Code Composer Studio (CCS) v7.0 on Arch Linux

People who like to tinker with electronics (at the level of programming MSP430 in assembly, no less) usually like to tinker with computers and operating systems too. With the release of Code Composer Studio (CCS) v7.0, Texas Instruments now has a Linux version of the programming tools that is (mostly) 64bit and does not require the multitude of 32bit libraries to run on 64bit Linux distributions (as was the case with CCS 6.x).

CCSv7 Running on Arch Linux

However, TI as usual provides instructions for running this useful program on Ubuntu only (LTS releases).  As I prefer to run Arch (I like the sense of total control of what gets installed and the ability to know exactly how my system works) I decided to give it a try. Here is how to successfully install and run CCS 7.x in Arch.

Step  1: Download CCS 7.x

Download the full version of CCS 7.x for Linux from this page. Scroll down and get the offline Linux version (it's about 700Mb). As stated in the notes, you will need to install the 32bit libc on your system.

Step 2: Install / Symlink Required Libraries

 Assuming you have the Multilib repository enabled, as root (or with sudo) execute the following command:

# pacman -S lib32-glibc

According to TI, this would be all you need but really this is not the case in an Arch system. While you don't need any other 32bit library, the installation program requires a few other libraries that may already be installed but are different versions than the ones required. These are:

• libncurses
• libusb

Additionally, running the ccs installer on a newly installed Arch GNOME system simply quits after a while without displaying an error message on the console and without presenting a GUI to the user. The installer requires libxss which you may install manually from the repository or let it get installed as a dependency of chromium. 

The specific version of libncurses required can simply be bypassed by symlinking the version already in the system:

# cd /lib

# ln -s libncursesw.so.6.0 libncurses.so.5

Libusb is a bit more tricky. You could symlink the required version it to the libusb already present in your system, but there is a more elegant way: simply install the Energia programming environment from yaourt:

# yaourt -S energia

This will install the correct version of the library and you will be good to go!

Step 3: Extract and Install CCS

There is no need to run the CCS installation as root; I recommend installing it in your home directory as a simple user. There is only a driver installation script you will have to run as root (or with sudo) after the main install completes.

Extract the archive (assuming it is in ~/Downloads):

$ cd /Downloads

$ tar xvpzf CCS7.0.0.00042_linux-x64.tar.gz

(Exact filename may of course vary as new versions are released)

Enter the directory created and run the install script:

$ cd CCS7.0.0.00042_linux-x64

$ ./ccs_setup_linux-x64

Want to program this Launchpad on Linux? You are out of luck.

The installer should come up shortly, without complaining about missing libraries (if it does, recheck your previous steps). Run the installation as usual, selecting all components you require (e.g. MSP430, MSP432, TI Arm and so on). Note that some MSP Launchpads are not supported in the Linux version, notably the cheap and cheerful Value Line Launchpad (pity). The older Stellaris launchpads are also not supported anymore (Time to get yourself a TivaC). 

Step 4: Install The Drivers

There is one final step: Get to the directory where you installed the program and run the driver installation script as root. Assuming you installed it in ~/ti, this is what it would like:

$ cd ~/ti/ccsv7/install_scripts

$ sudo ./install_drivers.sh

Now test your installation:

$ cd ~/ti/ccsv7/eclipse

$ ./ccstudio

You are done! Enjoy your new CCS on Arch. Happy New Year!

Intel Galileo: Linux, SSH, Static Addresses and Other Tips

I've recently acquired an Intel Galileo Gen 2 development board. Intel Galileo (just in case you haven't heard about it) is Intel's answer to the Arduino driven hobbyist community. In fact, Galileo is (or supposed to be) pin compatible with Arduino Uno: you can even use Uno shields on it. Galileo is programmed using the same IDE as the Arduino (albeit patched by Intel) and the sketches - once recompiled - are (supposedly) compatible: You can just move your LCD display circuit from your Uno to your Galileo, recompile and upload the program and it should work. Well, for the most part that is...

So why buy a Galileo instead of (or in addition to) a standard Uno or other AVR based Arduino? The Galileo is using an Intel SoC (System on-a Chip) running at no less than 400MHz. It is equivalent in power to some older Pentium CPU. There is more to it: There is 256MB of RAM on board, a micro-SD card slot, an Ethernet port and even a micro PCI-Express underneath for your WiFi card. Galileo is admittedly over-specified for an Arduino. And with good reason: Even without any SD inserted, Galileo runs a small Linux version inside it. You will notice it as it takes some good time to start and appear on the USB connection. And there is an even more complete Linux version available as an SD card image. Linux and Ethernet easily make this device an IoT (Internet of Things) development board.

How would you access the Linux part of Galileo? Let's examine two scenarios:

1. Access the built-in Linux. This is what runs on the Galileo if you don't insert any SD card.
2. Access the SD card Linux (Download it from here, along with any other utilities you need).

Accessing the Built-in Linux via Telnet

If you look at most of the Galileo tutorials you will find more than a few ways to access the Linux part of your Galileo:

• Use a serial cable and a (custom made) connector for the serial out pins (gen 2) or 3.5mm jack (gen 1). This is easy to do but most PCs lack proper RS-232 these days and I find USB to serial as a last resort scenario.
• Run a sketch to reconfigure the USB port as a serial terminal and use that for connecting. There are a couple of problems with this approach: although it will supposedly work on the built-in Linux, it stops accepting input after you connect through i.e. PuTTy. You will get a login prompt and a shell but that's about it. It won't execute any commands. It works a little better with the SD card version of Linux but there is another serious drawback: the sketch permanently reconfigures the USB port for serial connection and you lose the ability to upload sketches.

What if you really want to have both Linux *and* your sketches running at the same time?

It turns out that the built-in Linux does not have an ssh daemon, so we will have to use telnet. No big deal. And instead of going the serial route, let's try that nice Ethernet port. Ethernet is a lot more common these days: just plug your Galileo into your home switch then upload a sketch that looks like the following:

void setup() {
  system("ifconfig eth0 inet 192.168.0.10 netmask 255.255.255.0 up");
  system("telnetd");
}

void loop() {
}

Obviously, you will have to change 192.168.0.10 to an address that matches your home setup. If you are using DHCP (very common on home routers) just make sure you assign it a static address outside the scope of the DHCP server (I assume that as a good admin you've kept a couple of addresses from being automatically assigned, haven't you?)

You are now ready to telnet into your device! Use PuTTy from Windows or the command line telnet client from another Linux box. This is what you will get:

Login as 'root'. There is no password set.
Now that you are in, there is no need for the script to keep executing. Use the top command:

# top

Find the process id for /sketch/sketch.elf and just kill it (it usually is at the top of the list):

# kill 929

Or you could simply upload another sketch to execute. Unlike the USB serial approach, your upload ability is not affected in any way.

Play with the built-in Linux as long as you like: you will soon find out it is pretty limited in what it can do. And the moment you unplug your Galileo all the settings are lost: your sketches are erased and the nice static IP you've assigned is gone. You will have to rerun the above sketch after every reboot.

Sooner or later (probably sooner!) you will decide to migrate to the SD card version of Linux.

Accessing SD Card Linux via SSH and making settings permanent

So you've created an SD card image of Yocto Linux using Intel's instructions. And now you want to SSH into it.

• First, be patient. The SD card Linux may take more time to boot than the built-in version. You will know it has booted successfully when the USB connects to your PC (if you use Windows, it will make a sound)
• The SD card Linux provides an SSH service. Even better, you can make your IP address (and other settings) permanent.
• You will have to consider security if you leave your Galileo running and connected to the Internet.

As you can imagine, the script to get SSH access is even easier than before:

void setup() {
  system("ifconfig eth0 inet 192.168.0.10 netmask 255.255.255.0 up");
}

void loop() {
}

The SSH server is already running on SD card Linux, no need to restart it. Just assign your static IP and you are good to go! You can now login and make your settings permanent.

Use top and kill (like we did before) to stop the sketch from running. It is no longer necessary.  You may as well remove it so it won't run at next boot. We will make the configuration changes permanent anyway:

# rm /sketch/sketch.elf

The first thing you will want to do is assign a root password:

# passwd
Changing password for root
Enter the new password (minimum of 5, maximum of 8 characters)
Please use a combination of upper and lower case letters and numbers.
New password:
Re-enter new password:
passwd: password changed.
Change the /etc/network/interfaces file to assign your static IP address permanently. (It seems the only available editor is vi...) Find the following line:

iface eth0 inet dhcp

and change it to:

iface eth0 inet static
   address 192.168.0.10
   netmask 255.255.255.0
   gateway 192.168.0.250

The gateway part is only needed if you intend to give Galileo Internet access. You will also need to reconfigure /etc/resolv.conf for this (we will do it after a few more steps). The address on the gateway part is of course your home router's IP.

The above setting is not enough though: Galileo uses the Network Connections Manager (conman) to configure the network interfaces and the above file is ignored in this case. We will have to revert to the older method:

# cd /etc/rc5.d
# rm S05conman
# ln -s ../init.d/networking S05networking

Now reboot your Galileo:

# shutdown -r now

When it comes up again, just connect via SSH. No need to rerun the sketch as the changes to the files are permanent.

If you wish to connect your Galileo to the Internet, add this line to /etc/resolv.conf:

nameserver 192.168.0.250

Where 192.168.0.250 should actually be your router address. Or another available DNS server (if in doubt, use Google's DNS: 8.8.8.8).  For this to work, you must also have a valid gateway line in /etc/network/interfaces.
While still logged in as root, you may wish to change the weird 'clanton' hostname to something more cool. Just edit the /etc/hostname file and replace the contents with the name of your choice. Hostname will change on the next reboot.
If you wish to have a message of the day (motd) appearing at every login, create an /etc/motd file with your desired contents.

Securing your Galileo Linux

At a minimum, you will want to take a few security measures if you decide to connect your Galileo to the Internet:

• Give the root user a password. We've already done that.
• Disallow root logins via SSH.
• Create a standard user account for 'normal' use.

Creating a standard user acoount is easy:

# useradd -g root johndoe

(Johndoe is actually not a good choice for a username, but you get the idea!). It is best to make your user a member of the root group. The permissions on some devices (like /dev/null) are read-write for the root user and group only and will hinder your ability to use some commands like scp unless your account belongs to the root group. Give your new user a password:

# passwd johndoe

Changing password for johndoe
Enter the new password (minimum of 5, maximum of 8 characters) Please use a combination of upper and lower case letters and numbers.
New password:
Re-enter new password:
passwd: password changed.

This is a good time to check whether the new account works. Just open a new SSH connection and try to use the new user instead of root to connect. If it all works, continue by disabling the root login via SSH:

Edit /etc/ssh/sshd_config. Find the line that shows

PermitRootLogin yes

And change yes to no:

PermitRootLogin no

Or, alternatively, just comment out the entire line. The default setting for SSH is to not allow root logins.

Just reboot your Galileo and you are good to go:

# shutdown -r now

From now on, you will use your new user account to connect via ssh. You can always use:

$ su -

to switch to root when needed.

Sudo would have been a nice addition to this Linux version, but it is not available by default and I haven't researched package management yet!

You could create some fancy bash startup files for your account. You may also use my version (a slightly simplified version of what I use on my FreeBSD machines). Just login as the user you created and:

$ wget http://www.freebsdworld.gr/files/galileo-dot.tar.gz

$ tar xvzf galileo-dot.tar.gz

You may need to press 'A' to overwrite an existing file. Logout and login again to apply the changes.

Getting an LCD Screen to Work with Galileo

LCD screens are very popular with Arduino 'users' as they add a whole new dimension to projects. The 16 character, 2 line variant seems to be the most common and it is the one I currently have.

Since Galileo is Arduino Uno compatible, all you would have to do is move your LCD circuit from your Uno to Galileo (to the same GPIO pins), recompile your program for Intel and upload it. Well, supposedly.

Because I followed these exact steps and got just a blank LCD staring at me. Apparently you need to update the LCD Driver libraries with the ones found here.

After unzipping the file, replace all the contents of the libraries/LiquidCrystal folder (in your Arduino installation folder) with the contents of the archive.

Rebuild and upload your project - your screen should now be working!

Happy coding!

MSP430/ARM Development on Linux: Installing CCS 6.0

Code Composer Studio (or CCS as it is widely known), is Texas Instruments' own development tool for their series of MCUs like the well known MSP430 and the Stellaris/Tiva (ARM based) series. Texas Instruments also provides a number of inexpensive development / demo boards known as launchpads. I am the happy owner of two of them: the MSP430 one (F5529) and the ARM based Stellaris launchpad (LM4F120).

Texas Instruments provides some instructions on installing CCS 6.0 on Linux. We will provide additional instructions for installing Tivaware (or stellarisware).

Basic Install

Install your favorite *buntu variant. I've chosen Xubuntu 14.04 since it is lightweight enough and the speed is tolerable on my tiny 2009 Acer Netbook. (While I have beefier machines for development, the netbook is very convenient as it fits perfectly on my rather small electronics workbench).

Make sure to install the updates, either during installation or immediately afterwards by running:

sudo apt-get update; sudo apt-get upgrade

Downloading CCS6.0

Download CCS6.0 for Linux from this page:

CCS6.0 for Linux

It is preferable to download the full version rather than the web based installer.

Installing Dependencies

Before running the installation program, some dependencies need to be installed. In general the instructions in TI's wiki apply:

sudo apt-get install libc6:i386 libx11-6:i386 libasound2:i386 libatk1.0-0:i386 libcairo2:i386 libcups2:i386 libdbus-glib-1-2:i386 libgconf-2-4:i386 libgdk-pixbuf2.0-0:i386 libgtk-3-0:i386 libice6:i386 libncurses5:i386 libsm6:i386 liborbit2:i386 libudev1:i386 libusb-0.1-4:i386 libstdc++6:i386 libxt6:i386 libxtst6:i386 libgnomeui-0:i386 libusb-1.0-0-dev:i386 libcanberra-gtk-module:i386

Some of these are already installed, apt-get will inform you about this.

Note the wiki refers to the 64bit version, but CCS runs without any problem in Ubuntu 32bit too (as you may have noticed, all the above dependencies are 32bits anyway). Create the required symbolic link:

sudo ln -s /lib/i386-linux-gnu/libudev.so.1 /lib/libudev.so.0

Running the Installation

Change to your Download folder (or wherever you placed the downloaded CCS6.0 archive):

cd ~/Downloads

Extract the files:

tar xvzf CCS6.0.1.00040_linux.tar.gz

(replace with actual filename of downloaded file, may differ if a new CCS version is released)

Change to the folder of the extracted files and run the installer:

cd CCS6.0.1.00040_linux

./ccs_setup_6.0.1.00040.bin

There is no need to run the installer as root (with sudo), unless you wish to install it for multiple users. Otherwise, just run it as a standard user and install CCS in a subdirectory of your home directory. After accepting the license, you will be prompted to select an installation directory. Assuming your username is 'user', install CCS to /home/user/ti (the location is automatically suggested by the installer).

You will then be greeted by the processor support dialog:

It is wise to select all the MCUs that you intend developing for. For this example we selected MSP and Tiva/Stellaris development:

 

You won't have to change the default emulators selected in the next dialog. In the App Center dialog, select at least the MSP430Ware:

Installing the Drivers

After the CCS installation is complete and before plugging in your USB launchpad, install the necessary drivers:

cd ~/ti/ccsv6/install_scripts

sudo ./install_drivers.sh

Running CCS for the First Time

The first time you run CCS, you will be asked to select a workspace (and maybe set it as default). We have chosen /home/user/tidev here but you are welcome to choose your own or accept the default. Make a note of this as you will need it later.

On the first run the App Center will automatically download any options you selected during install (like the MSP430ware) and will also update other components. If you only intend to develop for MSP430, your setup is now complete. For Stellaris/Tiva, read on.

Installing Tivaware (Stellarisware)

Developing for tiva or stellaris requires the Tivaware library and some additional settings in CCS. Note that you can use Tivaware to develop for stellaris (LM4F devices) and there is no need to install stellarisware:

Download Tivaware from TI:

Tivaware download page

If for some reason you prefer stellarisware (for example, developing for LM3S devices):

Stellarisware download page

Tivaware is provided as an EXE file, but is actually a self extracting ZIP. Unzip to a subdirectory of your CCS6.0 install path:

cd ~/ti
mkdir tivaware
cd tivaware
unzip ~/Downloads/SW-TM4C-2.1.0.12573.exe

(the actual filename may vary)

There are several ways to include tivaware in your projects. In order to minimize required settings on each project, create a vars.ini file in your workspace. Remember this is /home/user/tidev in our example:

cd ~/tidev
vi vars.ini

(obviously, use your favorite editor instead of vi to create this file)

A single line is needed:

TIVAWARE_INSTALL=/home/user/ti/tivaware

 

Configuring Your Project for Tivaware

The vars.ini file creates an environment variable for the /home/user/tidev workspace (where the file is saved). To configure your project to use tivaware successfully:

• Import the vars.ini file as source for CCS Build variables
• Add an "include files" path to the compiler using the TIVAWARE_INSTALL variable
• Add the driverlib.lib file to the project.

Go ahead and create a new project (File => New => CCS Project). Use the following screenshot as a guide:

When finished, select File => Import => Code Composer Studio => Build Variables:

Select the vars.ini file previously created:

Next, right click on your project name in the project explorer. Select properties, ARM Compiler, Include Options and add the following directory path:

Finally, add (actually, link) driverlib.lib to your project. Right click on your project name in the project explorer and select add files:

The full path for driverlib.lib as shown: /usr/ti/tivaware/driverlib/ccs/Debug. On the next dialog, select link to file:

 

We are done! Here is our BlinkTheLed project (from the Tiva Workshop workbook), successfully built:

 

Since the drivers for the in-circuit debug interface are installed as well, you can actually connect your launchpad and run a debug session on the device.
And now you can continue running your development environment on your lean and mean Linux machine. Happy coding!