existing-network.
For more information about, please take a look at Docker Compose Documentation.