CONCURRENT DETECTION OF SOFTWARE AND HARDWARE DATA-ACCESS FAULTS

A new approach allows low-cost concurrent detection of two important t ypes of faults, software and hardware data-access faults, using an ext ension of the existing signature monitoring approach. The proposed app roach detects data-access faults using a new type of redundant data st ructure that contains an embedded signature. Low-cost fault detection is achieved using simple architecture support and compiler support tha t exploit natural redundancies in the data structures, in the instruct ion set architecture, and in the data-access mechanism. The software d ata-access faults that the approach can detect include faults that hav e been shown to cause a high percentage of system failures. Hardware d ata-access faults that occur in all levels of the data-memory hierarch y are also detectable, including faults in the register file, the data cache, the data-cache TLB, the memory address and data buses, etc. Be nchmark results for the MIPS R3000 processor executing code scheduled by a modified GNU C Compiler show that the new approach can concurrent ly check a high percentage of data accesses, while causing little perf ormance overhead and little memory overhead.