Desenvolvimento de um I-IP para o monitoramento da atividade do sistema operacional em processadores multinúcleos

Abstract

The use of Real-Time Operating System (RTOS) became an attractive solution to design safety-critical real-time embedded systems. At the same time, we enthusiasti-cally observe the widespread use of multicore processors in an endless list of our daily applications. It is also a common agreement the increasing market pressure to reduce power consumption under which these embedded, portable systems have to operate. As the major consequence, these systems are becoming more and more sensitive to transi-ent faults originated from a large spectrum of noisy sources such as conducted and radi-ated Electromagnetic Interference (EMI) and ionizing radiation (single-event effect: SEE and total-ionizing dose: TID). Therefore, the system s reliability degrades. In this work, we discuss the development and validation of an Infrastructure-Intellectual Prop-erty (I-IP) able to monitor the RTOS activity in a multicore processor system-on-chip. The final goal is to detect faults that corrupt the task scheduling process in embedded systems based on preemptive RTOS. Examples of such faults could be those that pre-vent the processor from attending an interruption of higher priority, tasks that are strict-ly allocated to run on a given core, but are running on another one, or even the execu-tion of low-priority tasks that are passed over high-priority ones in the ready-task list maintained on-the-fly by the RTOS. This I-IP, namely RTOS-Watchdog, was described in VHDL and is connected to each of the processor CPU-Addresses busses. The RTOS Watchdog has a parameterizable interface to easily fit any processor bus. A case-study based on a multicore processor running different test programs under the control of a typical preemptive RTOS was implemented. The case-study was prototyped in a Xilinx Virtex4 FPGA mounted on a dedicated platform (board plus con-trol software) fully developed at the Computing Signals & Systems Group (SiSC) [1] of the Catholic University (PUCRS). For validation, the whole system was exposed to combined effects of EMI and TID. Such experiments were performed in several steps, part of them carried out at PUCRS, Brazil, and part at the Instituto Nacional de Tecnología Industrial (INTI) and Centro Atómico, both located in the city of Buenos Aires, Argentina. The obtained results demonstrate that the proposed approach provides higher fault coverage and reduced fault latency when compared to the native fault detec-tion mechanisms embedded in the kernel of the RTOS.