Bureaucrats, dave_user, Administrators
4,650
edits
Changes
→Introduction
One of the most challenging problems related to embedded Linux systems is the so called post-mortem analysis. Post-mortem is a Latin expression that means "after death". In this context, death is meant as an event after which the system becomes unstable or even gets stuck. Therefore, post-mortem analysis refers to the tasks carried out after the occurrence of such an event to figure out its root cause
Even worse, post-mortem analyses are yet harder when these events occur randomly and it is apparently impossible to trigger them in a controlled fashion. SometimesIn spite of thorough testing at qualification stage, unfortunately, these situations may even occur when the system has already been deployed on the field and is used by end customers making the analysis amazingly troublesome.
Several techniques are available for post-mortem analysis. Software tools, hardware tools, or a combination of both can be leveraged. This article is the first of a series of Technical Notes (TN) describing in more details some of these techniques. Some Interestingly, some TN's refer to real-world cases in which DAVE Embedded Systems put in field deployed its expertise to support several customers reporting on-field failures they were able unable to analyze with traditional debugging tools. It is worth remembering that the analysis described in this series generally made no assumption about the root cause domainand approaches. In other wordsthese cases, often information reported by customers were are necessarily so limited and fragmented that was is impossible to determine a priori if the root cause was is software or hardware. Thus, no assumption about the root cause domain can be made and engineers need to be very open-minded to consider every possible cause.
