Skip to content

Latest commit

 

History

History
367 lines (237 loc) · 13.3 KB

README.md

File metadata and controls

367 lines (237 loc) · 13.3 KB

Remember that fread is a work in progress. Some documentation is still missing for how to compile Xorg drivers and how to bundle the userland into an ext4 file.

This is guide for how to compile fread from scratch using a vagrant machine. This may be of use to developers but if you're looking to get started using fread you should look at the precompiled images (ToDo) instead. All of the components described here are also available individually as pre-compiled packages. See the "Pre-compiled" section near the bottom of this document.

Table of contents

Overview

What you need to cross-compile:

  • The fread initrd
  • The fread kernel
  • The kexec tool
  • The fread userland

The initrd (initial ramdisk) is a very minimal linux userland that has just enough functionality to mount what needs mounting, chroot and boot into the real userland which is a complete, but minimal, debian-based system.

Hardware:

  • Kindle 4 NT (non-touch) with serial console
  • USB to 1.8v serial cable

Currently only the Kindle 4 NT has been tested.

Getting a serial console requires opening the Kindle and a bit of soldering. Do not try to use a 3.3v serial cable, it will fry your kindle.

Currently the only tested method for booting into fread is to let the Kindle boot into the normal Kindle operating system and then issue commands via serial console.

It is certainly possible to boot fread without serial console but it would require modification to the kindle's original operating system and has not yet been implemented.

Cross-compile environment using vagrant

We're using a vagrant vm to compile everything. It greatly simplifies things when all developers are building on the same system.

For some reason many of the cross compilers I've tried create a kernel that simply won't boot on my Kindle 4 NT. No error messages appear after running kexec -e, just... nothing. If anyone has any idea why this could be please let me know!

The cross-compilation toolchain built into Ubuntu 12.04 generates a bootable kernel so for now that's what we're using. For other parts of the system you can use whichever cross compiler you prefer.

The version of vagrant in Ubuntu 14.04 is a bit old and won't actually work for our purposes. If you have a newer Ubuntu system then you may be able to just do:

sudo apt-get install vagrant

If you get errors when running vagrant up then install the newest version instead:

sudo bash -c 'echo deb http://vagrant-deb.linestarve.com/ any main > /etc/apt/sources.list.d/wolfgang42-vagrant.list'
sudo apt-key adv --keyserver pgp.mit.edu --recv-key AD319E0F7CFFA38B4D9F6E55CE3F3DE92099F7A4
sudo apt-get update
sudo apt-get install vagrant

The following will install a Ubuntu 12.04 32-bit virtual machine with everything needed to compile the fread kernel:

git clone https://github.com/fread-ink/fread-vagrant
cd fread-vagrant/
vagrant up

Note for Fedora users. The virtual machine requires the virtualbox provider. Vagrant on Fedora comes with a different default provider libvirt, which will fail running the virtual machine. You need to append --provider=virtualbox to have a successful start:

vagrant up --provider=virtualbox

If this fails because the virtualbox hashicorp/pecise32 image is no longer available from hashicorp's website then you can find a backup of it here.

If the package repository for Ubuntu 12.04 is unavailable then you can find a backup of the required packages here.

To get a shell on the virtual machine just ensure that you're in the fread-vagrant dir and run:

vagrant ssh

The /vagrant dir is shared between the VM and your actual system such that any changes to /vagrant/ happens in the fread-vagrant/ dir.

For the rest of this guide make sure you're NEVER actually compiling anything directly in the /vagrant dir since that directory does not support hard links.

If you need to get back into the machine after rebooting just:

cd /where/you/put/fread-vagrant/
vagrant up
vagrant ssh

If you need to stop the VM use:

vagrant halt

but I don't like vagrant!

Alright just look at the bootstrap.sh file in the fread vagrant repo and set up your own Ubuntu 12.04 system with the right dependencies installed :)

Compiling the initramfs

The first thing to do is compile the initramfs (initial ram filesystem) since we'll be compiling a kernel with a built-in initramfs.

We'll use buildroot to compile a minimal buildroot that uses uclibc and busybox.

First you need a buildroot that's been configured to build an iMX 5 initramfs:

cd ~/
git clone https://github.com/fread-ink/fread-initrd
cd fread-initrd/
./fetch.sh 

Compile the initramfs. This will cause buildroot to download several packages and will probably take several hours:

./build.sh

Now create the initramfs file:

sudo ./bundle_initramfs.sh

The generated initrd will be at ~/fread-initrd/initrd.cpio.

Compiling the kernel

First download the kernel:

cd ~/
git clone https://github.com/fread-ink/fread-kernel-k4
cd fread-kernel-k4/linux-2.6.31

Copy the config file and configure it with the initrd that you have just built.

cp ../config .config
sed -i '/^CONFIG_INITRAMFS_SOURCE=/s@=.*$@='"\"~/fread-initrd/initrd.cpio\""@g .config

Now compile the kernel:

./build.sh

This will take a while but not nearly as long as the buildroot compile. You will need to copy the compiled kernel and modules out from the virtual machine. Copy the entire OUTPUT/ directory to /vagrant/KERNEL. The files will then be available in fread-vagrant/KERNEL after logging out from the virtual machine.

Compiling kexec

kexec is the tool used to load a new linux kernel without rebooting. This is currently the method used for booting into fread from the original kindle operating system. This has the advantage of not having to modify anything related to the original operating system so you don't brick your device during development.

Unfortunately kexec has to be compiled with the same glibc version, kernel headers and compile-time parameters used in the original Kindle operating system since it will be executed before fread has loaded.

Compiling kexec requires a different cross-compile toolchain.

To build this toolchain:

git clone https://github.com/fread-ink/fread-native-cross-compile
cd fread-native-cross-compile/
./build_k4.sh

again this will take a long time. If you encounter odd build errors that just say something "killed" with no reason given then your VM is running out of memory. Use vagrant halt to shut down the VM, then edit the line vb.customize ["modifyvm", :id, "--memory", "1024"] to e.g. 2048 instead of 1024, restart the VM with vagrant up and try again.

Then build kexec using the toolchain:

source env_k4.sh
./build_kexec.sh

Log out and then log back in before continuing.

Building the userland

See the fread-userland readme file for info on how to build a working userland.

Installing fread

Copying the files

Put the following files in a directory called fread/:

  • uImage: The fread kernel (including initramfs)
  • kexec: The command line utility used to switch the running kernel
  • fread.ext4: The fread userland

Connect your e-reader using USB, mount it as a storage device and copy the fread/ directory to the root of the e-reader's filesystem.

NOTE: THE NAMING AND LOCATIONS OF THE FILES AND DIRECTORY IS IMPORTANT!

Enabling usb networking

You can enable usb networking by adding a file named config.txt to the fread/ directory with the contents:

USB="ethernet"

WARNING: There is currently a bug in usb networking that causes the system to crash if the usb cable is not plugged into a computer when usb networking initializes. Remember to plug in the usb cable before booting to fread.

If you are on a Debian/Ubuntu (and probably on many others) then everything should auto-configure (assuming you have a DHCP client enabled) and after bootup is complete you should see a new network interface appear (possibly called usb0) with an assigned IP address in the 192.168.15.x range. You can then connect over ssh:

The default root password is fread.

Booting into fread using the KUAL extension

This method only works if KUAL is already installed in your kindle.

Somewhere on your computer, clone the fread KUAL extension:

git clone https://github.com/fread-ink/fread-kual-extension

Then copy the extensions/ directory to the root of your kindle's fat32 filesystem:

cp -a fread-kual-extension/extensions /path/where/e-reader/is/mounted/

Now on your kindle, access the KUAL app and select the "Switch to fread" menu entry to boot fread.

Booting into fread using root console

If you have console access, log in as root on the e-reader. Ensure that the e-reader has been unmounted/ejected from your computer (if it is plugged into usb) and then boot into fread like so:

cd /mnt/us/fread/
./kexec --type=uImage -f ./uImage

If the /mnt/us directory is empty then cd out of it, unmount/eject your kindle from the computer to which it's attached and cd back into /mnt/us.

Logging in

The default root password is fread.

Getting internet access via usb networking

Whichever computer you connected to e-reader should have received an IP from the fread.ink DHCP server in the 192.168.15.x range. On that computer, check which IP you got using:

ip addr show dev usb0 | grep 192.168.15

Replace 192.168.15.x with the actual IP in the following commands.

On the connected computer, run the following as root (assuming you're running GNU/Linux system):

echo "1" > /proc/sys/net/ipv4/ip_forward
iptables -P FORWARD ACCEPT
iptables -t nat -F POSTROUTING
iptables -t nat -A POSTROUTING -o wlan1 -j MASQUERADE

Note that this is not at all secure, so if you care about those things you should probably learn how to set up better firewall rules.

Now on fread do:

cd ~/
./internet.sh 192.168.15.x

Now you should be able to ping fread.ink.

Getting X working

Neither startx/xinit nor automatic updating is working yet, but you can test things out.

You will need to get internet connectivity working to install xeyes (or some other test program). Look at the section in this README called Getting internet access via usb networking.

cd ~/
apt update
apt install -y xeyes
./x.sh & # start X
DISPLAY=:0 xeyes &

Now X and xeyes will be running but since we don't have display out-updatig working you will have to trigger a display update manually:

echo "direct" > /sys/devices/platform/mxc_epdc_fb/mxc_epdc_update

You should now see two eyes at the top left of your display.

Old-style framebuffer graphics

This method will give you a (somewhat) normal framebuffer device but will not work with the mxc xorg driver. You can use this method if you have a non X application that works directly with framebuffers rather than

modprobe -r mxc_epdc_fb # this module is loaded at boot but with wrong arguments
modprobe eink_fb_waveform
modprobe eink_fb_hal
modprobe mxc_epdc_fb dont_register_fb=1
modprobe eink_fb_hal_fslepdc bootstrap=0
modprobe eink_fb_shim

Manually triggering a display update:

echo "1" > /proc/eink_fb/update_display

Thanks to

  • NiLuJe for KindleTool and for his amazing work on Kindle cross-compilation!
  • The mobileread e-reader hacking community!

Pre-compiled

The code described in this document is available as pre-compiled packages here:

or you can download the entire operating system in a ready-to-use package:

  • fread (ToDo)

Copyright and license

Unless otherwise stated everything in this repository Copyright 2016 Marc Juul and licensed under the GPLv3.

Disclaimer

Kindle and Lab126 are registered trademarks of Amazon Inc.

Kobo is a registered trademark of Kobo Inc.

Nook is a registered trademark of Barnes & Noble Booksellers Inc.

E Ink is a registered trademark of the E Ink Corporation.

None of these organizations are in any way affiliated with fread nor this git project nor any of the authors of this project and neither fread nor this git project is in any way endorsed by these corporations.