Would like to help out: e-ink support, 4.4 Super Long Term Support Kernel

[lvw5264]

I noticed your post on mobileread after googling for mainline linux kernel deployment on a Kindle Paperwhite 3, and it is awesome that you've already got Fedora working... just without e-ink display working. It's a lot easier for people to find these sorts of things on Github rather than random old forum posts so I'd like to help organize an effort here.

So I'd be willing to help with building a mainline kernel for your Kindle Paperwhite 3, or on another ereader, I have the Kobo Aura H2O myself, and a Kindle Paperwhite 1. In particular there's efforts like the following to make these readers more open, such as the following:

https://hackaday.com/2018/07/25/hope-xii-a-foss-operating-system-for-e-readers/

https://libreplanet.org/wiki/Group:Hardware/research/e-readers

Almost all ereaders use Freescale ARM CPUs Freescale CPUs should theoretically have a high degree of mainline Linux kernel support thanks to the manufacturer's aim for long term support and such. However, that does not really extend to the manufacturer's implementation of the e-ink drivers of course, those are just tied to the fire and forget configuration they have there.

One prime motivation for me is just to have some sort of Linux terminal displaying to the screen on these relatively cheap e-ink devices. This way, instead of buying a $500 HDMI e-ink display, multiple cheap ereaders could make up one big e-ink screen in aggregate. However, X Windows is not really ideal for e-ink screens, which is why these manufacturers develop all these other solutions to display to them. Therefore, users in the ereader homebrew community have put together these open source libraries to help display to e-ink screens:

https://github.com/NiLuJe/FBInk

https://github.com/kisonecat/fbpad-eink

Regarding open source eink reader software, there's loads.

https://github.com/NiLuJe/koreader

And depending on what we figure out, we may not necessarily have to use these proprietary encumbered ereaders, as there are efforts to make fully open source ereader hardware. I've made one of these, works decently enough though could really use a bigger chip to run Linux like the Giant Feather board:

https://hackaday.io/project/168761-the-open-book-feather

There's also this thing with a bigger display:

https://www.crowdsupply.com/e-radionica/inkplate-6

Also, with Raspberry Pis connecting to raw e-ink screens using SPI, someone has made a VNC Client for it. This is probably the most effective way to use fresh e-ink screens without too much hacking, but it becomes quite a fragile hardware DIY solution that a software focused modder may shy away from.

https://github.com/joukos/PaperTTY

Regarding kernels, something that was recently released was the Civil Infrastructure 4.4 Super Long Term Support kernel. This allows the 4.4 kernel (as well as 4.19 and other 5.x kernels) to have a 10 year or even 20 year support cycle for security patches, crucial for these e-ink devices where driver support is spotty.

https://www.cip-project.org/blog/2020/08/17/cip-kernel-team-helping-cip-sustain-industrial-grade-systems

We could try to pick a Kindle or Kobo device that has a kernel near that version, and try to compile kernels that work with that and the e-ink display drivers.

Also just put all your stuff from the google drive into a github repository or even its wiki, its much easier to browse.

https://drive.google.com/drive/folders/1S1JA8dKjbLpOqcOjbrgT0qoBYRqsj0gk

[bdm310]

I'd be very happy to work together and get something supported!

It looks like we've done a lot of the same reading. I picked specifically the 7th gen Paperwhite 3 due to maturity, touch sensor quality, availability, cost, and ability to find teardowns before buying. I also figured I'd hit the biggest market of interested users with a Kindle device. If you think there's a better hardware option, I'm open to other stuff. Like you said, they're all largely based around the i.MX6 SoC with the epaper module. I can do the same hardware reverse engineering that I did for the Paperwhite in a weekend.

This has been a weekend project and I haven't had time to put in to it for a few months now, so agreed that my organization could be better overall. I'll put some thought in to it when I open this project back up.

I'm swamped with career work for the short term future but I wanted to get a response to you and keep the conversation going. This stalled at integration of epd drivers after my last sprint.

[lvw5264]

You went to some pretty advanced extremes to tear down the kindle, to the point at which it seems like it might be better off just to either build an open hardware PCB from the start by making a bigger screen version of this hand solderable surface mount device:

https://github.com/joeycastillo/The-Open-Book

Or as you said, based on your research that specific kindle or its screens was most common, well we could just do the same as this and reuse Kindle screens on open hardware:

https://www.crowdsupply.com/e-radionica/inkplate-6

Or building a more stable PCB to display to the screen using known SPI methods especially using long supported mainlined open source hardware such as with that new Raspberry Pi 4 compute module that's come out, rather than having to force the kindle to do what we want on new kernels.

https://github.com/joukos/PaperTTY

https://www.raspberrypi.org/products/compute-module-4/?variant=raspberry-pi-cm4001000

Of course though for the general public the dream of just reusing the Kindle's board and running mainline Linux on it and just displaying to the epaper display is what a software developer really wants rather than having to hand solder any tiny surface mount components. Have you managed to try using the following libraries or tools to display to the e-ink screen through known methods, or the stuff from fread.ink? Or is it possible to bring in stuff from or use other kindles that happen to be on Super Long Term Support versions (4.4, 4.19), if as long as we make sure to use the old waveforms?

Is there some proprietary or open source but esoteric driver that sticks the e-ink display system to older kernels (which might not be that surprising), or have you just not tried the e-ink display part yet? I do know you need to extract the waveforms, but fread.ink does talk about that.

https://github.com/NiLuJe/FBInk

https://fread.ink/

https://hackaday.com/2018/07/25/hope-xii-a-foss-operating-system-for-e-readers/

If you haven't tried, it might be worth contacting these people with your efforts and doing a full team collaboration. Or just sharing your work, as it seems like your stuff ended up in a small difficult to find corner in the internet and they might not have noticed.