-
Notifications
You must be signed in to change notification settings - Fork 8
Minutes 2 March 2023
Paul Albertella edited this page Mar 2, 2023
·
1 revision
2023-03-02 Host: Paul Albertella Participants: Pete Brink, Kent Nelson, Raffaele Giannessi, Gabriele Paoloni, Elana Copperman Agenda: Discuss use of STPA
Goal of using STPA:
- Identify the negative outcomes that we want to avoid (losses)
- Identify the system / environment conditions that can lead to these losses
- By environment we mean things outside the system that either interact with it or
- Identify the (subset) of functionality provided by Linux that is pertinent to the safety goals that we are trying to achieve
- Help us define safety requirements for Linux and for the other components of the system that it interacts with
Issues:
- Pete: There may be things within Linux that don’t do what we need them to do, which may interfere with our objectives
- Paul: Goal is to first identify what is directly relevant to the safety goals, and then form this, identify what else might interfere with that
Control hierarchy approach
- Focus on controlled process that we care about with respect to safety
- Identify how that process is controlled to avoid or mitigate the losses
- Identify what criteria (constraints) need to be satisfied in order to avoid or mitigate the losses
- Understand what role our software has in:
- controlling that process, or
- controlling intermediate processes that control it
- enabling other controllers or processes to control it, or
- providing the means for controllers and controlled processes to receive information (feedback) that they need to perform their specified role
- Control hierarchy is likely to involve many cooperating controllers and controlled processes
- The process that is ultimately connected with safety (this thing that can hurt people) is unlikely to be directly connected to our component
- Control structure helps us to understand how (and through which other components of the system) we interact, and what our responsibilities are
Loss scenarios
- What can lead to a loss for a specific control action in a specific context
- i.e. for a particular pairing of controller and controlled process as part of the larger control hierarchy
- See diagram in STPA guidance document
What to look at next in OSEP?
- Stack memory
- CPU scheduling
- This is a popular choice
- Other hardware interaction (e.g. i2c, ethernet)