Hunting Hurricanes

RT-Linux is a patch which gives Linux many of the most important features needed by real-time programmers and embedded-system designers. It is implemented as a set of modules which can be installed and removed using insmod and company. You also use insmod to install any real-time code you write. RT-Linux programs execute in the kernel space and have full access to the system hardware and kernel code as well.

We've done a considerable amount of development using Turbo-C and DOS in the past, and it is truly amazing how infrequently we had to reboot Linux during development of the SRA. Back under DOS, we usually had to reboot several times per day. With Linux, we had to reboot only three or four times during the entire development period.

Figure 10. SRA Memory Usage

Once the RT-Linux programs/modules capture the data, they must be written to storage and displayed for the system operator. We accomplish this by using shared memory. The SRA has 64MB of RAM and we configured the kernel to boot using mem=61m which causes the kernel to manage only the lower 61MB, leaving 3MB untouched. It is this 3MB that we use for real-time data capture and as a common communication buffer area between RT-Linux modules and normal user-space programs. Figure 10 depicts the SRA memory usage.

We wrote a single C program (rgc.c) which provides most of the interface between Linux user mode and RT-Linux. This program is a simple command-line style program with tons of commands to read and write data space in common between RT-Linux and user space. Most of our Tcl/Tk scripts merely open a pipe to this program and use it to pass commands and extract data from the system. The program can also be used directly from the command line. This makes development and debugging simpler.

One of the run-line options to rgc causes it to loop, testing for data to be written to disk. If no data are ready, the program sleeps for one second. If data are ready, they are extracted and written to the specified disk file.

We use up to five laptops on the SRA at once: three for collecting GPS data (one laptop for each GPS receiver) and two for control and display of real-time SRA data. A personal laptop is used for control, and if we're both on the flight, we can both run several instances of the same display programs using another personal laptop. We each have our favorite color-bar for the image of the sea. We'll frequently use one machine to control the SRA and the other to write or modify display or system software as we're flying. The laptops are Chembook 9780s. Each has a 4GB internal hard drive and a modular 6.4GB drive (in place of the floppy), a 14.2'' XGA LCD display, PCMCIA Ethernet card and a 233MHz Pentium-Pro CPU.

Each of these machines dual boots either Red Hat Linux 5.1 or MS Windows 95. To use the laptops as X terminals, we boot Linux, then run the Xfree86 server. We run the X server such that the laptop becomes an X terminal for the SRA data system. This puts most of the burdensome display processing on the laptop processor, since the X server seems to be where the CPU cycles go. There are two ways to cause X to act as an X terminal. The first is:

X -query

and the second:

X -indirect

You can even have two or more X servers running at once. Here's an example:

X -query first-machine
X :1 -query second-machine
X :2 -query third-machine
X :3 -query fourth-machine

You can get a local X server going with the command:

startx -- :4