Raspberry Pi WLAN setup(berrylan.org) |
Raspberry Pi WLAN setup(berrylan.org) |
Long story short:
1. RPi broadcasts via BLE the wifi networks it can see
2. A ReactNative App lets the user choose the wifi network to connect to, pre-selected is the wifi network which the mobile is connected to
3. User enters wifi password
4. App sends it to RPI, still via BLE
5. RPi responds with IP address when successfully connected or with an error, all via BLE.
6. App can now talk to Pi via IP
7. BLE server on Pi shuts down. Done.
Advantage: user never needs to mess with changing to any ad-hoc wifi. Awesome UX.
This is a very similar situation to being able to access the device on the LAN over tcp/ip and needing to pass it a secret.
I own a company and we are currently developing a hardware tool for attendance tracking at child care centres. We use the Pi and our plans for shipping these things was to ask our customer what their SSID and password were ahead of time and hard code them in. This tool could solve our problem. Additionally, if the user changes their SSID or password at any time, the Pi won't be able to connect. We use a tool called Dataplicity to remote into the Pi and could change the password ahead of time, however if the user changes the password before we can do that, we lose all ability to connect to the unit.
I wrote exactly the same thing as a library, to allow user setup of the WiFi network from within an app as part of a hardware prototype - main difference being it was just a Go daemon running on the Pi. Didn’t end up using it or open-sourcing it but I will re-visit that, since it looks like there are a couple of use-cases.
https://www.raspberrypi.org/documentation/configuration/wire...
With that, those whose networks require whitelisting by MAC address who are trying to set up a headless Pi could boot it once to get it to write its MAC address to the SD card, whitelist that MAC address and assign it a static IP address, create /boot/wpa_supplicant.conf, and boot again to get the Pi on their network.
I'll definitely give this a try.
Code is highly modular and maintainable.
https://blog.thewalr.us/2017/09/26/raspberry-pi-zero-w-simul...
Prior to stretch the wpa_supplicant.conf file needed to be modified in the main Raspbian partition which is inaccessible on the unix machine without setting up a virtual linux box and creating a vb disk image linked to the sd card. It was a bit of a pain.
With Stretch you can just add a wpa_supplicant.conf file and an empty ssh file to the boot partition which is accessible and you should be ready to go.
This still solves the pain point of there being lots of outdated headless instructions out there telling you to modify wpa_supplicant.conf in /etc/wpa_supplicant on the main partition. Would have saved me a ton of time that's for sure!
Looks nice and clean as well, design wise.
Typically I actually just mount the partitions and chroot with qemu-arm, upgrade packages, and do all of the setup before the SD card is ever inserted in a real Pi. proot[0] makes all this a walk in the park.
mount /dev/sdX2 /mnt
mount /dev/sdX1 /mnt/boot
proot -q qemu-arm -S /mnt
> Download the BerryLan flavoured Raspbian image and flash the SD Card. We recommend Etcher to do that.
I’m curious. Why would anyone recommend people download and run a multi-100mb electron app instead of just using dd?
Especially for projects like this aimed at medium technical people.
My first idea was to put an OLED and a rotary encoder on the box. That way they can do a special dance on the button to enter setup mode, scroll through a list of wifi networks, and then tediously scroll through letters to enter the wifi password. I found a 1" OLED for cheap, and already needed a button for other things, so this was a cheap, effective solution.
But having to tediously type out a password with a knob sounded very user unfriendly.
Then I realized I could bring out a USB port, and they could plug in a USB keyboard! So I worked on that for bit. It's a lot easier to use. And the whole keyboard + tiny OLED idea was kinda fun.
But then I remembered the idea of doing a WiFi AP dance. That's where the device broadcasts as an AP, you connect with your phone, browse to a webpage to tell the device your Wifi network's info, and done! It's a solution that doesn't require an OLED or for me to bring out a USB connection. Cheaper and less mechanical engineering.
To that end I found Mozilla's IoT project which, conveniently, has RasPi compatible code for doing just that dance: https://github.com/mozilla-iot/gateway-wifi-setup (By the way, Mozilla's IoT project is neat and worth checking out).
That setup is quite similar to OP, except it doesn't require a separate app. Its downside is requiring the user to mess with the wifi settings on their phone. Not a huge deal (for my target audience) but can be annoying. (Hence why OP's project uses the Bluetooth connection; Smart!)
I've been tinkering with this on and off for about two weeks. While I wasn't thrilled with the solution, it was at least usable, cheap, and easy. ... then I realized I was dumb.
I have a camera on the device. I could just have them go to a website, punch in their wifi info, encode that as a QR code, and have them show it to the camera... (It'd be a secure website I set up, so no security hygiene problems). Same benefits of being cheap and easy, while being easier to use. Using `pyzbar` to do real-time QR detection I had something working within an hour. Just hold your phone in front of the camera and it'll automatically detect, decode, and setup the wifi settings.
The only downside to this approach is that I can't conveniently display a list of nearby wifi networks; so they have to type their SSID manually.
Figured I'd share my comedy of errors in case someone is looking for ideas for wifi setup solutions. Obviously if you're just setting up a RasPi yourself it's easier to just pop a wpa_supplicant.conf on the SD card. (Though I wish Raspbian also had a way to configure hostname, authorized_keys, and disable ssh password authentication through boot files. Only way to do that today is by mounting the root filesystem which isn't convenient on, e.g., a Mac)
You could make the same argument for other assisted headless installations: speakers, smart plugs, etc.
That's true, but in those devices can be isolated from the internet by a firewall. Berrylan can't be, or it loses its utility. In order to talk to the app, Berrylan has to send packets to and receive packets from the internet. Also, even after the setup, most RaspPi applications entail having the computer continue to have access to the internet. So the situation is much risker with the RaspPi.
https://github.com/RPi-Distro/pi-gen/pull/207
The minor downside is that you have to build the image yourself, but fortunately that is a single command once docked is installed.
I don't even understand the idea that it could be bad for newbies to maybe mess up. The whole point of the RPi was to give students a small, cheap, easy to repair computer with low-level features to learn with.
https://github.com/cimryan/teslausb
(Props to Cim and Ray who are doing a ton of work making the project go, also!)
The goals/advantages, vs just using a USB thumb drive:
- will connect to your wifi network (when detected) and copy off your saved dashcam files (over CIFS, SSH, others) - fixes any FAT filesystem errors on the virtual USB drive (presented to the host using g_mass_storage) automatically; some Tesla software versions had/have a bug that left the filesystem dirty on shutdown, rendering some thumb drives unreadable until formatted again
- (in process testing) uses BTRFS to keep the drive online while taking snapshots to do the file copies (and other neat things I have planned, like extending the 1-hour rolling buffer that Tesla uses by copying off files older than 1 hr into a secondary backing store on disk to be copied off later)
- automated (headless) setup by putting a configuration file in /boot (including wifi setup)
- quite a few more ideas to come; love to hear more brainstorms too. Maybe with a little more work it's worth its own Show HN? :)
I've learned a lot about how little I remembered about bash, how Raspbian images are built, setting up Jenkins for automated builds, even controlling the LED on the Pi to show progress of headless setup. (Sometimes it's the little things...)
On the to-do list was something exactly like this post: provide a way to get to the files if you are away from your home network, and/or don't have a computer with a USB or SD reader handy. Say, if you had just gotten in an accident and wanted to get to the files pronto. So, a way to trigger the Pi popping into AP mode was my next area of exploration; timely! I was actually exploring having your phone make a BT connection that would cause the host to switch configs (either automatically, or using BT PAN. (And a web UI to grab files, make settings changes, etc.) Good thoughts on security in some of the comments, and glad to see others have thought through this.
Thanks to OP and others; I think there's a solution one way or another in several of these project ideas! And, as customary, PR's welcome :)
It gives you the ability to basically control what your Pi does (provisioning included) via an API.
Just started using rpi with my underwater robots. Obviously the wifi cuts off when they dive, but doesn't reconnect always when they come back. Will definitely try this out.
[1] https://www.raspberrypi.org/documentation/configuration/wire...
On the other hand, I could see this being useful as an additional software package that gets run when the Pi can't find or join its usual network.
* Allows usage from laptops
* No app to install
This could be integrated to several stock pi images from other projects, and I think that's where this really has potential. There's WiFiManager [1] for the esp8266, but this is the first time I came across something similar for the pi.
Just saying perhaps including the word "Bluetooth" once in the description would help. Looking at this page the first time it wasn't so obvious what was going on or what it does.
Why would you do that when you can just drop a wpa_supplicant.conf on /boot?
I could be wrong, but I am fairly sure these phones only have a single Wi-Fi module.
[1] https://www.androidpolice.com/2018/11/03/pixel-3-3-xl-suppor...
Another use is to prevent accidents. Suppose you have multiple networks. For example, at work we have separate networks for machines that handle different kinds of sensitive data, with each network having the appropriate incoming and outgoing access rules to protect the kind of sensitive data on that network.
In such a situation, MAC whitelisting can protect against someone accidentally plugging a machine into the wrong network. Someone goofs and plugs a machine with customer financial data into a network that has unrestricted outside internet access? It's not on the whitelist for that network, doesn't get an IP address, and they quickly notice something is wrong.
In a home WiFi setting, MAC whitelisting could be used for partial sharing of your WiFi with people who don't know much about networking. Maybe I want to allow my neighbor, who has no idea what a MAC address is, to use my WiFi with his phone, but not to use it with his other devices. I could give him the access info and then use MAC whitelisting to just allow his phone on.
Finally, even if MAC whitelisting is not being used it would still be nice to be able to find out the Pi's MAC address easily, so that you can assign it a static IP address on your WiFi network.
Yes, you can find it out by letting it get a dynamic address on your network, going into the router settings and looking at the device table and figuring out which unexplained device is the Pi, and noting the MAC address.
Then you can configure the router to assign that MAC address a static IP address. In a world where consumer router firmware didn't often suck that would be fine. I've run into routers that once they have assigned a dynamic IP to a MAC address make it difficult to assign it a static IP address.
It's a lot easier if you can do the static IP assignment in the router before the Pi has ever connected to that network.
I’ve had this problem multiple times and find this approach really nice (and wish it shipped with raspbian!)
Edit: forgive me, I see that's not what you meant at all. Or maybe it was? Either way, the point is valid.