Getting Started with Embedded Linux: Part Nine - Self-hosted development -

Getting Started with Embedded Linux: Part Nine — Self-hosted development


As we discussed in the last installment, there are two common models for Embedded Linux development. These are cross-platform development and self-hosted development. In this installment, we are going to explore self-hosted development using the Raspberry Pi as a target.

Self-hosted development is an offshoot of developing on a host system, similar to what we have been doing with our VM installation in previous installments of this series. But there are several differences since the target system, like most embedded systems, has limited memory and a relatively slow processor.

Install Raspberry Pi
The Raspberry Pi (Model B*) is an inexpensive ($35) single-board computer (SBC) with a 700MHz Broadcom ARM processor, 512Mb RAM, high definition video output supported by a GPU, USB connectors for keyboard and mouse, an SD memory card used for the file system, an Ethernet port, as well as significant expansion ability. There are many other single-board computers on the market, some of which may be better suited to a particular application. The RPi (as it is sometimes called) was designed for educational use, so it fits well with the purposes of this column. If you need an board for a commercial application, I recommend that you look at other SBCs which are designed for the rigorous environment required by most embedded systems and are better tailored to your application's specific requirements. The RPi has a tremendous support base, with a dedicated website (, magazines, many online blogs and how-to guides, and even videos.

I followed the Quick Start Guide on and its recommendations to use the pre-installed NOOBS SD card and install the Raspbian distribution, based on the popular Debian distribution. First, I connected the RPi as shown in this photo.

I followed the instructions on using the NOOBS SD card, which displays a number of available distributions when the RPi boots from the SD card. I selected Raspbian. This was mostly uneventful, but it takes a while. After Raspbian is installed, it needs to be configured. I made the mistake of specifying All Locales, rather than just the US Local, where I live, which took longer to configure than installing the distribution. Once Raspbian was done configuring, I was able to log in using the username “pi” and password “raspberry.” I discovered that the USB keyboard I was using was not being recognized correctly, and some keys were displaying wrong, such as showing the British pound symbol (£) rather than the character on the key cap. There may be some way to fix this using one of the Debian configuration tools, but my fix was to edit the default keyboard configuration:

sudo nano /etc/default/keyboard

Change XKBLAYOUT="gb" to specify the correct country code, in my case, us .

To complete setting up the RPi, I needed to configure the Ethernet port. By default, Raspbian gets an IP address using DHCP. My DHCP server is configured to give the RPI the same IP address each time it submits a request. If you don't configure your DHCP server to assign the same address, it may assign a different address whenever the RPI starts, which can be confusing. Alternately, you can assign a static IP address to your RPI by following instructions found here:

Here's the RPI booted to a console login screen, powered by a small USB battery courtesy of the Yocto project:

Development on the Raspberry Pi
When you log in to the RPI, you'll discover that it is a pretty complete distribution. In keeping with its educational goals, you will find that most of the development tools have already been installed. It normally boots up into command line mode, but you can start a GUI by running “startx.”

Let's build the simple driver from Part Seven. Raspbian comes with the Midori and NetSurf web browser. I used NetSurf to visit and navigate to where I could download the ZIP file containing the source code for myrandom.c. []. I saved it in a subdirectory named myrandom and used “unzip” to unpack it.

Using vi or nano, create a new Makefile in the same directory like the one listed in Part Six, replacing “lkm.o” with “myrandom.o.” (I don't recommend cutting-and-pasting the text from the web page. That picks up hidden characters.) Then I typed in the command “make.” And the response was “Make: Nothing to be done for 'all'.” You might remember from Part Six where I mentioned that you might need to install the kernel development files for your stock kernel. If these are not installed, make doesn't know what to do.

We can use Debian's apt-get utility to install the kernel headers and build directory. But first, I wanted to make sure that I was running the latest version of Raspbian. So I first executed the command “sudo rpi-update”, waited for it to complete updated files, then I rebooted the Raspberry Pi.

There are a number of web pages which discuss installing the Linux headers and the kernel sources. An easy way is to use the rpi-source script available on GitHub. To install, I followed a slightly modified set of the instructions on the rpi-source wiki []:

$ wget$ sudo mv rpi-source /usr/bin/rpi-source$ sudo chmod +x /usr/bin/rpi-source $ rpi-source -q --tag-update

Before you run rpi-source, install the ncurses headers. The kernel build scripts require these headers. They can be installed using apt-get:

$ sudo apt-get install libncurses5-dev

Now run rpi-source:

$ rpi-source

As warned on the wiki, rpi-source generated a complaint that the installed version of GCC was different from the version needed to build the kernel. I followed the instructions and added the jessie source repository, and installed gcc-4.8:

$ sudo cat “deb jessie main contrib non-free rpi” > /etc/apt/sources.list.d/jessie.list$ sudo apt-get update$ sudo apt-get install -y gcc-4.8 g++-4.8$ sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.6 20$ sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.8 50$ sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-4.6 20$ sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-4.8 50

After updating GCC, I re-ran rpi-source. This ran for a while and downloaded almost 120Mb of sources. Luckily, I have a fast Internet connection. Rpi-source then built the support programs needed to build kernel modules. When this was done, rpi-source had created a directory under /home/pi, created a link in /home/pi/linux pointing to it, and updated /lib/modules/`uname -r`/build to point to /home/pi/linux.

When all this setup was complete, I was able to revisit Part Seven, and follow the instructions there:

$ cd ~/myrandom$ make$ sudo insmod myrandom.ko$ sudo grep myrandom /var/log/messages

As before, I created the /dev/myrandom node with the correct major number (not the same as before) listed in /var/log/messages. I ran the same small test of the driver. And it worked!

The Raspberry Pi, like similar Single Board Computers (SBCs) such as the BeagleBone and many others, act a lot like an underpowered laptop computer, with a separate keyboard and monitor. It's certainly possible to do application or development of Linux Kernel Modules on an SBC. The advantage is that it is native development, like working on your desktop system. The disadvantage is that the resources available on an SBC are often limited, which can result in long build times or running out of space on the flash card.

Next time, we'll talk about a hybrid approach, where we build the Linux kernel for the Raspberry Pi on a fast host system, using cross-development tools.

* A new version of the Raspberry Pi, the Model B+, was announced as this article was being written. It looks like a good enhancement to the Model B, with a microSD card rather than the larger SD card, and additional USB ports. There should be no differences in using the Model B and B+.

Michael Eager is principal consultant at Eager Consulting inPalo Alto, Calif. He has over four decades experience developingcompilers, debuggers, and simulators for a wide range of processorarchitectures used in embedded systems. His current and former clientsinclude major semiconductor companies and systems developers. Michaelhas been a member of the ISO C++ Standard Committee and ABI Committeesfor several processor architectures. He is chair of the DebuggingStandards Committee for DWARF, a widely used debug data format. He isactive in the open-source and Linux communities.

Related links:

2 thoughts on “Getting Started with Embedded Linux: Part Nine — Self-hosted development

  1. “@Cpetras991: nMaybe RT_PREEMPT is currently not the best option since one of the main developers will be working on the realtime patches only on a “hobbyist basis” due to lack of funding.[1]nnBesides it's a patch, which means it's not in the upstream

    Log in to Reply

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.