Embedding Linux to Control Accelerators and Experiments

A scientific laboratory in Europe depends on Linux for controlling equipment used in their research.
TACO

To provide a unified software interface to many different pieces of hardware with as many different kinds of hardware interfaces, we have developed an object-oriented toolkit called TACO for doing just that. It is also used at the Hartbeesthoek Radio Astronomy Observatory in South Africa (see Resources) for controlling a radio telescope, and will be used at the FRM II research reactor in Germany for controlling future experiments.

TACO unifies the software interface with the hardware and the way control software is written by providing a framework for developing control objects. Each control point is implemented in a Device class as a method. At runtime, the Device classes create as many copies of Device objects as there are pieces of hardware, and export their functionality onto the network. The Device objects are served by Device servers, which are stand-alone processes under Linux. Clients (which can be graphical GUIs or other Device objects) access the Device objects via the Device Server API (DSAPI).

The network addressing is managed by a database (implemented with gdbm) and the network protocol by the ONC RPC (from Sun, which is now part of glibc). RPCs make network calls look like local procedure calls. The ONC RPC is the basis of NFS. This means it can be found on all systems where NFS has been ported, i.e., just about everywhere. It requires one extra process to be running per host—the portmapper, which manages the mapping of program number to local port numbers. Performance overhead induced by an RPC over the network is on the order of a few milliseconds. The memory footprint is less than 100 kilobytes per server, linked with shared libraries (TACO API's, glibc and pthreads).

The DSAPI implements various flavours of network calls needed for doing controls—synchronous (client blocks waiting for the answer), asynchronous (client continues immediately and retrieves answer later) and events (client registers interest in an event and gets sent events asynchronously). It also implements a large number of standard types, timeouts, UDP and TCP protocols and stateless connections. Device classes can be written in C or C++. The database is used to store and retrieve device-dependent information. Interfaces to high-level scripting languages such as Tcl, LabView, SPEC and Mathlab are available. Clients can also be written in C, C++ or Java.

Because of its simple approach, TACO scales easily. It has been used for laboratory systems consisting of a single device right up to an entire synchrotron which consists of more than 10,000 devices. We use TACO to control everything from simple digital I/O to entire data acquisition systems.

We are presently working on the next generation of TACO called TANGO (TAco Next Generation Objects). In TANGO, CORBA will replace ONC RPC, and it will be possible to write Device classes in Java as well as C++.

Examples of Embedded Controllers

We started using Linux in embedded controllers approximately one year ago. To date, we have built the following controllers.

Serial Line Controller

We have traditionally used a VME-based board with 16/18 ports to control RS-232 and RS-422 serial lines. We have thousands of serial lines today to control vacuum devices, power supplies, PLCs, etc. As the number of devices requiring serial lines increases each day (and in some crates, we have reached the maximum number of serial line VME boards), we started looking for a new system, which would be cheaper than the existing VME-based system and would be PC-based. (Many of the serial line devices are certified only on PCs.) That is how we came to the PC/104. Cost is not the only reason for changing: we also wanted to implement new technologies, like web support, for better maintenance and configuration of the serial lines. We wanted as many serial lines as the VME-based system; therefore, we looked for a PC/104 board with eight ports which we could stack. Unfortunately, the choice of boards with more than 4 ports is limited. We finally chose the Parvus Octal Serial card.

This board is a so-called “dumb multiport serial” board, based on two 16554 UART (which are compatible with the standard 16650 UART) plus a programmable gate array for the address and UART register mapping on the ISA bus. As this board is quite new, we encountered some problems with the on-board gate array (bugs and limitations of programming) but we also found workarounds. To integrate this board in our Linux TacoBox, we used the standard serial line driver of the kernel (2.0.35). It implements IRQ sharing; thus, we consume only one IRQ for the eight ports of a Parvus board.

Before accessing the extended serial lines through the system device descriptors /dev/ttyS, we have several setup steps. All are processed from a single homemade script. Therefore, to install one (or more) Parvus boards, we need to add a single line to the booting scripts. We added a call to our script in the /etc/rc.d/serial file.

The first step is to set up the Parvus card gate array. We wrote a C program to access the gate array registers. The script picks up all the parameters needed for the call to this program and for configuring the serial lines from an easy-to-understand and modify ASCII file. Then the device descriptors /dev/cuaxx and /dev/ttySxx are created. Since we use the standard driver, the major numbers are the same as the ones used for the CPU board serial lines (4 for ttySxx and 5 for cuaxx). As with any other dumb multiport card, the minor numbers start from 64. At this step, we also change the access permissions to the device descriptors to allow any user to use the serial lines.

The last step is configuring the serial ports using the helpful tool setserial. We set up each port of a board with a different address but with the same IRQ; for example, addresses 0x100, 0x108 and IRQ 12 for one board and addresses 0x140, 0x148 and IRQ 9 for the other board. At this point, the serial ports are fully integrated in the operating system and can be used from any program.

Figure 4. Manuel (aka RastaMan) after a heavy day of debugging the serial line TacoBox, feeling like he has serial lines coming out of his head.

For our embedded controller, we ported the existing serial-line TACO device server from our commercial OS to Linux. The new device server uses POSIX calls to interface with the serial-line driver and should therefore work with any OS supporting POSIX. The compatibility of the two device servers and the use of TACO allows us to interchange a VME-based system with a PC/104-based system without changing the client application.

______________________

Webinar
One Click, Universal Protection: Implementing Centralized Security Policies on Linux Systems

As Linux continues to play an ever increasing role in corporate data centers and institutions, ensuring the integrity and protection of these systems must be a priority. With 60% of the world's websites and an increasing share of organization's mission-critical workloads running on Linux, failing to stop malware and other advanced threats on Linux can increasingly impact an organization's reputation and bottom line.

Learn More

Sponsored by Bit9

Webinar
Linux Backup and Recovery Webinar

Most companies incorporate backup procedures for critical data, which can be restored quickly if a loss occurs. However, fewer companies are prepared for catastrophic system failures, in which they lose all data, the entire operating system, applications, settings, patches and more, reducing their system(s) to “bare metal.” After all, before data can be restored to a system, there must be a system to restore it to.

In this one hour webinar, learn how to enhance your existing backup strategies for better disaster recovery preparedness using Storix System Backup Administrator (SBAdmin), a highly flexible bare-metal recovery solution for UNIX and Linux systems.

Learn More

Sponsored by Storix