RTEMS Classic API Guide (5.3).

Copyrights and License

© 2017 Chris Johns
© 2017 Kuan-Hsun Chen
© 2015, 2020 embedded brains GmbH
© 2015, 2020 Sebastian Huber
© 2011 Petr Benes
© 2010 Gedare Bloom
© 1988, 2018 On-Line Applications Research Corporation (OAR)

This document is available under the Creative Commons Attribution-ShareAlike 4.0 International Public License.

The authors have used their best efforts in preparing this material. These efforts include the development, research, and testing of the theories and programs to determine their effectiveness. No warranty of any kind, expressed or implied, with regard to the software or the material contained in this document is provided. No liability arising out of the application or use of any product described in this document is assumed. The authors reserve the right to revise this material and to make changes from time to time in the content hereof without obligation to notify anyone of such revision or changes.

The RTEMS Project is hosted at https://www.rtems.org. Any inquiries concerning RTEMS, its related support components, or its documentation should be directed to the RTEMS Project community.

1. Preface

In recent years, the cost required to develop a software product has increased significantly while the target hardware costs have decreased. Now a larger portion of money is expended in developing, using, and maintaining software. The trend in computing costs is the complete dominance of software over hardware costs. Because of this, it is necessary that formal disciplines be established to increase the probability that software is characterized by a high degree of correctness, maintainability, and portability. In addition, these disciplines must promote practices that aid in the consistent and orderly development of a software system within schedule and budgetary constraints. To be effective, these disciplines must adopt standards which channel individual software efforts toward a common goal.

The push for standards in the software development field has been met with various degrees of success. The Microprocessor Operating Systems Interfaces (MOSI) effort has experienced only limited success. As popular as the UNIX operating system has grown, the attempt to develop a standard interface definition to allow portable application development has only recently begun to produce the results needed in this area. Unfortunately, very little effort has been expended to provide standards addressing the needs of the real-time community. Several organizations have addressed this need during recent years.

The Real Time Executive Interface Definition (RTEID) was developed by Motorola with technical input from Software Components Group [Mot88]. RTEID was adopted by the VMEbus International Trade Association (VITA) as a baseline draft for their proposed standard multiprocessor, real-time executive interface, Open Real-Time Kernel Interface Definition (ORKID) [VIT90]. These two groups worked together with the IEEE P1003.4 committee to ensure that the functionality of their proposed standards is adopted as the real-time extensions to POSIX.

This proposed standard defines an interface for the development of real-time software to ease the writing of real-time application programs that are directly portable across multiple real-time executive implementations. This interface includes both the source code interfaces and run-time behavior as seen by a real-time application. It does not include the details of how a kernel implements these functions. The standard’s goal is to serve as a complete definition of external interfaces so that application code that conforms to these interfaces will execute properly in all real-time executive environments. With the use of a standards compliant executive, routines that acquire memory blocks, create and manage message queues, establish and use semaphores, and send and receive signals need not be redeveloped for a different real-time environment as long as the new environment is compliant with the standard. Software developers need only concentrate on the hardware dependencies of the real-time system. Furthermore, most hardware dependencies for real-time applications can be localized to the device drivers.

A compliant executive provides simple and flexible real-time multiprocessing. It easily lends itself to both tightly-coupled and loosely-coupled configurations (depending on the system hardware configuration). Objects such as tasks, queues, events, signals, semaphores, and memory blocks can be designated as global objects and accessed by any task regardless of which processor the object and the accessing task reside.

The acceptance of a standard for real-time executives will produce the same advantages enjoyed from the push for UNIX standardization by AT&T’s System V Interface Definition and IEEE’s POSIX efforts. A compliant multiprocessing executive will allow close coupling between UNIX systems and real-time executives to provide the many benefits of the UNIX development environment to be applied to real-time software development. Together they provide the necessary laboratory environment to implement real-time, distributed, embedded systems using a wide variety of computer architectures.

A study was completed in 1988, within the Research, Development, and Engineering Center, U.S. Army Missile Command, which compared the various aspects of the Ada programming language as they related to the application of Ada code in distributed and/or multiple processing systems. Several critical conclusions were derived from the study. These conclusions have a major impact on the way the Army develops application software for embedded applications. These impacts apply to both in-house software development and contractor developed software.

A conclusion of the analysis, which has been previously recognized by other agencies attempting to utilize Ada in a distributed or multiprocessing environment, is that the Ada programming language does not adequately support multiprocessing. Ada does provide a mechanism for multi-tasking, however, this capability exists only for a single processor system. The language also does not have inherent capabilities to access global named variables, flags or program code. These critical features are essential in order for data to be shared between processors. However, these drawbacks do have workarounds which are sometimes awkward and defeat the intent of software maintainability and portability goals.

Another conclusion drawn from the analysis, was that the run time executives being delivered with the Ada compilers were too slow and inefficient to be used in modern missile systems. A run time executive is the core part of the run time system code, or operating system code, that controls task scheduling, input/output management and memory management. Traditionally, whenever efficient executive (also known as kernel) code was required by the application, the user developed in-house software. This software was usually written in assembly language for optimization.

Because of this shortcoming in the Ada programming language, software developers in research and development and contractors for project managed systems, are mandated by technology to purchase and utilize off-the-shelf third party kernel code. The contractor, and eventually the Government, must pay a licensing fee for every copy of the kernel code used in an embedded system.

The main drawback to this development environment is that the Government does not own, nor has the right to modify code contained within the kernel. V&V techniques in this situation are more difficult than if the complete source code were available. Responsibility for system failures due to faulty software is yet another area to be resolved under this environment.

The Guidance and Control Directorate began a software development effort to address these problems. A project to develop an experimental run time kernel was begun that will eliminate the major drawbacks of the Ada programming language mentioned above. The Real Time Executive for Multiprocessor Systems (RTEMS) provides full capabilities for management of tasks, interrupts, time, and multiple processors in addition to those features typical of generic operating systems. The code is Government owned, so no licensing fees are necessary. RTEMS has been implemented in both the Ada and C programming languages. It has been ported to the following processor families:

  • Adapteva Epiphany

  • Altera NIOS II

  • Analog Devices Blackfin

  • Atmel AVR

  • ARM

  • Freescale (formerly Motorola) MC68xxx

  • Freescale (formerly Motorola) MC683xx

  • Freescale (formerly Motorola) ColdFire

  • Intel i386 and above

  • Lattice Semiconductor LM32

  • NEC V850

  • MIPS

  • Moxie Processor

  • OpenRISC

  • PowerPC

  • Renesas (formerly Hitachi) SuperH

  • Renesas (formerly Hitachi) H8/300

  • Renesas M32C

  • SPARC v7, v8, and V9

Since almost all of RTEMS is written in a high level language, ports to additional processor families require minimal effort.

RTEMS multiprocessor support is capable of handling either homogeneous or heterogeneous systems. The kernel automatically compensates for architectural differences (byte swapping, etc.) between processors. This allows a much easier transition from one processor family to another without a major system redesign.

Since the proposed standards are still in draft form, RTEMS cannot and does not claim compliance. However, the status of the standard is being carefully monitored to guarantee that RTEMS provides the functionality specified in the standard. Once approved, RTEMS will be made compliant.

This document is a detailed users guide for a functionally compliant real-time multiprocessor executive. It describes the user interface and run-time behavior of Release 4.10.99.0 of the C interface to RTEMS.

2. Overview

2.1. Introduction

RTEMS, Real-Time Executive for Multiprocessor Systems, is a real-time executive (kernel) which provides a high performance environment for embedded military applications including the following features:

  • multitasking capabilities

  • homogeneous and heterogeneous multiprocessor systems

  • event-driven, priority-based, preemptive scheduling

  • optional rate monotonic scheduling

  • intertask communication and synchronization

  • priority inheritance

  • responsive interrupt management

  • dynamic memory allocation

  • high level of user configurability

This manual describes the usage of RTEMS for applications written in the C programming language. Those implementation details that are processor dependent are provided in the Applications Supplement documents. A supplement document which addresses specific architectural issues that affect RTEMS is provided for each processor type that is supported.

2.2. Real-time Application Systems

Real-time application systems are a special class of computer applications. They have a complex set of characteristics that distinguish them from other software problems. Generally, they must adhere to more rigorous requirements. The correctness of the system depends not only on the results of computations, but also on the time at which the results are produced. The most important and complex characteristic of real-time application systems is that they must receive and respond to a set of external stimuli within rigid and critical time constraints referred to as deadlines. Systems can be buried by an avalanche of interdependent, asynchronous or cyclical event streams.

Deadlines can be further characterized as either hard or soft based upon the value of the results when produced after the deadline has passed. A deadline is hard if the results have no value or if their use will result in a catastrophic event. In contrast, results which are produced after a soft deadline may have some value.

Another distinguishing requirement of real-time application systems is the ability to coordinate or manage a large number of concurrent activities. Since software is a synchronous entity, this presents special problems. One instruction follows another in a repeating synchronous cycle. Even though mechanisms have been developed to allow for the processing of external asynchronous events, the software design efforts required to process and manage these events and tasks are growing more complicated.

The design process is complicated further by spreading this activity over a set of processors instead of a single processor. The challenges associated with designing and building real-time application systems become very complex when multiple processors are involved. New requirements such as interprocessor communication channels and global resources that must be shared between competing processors are introduced. The ramifications of multiple processors complicate each and every characteristic of a real-time system.

2.3. Real-time Executive

Fortunately, real-time operating systems or real-time executives serve as a cornerstone on which to build the application system. A real-time multitasking executive allows an application to be cast into a set of logical, autonomous processes or tasks which become quite manageable. Each task is internally synchronous, but different tasks execute independently, resulting in an asynchronous processing stream. Tasks can be dynamically paused for many reasons resulting in a different task being allowed to execute for a period of time. The executive also provides an interface to other system components such as interrupt handlers and device drivers. System components may request the executive to allocate and coordinate resources, and to wait for and trigger synchronizing conditions. The executive system calls effectively extend the CPU instruction set to support efficient multitasking. By causing tasks to travel through well-defined state transitions, system calls permit an application to demand-switch between tasks in response to real-time events.

By proper grouping of responses to stimuli into separate tasks, a system can now asynchronously switch between independent streams of execution, directly responding to external stimuli as they occur. This allows the system design to meet critical performance specifications which are typically measured by guaranteed response time and transaction throughput. The multiprocessor extensions of RTEMS provide the features necessary to manage the extra requirements introduced by a system distributed across several processors. It removes the physical barriers of processor boundaries from the world of the system designer, enabling more critical aspects of the system to receive the required attention. Such a system, based on an efficient real-time, multiprocessor executive, is a more realistic model of the outside world or environment for which it is designed. As a result, the system will always be more logical, efficient, and reliable.

By using the directives provided by RTEMS, the real-time applications developer is freed from the problem of controlling and synchronizing multiple tasks and processors. In addition, one need not develop, test, debug, and document routines to manage memory, pass messages, or provide mutual exclusion. The developer is then able to concentrate solely on the application. By using standard software components, the time and cost required to develop sophisticated real-time applications is significantly reduced.

2.4. RTEMS Application Architecture

One important design goal of RTEMS was to provide a bridge between two critical layers of typical real-time systems. As shown in the following figure, RTEMS serves as a buffer between the project dependent application code and the target hardware. Most hardware dependencies for real-time applications can be localized to the low level device drivers.

RTEMS Application Architecture

The RTEMS I/O interface manager provides an efficient tool for incorporating these hardware dependencies into the system while simultaneously providing a general mechanism to the application code that accesses them. A well designed real-time system can benefit from this architecture by building a rich library of standard application components which can be used repeatedly in other real-time projects.

2.5. RTEMS Internal Architecture

RTEMS can be viewed as a set of layered components that work in harmony to provide a set of services to a real-time application system. The executive interface presented to the application is formed by grouping directives into logical sets called resource managers. Functions utilized by multiple managers such as scheduling, dispatching, and object management are provided in the executive core. The executive core depends on a small set of CPU dependent routines. Together these components provide a powerful run time environment that promotes the development of efficient real-time application systems. The following figure illustrates this organization: