Assembly-Level Pre-injection Analysis for Improving Fault Injection Efficiency

Light-Weight Techniques for Improving the Controllability and Efficiency of ISA-Level Fault Injection Tools

“…In most tools the location and the time for fault injection are chosen randomly from the complete faultspace [23], which is typically extremely large. The statistical implication of this is that the cost of obtaining appropriate confidence levels of the dependability measures becomes unnecessarily high.…”

“…One way of injecting faults in a more efficient way is to use Barbosa et al's pre-injection analysis [23] to avoid injecting errors into "dead" 1 registers or memory locations in order to decrease the time it takes to run a fault injection campaign. This analysis is called pre-injection analysis as it uses knowledge of program flow and resource usage to choose the location and time where faults should be injected, before any experiment is performed.…”

“…According to Barbosa et al [23], the pre-injection analysis reduced the fault space two orders of magnitude for the registers and four to five orders of magnitude for the memory locations. Tools such as GOOFI-2 [12] and FAIL* [8] use the inject-on-read fault injection technique.…”

“…The first one places the fault injection code in exception-handling routines intended for debugging, while the second one injects faults by instrumenting the executable file with fault injection code before it is downloaded to the target system. GOOFI-2 uses a pre-injection analysis [23] to implement an ISA-level fault injection technique known as inject-onread that selects the time of the injection so that errors are only injected in a register or a memory word immediately before it is read. In other words, all faults targeting a specific bit of a given register or memory word, from the time the register or the memory word is written into until it is read, are considered equivalent.…”

“…As mentioned in Section 2, GOOFI-2 uses a preinjection analysis [23] that selects the time of the injection so that errors are only injected in a register or a memory word immediately before it is read. Even though the pre-injection analysis would considerably reduce the fault space, the faultspace is still quite large when targeting the complete execution of programs under test.…”