Re-purpose your $30 Pocket-Router as an AWS-IoT-Edge-Device

 

This blog is about upgrading default firmware of GL.iNet-MT300N-V2 pocket router with a custom built  Openwrt-Linux Image which includes a AWS-IoT-Device-SDK and aws-iot-pubsub-demo application.

With this demo image, you can turn your low cost Gl.iNet-MT300N-V2 pocket router into an AWS-IoT-Edge device which can feed local sensor data to the AWS-Cloud and it opens up many interesting use cases for developing various IoT applications.

As an average user, here are 3 easy steps which can turn your low cost pocket router into an AWS-IoT -Edge-Device,

1. Overwrite OEM firmware with custom gl-mt300nv2-awsiot-demo.bin
2. Using webUI, upload AWS-Cloud generated Device-certificate/Private-Key and set the Endpoint.
3. Reboot the device and watch Hello-World messages being published to your Aws-IoT-Core.

Step-1: Prepare the Setup as shown in this picture below and Power-ON(wait for LED to stop blinking)

Step-2: On your Browser PC, open URL: http://192.168.8.1/cgi-bin/luci/admin/system/flashops (or you may be guided through a different onboarding process of GL.iNet's webUI)

Step-3: Find the sw upgrade menu on the webUI and upgrade the OEM sw with gl-mt300nv2-awsiot-demo.bin (make sure to disable "keep-settings" menu - we want to start with default settings)

Step-4: Wait for about 2minutes(till LED's on pocket router stops blinking)

Step-5: On you PC, Remove the network cable and re-plug so that your PC gets new IP starting with 192.168.20.x

Step-6: On your Browser PC, change the ULR to http://192.168.20.1 and you shall see the following:

Linux-PREEMPT_RT Scheduler latency measurement with Oscilloscope

For those exploring Realtime capabilities of Linux using Preempt-RT patch, Here is a bootable sdcard-image for Beaglebone-Black(or Green) with fully preemptible kernel. This image includes all required libraries/utils/binaries and kernel-debug-capabilities required for evaluating Linux-Preempt-RT behavior. As a bonus, this image includes a gpio-test-program that mirrors an input signal to a gpio-output pin for measuring the scheduler latency using an Oscilloscope(Thanks to Maxime Chevallier of Bootlin for providing this demo-utility during PREEMPT_RT Training).

Setup Diagram: 

Required Items:

1. Beaglebone-Black or Green
2. +5V adapter
3. micro-sdcard(1Gb or higher)
4. Programmable-signal-generator-board
5. 2-Channel-Oscilloscope
6. Jumper-wires

Instructions:

1. Prepare the setup as shown above(As a starter, set signal generator frequency to 100Hz with 50% duty)
2. Download sdcard-beaglebone-preemptrt-demo.img.xz (image size is just under 25MB)
3. Write sdcard-beaglebone-preemptrt-demo.img.xz to micro-sdcard using Balena-Etcher-Application.
4. Remove the micro-sdcard from your PC and insert in Beaglebone and power ON the +5V
5. After Linux boot, you will be able to see two square waves with a small phase difference between them(yellow waveform is the input to beaglebone and blue waveform is output from the beaglebone).
6. Phase difference between yellow and blue signal indicates the latency introduced by scheduler while scheduling preemptrt-gpiotest program
7.  On startup, /etc/init.d/S99PreemptrtGpioTester invokes "chrt -r 99 /usr/sbin/preemptrt-gpiotest gpiochip0 28 gpiochip0 17 h p" - which means starting preemptrt-gpiotest utility with priority 99. 

What is the purpose of this image?

For a quick evaluation of Linux-PREEMPT_RT behavior(or as a training material) one case use this readymade image to avoid long build process from sources using buildroot.

Using this image, you can study and tune your application to have deterministic timing behavior of the Linux scheduler at different load conditions. preemptrt-gpiotest program included in this image, mirrors a gpio-input to an output and the phase difference between these two signals gives an idea on how fast scheduler is able to give the required context to your application and the jitter(when measured with persist mode of the scope) on the output signal indicates the best and worst case scenario of the scheduler latency. Following picture shows best and worst case latency of the scheduler when subjected to different load on the hw(e.g network-traffic/interrupts/etc). For further details on this topic, check here to access Bootlin's PREEMPT_RT training material.

How can I build this bootable Linux-Image for beaglebone from sources?

Detailed build instructions are available here on my git-repository.

How do I modify and run my customized preemptrt-gpio utility on this beaglebone's target Linux?

Cross-compile instructions are available here on my git-repository.